• We already know that Brexit will have no impact on the City’s core wholesale business. This piece explains why, despite this, some banks have announced Brexit related job losses and why those announced losses are so trivial. The fascinating answer is found in an analysis of the relatively dry subject of UK bank lending to EU customers.
  • Deposit taking from and financial product sales to certain EU customers from the UK require an EU financial services passport. Some commentators have circulated sensationalist estimates of the income and jobs that will be lost if, post-Brexit, UK located banks lose their ability to use this passport to provide such services to EU customers.
  • Thankfully, very detailed data on the cross-border financial claims of banks located in the UK are available from the Bank for International Settlements (BIS), as well as the financial statements and other filings of individual banks with UK operations, which enable us to quantify this risk.
  • Sadly, the scaremongering commentators seem to have made scant use of these data, which flatly contradict their alarmist stories.
  • Oliver Wyman’s high profile estimate of £23-27bn revenues at risk from Brexit in this area shrinks to £7.5bn when adjusted for activities, instruments and clients not requiring EU authorisation. Most of these £7.5bn require minimal operational adjustments to continue to be earned in the UK.
  • Contrary to the assumptions circulated by many in the mainstream media, BIS data show no discernible impact from the introduction of the EU financial services passport on the claims of banks located in the UK on EU customers.
  • Our analysis does however reveal that a few banks in the UK do have small EU retail banking activities which are affected by Brexit. These retail operations fully explain the small job losses attributed to Brexit.
  • Critically, Brexit related job relocation announcements so far already assume a worst case loss of passporting scenario. Contrary to ill-informed press commentary, the trivial job losses (<1% of financial services jobs in London) from these peripheral retail operations represent the full impact of Brexit on the City.


Executive Summary

  • Management consultants Oliver Wyman (OW) estimate that UK financial services firms earn an eye catching £23-27bn from “wholesale business related to the EU” in their “banking” activities.
  • That’s about 20% of OW’s estimate of total Banking Revenue earned by firms in the UK, leading many to extrapolate from that figure to a now common assumption that 20% of the City of London’s revenues are threatened by Brexit. This post demonstrates that this figure is a wild exaggeration.
  • OW includes a broad range of activities in its definition of “banking”: Sales and trading, Investment banking, Retail and business banking and Private banking and wealth management. We can use OW’s figures to estimate the following EU related revenue breakdown for these different activities (the £25bn “Total Banking Revenue” figure is simply the mid-point of OW’s £23-27bn range):

EU related revenues by segment

  • In Retail and business banking, although there is no restriction preventing non-EU banks from lending to EU customers, there is on their taking deposits from them. Except in the case of large companies, most banks will only be confident in lending to customers if they can also hold their deposit. Therefore, the de jure restriction on deposit taking constitutes a de facto restriction on lending to such EU customers.
  • Most Retail and business banking fees are derived from products like Swaps whose sale constitutes financial advice, which can only be provided to EU SME customers from within the EU or by companies holding an EU financial services passport. Private banking, similarly, constitutes financial advice to individual customers. These revenue streams are therefore also at risk from Brexit.
  • Wholesale Sales and trading, by contrast, is carried out with banks and institutional investors, while Investment banking (which is also commonly referred to as “corporate finance”) is typically advice or assistance, on issues such as capital raising or mergers and acquisitions, given to companies in connection with their ordinary course of business. Both services can already, as things stand, be provided to EU customers by third country providers, and neither requires any EU authorisation, reducing the estimated Brexit impact by £10bn:

activity reduction

  • Retail and business banking revenue is constituted by net interest income (NII) and fee income. NII depends on two things: interest bearing assets (IBAs) and the net interest margin (NIM) earned on those assets. Fee income can, similarly, be estimated based on customer assets and the fee income margin expressed as a percentage of those assets.
  • Competitive pressures mean that most banks will earn fairly similar returns on their assets. We use the UK headquartered banks’ NIM and fee income as a percentage of customer loans as representative of the returns that most banks will earn on their EU assets:


  • As the table above shows, IBAs consist of loans to customers, loans to banks and various kinds of securities, mainly bonds. Only loans to EU customers are potentially affected by Brexit, since EU banks are free to borrow from banks across the world, and EU securities are freely traded by investors across the world.
  • By quantifying the different EU IBAs of UK located banks and applying our NIM and fee income percentages we can arrive at an accurate estimate of the Retail and business banking revenue at risk from Brexit.
  • OW only uses company reports and proprietary data to build a bottom up estimate of Retail and business banking income at risk from Brexit. We, in addition, use the detailed, comprehensive top down data from the Bank for International Settlements (BIS), which offers data on the EU IBAs of banks headquartered and located in the UK:

headquartered versus located

  • The BIS A.6 reports for claims of banks located in the UK are not consolidated and therefore include intragroup loans from a bank’s offices in the UK to its offices in the EU. They also, helpfully, break the claims they record down into loans to customers, loans to banks and securities, which enables us to identify which assets are at risk from Brexit. They give us a claims figure for all UK located banks on EU customers of $1,663bn at the end of 2015. OW’s report, written in 2016, based its estimates on 2015 data. Claims on EU entities of UK located banks have been declining steadily in dollar terms since 2015.
  • The BIS B.4 reports for claims of banks headquartered in the UK are consolidated, but not broken down by instrument. They give us a claims figure for all UK headquartered banks on EU customers of $641bn at the end of 2015. With a few small exceptions (discussed below), the $641bn recorded for UK headquartered banks in the BIS B.4 reports will also be included in the $1,663bn recorded in the BIS A.6 reports (if it’s headquartered in the UK, it’s usually also located in the UK). Subtracting the ~$600bn UK headquartered claims from the ~$1.6tn UK located claims leaves us with ~1 trillion in claims held by foreign banks located in the UK.
  • We can compare BIS claims data with individual bank disclosures of their EU exposure in their reports and accounts and other filings, in order to estimate the weight of UK, US and other banks in the total exposure of UK located banks to the EU. Quantifying the EU assets of the bigger banks – which make up the majority of the BIS’ aggregate asset figure – also helps us assess the risk to those assets from Brexit, based on those individual banks’ particular circumstances.
  • Exposure is similar to claims but there are important differences. Loan claims are stated pre-credit provisions by the BIS, while loan exposure is stated post-credit provisions by individual banks. BIS securities claims include equities whereas individual banks’ statements of EU securities exposure exclude equities:

Claims versus exposure

  • All four UK headquartered banks give explicit disclosure on their exposure to loans to EU customers. These four banks on their own account for just over a third of the total UK located claims on EU clients recorded by the BIS. Lloyds, RBS and Barclays also give detailed disclosure on their exposure to other European IBAs, whereas for HSBC we need to make a few reasonable assumptions for the bottom up exposure numbers from the four banks’ annual reports to reconcile with the $539bn exposure number implied (in the second line of the table above) by the BIS’ top down claims figure of $641bn:

EU exposure banks BIS complete

  • This results in a provisional estimate of £4.4bn banking revenue at risk from Brexit earned by the four UK headquartered banks on a consolidated basis, as demonstrated in the following table. The aggregate 3.13% return on loans to customers will be used to estimate the income at risk from Brexit earned from the EU assets of the other banks analysed in this report:

UK bank EU income

  • The biggest contributor to the remaining ~$1 trillion of claims on EU entities, held by foreign banks located in the UK, are the big five US banks. They give detailed disclosure on their EU assets, albeit less detailed than the four UK headquartered banks. A few reasonable assumptions (explained in great detail in the report) lead to an estimate of $545bn EU IBAs (versus $539bn for the UK headquartered banks) including $145bn loans to customers (versus $209bn for the UK headquartered banks):

Top 5 EU exposure breakdown adjusted total

  • Two of the US banks give useful disclosure on their exposure to non-bank financials or NBFIs, which include, among others, insurers, brokers and fund managers. EU NBFIs can borrow from UK located lenders without a passport, since they are financial companies themselves, and therefore loans to NBFIs are not impacted by Brexit. This leads to an estimate of £1.8bn in revenue at risk from Brexit for these five US banks:

US bank adjusted loans to customers

US income estimate

  • Combining BIS data with the data on the nine US and UK banks analysed reveals that just nine UK and US banks represent over two thirds of UK located claims on EU clients. The remaining $472bn of exposure, just under a third of the total, is held by other foreign banks located in the UK, including, in the main, Swiss, Japanese, Canadian, Middle Eastern and Asian banks:

Other countries exposure

  • What interests us in these UK located “other banks'”$472bn of EU exposure is their EU customer loans. The BIS A.6 data’s breakdown of claims by UK located banks on EU entities by type of claim allows us to estimate the customer loans contained within the other UK located banks’ $472bn EU exposure. Loans to customers, which the BIS refers to as “Loans and deposits, non-banks,” are displayed on the second line of the following table. The BIS data, which OW surprisingly chose not to use, give us a figure for those loans which is more precise and comprehensive than any survey which OW or any of its competitors could perform.  Subtracting the UK and US loans to EU customers, calculated above, from the BIS’ total EU loans figure, below, allows us to begin to estimate the EU loan exposure contained in the “other banks'” total $472bn EU exposure:

claims exposure total

  • The BIS also gives the following country breakdown of these $482.5bn loans and deposits, non-bank claims. Surprisingly, the top three countries for UK located bank loans to EU customers are the Netherlands, Luxembourg and Ireland, which represent over half of all such loans. The only explanation for this anomaly is that a significant amount of the lending by UK located banks to EU customers is to tax driven structures, typically held by large companies to whom UK located banks can lend without EU membership. OW missed this crucial point because it didn’t use the BIS data:

UK nonbanks

  • The A.6 reports’ breakdown of EU claims by type of claim also allows us to better reconcile EU securities exposure reported by individual banks with the BIS figures. The low number for securities in the BIS A.6 report ($342bn claims implying $239bn exposure, as detailed above) clearly indicates that foreign banks located in the UK hold most of their EU securities in their headquarters and not their UK offices, meaning adjustments need to be made to the exposure numbers in the UK  headquartered and particularly the US individual company reports to make them comparable with the A.6 data:

securities adjusted

  • In addition, $51bn of the loans to EU customers in UK headquartered bank disclosures are made by foreign subsidiaries such as HSBC France and will therefore not be included in the BIS’ A.6 report of loans to EU customers by UK located banks. This reduces the UK headquartered banks’ loans to EU customers exposure recorded in the BIS A.6 reports from the consolidated figure of $209bn, calculated above, to $158bn.
  • A further, important adjustment needs to be made for the intragroup loan figures in the A.6 reports. Loans from UK located banks to their EU offices are recorded as cross-border intragroup loans to banks by the BIS A.6 report, while the loans to EU customers or other use of funds by those EU offices are recorded in the BIS’ domestic data.
  • Based on detailed BIS disclosure we estimate there were $435bn of such intragroup loans from UK to EU banking offices at the end of 2015, some of which should be reclassified as loans to customers, increasing the Brexit impact estimate.
  • This results in the following reconciliation of BIS and individual company data:

comparable BIS and company accounts

  • Estimating how much of the intragroup loans are used to make loans to customers is a critical final step required to calculate the total value of EU IBAs at risk from Brexit. To help frame this analysis, we examine the EU exposure of ten large Swiss, Japanese and Canadian banks, as well as Wells Fargo. We estimate their EU loan exposure at $89.5bn, based on detailed analysis of their consolidated reports and accounts, Pillar III disclosures (in the case of the Swiss banks) and Companies House filings (in the case of Sumitomo Banking Corporation Europe):

other countries individual banks

  • We combine this information with the data from the UK and five US banks, detailed above, in the chart below. It demonstrates that only ten banks of the twenty analysed (which are the largest non-EU headquartered banks in terms of EU exposure) have EU loan exposure above $10bn, and that most banks outside the top ten have EU loans of $1-5bn. Our examination of a sample of smaller banks outside of these twenty banks confirms that none of them have more than $1bn in loans to EU customers. Effectively, we are confronted with a “tail” of small loans outside the biggest banks:

pareto chart all loan to customers in decreasing size

  • The amount of intragroup loans which are ultimately used to make loans to EU customers will decide the estimated length of that tail of small loans. Intragroup loans include loans to subsidiaries and loans to branches. Intragroup loans to subsidiaries fund the whole subsidiary’s balance sheet, not just its loans to customers. Intragroup loans to branches, however, will only be used to fund loans to customers.
  • Although the EU financial services passport makes it possible to establish a branch in another EU country without the need to establish a subsidiary, and subsidiaries increase capital and buffer liquidity requirements for banks, there are many advantages to subsidiaries, in particular shielding the rest of the banking group from problems in any of their subsidiaries. In addition, these subsidiaries and the financial payment networks they command were often acquired or built up long before the EU financial services passport was even dreamed of; banks’ attitude is “if it ain’t broke, don’t fix it,” leading to persistent use of subsidiaries.
  • Data from Allen and Gu (2011) and individual bank disclosures confirm the continued use of an often complex network of subsidiaries, with subsidiaries often simultaneously borrowing from and lending to other parts of the same banking group. On the basis of all this information we assume that two thirds of EU intragroup lending is to subsidiaries with the balance to branches, leading to an estimate that $231bn, or just over half, of the $435bn cross-border intragroup loans are used to make loans to customers and picked up as such in the BIS’ domestic loans to customers data:

Intragroup calc

  • This $231bn claims figure is equal to $229bn exposure (after deduction of credit provisions as described above) and results in the following increased estimate of UK located bank loan exposure to EU customers:

Remaining claims and inter company loans adjusted

  • This in turn allows us to present a final, comprehensive picture of UK located claims on and exposure to EU entities (total exposure is $1,554bn rather than $1,556bn in the table above because of the $2bn credit provisions on the $231bn intragroup loans to banks claims reclassified as loans to customers). The key figure is the “Loans and deposits, non-banks” adjusted exposure figure of $707bn (on the second line of the following table), which you can reconcile with the chart above by adding the original $478bn of loans, non-bank (in dark green) to the $229bn intragroup loans to banks reclassified as loans to customers (in light green):

adjusted BIS claims

  • The implication of this reclassification is that there is a very long tail of around 237 UK located banks with small loan exposure to EU customers, averaging $1.3bn. This is fully consistent with the tail of small loans, observed in the chart above, in the banks outside the top ten in terms of EU exposure:

whole market distribution

  • This tail essentially consists, we believe, of European banks which booked some of their loans in the UK, and emerging markets banks lending to EU subsidiaries of their domestic customers. Neither of these groups of banks’ loans are at risk from Brexit since the EU banks have EU subsidiaries and other structures necessary to keep servicing EU customers and the emerging markets banks are not even dealing with EU customers in the first place.
  • This allows us to complete the picture of loans to EU non-financial customers and NBFIs by the different UK located banks, leading to an estimate of £10.4bn income at risk from Brexit (the figure for the UK headquartered banks has reduced from the £4.4bn provisional estimate above to £2.7bn, due to the adjustments, described above, for the $51bn non-UK located loans as well as the $31bn loans to NBFIs, estimated in this table using the BIS A.6 data):

Total exposure income

  • This is just over two thirds of the OW £14bn Retail and business banking estimate and leads to an estimate of £11.4bn of Banking Revenue at risk (£10.4bn Retail and business banking plus £1bn Private banking), less than half of OW’s original £25bn estimate:

exposure reduction

  • This estimate includes NII and fees earned from larger companies. Banks are happy to lend to larger companies without holding their deposit. Larger companies also have professional finance departments and do not require the same investor protection as individuals or SMEs, meaning banks do not have to be authorised in the EU to sell them products like Swaps. Banking revenue from such large companies is therefore not at risk from Brexit.
  • Using company disclosures to estimate loans to individual EU customers by UK located banks (mainly personal mortgages), central bank data on the share of larger companies as a percentage of total loans to corporates and BIS data to estimate the tax-driven loans to subsidiaries of foreign companies in the Netherlands, Luxembourg and Ireland we can arrive at the following picture (the $493bn total loans to non-financial customers figure on the first line of the following table is also found in the previous table):

size effect

  • Our estimate of Retail and business banking income at risk from Brexit has shrunk to £6.5bn, under half of the £14bn estimated by OW. OW’s initial £25bn estimate of Banking Revenues at risk from a loss of passport has been compressed to £7.5bn, well under one third of the original:

final reduction

  • The root of OW’s mistake was its vague reference to “EU related” revenues. We speculate that OW probably asked the banks with which it has relationships to tell them what their EU related assets were and multiplied that number by an asset margin. Our detailed analysis shows that a significant proportion of “EU related” Banking Revenues does not require EU membership.
  • Even this lower number is very likely overstated. Of the £10.4bn revenue impact estimated before adjusting for larger company exposure, £4.6bn was attributable to the tail of smaller banks outside the top twenty in terms of EU exposure, most of which is probably attributable to European and emerging markets banks whose EU revenue is not impacted by Brexit. Because these banks do not disclose this exposure separately (it will be classified as exposure to the relevant EU country by the EU banks and as domestic exposure by the emerging markets banks, regardless of whether it is booked in London or elsewhere) we cannot quantify it. Nor can we determine how much of it is to larger companies or tax structures. Out of prudence, we have therefore fully included it in our estimates. However, if this tail of banks largely consisted, as we suspect, of EU and emerging markets banks, and a similar proportion of the £6.5bn Retail and business banking income – adjusted for larger company exposure – at risk from Brexit were attributable to this tail of banks as was the case with the £10.4bn – not adjusted for larger company exposure – the Brexit impact would reduce by 44% (£4.6bn/£10.4bn) to £3.6bn and the total Brexit impact on Banking Revenue would be a trifling £4.6bn.
  • Whatever the number is, it overstates the impact on UK located banks from Brexit. To keep any EU related NII at risk from Brexit, all the banks need is a subsidiary through which they can accept EU deposits. Many UK located banks already have such fully capitalised subsidiaries, and the value of the deposits banks need to hold is only a few percentage points of the loans they write. The aggregate incremental loss of annual revenue from the equity required to fully capitalise these subsidiaries is a derisory £12m. As for the fee income, a large proportion of it is already generated by salespeople located in the EU, near their customers.
  • In terms of jobs, the only employees who need to move to or be hired in the EU are the staff required to maintain any new banking subsidiary, as well as any staff who happen to provide financial advice to EU customers from the UK. Not many staff are required to set up a subsidiary, and most of the relocation announcements from UK located banks have been small because not many of their employees provide financial advice to EU retail customers. It seems that such cross-border provision of retail financial advice is the exception, not the rule:

job losses

  • The assumption, commonly cited in the media, that 20% of City activities are at risk from Brexit seems to have been influenced by job relocation announcements by HSBC and UBS in January 2017, amounting to 1,000 jobs each or 20% of their workforce. UBS later revised that figure to 250, confirming in its announcement that only staff providing financial advice needed to move. Deutsche, JP Morgan, BNP Paribas and Société Générale also seem to be relocating a bigger number and/or higher percentage of their workforce:

job losses HSBC et al actual

  • The higher relocation figures announced by these banks is in part due to their larger exposure to private banking and their combination of a substantial book of EU SME lending with financial product engineering expertise. However, it is also due, we believe, to their active decision to create specialist product desks to sell financial services to their EU SME and private banking customers from London, rather than through local teams. These banks’ higher job loss announcements are due to the idiosyncratic nature of their book of business and the way they have centralised their product offering in London. It is not representative of the fate awaiting the City as a whole, contrary to the consensus assumption.
  • Unlike the banks for which job losses are reported in the tables above, most City firms don’t have EU retail exposure (the banks announcing Brexit related job losses only represent ~8% of the roughly 1m financial services jobs in London). The ~4,000 job relocations detailed in these tables represent less than 1% of total financial services jobs in London.
  • Commentators disappointed by the low level of job losses so far (such as EY and Reuters) have argued that they are merely “the tip of the iceberg.” But all the larger and many of the smaller job loss announcements so far include both subsidiary maintenance and trading jobs, in other words the only jobs which need to move because of Brexit. Indeed, the job loss announcements state clearly that they are made on the assumption of a worst case scenario in which UK located firms lose access to the EU financial services passport. Therefore, contrary to what those commentators have speculated, these announcements represent the maximum Brexit related job losses to be suffered by the banks concerned.
  • Taking all these points together we can estimate that, combined, annual corporate and income tax loss due to Brexit could amount to a paltry £405m (~4% of the UK’s annual net post-rebate contribution to the EU’s budget).
  • Looking at the BIS data over history, we do see a substantial increase in the claims of UK located banks on EU entities following the introduction of EU legislation seeking to harmonise the provision of financial services. Some might take this as evidence that the financial services passport has actually had a tangible impact. However, we see a very similar increase in total cross-border claims across the globe at the same time. That can’t be because of the passport.
  • In fact, the fairly synchronised growth in both UK located claims on the EU and total cross-border claims across the globe are driven by global macro factors, particularly China’s accession to the WTO at the end of 2001 and the global financial crisis in 2008. As a percentage of total cross-border claims, UK claims on EU entities have been declining since the series began (as shown in the title chart), and show no discernible influence whatsoever from the introduction of any EU financial services legislation.
  • In theory, the financial services passport should have led to more branches and fewer subsidiaries, and to more lending from UK located banks to EU customers. In practice, banks usually find a way to make the loans for which their business model is designed despite obstacles from the regulatory environment, and many other factors influence their business decisions.
  • Using the UK market as an example, Sweden’s Handelsbanken has grown there thanks to the passport, but the Benelux’s ING quit the UK despite benefiting from the passport, whereas US-inspired Metro Bank has grown rapidly with no influence from the passport. In practice, business model and environment are much more important influences on whether or not a bank expands in a particular country than the EU financial services passport.
  • OW’s report purportedly focuses on wholesale financial services revenue, but its Brexit impact estimates include activities which – the clue is in the name – are retail activities: Retail and business banking and Private banking. This seems to be a rhetorical subterfuge by OW in response to the fact that nearly all of the City’s export business with Europe is wholesale business which doesn’t need EU authorisation, whereas the only financial services business requiring EU authorisation is of a retail nature and mainly domestic.
  • OW seems to have used this confusion to hide the fact that Brexit only impacts the very peripheral area of cross-border retail financial services provision, and to make it appear, falsely, that the City’s critical wholesale business is at risk from Brexit.
  • In fact, Article 63 of the EU’s constitution prohibits any restriction on the flow of wholesale finance outside the EU. Permitted or not, any such restriction would be disastrous for the EU. Behind OW’s rhetorical ploy lies a fundamental contradiction between the EU’s commitment to and vital dependence on the free movement of capital – and its misguided pretence that it can punish the UK financial services industry for Brexit.


£23-27bn! Wow!

On page 6 of its report “The Impact of the UK’s Exit from the EU on the UK Based Financial Services Sector” (reviewed in another post), Oliver Wyman (OW) estimates that UK financial services firms earn £23-27bn in income from “international and wholesale business related to the EU” in their “banking” activities:

23 to 27 chart

£23-27bn is a lot of money. The £23-27bn EU related revenues classified by OW as generated in the “banking” sector are larger than the EU related revenues of the other three sectors analysed by OW put together. Even the Eurosceptic Open Europe think tank, in its own more measured piece on the impact of Brexit on UK financial services, makes much of this figure (p 19):

Open Europe 23 to 27

Open Europe uses the £23-27bn figure as evidence for the commonly expressed assumption (cited elsewhere and discussed below in this piece) that “around 20% of [City] revenues are generated via the [EU financial services] passport.” Our other work on the (non-)impact of Brexit on the City would lead us to be extremely sceptical of this 20% figure. It is therefore important to investigate any figure, such as OW’s £23-27bn estimate, which seems to support it.

Unfortunately, OW doesn’t explain how it arrives at this estimate. However, luckily, there is a considerable amount of useful, publicly available data available on this subject:

  1. The Basel based Bank for International Settlements (BIS), an international bank owned by 60 central banks, has been collecting detailed data on the cross-border claims of banks throughout the world for decades. The BIS can tell you what the claims of banks located in Belgium were on entities in Botswana at the end of 2016 (for the record, the figure was $57bn). For the purpose of assessing any impact of Brexit on banking activities in the UK, it holds data for the claims on EU entities of both UK headquartered banks, such as Barclays, and, importantly, international banks transacted from their offices in the UK, such as loans to a German corporate by JP Morgan from its branch in Canary Wharf;
  2. UK headquartered banks’ annual reports and accounts, US banks’ SEC 10k reports and the annual reports and accounts of Swiss, Japanese, Canadian and other banks that transact a significant amount of international business from London and elsewhere in the UK, all disclose details regarding the geographical source of their revenues and location of their assets, which are helpful in determining any impact of Brexit on banking activities in the UK. These banks also often publish reports to the European Banking Authority and other regulatory disclosures such as Pillar III reports which sometimes contain relevant additional country data.

These data allows us to do three important things:

  1. Assess whether OW’s estimate of £23-27bn is reasonable;
  2. Assess how much EU related banking revenue in the UK, reasonably estimated, really requires an EU financial services passport to stay there;
  3. Assess how many jobs and how much tax and other income will be lost in the banking sector by the UK due to any removal of the EU financial services passport.


What is “banking” activity?

“Banking” activity can mean many things, so it is important to clarify the specific activities included by OW under this general term. OW (p 4) helpfully provides the following breakdown for total banking revenue (i.e. domestic revenue and all international revenue, including from the EU) earned by bank branches located in the UK (and therefore including the earnings of foreign banks with UK branches):

banking income.PNG

The first thing to note is that not all of what OW calls “banking activity” is what is commonly thought of as “banking,” namely money earned on loans and deposits. OW also includes “sales and trading,” “investment banking” and “private banking and wealth management.” This matters because each of these activities is impacted differently by Brexit (if at all).

Here are the revenue figures from OW for each sector of banking activity expressed as a percentage of total banking revenue. Where OW provides a range I have used the mid-point of that range:

banking revenues percentage of total

Note that, for reasons not explained by OW, the sum of the total revenue based on these mid-point values (£109bn) is almost at the very bottom of the range for total revenue (£108-117bn) given by them. If you add the figures at the bottom of the range for each individual activity you get £103bn – versus the £108bn given by OW as the bottom of the range for total revenues. And if you add those at the top of the range for each individual activity you get £115bn – versus the £117bn given by OW as the top of the range for its total revenue figure. So the range for total banking revenues based on the sum of the individual sectors should be £103-115bn, not  the £108-117bn printed in bold by OW. The “£108-117bn” total range is the headline figure that most readers will take away from that table (as did Open Europe). Without explanation, OW seems to have inflated that headline figure to make it more eye catching. That’s not a good sign.


Sector breakdown of Banking Revenues derived from EU customers

In their figure 3 and banking sectors table, reproduced above, we saw OW break down total banking revenue in two ways: by sector (Sales and trading, Investment banking etc.) and by region (Domestic, EU, International ex EU). What they don’t do is break down the regions by sector, for example telling you what proportion of EU Banking Revenue came from Sales and trading activities, what proportion came from Private banking, etc. We, however, need to estimate the sector composition of EU related revenues so we can estimate the potential impact of Brexit on each sector’s EU revenue stream. In the absence of more precise information from OW, the best we can do is to use the breakdown of total Banking Revenues by sector and apply it to EU Banking Revenue.

This is not necessarily entirely accurate because the breakdown of EU Banking Revenues by sector is not necessarily the same as the one OW gives for total Banking Revenues (and which we displayed above). Private banking and wealth management, for example, is a service that tends to be provided locally and is hard to “export”: if you want to become a client of London headquartered Fleming Stonehage you would typically open an account with them in London or one of their other eleven international offices. In other words, Private banking revenues will tend to be more domestic than those of other banking activities. Accordingly, you might expect only 1-2% of the UK’s EU related (and other international) Banking Revenues to be generated by Private banking and wealth management, compared to the 5% share estimated by OW for total (i.e. including domestic) Banking Revenue. However, given the fact that total Banking Revenue is dominated by two large sectors (Retail and business banking and Sales and trading combined represent 85% of total revenues), it is unlikely that the breakdown of EU Banking Revenue by sector will be sufficiently different from the breakdown for total Banking Revenue to have a big effect on the assessment of the impact of Brexit on EU Banking Revenue. So although this method isn’t exact, the impact of this lack of exactitude will not be significant.

This allows us to estimate the EU related revenue that banks located in the UK specifically earn in each sector of banking activity, according to OW:

EU related revenues by segment


The total revenue of £25bn at the bottom of the first column is the mid-point of OW’s £23-27bn range for EU related banking revenue in the chart above. Looking at Sales and trading to illustrate how this table works, 28% of total Banking Revenue is derived from Sales and trading (£30bn/£109bn), as shown in the previous table. Applying that specifically to EU related Banking Revenue gives you an estimate of £7bn EU related Sales and trading revenue (28% x £25bn).

Whether or not the figures we derive from OW in this way are an accurate estimate of the real banking revenues at risk from Brexit is the very question this post will try to answer.


Which of these activities can Brexit impact?

£25bn is only a figure for the income which might, potentially, be impacted by Brexit. To understand what the impact of Brexit is really likely to be, we need to look at each activity specifically, to understand how it is affected by EU Single Market membership, or what use it makes of the EU financial services passport, and what steps it might have to take in order to continue carrying out those activities which are dependent on EU authorisation.

Sales and trading: no impact from loss of passport or EU Single Market access

As discussed elsewhere, wholesale sales and trading, where a fund manager or a bank is the customer, does not require a European financial services passport. If French fund manager Carmignac can trade futures in Chicago or equities in Taipei, how on earth is the EU going to stop them trading equities or Swaps in London? Therefore, none of these £7bn estimated revenues requires any passport. The impact of losing the passport on this segment of wholesale banking activity is nil.

Investment banking: negligible impact from loss of passport or EU Single Market access

Investment banking, or corporate finance, is typically provided to larger companies. A small chain of opticians looking to acquire a group of optical practices in North West France is going to use their accountant to provide a valuation, rather than Goldman Sachs in London. If you are a small tech start up, you will typically use friends and family, local venture capitalists and crowd funding to start your business. It is only after that business reaches a certain size that you will float a portion of it on London’s Alternative Investment Market (AIM). If a small Spanish biotech company wants to float on AIM, Citigroup doesn’t need a passport to carry out the flotation. In fact, companies from as far as China and Russia are frequently floated on the London Stock Exchange, including small companies on AIM. There are plenty of US businesses with quotations on European exchanges and plenty of European businesses quoted on Nasdaq or the NYSE. EU membership is completely irrelevant in all of this.

The only case in which investment banking services might require a passport is if they were deemed to constitute investment advice. Investment advice to an EU retail entity is a regulated activity which has to be carried out from within the EU and requires authorisation in the customer’s jurisdiction, either via the passport or establishment of an authorised subsidiary in that jurisdiction. However, if an investment bank presents a company with a potential acquisition in their own sector (e.g. a chain of optical practices buying another chain of optical practices), that would not be considered investment advice but information regarding a transaction that forms part of that company’s ordinary course of business. Virtually all transactions in which investment banking practitioners are involved are such “ordinary course of business” company to company transactions. Even in the case of an an investment bank presenting an individual owner of a group of optical practices with the potential acquisition of another chain of optical practices, that too would generally not be deemed to be investment advice but information regarding a transaction that forms part of that individual’s ordinary course of business. And even when corporations or very wealthy individuals like Rupert Murdoch diversify, conglomerate style, into different industries, their size and knowledge of their own industry is enough for the investment bank’s help with the transaction not to constitute investment advice.

It is very rare for an inexperienced individual or small company to have the means or the appetite to invest large sums in a business they don’t understand, in which case the investment bank presenting them with the opportunity might be considered to be providing them with investment advice. Even then, practices such as Angel Investing are designed to make such transactions possible. Angel Investing generally involves people who have made money in one business investing in another. The way it works though is that investors register on an Angel Investment platform where they self-certify as able to invest on the platform and agree to “caveat emptor” terms and conditions according to which they accept that the platform is not providing them with investment advice. The companies which seek to raise funds on the platform post investment pitches on it according to a predetermined template which ensures a minimum level of information. It is then up to the investors to search on the platform for investments they might like and, if they so choose, to approach the company seeking to raise the funds. The crucial point is that the investor has taken the decision to approach the company. That means that the material provided on the platform by the company seeking to raise funds is not investment advice.

Investment bankers will sometimes act for the seller in an Angel Investment transaction, particularly the larger ones. But the material they provide will not constitute investment advice because of the way in which contact was initiated by the potential investor. Alternatively, if an investment bank is retained by a buyer to value a potential acquisition, that is not investment advice either because the buyer has expressed an interest in the business already and asked the investment bank for an opinion to help them make a decision. I have advised on three such deals, on behalf of both sellers and buyers, one of which resulted in an investment, without any need for regulatory authorisation on my part. To take the most extreme case, if an investment bank were to be confronted with a potential buyer for a business they were selling who was very wealthy but clueless (a lottery winner or an heir to a fortune, say), they could sell him (or her) the business without the authorisation to provide investment advice by asking her (or him) to conduct negotiations through a local accountant of their choice, which would be the entity providing the regulated investment advice.

The shares or other financial instruments routinely issue to raise capital for companies are sometimes available to individual retail investors. UK located investment banks might not be able to issue securities to individual retail investors in the EU post-Brexit. However, the issues with retail components are infrequent and the retail component of any issue is typically very small; most of the paper is taken by institutional investors. In the even smaller sub-set of issues to retail investors in the EU, most such issues are carried out by a consortium of banks, one of which has a retail presence in the country where the shares or other instruments are going to be listed or traded. The authorisation to market the issue to individual retail investors will be secured by that local member of the consortium.

In short, the impact of losing the passport on this segment of wholesale banking activity is negligible.

Private banking and wealth management: requires the passport

Private banking is a service provided to high net worth individuals. Although these individuals will often fall under the category of eligible party, meaning a financially experienced individual to whom more complicated financial products can be sold, they are still individuals. Therefore, such people, if based in the EU, could not receive financial advice from London without a passport. Moreover, even if the private bank had a subsidiary in the EU, the person providing the advice might have to provide it from within the EU in order to be eligible to provide that advice to the EU client. However, as we have discussed elsewhere, such high net worth individuals are very mobile. They are the sort of person who hops on a plane to open an account in Geneva or Monte Carlo. That person can very easily hop on a plane to the UK in order to have their personal affairs managed there. If they have enough money they can even establish a family office as an independent entity in London to manage their wealth. Of course, that is an extra inconvenience that they wouldn’t have had before, but with it obviously comes the advantage of having their money offshore, away from the European regulators. Quantifying how much money is managed in the UK for such “flexible” customers is difficult and so we will simply assume that all private banking and wealth management business carried out for EU clients from the UK requires the passport. Because this is (as explained above) already a highly localised activity, the estimate of £1bn in revenues earned from EU clients in this sector is probably a little high in any case.

Retail and business banking: requires a passport except for large companies

Banking is a highly regulated activity. As discussed elsewhere, banks are entrusted with other people’s money. They lend out what they borrow and are therefore highly leveraged. They are also vital to the functioning of the economy through their provision of credit. It is understandable, therefore, that the EU would not want any bank from anywhere to be allowed to take deposits, write mortgages or underwrite commercial credit in the EU. They want them to be in the EU and regulated by an EU country. That’s fair enough.

Technically, there is no explicit regulatory requirement for a bank to be in the EU to lend to a European borrower. There is a however a hard requirement for any bank that wants to accept deposits from EU customers to be incorporated and regulated in the EU. This restriction on deposit taking from EU savers indirectly impacts banks’ ability to extend loans to EU borrowers. Except for business customers that are large enough to sustain multiple borrowing relationships, most banks making business loans also want to hold the deposit of the business to which they are making any loan. This gives them greater control over and visibility of the customer’s cash flows, and prevents them from entering into conflict with other banks in case of any credit issues. In basic, practical terms, if the bank controls the account into which the customer’s sales get deposited, they can, conveniently, pay themselves the interest and principal repayment on the loan they have extended to that customer with the money in that deposit account. If another bank were to hold that deposit, and payments to suppliers or staff led to there being insufficient money in the account to make payments on the loan, the direct debit in favor of the lending bank would fail. If the lending bank holds the deposit, the company would find that it couldn’t pay its staff and suppliers if enough money wasn’t left to make payments on the loan. Therefore, for smaller and even some medium sized businesses, banks will not lend to a customer unless they can also hold their deposits, which means that the de jure requirement that a bank must be incorporated in the EU for deposit taking serves as a de facto requirement that it also must be incorporated in the EU in order to be able to lend.

In addition, loans such as mortgages involve a form of financial advice. Individuals and small businesses take on financial risk when they borrow, and regulators want the provision of such credit to take the customer’s financial health into account. The Bank of England, for example, compels banks in the UK to ensure the mortgages they write meet certain affordability tests. In addition, given the compression of the spread banks can make on lending and the amount of equity they need to hold in order to extend any loans, banks are keen to earn fees from customers related to the loans they provide, for example by arranging interest rate Swaps or FX hedges (as we shall discuss below). This too involves the provision of a form of financial advice, mainly to corporates. The provision of financial advice is, as previously stated, a regulated activity, and so any bank wishing to extend certain kinds of loans (such as personal mortgages) or to sell certain kinds of financial products (such as Swaps) to certain EU customers must be incorporated and regulated in the EU to do so. In short, even though you don’t need to be incorporated and regulated in the EU to lend to EU customers, you do need this in order to write certain kinds of loans deemed to involve investment advice (mortgages) or to carry out a variety of ancillary services (advice on Swaps, etc.) which are intrinsic to enabling the bank to make money from the lending relationship.

This is particularly the case in retail and business banking. By contrast, banking between banks (mainly inter-bank loans and deposits) or with governments, clearly, does not require any passport. Even in business banking, very large companies, like Akzo Nobel or Mercedes, have subsidiaries all over the world and their own finance departments – which are often more sophisticated than those of many investment banks. Banks in London do not require an EU financial services passport to provide banking services to such customers, for example to sell an interest rate Swap to Nokia or an FX hedge to Peugeot. Moreover, large corporations are big enough to sustain multiple banking relationships, and there is therefore no need for banks to hold their deposits in order to lend them money, as they might insist on doing with smaller customers.

In conclusion, banks require a passport to offer deposits and financial advice only to individuals and companies outside the large companies. It is here that an inconsistency in OW’s classification must be highlighted. Individuals and small businesses are retail customers. Retail customers, by definition, are not wholesale customers. Therefore, one must assume that OW’s use of the term “wholesale” (discussed in detail below) is a little wayward. They have in fact included retail business, i.e. lending to individuals and small businesses, in their EU related “wholesale” income estimates.


£23-£27bn has shrunk to £15bn

From our analysis of the impact of the passport on the different banking activities above, we can see that only Retail and business banking and Private banking and wealth management are at risk from the loss of the passport.

EU revenue potential impact

activity reduction

That amounts to £15bn, which is still a significant sum.

The next question that needs to be asked is whether that estimate, and in particular its largest component, the £14bn of income in Retail and business banking, is reasonable.


Components of retail and business banking

Here is OW’s definition of retail and business banking (p 18):

retail and business banking definition

Note that OW excludes “syndicated lending” revenues from its retail and business banking estimate, and includes them in investment banking. Syndicated lending takes place when a group of banks – the lending syndicate – come together to jointly offer a loan to a company. OW doesn’t explain why it classifies syndicated lending in this way, or how much this classification impacts its estimates for the two divisions. Arranging a syndicate is an investment banking service for which the arranging bank would earn a fee. But for the banks in the syndicate, especially those which are not its leading banks, any loan made as part of a syndicate is similar in its regulation and economic profile to the other business loans they make. It is possible that OW has only included the syndication fee in the investment banking segment, and left the interest income earned by banks in the syndicate in retail and business banking. In any case, if an EU company is large enough to be able to service a syndicated loan, then it will also be large enough to borrow internationally without the banks in the syndicate requiring a passport to lend to it. Therefore any interest income or syndication fees earned by banks from syndicated lending to EU customers which OW has classified as investment banking revenues will not require a passport.

Banking income, as defined by OW, has two essential inputs: assets (and liabilities) and return on assets (and liabilities). OW, rightly, includes liabilities in its estimate because banks make money on their liabilities through their deposit margin. If inter-bank rates (such as Euribor or Libor) are 0.20%, and the depositor is offered a rate of 0.05% on their deposit, then the deposit margin earned by the bank on that deposit is 0.15%. It’s what the bank would make if its business consisted solely of taking deposits at 0.05% and lending those deposits to other banks at the inter-bank rate. But banks don’t, obviously, just lend to other banks. On the other side of the balance sheet, banks also make a margin on their loans to customers, which is defined, technically, as the spread of the loan interest rate they charge over the inter-bank rate. Finally, as we will discuss in more detail below, banks earn interest on various interest bearing securities, and don’t only fund the various assets on which they earn interest through deposits but also a variety of other liabilities, including inter-bank loans and bonds issued by the bank. The net of the income they earn (on loans to customers and other interest bearing assets) and the cost of their funding (on deposits from customers and other interest bearing liabilities) is their net interest income (NII). Here, we show note 1 of RBS’ 2016 annual report (p 314), which helpfully breaks NII down into the interest income from different types of interest bearing assets and interest expense from different types of funding sources:

detailed NII RBS 314.PNG

A crucial figure for banks is their consolidated “net interest margin” (NIM). NIM is calculated by dividing NII  by the stock of the assets (i.e. loans and advances to customers, loans and advances to banks and debt securities) on which the bank earns interest, commonly referred to as  “interest bearing assets” (IBAs). Please note that although, as we recognised above, NII is, technically, earned on both IBAs and on certain liabilities like deposits, the NIM calculation simplifies things by presenting net interest income solely as a percentage of interest bearing assets. Effectively, the standard working assumption people follow in the banking industry is that NIM is a composite figure that includes both the interest margin on assets and on liabilities, expressed as a percentage of assets. Effectively, this convention simplifies things by assuming that all IBAs are funded by interest bearing liabilities, and that the interest margin on those liabilities is subsumed in the total consolidated NII. You could of course drill down and calculate the interest margin the bank makes on its IBAs and on its liabilities separately, but this would make an already very intricate calculation even more complicated without much improvement in the accuracy of our final estimate. We will therefore follow the conventional, simplified approach below, and estimate the EU related NII for banks in the UK based on their NIM and their EU located IBAs. Any EU deposits held by any bank we review will be assumed to fund those EU IBAs, and the margin those banks make on those deposits will be included in their NIM. A NIM calculation will be displayed in full, below, for the four largest UK banks.

As is well known, In addition to earning interest on their assets (and liabilities!) banks also charge fees, which include monthly account fees and transaction fees; OW offers credit card and payment services fees as examples. Transaction fees can vary from charges for each payment made through the bank to larger, more infrequent fees such as a loan arrangement fee. A significant portion of many banks’ income will be derived from arranging FX hedges or interest rate Swaps, often to cap a customer’s borrowing costs. The cost of such products can be added to the interest rate charged on a loan, or charged as a fee, or both. Note 2 (p 314) to RBS’ 2016 annual report offers a helpful illustration of the fees which might be included in OW’s estimate of Retail and business banking income:

RBS banking fees.PNG

The fees in the table will be included in Retail and commercial banking income as analysed by OW, with two exceptions: investment management and brokerage fees. Investment management fees will be included in either the Asset Management estimate in OW’s figure 3, above, or in the Private banking and wealth management segment of Banking Revenue. The wholesale portion of brokerage will, obviously, be included in OW’s Sales and Trading estimate. Turning to the other fees – those which are included in OW’s estimate of Retail and business banking income – the point is that all of these fees are, to different degrees, related to the size of a customer’s loans or deposits with the bank (banking fees are not earned on loans to banks or on securities which the bank holds on its own balance sheet). We will use loans to customers as a proxy for the assets on which banks earn fees.

Therefore, the key components of banking income are:

  1. Interest bearing assets (IBAs);
  2. Net interest income as a percentage of average interest bearing assets (NIM);
  3. Loans to customers;
  4. Fee income as a a percentage of average loans to customers.

Luckily, both NII as a percentage of IBAs and and fee income as a percentage of loans to customers are typically in a tight range for most banks (as we shall see in further detail below), due, largely, to the competitive nature of the industry and the commoditised nature of the products offered. Therefore, if you are able to get a good estimate of the value of the banking assets on which banks in the UK earn interest and commissions thanks to their EU membership, then you will be in a good position to assess whether OW’s £14bn estimate of retail and business banking revenue at risk from Brexit is reasonable.


What a bank’s balance sheet looks like

It is important to understand the composition of banks’ balance sheets because their different components are affected in different ways by the potential loss of a European banking passport as a result of  Brexit. As we alluded to above and will investigate in detail below, some EU located assets need a passport and some don’t. A seemingly innocuous but actually important feature of the OW report is that it estimates the amount of revenues generated by “business related to the EU.” This vague term, “related to the EU,” means generated from assets or by customers located in Europe. However, many assets and customers located in Europe do not require any passport. As we shall see, this is a crucial distinction and an important reason why OW’s £14bn estimate is misleading. It seems that OW has based its implied £14bn estimate on all assets and customers located in the EU, not just those which require a passport, and this has seriously inflated the headline number which an unsuspecting reader (and even readers you could expect to be a little more “suspecting,” like Open Europe) might glean from their report.

Here, as a good illustration of the different components of a bank balance sheet, are the assets on HSBC’s balance sheet, as presented in its 2016 annual report:

HSBC balance sheet

As you can see, there is a lot in there apart from “loans to customers,” which is eight lines down and amounted to $861.5 billion, or around a third of the total balance sheet, at end December 2016. On top of loans to customers you also have loans to banks, including $128bn to central banks and $88bn to other banks. But the total balance sheet is a whopping $2.4 trillion dollars. What else is there to make it so big?

  • First of all you have derivatives, with which HSBC (purportedly) hedges various exposures, typically interest rates and currencies, and which amounted to $290.9bn at December 2016, or just over one third of HSBC’s customer loans.
  • We also have reverse repurchase agreements (Repos), where HSBC buys a security, such as a bond, from another bank or financial institution, with an agreement to sell it back to them in the future on specified terms. In effect, a Repo is a short term loan by HSBC to the other bank or financial institution, which is secured by the asset that institution has sold to HSBC. Banks and financial institutions “Repo” their securities to increase their liquidity without selling the security outright. In simple terms, it’s as if the other bank or financial institution had pawned its asset to HSBC. But that’s a lot of explaining for a class of asset which only represents just under 7% of HSBC’s balance sheet.
  • HSBC’s balance sheet also contains a large chunk of securities of various kinds, mostly government and corporate bonds, as well as equities and other securities. These are listed on HSBC’s balance sheet under a number of headings: “Hong Kong certificates of indebtedness,” “trading assets,” “financial assets at fair value,” and “financial investments.” Together, these securities amount to $728bn – almost as much as customer loans.
  • Finally, there are various other assets on which the bank doesn’t earn any interest or fees, such as goodwill, tax assets or the property in which the bank has its branches.

These assets are all impacted by Brexit in different ways, and not all of their returns should be included in OW’s £14bn income figure:

  1. Customer loans require a passport (except for large companies like France Telecom or Siemens), as mentioned above;
  2. Loans to banks do not require a passport because, as mentioned above, EU banks are allowed to borrow from and lend to other banks across the world, not just European ones;
  3. Securities and Repo assets do not require a passport because these securities are freely tradeable and, as mentioned above in connection with Sales and trading, banks have the right to invest in securities across the world with minimal restrictions. If a bank located in the US can hold a German Bund on its balance sheet, why shouldn’t a bank located in the UK do so? The Repo agreement is with another financial institution and, as with loans to banks, EU banks can Repo their assets with banks all over the world;
  4. Derivatives on the balance sheet do not directly earn interest or fee income for the bank and so their returns will not be directly included in OW’s £14bn estimate. Instead, derivatives indirectly hedge the returns of IBAs and other parts of the balance sheet (when they’re not exploding in the banks’ faces, but that’s a different story). If the bank is actively trading derivatives on its own book (“Prop trading”) or for clients, the gains and losses or the commissions will be captured in the Sales and trading segment of Banking Revenue, not in the Retail and business banking segment. In any case, derivatives are either securities that the bank can trade across the world, or bilateral over the counter (OTC) contracts which it can enter into with other banks all over the world. In short, there is no direct income from and no impact of Brexit on the derivatives on a bank’s balance sheet;
  5. Assets like property which earn neither interest nor fees are obviously not included in OW’s £14bn estimate and not impacted by Brexit.


UK bank balance sheets and margins

Using this classification, we can look at the four large UK banks, including HSBC, in order to get an idea of what split there is in their balance sheets between these different types of asset, and what their asset margins are. The data is for December 2016 taken from the face of and notes to the reports and accounts. I have included Repos in Securities, since the bank is holding a security on the asset side of their balance sheet; they could also have legitimately been included in loans to banks, since the security is acting as security for a loan provided to a bank. Either way it doesn’t make much of a difference. IBAs, in the summary presentation below, consist of (i) loans to customers, (ii) loans to banks and (iii) securities (including Repos). The average IBA figure is the simple average of IBAs at end December 2016 and 2015.

Commission income excludes any commissions, such as asset management or broking, which OW doesn’t include in Retail and business banking (as illustrated above in connection with RBS’s commission income). Barclays’ breakdown of its commission income (note 4, page 290 of the 2016 accounts) only gives a figure for banking and investment management fee income combined, and Barclays does not, in any investor presentation I have seen, give any figures for investment management fees in isolation, which would allow us to calculate the banking fees which are relevant to this discussion. All data are in millions of pounds Sterling except HSBC’s, which are in millions of US dollars. “Total return on IBAs” is equal to NIM (“Net interest income/Avg IBAs”) plus banking fee income as a percentage of average loans to customers (“banking fee income/Avg customer loans”).


This table shows that NIM is in a tight range from 1.25% for Lloyds to 1.65% for RBS  and that banking fee income/average loans to customers ranges from 0.37% for Lloyds to 0.90% for HSBC (Barclays, as mentioned above, does not enable us to isolate banking fees). It is important to note however that the NIM is a blended margin across all IBAs. Loans to customers, because they are more risky and require more expertise and more of a franchise and network to write, will typically carry higher interest margins. We can quantify this using the detailed figures from RBS: by dividing the interest RBS earned on customer loans by its average customer loans, we can calculate the percentage interest rate RBS specifically earned on those customer loans. We then need to calculate RBS’ cost of funding, which is equal to RBS’ interest expense divided by the average liabilities on which it paid that interest. The interest rate RBS earns on its customer loans minus its cost of funding is equal to the NIM it specifically earns on customer loans. Here is the calculation:

customer loan NIM

As you can see, RBS’ customer loan NIM of 2.55% is 0.9% higher than its overall NIM of 1.65%, and 1.1% higher than the aggregate NIM of 1.45% for the four banks.

The table above shows that the composition of the assets of the four banks varies substantially. As the total balance sheet sizes also range widely, from around £800bn for Lloyds and RBS to $2.3 trillion (£1.9 trillion at December 2016 exchange rates) for HSBC, it will be easier to analyse the difference in composition of the four balance sheets by looking at those different assets as a percentage of total assets:

asset make up UK banks

The most salient difference between the four balance sheets is the percentage represented by derivatives. The value of these derivatives on any bank’s balance sheet will depend, among other things, on the proportion of its balance sheet a bank feels it needs to hedge and movements in the price of the assets which it is hedging. In simple terms, changes in the value of any derivatives contract or security are designed to be offset against the cash flows hedged by that derivative. In theory at least. Imagine Bank A has a loan to a Turkish bank in Turkish Lira which it has hedged by selling Turkish Lira for Sterling using a Swap. As the Turkish Lira devalues against Sterling, the Sterling value of that loan will decrease and the value of its Swap will increase. Imagine that bank B, on the other hand, has made a loan in Yen to a Japanese bank which it has hedged by selling Yen for Sterling using a Swap. As the Yen appreciates against Sterling the value of that loan will increase and the value of its Swap will be negative. The outcome however will be that the Sterling value of the two loans, net of the Swap, will be about the same for Bank A and Bank B. In both cases, any gains (or losses) on the currency in which the loan has been made will be offset by losses (or gains) on the FX hedge. Both banks will have effectively hedged the Sterling value of their foreign currency loan. But in terms of the overall balance sheet, Bank A will have a higher derivatives position relative to total assets as a result of its Turkish loan than Bank B as a result of its Yen loan. From time to time you might even find a Bank C whose business model is so blissfully simple that it doesn’t use any derivatives at all. Without delving too much into the detail of banks’ use of derivatives, which could easily fill several tomes and indeed many asylums with those writing them, the absolute size of any bank’s derivatives exposure, while important, can be determined by many arbitrary factors which do not discernibly impact the fundamentals of the way its balance sheet generates the income which is calculated by OW and with which we are concerned here.

For the purpose of our analysis of the make-up of these different banks’ balance sheets, and in particular how their IBAs might be affected by the loss of the EU passport, we can therefore strip the derivatives out. So although when we look at IBAs as a percentage of total assets we see a wide range, from 91% in the case of Lloyds to only 63.9% in the case of HSBC, if we look at IBAs to total assets ex derivatives, on the line underneath, we see a much tighter range, from 89.5% for Barclays to 96% for RBS.

This is not to say that derivatives are not important or indeed risky. It is the very fact that there is such a large amount of derivatives on these banks’ balance sheets, and that this exposure is so opaque, indeed is such a black box, that makes them very risky indeed. There may be a bomb hidden inside any of these four black boxes. The remark about derivatives blowing up in banks’ faces, above, was not entirely tongue in cheek. In the last financial crisis, derivatives exposure valued at a few tens of millions on many banks’ balance sheets rapidly escalated into billions of losses as the sub-prime market collapsed. Each one of the four banks under examination has a significant amount of derivatives on their balance sheet and a significant amount of derivatives risk. But that risk is not directly related to the narrow issue which concerns us here, which is the parts of the banks’ balance sheets which are affected by Brexit, or not.

Since, ex derivatives, IBAs/total assets is in a tight range, we can go on to look at the split of those IBAs between loans to customers, loans to banks, and securities:

IBA make up of UK banks

The important thing for our assessment of the £14bn in revenue is what percentage of total IBAs in the EU consists of loans to customers. It is, I repeat, only loans to customers which require a passport (and as mentioned above, of those loans to customers, only loans to companies other than large corporates such as Philips or Inditex, the owner of Zara); loans to banks and securities holdings do not require a passport. The loans to customers as a percentage of total IBAs are in a relatively tight range (23% points) from as low as 43.7% in the case of HSBC to as high as 66.5% in the case of RBS, or from less than half to around two thirds.

The table above shows that there are a lot of IBAs on the balance sheets of the UK’s four largest banks which are not the sort of customer loans which require a passport. This seems to have been forgotten in the analysis by OW and others of the impact of Brexit on Retail and business banking income.


Assets related to Europe: BIS and individual company data

Having used the accounts of the UK’s four largest banks to define a range for net interest income and banking fee income to IBAs and to loans to customers, and to understand the typical shape of a bank’s balance sheet, we can now try to quantify the IBAs which those banks and other banks operating in the UK hold on their books against European counterparties.

It is here that we discover a very surprising omission in OW’s report. These are the sources of information it uses for its estimate of Retail and business banking revenue (p 19):

retail and business banking sources

OW, understandably, uses company reports and proprietary data to build an estimate of EU related revenues from the bottom up, adding up all the individual items to get an approximation of the overall picture. But by restricting themselves to this – admittedly rich – data set, they are neglecting a hugely valuable resource: the BIS cross-border claims data series, which provides top down country level data for the financial claims of banks in the UK on the countries remaining in the EU. Our estimates will be based on both the publicly available bottom up data (going into much greater detail than OW) and the top down data provided by the BIS (and ignored by OW):

Bottom up and top down

The BIS organises its data according to many different criteria. However two of those are critical for this analysis:

  1. Where a bank is headquartered
  2. Where a bank’s office is located

These classifications allows us to use BIS data to calculate both the claims of UK headquartered banks – in other words the big four banks whose balance sheets we analysed above – on EU entities, as well as the claims of any bank, including foreign banks, on EU entities, which were made from one of their offices in the UK. The terms the BIS uses to distinguish these two is “UK headquartered” and “UK located.”

The BIS offers two types of reports, one of which breaks down the total claims on any country by the headquarter of the bank holding that claim (the B.4 tables) or the country in which the branch holding that claim is located (the A.6 tables):

headquarter versus location tables

There is a table A.6 and a table B.4 for almost all countries in the world and for each country in the EU. The UK is one of the largest holders of claims on EU entities in both reports. By adding the claims for all the EU countries (excluding the UK) reported to the BIS held by bank branches located in the UK (A.6) or headquartered in the UK (B.4), you can calculate the total amount of claims on EU entities of banks booked in their UK branches or of UK headquartered banks.

Here is a headline summary of the data for 2015, the reporting year from which OW derived its bottom up estimates:

headquartered versus located

The total for claims on EU entities of all bank branches located in the UK is the larger figure because it is the most inclusive. It includes most of the claims of UK headquartered banks (~$600bn), which (with some exceptions discussed in full below) are booked in their UK offices, as well as the claims of foreign banks with UK branches. In other words, table A.6 includes most of table B.4. The total ~$1.6 trillion of claims on EU entities of all bank branches located in the UK minus the ~$600bn claims of UK headquartered banks on EU entities leaves you with an approximation of the ~$1 trillion of claims on EU entities by foreign banks with UK located branches. We will analyse these data in detail below, including a number of adjustments to reconcile the A.6 and B.4 data, but for now this will do as a rough approximation.


How data are presented in the BIS A.6 report

The data in the A.6 report are not consolidated. This is very important. It means that it would include an intragroup loan from Citigroup’s branch in London to Citigroup’s branch in Belgium. The interest earned on such an intragroup loan would not contribute to Citigroup’s consolidated NII. Only the interest Citigroup earned on a loan from its Belgian branch to a third party customer would do so. Intragroup lending is typically used to allocate funds raised in one branch or subsidiary to other branches or subsidiaries which need those funds. Helpfully, the A.6 reports include a line item for the value of intragroup claims.

The A.6 report also breaks down the total claims of all banks reporting to the BIS (which is virtually all banks) on entities in any given country into:

  • loans to banks;
  • loans to non-banks;
  • debt securities;
  • other securities (including equities and others);
  • and unclassified claims.

Loans to non-banks are further split into:

  • loans to non-financial companies;
  • and loans to non-bank financial companies or “NBFIs” (these would include loans to insurance, fund management or brokerage companies).

By adding up each type of claim (i.e. loans to banks, debt securities etc.) for all the EU countries (excluding the UK) you can calculate the total amount of that type of claim on EU entities held by foreign banks. These data can be used, for example, to calculate that of the ~$8 trillion in cross-border claims of all banks in the world on EU entities, ~$2.27 trillion consisted of debt securities. The A.6 data therefore allow us to carry out a more precise measurement of the composition of the claims on EU customers, i.e. what proportion of those claims are debt or other securities, loans to banks, loans to non-bank financials and to non-financial customers (whether individuals or companies). This matters, because it is only loans to non-financial customers which require the passport. These A.6 data data can be cross-checked against the company level data available in order to arrive at a precise, detailed picture of lending from the UK to the EU which may be at risk from any loss of the EU banking passport.


Start with UK headquartered then look at the rest

For all intents and purposes, we can assume that the claims of UK headquartered banks on EU entities in the BIS B.4 tables are held by the big four quoted UK banks and that the EU claims of building societies and smaller banks will be negligible. For example the Co-op Bank, which had £14.7bn of customer assets (less than 5% of RBS) in 2015, disclosed exposure to Europe in that year on p 125 of its 2015 annual report of £0.5bn, just over 1% of Lloyds’ EU £39bn exposure, which (as we shall see) is the smallest of the big four in terms of EU exposure. CYBG, the £2.7bn valued holding company owning Clydesdale Bank and Yorkshire Bank, which had £28bn customer loans in 2015, disclosed £100m exposure to the European Investment Bank as its only Eurozone exposure in 2015 on page 167 of its 2016 annual report:


Standard Chartered, a £24bn valued UK headquartered bank focused on Asia, with loans and advances to customers of $94bn in fiscal 2015, had exposure to loans to customers in Europe (including the UK) and the Americas of only $282m, as detailed on page 36 of its 2016 annual report:

Standard Chartered

The four largest UK banks, each of which has significant exposure to the EU, report on that exposure in detail. It therefore makes sense to quantify that exposure before moving onto that of the many banks headquartered outside of the UK but booked in their branches in the UK.

The BIS includes the following chart in its report on international banking statistics at end-March 2017 (p 9):

B1 chart Q12017

The strapline indicates that this data is from “Graph B.1” (the member of staff from the BIS who responded to my request for information was very helpful in answering my queries and I can’t thank her enough for helping me navigate through and understand the rich data set on offer from the BIS). That means that this chart shows the claims of banks with headquarters in selected countries on the Euro area. For the purposes of our investigation, the brown GB line shows the claims of UK headquartered banks (the big four analysed above) on customers in the Euro area. The blue line for the US or the green line for Japan will include loans from US or Japanese headquartered banks booked via branches in the UK, and which are potentially impacted by Brexit. As previously stated, such claims will record the UK as the “reporting country” – but in the BIS’ A.6 table, not in the B.4 or B.1 tables.

As you can see from the chart, the exposure of UK headquartered banks to the EU has reduced dramatically over the last five years, practically halving in that period. This is the result of a number of factors including:

  • divestments of EU operations by UK headquartered banks, for example Barclays’ sale of its subsidiary Bank SAU to Barcelon based La Caixa in 2015. That transaction alone will have reduced claims of UK headquartered banks on EU entities, tracked by the BIS in the table above, by c. ~20bn or 3% of the amount of claims outstanding at the end of 2015;
  • the ~21% decline in the value of the Euro versus the dollar over the period;
  • the entirely rational reduction by UK banks of their holdings of EU sovereign debt, particularly issued by countries such as Greece, Ireland, Spain and Portugal (the “PIGS”), in the wake of the EU sovereign debt crisis.

The chart tells us that UK headquartered banks had roughly $600bn of claims on the Eurozone at the end of both 2015 and 2016, with a slight year on year decline in 2016. Remember OW’s report was written in 2016 and bases its analysis on 2015 numbers.

This allows us to do two things. First of all, we can compare the available bottom up numbers provided by banks headquartered in the UK for their exposure to EU customers with the top down total figure of around $600bn provided by the BIS, to see if they are consistent with each other. Secondly, if they are consistent, we can use the detail in the annual reports of the UK banks to establish what proportion of those claims require a passport.

Here are the figures from the B.4 tables for the claims of banks headquarted in the UK on each EU 27 country for 2015 and 2016:

B1 country detail

Note the BIS’ chart only tracks claims on the Eurozone, whereas we are concerned with claims on EU entities both within and outside the Eurozone. Looking specifically at the Eurozone figure we can see, satisfyingly, that the total is around $600bn in both 2015 and 2016, with a slight year on year reduction in 2016. This is entirely consistent with the chart above.

Next we can look at the data given by individual UK headquartered banks for their exposure to the EU.


Exposure versus claims

Banks provide a breakdown of their credit exposure to entities in Europe, ex the UK, in the credit risk sections of their annual reports. In principle, these exposures can be used as a good proxy for what BIS would classify as a financial claim on an EU entity. However there are differences to bear in mind. Barclays gives the following gloss of its definition of exposure (annual report 2016, p 172):

credit exposure

This means that Barclays, when calculating its exposure, reduces the value of any of its claims by any amount of those claims against which it has made a provision for credit impairment, and by the value of any cash collateral and of any claims from the entity to which it is exposed which can be netted off from its exposure. Here, by contrast, is how BIS defines a claim in its “Guide to the international financial statistics” (p 6):

BIS claim definition

What this means is that, whereas banks’ present the value of credit exposure net of credit provisions and other offsets, BIS’ presents the value of claims gross i.e. without reducing those claims for any provisions or other offsets.

Here are the 2015 credit provisions on the UK banks’ balance sheets as an absolute number, and as a percentage of 2015 customer loans, pre-provision, from the 2016 reports and accounts (as before, HSBC’s figure is in $ms and those of the other banks in £ms):


As you can see the percentage, at a little over 1% for the four banks in aggregate, is not very big and therefore, per se, the fact that banks calculate exposure post-provision and the BIS calculates claims pre-provision will not result in an important difference between the two figures.

Before looking at Barclays in more detail, it will be useful to examine how Lloyds presents its exposure numbers. Here is Lloyds Banking Group’s exposure definition (2016 accounts p 143):

definition of exposure

Loans and advances exposure figures are also stated net of provisions but without reference to the deduction of the offsets mentioned by Barclays. However, if we turn to note 52, page 262, we find the following statement of exposure:

credit exposure n 52 p 261.PNG

The “Maximum exposure” numbers here are consistent with the balance sheet sheet values for the corresponding line items, except that the exposure numbers do not include £1,213m and £67,697m of equity exposure  in “Available-for-sale” and “Trading and other financial assets,” respectively. Lloyds’ exposure numbers therefore do not include equity holdings, which are, however, included in the BIS numbers as “Other securities.” There is a significant £18.5bn offset reducing Lloyds exposure to derivatives, but these are not included in the BIS claims data we are using. Interestingly, there is a £6.3bn offset reducing Lloyds’ net exposure to loans and advances to customers from the £457,958m value of those loans, net of provision, on the balance sheet to an exposure figure of £451,627m. Lloyds’ footnote 2 to the Offsets column reads “Offset items comprise deposit amounts available for offset, and amounts available for offset under master netting agreements, that do not meet the criteria under IAS 32 to enable loans and advances and derivative assets respectively to be presented net of these balances in the financial statements.” These £6.3bn are, therefore, cash deposits with Lloyds by the customers to whom loans were made and which could be seized in the case of any default, while the £18.5bn will be assets such as cash held as collateral from and derivative liabilities toward any derivatives counterparties under master netting agreements. What we do not know, however, is whether Lloyds’ specific EU loan exposure, as presented in its risk report (see below), is reduced by any such offsets.

Here are Barclays’ total credit risk exposure figures (2016 report and accounts p 171) juxtaposed with the corresponding balance sheet figures for the line items in that statement of exposure. In Barclays’ annual report, the regional exposures add up to the total exposure figure, and are therefore calculated in a way which is consistent with the calculation of total exposure (none of the other UK banks provide such a reconciliation between regional and total exposure). I’ve added the total for the relevant balance sheet numbers at the bottom to assist comparison:

Barc credit exposure offset

Comparing each total exposure figure with the corresponding balance sheet figure, nearly all the numbers are identical, with the only exceptions being a relatively large difference for “Trading portfolio assets” and small differences for “Financial assets designated at fair value” and “Financial investments” (all highlighted by a red mark). The three line items where these differences are observable all consist of securities. The exposure value of securities for Barclays is equal to £192.1bn, versus a balance sheet value of £235.6bn. The reason the exposure figures for these securities is less than the balance sheet value is likely that, as with Lloyds, Barclays’ securities holdings on the balance sheet include equities which are not included in credit exposure. Barclays may well also have hedged some of the positions or borrowed from some of the entities emitting the securities concerned, but we can’t say for sure, as Barclays doesn’t specify the components of the £43.5bn adjustments which it makes to the £235.6bn figure for securities as recorded on its balance sheet in order to arrive at its £192.1bn exposure to securities. However, crucially, Barclays’ figure for credit exposure for loans and advances to customers is identical to the value of loans and advances, net of provisions, on the balance sheet. Therefore, although Barclays says it takes account of netting and cash collateral, unlike Lloyds this has no impact on its statement of exposure to loans and advances to customers. For loans and advances to customers, the difference between claims, as recorded by the BIS, and exposure, as reported by Barclays in its risk report, is simply the amount of credit provisions.

RBS also defines its exposure as balance sheet assets net of provisions (p 216):

exposure definition RBS

Its total exposure of £457.6bn at the end of 2016 compares with IBAs of £529.6bn on the same date. There is therefore a difference of roughly £72bn. Again, this is likely to be because of equity holdings on balance sheet not included in credit exposure. However, RBS’ total securities on balance sheet are valued at only £73bn, implying that virtually its entire securities exposure would be in the form of equities. RBS does not disclose enough detail to enable us to say for sure.

HSBC, on page 88 of its risk report, explicitly excludes equities from the numbers which it gives for its regional credit exposure (and which we will analyse below):

equities excluded from credit risk

In summary, we can assume that the European exposure disclosures of the UK headquartered banks exclude equities, and that this will be true of the other banks whose disclosures we analyse below. In terms of comparability with BIS numbers, the EU exposure numbers in individual bank disclosures therefore exclude equities which are included in the BIS numbers (classified as “other securities”).

Although we know the total equities holdings these banks hold, we have no data whatsoever on the amount of EU equities they hold. Moreover, the BIS B.4 tables we used above to calculate the claims of UK headquartered banks on EU entities recorded by the BIS do not break those claims down into “debt securities” and “other securities” (which are mostly equities) like the A.6 tables do, and therefore do not allow us to identify a BIS “claims ex other securities (which are mostly equities)” number which would be comparable with the EU exposure numbers disclosed by the four UK headquartered banks in the risk reports of their annual reports and accounts.

Since EU securities holdings are not impacted by Brexit, the difference between BIS claims and annual report exposure numbers due to banks’ EU equity holdings will not impact our forecast of any potential Brexit impact. However, these EU equity holdings by UK headquartered banks do interfere with our attempts to reconcile the bottom up company numbers from the four UK headquartered banks with the BIS B.4 data. So we need to estimate a value for EU equity holdings by UK headquartered banks which is included in the BIS claims number but not in the individual banks’ statement of exposure, in other words assume that in addition to the EU debt securities exposure reported by the UK banks in their country risk reports they also hold a certain amount of EU equities which are recorded by the BIS but not included in those risk reports.

In conclusion, I have made the following adjustments to estimate the EU exposure numbers implied by the BIS claims data:

  • To make loans to non-banks claims (BIS) comparable with loans to customers exposure (individual company reports), I deduct a 1% provision from BIS customer loans to non-bank claims to calculate the exposure to customer loans in the four individual UK banks’ accounts. The 1% figure assumed for credit provisions is a round average of the provisions to customer loans of the four UK banks analysed above. You could argue that Lloyds’ £6.3bn offset, which is deducted from balance sheet loans to arrive at loan exposure, indicates that, in addition to the 1% provision, I should deduct some equivalent offset from the UK located banks’ loans to EU customer claims figure, recorded by the BIS, to calculate UK headquartered banks’ exposure to EU customer loans. I haven’t done so as we have insufficient data to estimate this with any confidence. We don’t even know whether Lloyds’ statement of EU credit exposure is made after adjustment for any EU portion of this £6.3bn offset or not, since Lloyds does not provide any reconciliation between its EU credit exposure statement and its total exposure to loans to customers (as Barclays did). Moreover, it seems only Lloyds may have calculated its EU loans to customers exposure values in this way: we do have evidence that the figures we have available for Barclays and HSBC’s EU loan exposure have not been reduced by any such offsets, and there is no evidence that RBS’ loan exposure values have been reduced by such offsets.
  • To make loans to banks claims (BIS) comparable with loans to banks exposure (individual company reports), we can confidently assume there is minimal provisioning of loans to banks and that claims equal exposure.
  • To make securities claims (BIS) comparable with debt securities exposure (individual company reports) is more complicated: EU equity securities exposure of UK banks as a percentage of total securities recorded by the BIS is a bit of a guess. The best available benchmark is the split of total cross-border holdings of EU equity and debt securities recorded by the BIS in its A.6 report. That split was as follows end December 2015:

% securities

We can compare that with the implied numbers we have for the two UK banks in which we can observe a difference between securities on balance sheet and securities exposure at the total level (not just for the EU, as explained above):

equity %

The numbers show a significant difference between Lloyds and Barclays, highlighting how approximate this estimate is. However, the similarity in the ratio for Lloyds to the ratio derived from the BIS A.6 report for the EU as a whole offers some comfort. In any case, this estimate doesn’t affect our calculation of the revenue at risk from Brexit, as neither equities nor debt security holdings require any passport.

Here are the headline claims numbers, as reported by the BIS, with our estimate of credit provisions and equity holdings which, subtracted from those claims numbers, enables us to estimate the EU exposure figures for UK headquartered and located banks, as they would appear in those individual banks’ disclosures, which are implied by the BIS data:

Claims versus exposure

The full calculations will be displayed below as part of our estimates of the impact of Brexit on Retail and business banking income. As we saw above, the main difference is constituted by the equity holdings which have to be removed from the BIS numbers to make them consistent with those reported by individual banks. As we shall see below, international banks with London offices tend to hold most of their securities, including equity securities, at the parent company level and in their headquarters, rather than in their London offices, which is why very little of their EU equity securities claims are included in the A.6 report for UK located claims, from which our $1,663bn claims figure is taken. As a result, most of the equity related adjustment is attributable to the UK headquartered banks, all of whose EU equity holdings are included in the consolidated B.4 reports, from which our $641bn claims figure is taken.


Raw data

Up to now we have used the 2016 figures to present an up to date picture of asset margins and balance sheet compositions. Below we are going to look at the 2015 values for credit exposure which could have informed OW’s “EU related” banking revenue estimate and which we will attempt to reconcile with the estimated BIS exposure figure of $539bn (calculated above based on the BIS’ disclosed $641bn claims figure).

Here is Barclays’ exposure data in 2015, showing total financial exposure to Europe of £240.2bn, including £103.5bn of derivative instruments and £47.3bn of loans to customers (p 171, again):

Barclays EU

RBS gives a breakdown in its risk report for exposure to Ireland and the rest of Europe, which together are equal to RBS’ total EU ex UK credit exposure (annual report 2016, p 216):

RBS EU exposure

Although not as detailed as Barclays, this table breaks RBS’ exposure down sufficiently to distinguish exposure to banks and sovereigns, which doesn’t require a passport, from that to other sectors. What it doesn’t do, unfortunately, is break the exposure to corporate entities down into loans and securities (such as bonds). It is very likely, for example, that RBS holds some property bonds of EU entities like Unibail Rodamco. However, given that this corporate exposure amounts to only c. £15bn out of RBS’ total exposure to EU counterparties of £75bn, conservatively assuming that all of that exposure is through loans won’t make much of a difference.

Lloyds 2016 accounts provide a similar disclosure of its 2015 exposure to the Eurozone in its risk report (pps 143-144; this can be compared to related figures in its report to the EBA, which confirms Lloyds does not have any exposure to the EU outside the Eurozone):

Lloyds EU exposures

Lloyds EU exposures other.PNG

As with RBS, some of the “Other Financial Institutions,” and “Corporate” exposure may consist of securities, but I have conservatively assumed it is all in the form of loans.

HSBC, finally, offers a country breakdown, not for all assets, but for its gross (i.e. pre-credit provision) loans and advances to customers by region (p 104):

HSBC EU consumer loans

Total loans to the EU is the total for Europe, as defined by HSBC, minus the UK and Switzerland. This comes to $65.7bn in 2015 and represents 7.0% of gross loans and advances to customers for all countries where HSBC is active of $934,009m. However, what HSBC presents on its balance sheet is net loans. HSBC’s net loans figure equals gross loans minus credit provisions. Here is HSBC’s presentation of gross loans, provisions and net loans for 2015 on page 89 of its 2016 report and accounts:

HSBC provisions.PNG

HSBC discloses credit provisions for total loans to customers here, but doesn’t break those provisions down by geography. The $924,454m underlined is the figure for loans and advances you will see on the face of HSBC’s balance sheet for 2015 above. The “total gross amount” figure of $934,009 is the total of gross loans and advances to customers by country on page 104 of which Europe’s share, reproduced above (as defined by HSBC to include the UK and Switzerland), is $388,519m. The difference between the gross and net figures is the provisions or “impairment allowances,” which is the circled figure of $9,555m.

To estimate net loan exposure to EU entities we need to estimate the level of credit provisions as a percentage of gross loans for the EU. Fortunately, HSBC’s French and German subsidiaries are former acquisitions (HSBC purchased Crédit Commercial de France or “CCF” in France in 2000 and Trinkaus in Germany was acquired by the Midland bank in 1980 and became part of HSBC when it acquired Midland in 1992). Both have independent shareholders and report separate or “silo” accounts (p 265):

Investments in subsidiaries.PNG

These silo accounts give us a figure for the two subsidiaries’ customer loans both gross and net of provisions. Both entities report in Euros, so their figures need to be translated into dollars, for which I use the end 2015 exchange rate given on page 31 of HSBC’s 2016 annual report of €0.919 for one US dollar. When these entities are consolidated by HSBC plc in its accounts, which we have been analysing above, any intragroup balances will be eliminated. Therefore, the figures reported in HSBC’s French and German subsidiaries’ silo accounts will be slightly different from those recorded for them in the consolidated accounts.

HSBC France’s “Document de référence” (DDR) is a French equivalent to the US SEC’s 10k reports. I have used the figures from HSBC France’s 2016 DDR and HSBC Trinkaus’ 2016 annual report and accounts, which present the entities’ balance sheets on pages 192 and 94, respectively. Since we have figures for gross customer loans in both the consolidated accounts and the silo accounts, we can compare them, using HSBC’s declared exchange rate. This allows us to calculate the amounts eliminated on consolidation. The US dollar figure for credit provisions is simply the Euro figure for provisions disclosed in the subsidiary accounts, translated at the exchange rate given by HSBC. This gives us all the elements we need to calculate net loans to customers for France and Germany as included in the consolidated accounts.

We do not, however, have silo accounts for HSBC’s “other” EU exposure. To estimate “other” EU customer loans net of provisions I assume that provisions are equal to 1.02% of gross “other” EU loans, the same as for HSBC’s total loans:

HSBC EU net loans

We can combine these estimated figures with the published figures for HSBC France and HSBC Trinkaus to calculate HSBC’s loans and advances to EU customers, net of provisions, as they will be included on the face of HSBC’s consolidated balance sheet:

HSBC EU net loans

The figures in the “Loans to customers gross, consolidated accounts,” are those displayed in the “Gross loans and advances to customers by country” extract from the HSBC annual report above. The information and calculations in the table above result in an estimate of HSBC’s net loans and advances to EU customers of $64.8bn at the end of 2015. This number is HSBC’s equivalent of the loans to customers figures disclosed by the other three UK headquartered banks in their exposure reports, reproduced above.

Turning from loans to customers to other IBAs, HSBC unfortunately does not break down its loans to banks or its securities by geographical exposure, the way the other three banks do. We therefore need to make assumptions. I assume that the EU share of the three asset lines on HSBC’s balance sheet related to loans to banks – “cash and balances at central banks,” “items in the course of collection from other banks,” and “loans and advances to banks” – is 7.0%, the same as the EU’s share of loans to customers. Loans to banks tend to be used as buffer liquidity in any given country which can be tapped to increase the loan book to customers. The inter-bank lending market is one of the most accessible means of deploying capital internationally. For want of a better alternative, it seemed fair to assume EU share of loans to customers and loans to banks were the same. As we shall see below, this leads to an estimate of loans to EU banks of $13.6bn. HSBC does disclose its wholesale exposure to loans to banks in Europe, including the UK and Switzerland (p 95):

loans to banks

If anything, $13.6bn for EU located banks out of a total of $17.1bn for Europe including the UK and Switzerland is a larger proportion than what would be suggested by the weight of EU countries in Europe including the UK and Switzerland loans to customers, which is dominated by the UK.

We also need to make assumptions to estimate the value of HSBC’s exposure to claims on EU entities via debt securities. For such exposure, recorded on HSBC’s balance sheet as “trading assets,”  “financial assets at fair value” and “financial investments,” it is likely that the regional breakdown will be different from what it is for loans. The geographic spread of customer loans we discussed above depends on a bank’s penetration of particular markets; the reason HSBC’s top four markets for customer loans and deposits are the UK, Hong Kong, the US and France is that it has been operating in those markets a long time, or entered them by buying a large player (as with its acquisition of France’s CCF). However, many markets do not have deep and liquid securities markets, or if they do, these carry significant credit and currency risks. Government debt securities holdings therefore tend to be concentrated in more mainstream geographies than loans to customers or loans to banks.

HSBC does give us some regional information which, by a process of elimination, can enable us to limit the extent of our assumptions and therefore their scope for error. Here is note 15, page 229:

HSBC financial assets

“Financial investments” on the balance sheet are recorded as $428,955m, so a little lower than the total fair value figure in this table of $430,111, as the latter may include some mark to market unrealised gains not included in the balance sheet figure. Notwithstanding this minor difference, the geographical breakdown of the fair value figure given in the table of this note 15 is a useful indicator of the geographical breakdown of HSBC’s financial investment asset line as it is recorded on the face of the balance sheet. This detailed table of financial investments can be broken down into four buckets:

  1. US, UK and Hong Kong government and government related securities which, by definition, are not issued by EU entities;
  2. Other government securities, some of which will be issued by EU entities;
  3. Equities, which won’t be included in EU exposure figures;
  4. Other assets like corporate debt, some of which will be issued by EU entities.

Accordingly, we can exclude the US, UK and Hong Kong debt securities and equities and only apply our assumptions to the other debt securities. Any inaccuracy in our assumptions will only impact the smaller number constituted by HSBC’s holdings of these other debt securities.

The number resulting from the assumptions we make is one piece of the jigsaw of the claims of UK headquartered banks on EU entities. Helping us in this exercise is the fact that many of the other pieces of this jigsaw are already in place. HSBC’s claims on EU counterparties in the form of securities is both the last item we need in order to calculate the value of all of HSBC’s EU exposure, and to reconcile our bottom up calculation of the total exposure of UK headquartered banks to EU entities with the corresponding BIS B.4 claims figures.

Here is what we know or have estimated so far about HSBC’s EU exposure at December 2015:

HSBC EU ex securities

We can combine this partial data on HSBC with the complete data on the other three banks, and compare it to the total BIS figures for UK headquartered banks. To do so we need to estimate the breakdown of the total BIS numbers, which we can easily do using the numbers we have from the individual banks:

  • We calculate loans to customers claims by adding 1% credit provisions to the $209bn loans to customers exposure disclosed in the four UK banks risk reports, resulting in an estimated claims on EU customers figure of $211bn (technically we divide the exposure number by 0.99 but in practice it makes no difference with such a small adjustment).
  • The loans to EU banks disclosed in Lloyds’, RBS’ and Barclays’ risk reports and estimated for HSBC total $96bn. The BIS claims figure will be the same as there should be no meaningful credit provisions on loans to banks.
  • We can calculate total securities claims recorded by the BIS by taking the total claims on EU entities by UK headquartered banks of $641bn, disclosed by the BIS, and subtracting the loans to customers claims of $211bn and loans to banks claims of $96bn estimated above. The result is a figure for claims by UK headquartered banks on EU entities in the form of securities of $333bn. We assume, as outlined above, that 30% of those securities claims consisted of equity securities in order to arrive at the following picture:

exposure versus claims detail

The total value of UK headquartered bank exposure to EU customers is $539bn, based on the BIS recorded claims figure of $641bn and the assumptions detailed above.

We can now compare this detailed BIS exposure estimate for UK headquartered banks with the information given by each of those banks for their EU exposure. Lloyds’, RBS’ and Barclays’ exposure to loans to customers, loans to banks and securities in the EU can be converted from the pound Sterling figures disclosed in those reports into US dollars at the $/£ exchange rate at the close of 2015 from HSBC’s 2016 report and accounts used above. We also have dollar values from HSBC’s balance sheet for loans to EU customers and the estimate, calculated above, for loans to EU banks. This allows us to draw the following picture:

EU exposure banks BIS ex HSBC securities

(All data is in billions of dollars).

This implies that the figure for HSBC debt securities representing a claim on an EU entity, the figure with the question mark, has to be $89bn ($538.5bn minus $449.4bn) for the bottom up data from the UK banks’ annual reports to tally with the top down data from the BIS. That means, in turn, that the sum of the EU portion of the various lines of HSBC’s balance sheet marked with a question mark above also have to sum to $89bn. In other words, the percentage we assume for HSBC’s EU debt securities as a percentage of total securities holdings has to result in a total EU debt securities holding equal to $89bn.

We are, accordingly, going to first assume a figure for the percentage of HSBC’s other government securities (i.e. government securities ex the securities HSBC has told us in the notes to the accounts are issued by non-European governmental entities) which are issued by EU entities. We also need to assume a figure for what percentage of HSBC’s total non-government debt securities are EU non-government debt securities. We are going to assume that this percentage is roughly halfway between the 7% we observed for loans and the figure we assume for EU government debt securities. The government debt securities a bank like HSBC is prepared to hold will be heavily concentrated in stable developed market issuers, whereas its loans will be spread through the countries across the world in which it seeks to grow. The geographical spread of other assets such as repos or non-governmental trading securities will sit somewhere in the middle.

Here are the resulting estimates for the EU portion of trading assets (note 10, p 216), financial assets at fair value (n 13, p 226) and financial investments (presented above). The figure assumed for EU government debt securities as a percentage of other government debt securities is 21.6%:

HSBC EU trading assets and financial investments

This allows us to complete the table of HSBC’s EU exposure:

HSBC full EU picture

EU debt securities of $68,246m plus EU Repos of $20,931m equals HSBC EU debt securities exposure of $89bn. This allows us to complete the picture for UK headquartered banking claims on EU entities:

EU exposure banks BIS complete


Sanity checking our UK numbers

The key figure that allowed us to reconcile the bottom up with the top down data is the assumption that 21.6% of government securities held by HSBC, ex US, UK and HK, were issued by EU governments and governmental entities. We can benchmark this figure with data from the BIS for all banks in the world. The BIS records data for all cross-border holdings of debt securities issued by all countries. By deducting the UK, Hong Kong and the US from the total, and looking at what percentage of the residual is composed of EU securities, we can arrive at a benchmark percentage with which we can compare our assumed figure for HSBC. We need to caveat that the BIS’ figures are for cross-border holdings, whereas HSBC’s holdings of UK government bonds will count as domestic holdings in BIS’ classification, and that BIS’ figures are for all debt securities, not just government bonds, even though government bonds can be expected to constitute a sizeable proportion of total cross-border holdings of debt securities.

BIS EU ex UK debt securities percentage other

The EU government debt securities percentage of 21.6% of government securities ex the UK, US and HK, assumed for HSBC, is less than half the equivalent figure of 57.8% for global banks’ holdings of EU debt securities as a percentage of total debt securities ex UK, US and HK. That figure however includes intra-EU bond investments, for example holdings of Finnish government bonds by a Dutch bank. Eurozone banks are natural investors in Euro denominated bonds, and their investments therefore boost the percentage of Europe ex UK debt securities held by all international banks. The figure we estimated for HSBC’s securities holdings, above, would imply a greater proportion of HSBC’s sovereign portfolio was invested in Canada (which is its sixth biggest market, just after mainland China) and Australia (which is larger for HSBC than Germany and Switzerland combined) relative to Europe. Given HSBC’s international business mix this is a reasonable assumption. HSBC is more international than the average European bank and can therefore be expected to be less exposed to European government securities than they are.

We can also sanity test our assumptions by comparing the composition of HSBC’s EU assets which result from them with the percentages for the other three banks, which are based on disclosed numbers:

asset make up UK banks EU exposure

As we can see, RBS and Barclays are similar, with 35-40% of the exposure in loans to customers and 44-48% in exposure to securities, with the residual made up of loans to banks. Lloyds has the smallest exposure to Europe, some of which, in Ireland for example, is in run-off. Europe is not a core business for the bank, which is focusing on the UK. Government bonds are held by banks in part to satisfy parent company or subsidiary reserve ratios and capital requirements. Lloyds didn’t have a subsidiary in the EU pre-Brexit, and it therefore didn’t hold government securities for that purpose.

Compared with the disclosed figures for these banks, the figures we estimated for HSBC – in order to reconcile the bottom up annual report numbers with the BIS B.4 claims data –  result in a lower percentage of its European assets in loans to banks and higher percentage in debt securities than its peer group. Given that our estimate of loans to banks was based on the disclosed figure for loans to customers as a percentage of total loans, and that that number is low compared to the other banks, it may simply be that our estimate of loans to banks was too low and therefore the residual in securities was too high. However, we saw above that our estimate for loans to EU banks looked very high as a percentage of total wholesale loans to European banks including the UK and Switzerland; increasing our estimate of loans to EU banks to reduce the EU securities estimate would only further exacerbate this anomaly. Alternatively, if our estimate of $100m EU equity securities holdings by UK headquartered banks is too high (low) it means that we would have to increase (decrease) our estimate of HSBC EU debt securities holdings to make the bottom up numbers tie up with the top down numbers. Or it could just be that HSBC really does have much lower relative exposure to banks in the EU than its UK peers. While it is impossible to pinpoint the reason for this difference exactly, it isn’t particularly egregious.

We can also sanity check by comparing the breakup of each banks’ consolidated IBAs (i.e. IBAs for all countries, brought forward from above and marked “tot” for total) with the breakup for EU IBAs:

asset make up UK banks EU vs tot exposure

This demonstrates that the difference between the percentage of HSBC’s EU IBAs and total IBAs is not out of line with the other three banks. The weight of loans to customers in EU IBAs is lower than it is in total IBAs across all four banks, with the difference ranging from only 5% for HSBC (43.7% loans/IBAs for HSBC as a whole versus only 38.7% for HSBC in the EU) to 27% for RBS (66.5% for RBS as a whole versus only 39.2% for RBS in the EU), with Barclays at 16% and Lloyds around 9%. This is likely because the UK banks have stronger franchises in the UK (as HSBC also does in Hong Kong), and are therefore able to devote a higher proportion of their balance sheets there to higher return loans to customers.

When comparing securities as a percentage of UK headquartered bank EU IBAs with securities as a percentage of total UK headquartered bank IBAs, we should bear in mind that the EU number excludes equity securities whereas the total number includes them. If we compare HSBC’s EU debt securities/IBAs ratio to the securities/IBAs ratio for its total balance sheet, we find that the difference is not out of line with that observed in other banks. RBS’ EU debt securities as percentage of IBAs of is 47.5%, 33.7% higher than its total securities as a percentage of total IBAs, which compares with HSBC’s estimated figure of 53.1% for EU debt securities which is 8% higher than the percentage for total securities – a significantly smaller gap than the equivalent one disclosed by RBS. So the greater weight of EU debt securities to EU IBAs for HSBC is consistent with the greater weight of total securities to total IBAS on its consolidated balance sheet.

Finally, HSBC’s French and German subsidiaries’ silo accounts give us details on the other IBAs held by those two entities. Because they constitute the large majority of HSBC’s EU loans to customers and IBAS, we can use their silo reports to sanity check the figures we have estimated for HSBC’s total EU exposure. In so doing, it should be born in mind that these are full consolidated accounts, not exposure reports. They therefore report a more comprehensive amount of data than the exposure reports, and include equity securities holdings which do not appear in the exposure reports, as discussed above.

One interesting point to emerge from HSBC France’s DDR, which helps shed light on the large percentage of securities in its total EU IBAS, is that HSBC France’s insurance activities hold a significant volume of securities (p 175):

HSBC insurance securities.PNG

HSBC France’s insurance division holds €21,370 (~$23bn) of securities. HSBC discloses on page 123 of its annual report that its European insurance division’s holdings of securities were ~$24bn, so these securities holdings in HSBC France almost represent the totality of HSBC’s EU insurance division’s securities holdings. Apart from the small derivatives exposure of €149m, these securities will be included in the BIS statistics for financial claims of UK headquartered banks on EU entities. €5.6bn of that exposure is equity exposure which would not be included in HSBC’s statement of credit risk exposure and which is part of the estimated $100bn of equity exposure we have excluded from the estimated BIS recorded $333bn total claims of UK headquartered banks on EU entities held in the form of securities in order to arrive at our $233bn estimate of of UK headquartered banks’ exposure to EU debt securities. These HSBC France insurance securities holdings represent, converted to dollars, 7% total UK headquartered banks’ $333bn EU securities claims recorded by the BIS and 18.3% of HSBC France’s IBAs.

We can go on to juxtapose the figures for HSBC France and Germany to those we estimated for HSBC’s EU exposure as a whole in order to calculate what they imply for HSBC’s “other” EU exposure and see if this is reasonable. In other words, we will subtract HSBC France and HSBC Trinkaus from the total estimate for HSBC EU, and analyse the residual, which consists of HSBC’s loans to customers and banks in other EU countries and EU securities held both in “other” EU countries and in its headquarters, displayed in the “Other etc” column below.

The BIS in its guide to locational banking statistics writes that for holdings of debt securities “allocation to the counterparty country is recommended to be done according to the residence of the issuer” (p 9). Therefore, if some of the debt securities held in HSBC France or HSBC Trinkaus are not issued by entities resident in the EU, then those securities should not be included in the BIS statistics for claims on EU entities. Similarly, loans by HSBC France and HSBC Trinkaus to banks outside the EU would not be included in the BIS statistics for claims on EU entities. We therefore need to remove any non-EU exposure from the silo accounts to make them comparable with the consolidated accounts.

HSBC France’s credit risk report on page 113 shows a breakdown of credit exposure into exposure to France, other EU and outside the EU.  The table shows that there is no exposure to banks (and negligible credit exposure) outside the EU. In its risk report on exchange risk (p 159) HSBC France confirms that its exchange rate risk is tied to investments in “subsidiaries, branches and affiliates” and not to securities or IBAs. We do however need to remove HSBC France’s equity holdings to make the numbers comparable with HSBC’s consolidated EU securities exposure, which excludes equities (the information is from HSBC France’s 2016 DDR n 8 p 228, n 11 p 246 and n 13 p 253):

HSBC France debt securities

HSBC Trinkaus for its part records small amounts of deposits received from foreign banks on the liabilities side of its balance sheet in note 31 (p 119), but nothing on the asset side. Its credit risk disclosures (p 58) show that 6.7% of 2015 credit exposure is to non-EU countries and 1.8% other EU countries. Credit exposure in 2015 amounted to ~€31.4bn, however, more than the entire HSBC Trinkaus balance sheet, and is therefore calculated in a different way to net loans and advances on the face of the balance sheet. We therefore can’t use the numbers disclosed – as these are disclosed on a very different accounting basis from the balance sheet figures – but we can use the percentages as a guide. The ex-Germany loans revealed by this disclosure should in theory be included in the $75m consolidation eliminations between the Trinkaus figures and the HSBC consolidated German exposure figures, calculated above. As we shall see below, the loans ex-EU figure calculated using the 6.7% as guide ($526m) is greater than the Trinkaus consolidation elimination figure ($75m), suggesting the estimate using the 6.7% figure is too high. However, given we are dealing with small figures (Trinkaus’ total IBAs are just over an eighth of HSBC France’s), the impact of any error on our overall calculations will be behind the decimal point.

Note 57 (p 159) discloses that €2,883m of assets were in foreign currencies. Some of these will be loans in foreign currencies, some will be equities and the rest will be debt securities. We don’t know what percentage of Trinkaus’ EU securities are equities, but we can use the 10% of Trinkaus’ total securities holdings consisting of equities, calculated here, as a proxy:

Trinkaus equity %

We can accordingly perform the following calculation:

Trinkaus EU securities

In short, only one adjustment for equity holdings is needed for HSBC France, but Trinkaus, as an internationally active investment bank, seems to have just under half of its securities holdings outside of the EU.

Combining the loans and advances to customer numbers calculated above for HSBC’s exposure to EU entities as a whole with HSBC France’s accounts translated into dollars and HSBC Trinkaus’ accounts adjusted for ex EU securities and translated into dollars, you get the following picture:

HSBC other EU IBA composition

The loans to customers in “Other” EU will consist mainly of loans made direct from HSBC plc via branches in the rest of the EU or from branches in London, without going through any active subsidiary. The securities and loans to banks represent the loans to EU banks and investment in EU debt made by HSBC’s central liquidity management desk. Securities may appear large compared to loans to customers, but that is because we are comparing a small loan book to a large central securities investment function. In any case, the resulting makeup of “Other” Europe’s IBAs is not dissimilar to that of HSBC France. In other words, the composition of HSBC’s “Other” Europe IBAS implied by our assumptions for HSBC’s total EU exposure is perfectly reasonable.

Taking a step back, it is interesting that if we try to reconcile BIS and individual account figures, the implied holdings of EU debt securities by HSBC is below the percentage of EU securities held by all international banks, but leads to a somewhat higher percentage of IBAs held in EU securities relative to HSBC’s UK peers. Hopefully this little excursion through bank balance sheets will have demonstrated that they are influenced by a variety of factors and as a result come in a variety of shapes. In that context, the outcome of our assumptions for HSBC’s EU exposure are not unreasonable.

Note that the figure for HSBC’s gross loans and advances to EU customers, which is the only one that matters in terms of estimating the impact of Brexit, is disclosed by HSBC. The only thing we had to do was estimate the level of provisioning for the relatively small proportion of its EU loans outside France and Germany, which, at around 1% (based on the figure for HSBC’s total loans and advances to customers), is hardly going to impact the result. So this is a solid number for which minimal estimation was required.


UK conclusions

There are three key take-away from this analysis:

  1. BIS figures seem satisfyingly consistent with the figures reported by individual banks;
  2. Out of the $539bn exposure ($641bn claims) only $209bn or around 40% was in the form of loans and advances to customers, for which a passport may be required;
  3. A large proportion of BIS recorded claims are in the form of freely tradeable equities.

To calculate the income impact of those loans to customers I have applied each bank’s revenue margin to its EU assets, adding 1% to their NIM to reflect the higher margins on loans to customers, since we are dealing exclusively with loans to customers here, and not other, lower margin assets. This is only an approximation, of course, since EU margins may be different from overall margins. This approximation will likely exaggerate the income estimated to be earned by these banks in Europe, as it would be reasonable to expect that their overall margins, dominated by the UK where they enjoy dominant market share, are higher than their EU margins, where their market shares are often small. To estimate Barclays’ total return on IBAs figure I have assumed their fee income was 0.62% of IBAs, like RBS’ (RBS fee income/average loans to customers score was roughly half-way between that of the other two peer banks). Since we are trying to calculate a figure consistent with OW’s report, based on 2015 revenues, I have used the 2015 $/£ average rate of 1.48 in this calculation.

UK bank EU income

Based on known UK headquartered banking asset margins and disclosed loans to EU consumers in their balance sheets, and consistent with BIS data, we can estimate with some confidence that they earned $6.5bn of banking income, or £4.4bn at 2015 exchange rates. Although by no means an insignificant number, this is less than a third of the £14bn revenue estimated for the Retail and business banking sector by OW.


UK bank exposure to the EU in structural decline

Note that, at current exchange rates, the impact would be higher (£5.3bn), but since 2015 the assets of UK headquartered banks in the EU have changed too, so you would need to take all changes into account to be consistent. Looking at the latest available company data, UK bank loans to EU consumers declined over 10% in 2016 versus 2015 in dollars, following the same trajectory as the BIS data, with three out of four banks reporting declines and none an increase. Note Barclays’ $12bn reduction in loans to EU consumers happened after its sale of SAU to La Caixa.

UK bank EU customer loans yoy

This trend is likely to continue, as a number of markets are identified by the UK banks as non-core; Lloyds for example describes £4.5bn of Irish loans as being in run-off (p 141):

Lloyds Ireland run off.PNG

Barclays was also shedding EU assets pre-Brexit (n 44 p 365):


And RBS also announced a number of small disposals in Europe due to complete after the balance sheet date on which OW based its forecasts. Note 18 (p 352):


RBS sold its stake in Visa Europe in 2016 to Visa Inc., like the many other banks who were also Visa Europe shareholders. Although this will not affect RBS or the other banks’ loans and advances to customers, it further reduced the claims of UK headquartered banks reported by the BIS (p 13):

Visa Europe

The 2016 decline in assets, in dollars, more or less cancels out the impact of sterling devaluation on the translation of income, so the £4.4bn figure is representative of both the relevant figures for OW’s report and the potential ongoing impact of a loss of EU passport on the big four UK banks.

Furthermore, total announced sales and run-offs of Lloyds and Barclays to take place after 2016 amount to £9.1bn, which was $11bn at end 2016 exchange rates, or a further  5.5% of total loans and advances to EU customers by UK headquartered banks in 2015.

The reason this decline is structural is that banks across the world are in the process of strengthening their balance sheets to meet the tighter capital requirements imposed by Basel III and the prudential cushions applied to them by national regulators. Gone are the days of Pan-European empire building. They therefore have less leeway in their balance sheet to sustain small operations across multiple countries. The only exception to this trend of concentrating the balance sheet are markets with strong growth and high return potential. The EU is on the whole a competitive, mature market and therefore does not meet those criteria.


Not all companies need a passport

Note that this £4.4bn is a maximum. That’s because not all loans to customers require a passport. As discussed above, large customers like H&M or Benneton are able to borrow internationally. We haven’t, at this stage, attempted to quantify what percentage of UK headquartered loans to EU customers are to such large corporates. Moreover, you may have noticed in Lloyds disclosures that they had just over £2bn of exposure to “Other” financial institutions. These non-bank financials (NBFIs), such as insurance companies or mutual funds, are regulated financial companies and can, by virtue of their everyday exposure to and familiarity with financial markets, buy Swaps and other financial products from banks outside the EU. At this stage we are unable to do anything other than guess what percentage of the loans made by UK headquartered banks to non-bank EU customers are to NBFIs, but our analysis below of the US bank balance sheets and of the BIS data on UK located claims on EU entities will provide further pieces of the jigsaw which will allow us to quantify this exposure with a bit more precision.


UK branches of non-UK banks

So far we have focused on the ~$600bn claims of the big four UK headquartered banks. We now need to turn to the ~$1 trillion claims of the non-UK banks via their UK located branches. To do so, as discussed above, we need to look at the BIS data from its A.6 reports for claims of  branches located in the UK on EU entities. Here are the figures for total claims (which we summed up to arrive at the ~$1.6 trillion headline figure used in our simplified table above):

A6 country detail.PNG

There are some interesting country level differences with the B.4 tables above, with France the biggest recipient of bank funding from UK located bank branches in the previous table, whereas Germany was the biggest recipient of funding from UK headquartered banks in the table presenting the B.4 data.

We can subtract the claims of UK headquartered banks (the $641bn taken from the BIS B.4 reports and presented above) for each country from the (with one small exception) larger figure for claims of bank branches located in the UK on that country (the $1.66 trillion taken from the BIS A.6 reports and presented in the previous table). Here is the result:

BIS A6 B1 residual.PNG

The sum of the claims booked in the UK branches of non-UK headquartered banks for these different countries is what we sum up to arrive at the ~$1 trillion dollar used in our simplified table above. Note the claims of UK headquartered banks on Malta are higher than those of UK located banks on Malta (the one small exception alluded to above), possibly because there are many British citizens in Malta who may have accounts with the Malta branch of a UK bank (however, the only UK bank listed as having a branch in Malta is HSBC).


The big five US investment banks

This $1 trillion will be comprised of the claims of a myriad of banks and trying to reconstitute it from the bottom up in its entirety, as we did for the four UK banks, would be an impossible task. We can, however, begin by looking at the disclosures given by the top five US investment banks in their SEC 10ks or annual reports and accounts, which one would expect to constitute a large part of this ~$1 trillion in cross-border bank claims, and pretty much all of whose EU loan exposure will be booked in their branches in the UK.

To put these banks’ EU related disclosures into context, we should first look at their overall balance sheets, as we did with the UK banks. The assets on Goldman’s balance sheet, reproduced here from page 112 of their 2016 SEC 10k, demonstrate some important differences between US investment bank balance sheets and those of their UK peers:

Goldman BS.PNG

  1. Rather than loans to banks Goldman has “cash and cash equivalents” of $121.7bn. These are presumably deposited with Goldman or another bank, but their designation as cash demonstrates that the primary purpose of this cash is to help the business function; it isn’t part of an interest margin and funding strategy as it would be for a conventional bank.
  2. You can immediately see that “loans receivable” from customers at $49.7bn are a mere 5.8% of Goldman’s 2016 total balance sheet. Lending to customers is a much smaller part of what Goldman and other investment banks do than it is for conventional banks.
  3. Almost equal in size at $47.8bn is the line “customers and counterparties.” An example of what this represents might be money owed to Goldman by clients for securities purchased through Goldman’s trading desk but not settled on the balance sheet closing date. These amounts are not IBAs. But they will be classified as financial claims by the BIS (as confirmed by the helpful BIS staff member). For want of a better option, I include these in loans to banks as they are mostly owed by banks and other financial customers.
  4. Repos at c $300bn or around a third of total assets are a much bigger part of Goldmans’ and most US banks’ balance sheets than they are for their UK counterparts.
  5. One notable absence from the face of all US bank balance sheets, except for the small figure in Bank of America’s, is a line for derivatives. As this extract from Citigroup’s 2016 10k (n 22, p 239) shows, US banks record their derivative assets and liabilities on the balance sheet net of significant offsets, including $519bn of netting agreements in this case. This significantly reduces their size on US banks’ balance sheets, relative to what it tends to be on a UK or European bank’s. You can find the derivatives balance sheet values in the notes to the accounts, but I have not included the small amounts unless they appear on the face of the accounts, to assist comparison of my figures with the figures as they appear on the face of the accounts.

Citigroup derivatives

Here are the big five US banks’ 2016 consolidated balance sheets in millions of dollars presented with the same template as used above for the UK banks:

US bank asset breakdown

The balance sheets are not wildly different in size from those of their UK counterparts. Here is their composition in percentages, with the aggregate balance sheet of the UK banks added for the purpose of comparison:

US bank asset breakdown %.PNG

Morgan Stanley and Goldman have much smaller conventional banking activities than the other three. That’s why loans to customers are such a small part of their balance sheet, as highlighted above in connection with the Goldman balance sheet. It therefore makes sense to compare the UK banks to the three other US banks which are not so “investment banking centric.” This comparison is displayed in the table below. It reveals that Citigroup and JP Morgan, with more long standing and substantial involvement in conventional bank lending than Goldman and Morgan Stanley, have similar balance sheet compositions to RBS and Barclays, while Bank of America, which is less exposed to investment banking (despite its acquisition of Merrill Lynch), is more similar to Lloyds in terms of customer loans as percentage of IBAs (though note Lloyds’ reliance on inter-bank lending versus Bank of America’s on securities, for the balance of their IBAs).

US UK balance sheets detailed comparison

In other words, adjusted for business mix, the make-up of these three “non-investment banking centric” US banks’ IBAs is not very different from that of their relevant UK peers.


US big five EU exposures

All five of these banks give information on the size of their largest country risk exposures, which include EU countries. Some banks report information on their exposure to the top twenty countries outside the USA, others information on their exposure to any country which represents more than a certain percentage of their total balance sheet (above 1% of consolidated assets in the case of Morgan Stanley). The countries, and above all EU countries, featuring in these tables varies from bank to bank and from year to year, depending on their activity at the time. Here is the raw data as presented in their SEC 10ks or their reports and accounts. As with the UK headquartered banks above, we will be looking at balance sheet positions in 2015, the year on which OW bases its estimates.

Citigroup (2016 SEC 10k, p 118):


JP Morgan (2015 annual report and accounts, p 141):

JP Morgan.PNG

Morgan Stanley (2015 SEC 10k, p. 257):

Morgan Stanley country exposure.PNG

Goldman Sachs (2015 10K, p 214):


Bank of America (2015 SEC 10k, p. 87):


These disclosures allow us to add up the exposure to the EU countries listed above to arrive at a preliminary estimate for the banks’ total EU exposure. This preliminary estimate will be understated, as each bank will have exposure to other EU countries not large enough to be listed in the tables. To extrapolate from the EU exposures disclosed in these tables for the banks’ largest countries of exposure to the total EU exposures I have used the following principle. Each of these five banks lists between three and seven EU countries in their list of top country exposures in 2015. Looking at the BIS data above, we can see that, for international banks as a whole, the top three to seven EU countries represent a large percentage of total claims on EU counterparties:

Top BIS countries as percentage of total

We can use this to estimate the total exposure of each bank by grossing up its disclosed exposure by the percentage of total claims represented in the BIS A.6 tables by the number of countries included in their disclosure:

Top five EU exposures.PNG

The big five US investment banks disclose $397bn of EU exposure to their largest EU countries which, we can extrapolate, represents around $545bn of exposure to all EU countries. If anything, this is a generous estimate. The BIS data covers all banks in all countries. They are therefore the aggregate of some banks active in some countries and other banks active in others. Very few banks in the BIS data will lend to all the borrower countries on which it collects data. For example, many Scandinavian banks have operations in the nearby Baltics, but hardly any exposure to Portugal. Most banks in Spain have some exposure to Portugal, but hardly any to the Baltics. Therefore, the top countries will be a lower percentage of total in the BIS data than they will be for any individual bank. That would suggest that US banks exposure outside of the countries they disclose is smaller than what we have estimated using the BIS data as a guide.

The numbers do not seem at all unreasonable at the headline level however. $545bn is almost equal to the four UK banks’ $539bn EU exposure. To put this into further perspective we need to compare this figure to the total claims on EU entities of non-UK headquartered banks located in the UK and reported by the BIS. We can see that the top five US investment banks’ estimated exposure of $545bn represents just under 55% of the of $992bn total exposure for non-UK headquartered banks with branches in London (estimated above from the BIS claims figure of $1,022bn), which further underlines the importance of these five entities. The four UK and the top five US banks together ($1,085bn) represent ~70% of total BIS exposure of $1.55 trillion, suggesting that the distribution of claims on EU entities among foreign banks follows something like a Pareto law.

In order to calculate the income earned from the EU exposure held by the top five US banks which might be under threat as a result of Brexit we need to understand more about their composition, as we did with the UK banks. The US banks break their EU and other country exposure down into similar buckets to those used by the UK banks. But there are important differences. Goldman, Citigroup and Morgan Stanley divide their exposure by counterparty, not by instrument. They accordingly group their exposure into three broad counterparty risk buckets: banks, governments and “other”; Morgan Stanley and Citi further divide their “other” exposure into NBFIs and other “other”, because of the large size of their NBFI exposure (discussed further below).

The figure for exposure to banks should correspond broadly to the “loans to banks” category we saw above with the UK banks, but may also include bonds issued by banks. Government exposure will almost exclusively be via securities, but may include loans to government or government guaranteed entities. “Other” covers a multitude of sins. It will include loans to customers, but also the receivables from customers and counterparties we discussed above in connection with Goldman Sachs, for example money owed by a French pension fund for a security purchased through Goldman’s trading desk in London will be included here. A larger chunk of the “other” exposure will include non-governmental securities such as corporate bonds. Goldman’s “other” EU exposure of $35bn compares with the total customer loans on its consolidated balance sheet of only $45.4bn in 2015. If Goldman’s $35bn “other” EU exposure was all loans to customers, then 77% of Goldman’s total loans to customers would be to EU entities. That seems very unlikely.

Bank of America and JP Morgan, meanwhile, split their exposure into loans and either securities, in Bank of America’s case, or trading and investment in JP Morgan’s case (which is broadly the same thing as securities), but, unhelpfully, they don’t split their loan exposure into loans to banks and loans to other customers. Note that only Bank of America, which has the smallest EU exposure of the big five, has more loans to customers on its balance sheet than securities.

Here is a summary table, in billions of dollars, based on disclosed and identified EU exposure (not on extrapolated total EU exposure):

Top 5 EU exposure breakdown

Loans (for JP Morgan and Bank of America) and “Other” exposure (for Citigroup, Morgan Stanley and Goldman) are only a starting point to estimate the assets which might be affected by Brexit. To make our estimates more precise we need to make a couple of assumptions:

  1. In the case of Citigroup, Morgan Stanley and Goldman Sachs, what percentage consists of “other” exposure consists of loans to customers and what percentage of non-government securities and receivables from trading customers?
  2. In the case of JP Morgan and Bank of America, what percentage of “loans” are loans to customers and what percentage loans to banks?

To estimate that, we will take the figures which each bank discloses for these line items for the bank as a whole, and apply the percentages to their “other” EU exposure or their EU “loan” exposure, as applicable.

Citigroup, Morgan Stanley and Goldman disclose their holdings of sovereign debt as part of their liquidity reserve disclosures on pages 89, 60 and 189 respectively of their 2016 10ks.  Using this data allows us to produce the following estimate of the breakdown of “other” exposures between loans to customers on one hand and non-governmental securities and receivables from trading customers on the other (the consolidated balance sheet data, like the EU exposure data, is from 2015 and therefore will not match the 2016 data displayed above):

loans as a percentage of other exposure

Using Citigroup as an example, total securities disclosed on their balance sheet amount to $804bn. If we subtract the $307bn of government security exposure in their liquidity report from this we are left with $497bn non-governmental securities. Add $605bn of loans to customers and $28bn receivables from trading customers to this and you get $1.13 trillion of balance sheet items which would be classified as “other” in their EU exposure disclosures. Customer loans are 53.6% of that exposure ($605bn/$1.13 trillion). This allows us to estimate that 53.6% of Citigroup’s $55bn “other” EU exposure consists of loans to customers, which is $29bn (53.6% x $55bn). If we use the overall balance sheets as a guide in this way, we can estimate that of the $134bn “other” EU country exposure reported by the banks, only $39bn is loans to customers and the rest is holdings of non-governmental securities and trading receivables.

To sanity check this estimate we can compare its results to an alternative calculation. The consolidated balance sheets of these three banks allows us to calculate a ratio between government securities on one hand and other securities and trading receivables on the other. By applying this ratio to the disclosed EU government exposure, we can estimate a figure for EU non-government securities and trading receivables. We can subtract that figure from the disclosed “other” exposure to estimate EU loans to customers:

loans as a percentage of other exposure alternative calc

This alternative calculation yields an even lower combined number for the three banks, supporting the hypothesis that loans to customers are a small part of “other” EU exposure for these three banks. In order to be generous to the theory that significant revenues are at risk from Brexit, when calculating the impact of Brexit on the EU retail and business banking revenues of these three banks I will take the highest estimate for loans to customers from these two alternative calculation methods, in other words I will retain the figures of $29bn and $3bn for Citigroup and Goldman, respectively, from the first calculation method, and the figure of $29bn for Morgan Stanley from the second calculation method. This is generous, as $29bn EU loans to customers would represent a full 40% of Morgan Stanley’s total $73bn loans to customers. The reason the EU loans to customers estimate is so high for Morgan Stanley using this calculation is that they have low exposure to EU governments. Although this Morgan Stanley loans to customers exposure is perhaps the estimate with which I am least happy, it does not materially influence the final impact calculation as the increase in the estimate for Morgan Stanley and the US banks merely reduces the estimate for the impact attributable to banks outside the four UK headquartered banks and the US top five banks. The result is summarised here:

Loans in other exposure summary

According to this calculation, just under half of the three banks’ “other” exposure is in the form of loans to customers. This is despite the fact that Morgan Stanley and Goldman, who do very little customer lending, represent a sizeable proportion of the big five other exposure to EU entities, due to the generous assumptions used to calculate the Morgan Stanley figure.

Turning to JP Morgan and Bank of America, we can simply take loans to customers as a percentage of loans to customers and banks for the balance sheet as a whole, and apply this percentage to the figures they disclose for EU loans:

Loans to customers percentage loans

These figures suggest that of the $68bn EU loan exposure disclosed by JP Morgan and Bank of America, $50bn was loans to customers.

We can use these calculations to present detailed figures for the individual banks, using the same template as for the UK banks, based in the first instance on the $397bn disclosed exposures, to make it easy to match them with the figures above. Below the total IBAs figure I have, for the sake of clarity, provided a reconciliation of the different types of IBAs with the figures disclosed by the banks in their country exposure disclosures and the estimated figures for loans to customers calculated above:

Top 5 EU exposure breakdown adjusted disclosed

We can then translate these figures for disclosed exposures into the figures for total exposure using the same methodology as used above for the total IBAs, to arrive at a breakdown of the total exposure figure of $545bn:

Top 5 EU exposure breakdown adjusted total

In summary, we have estimated with more precision that $145bn of the US banks’ exposure to EU entities is really at risk from Brexit, after classifying ~$20bn of loans included by JP Morgan and Bank of America in their “loan” exposure” as loans to banks and ~$110bn of Citigroup, Morgan Stanley and Goldman’s “other” exposure as non-government securities and trading receivables.

As a sanity check, as we did with the UK banks, we can compare the percentage breakdown of IBAs implied by these estimates for each bank’s EU exposure to the percentage breakdown of IBAs for each bank as a whole:

Top 5 EU exposure breakdown % EU and tot

And for the big 5 banks combined:

Top 5 EU aggregate exposure breakdown % EU and tot

The gap between loans to customers as a percentage of total IBAs at the EU and total level is 7.8% (26.7% EU versus 34.5% total) and ranges from 29.6% higher for Morgan Stanley (39.6% EU versus only 10% total) to 16.6% lower for Citigroup (21.5% EU versus 38.1% total). These differences are not unreasonable in themselves, and compare, as discussed above, to a gap of up to 27% for RBS and 16.5% for Barclays. The UK gaps are based on hard, disclosed figures and are bigger than those we have estimated for the US, which strongly supports the view that our estimates do not underestimate EU loans to customers as a percentage of total EU exposure for the big US banks. As we noted with respect to the UK banks, loans to customers are the hardest assets to accumulate, as they require a network, franchise, expertise and critical mass not required by other types of IBAs. It is therefore perfectly reasonable for the top five US banks’ EU customer loans to be lower as a percentage of total EU IBAs than customer loans for their total balance sheet, which is dominated by the US where they have long established franchises and large branch and intermediary networks.

Making loans to customers is a business with significant overheads. You need credit assessment, compliance, default management and customer servicing. If you are too small, you won’t have the credit experience to assess future loans or enough diversification in your current book. It therefore makes perfect sense that the further down the size scale you go the smaller the percentage of your IBAs will be in loans to customers. Since the US and UK banks are smaller in Europe than they are in their home markets, it makes perfect sense for loans to customers to represent a smaller percentage of their IBAs there.

The fact that Morgan Stanley and Bank of America’s EU loans representing a greater proportion of EU IBAs than total loans do of total IBAs is certainly an anomaly, but could be explained by the fact that Morgan Stanley has both small loan and EU exposure overall, meaning that a few large loans in the EU could have a big impact on the relative weight of loans in the EU relative to the bank as a whole, while Bank of America may have made a commercial or investment decision to expand the loan book in the EU without building up other EU exposure. In any case, this anomaly merely results in an increase in our estimate of the impact of Brexit on the income earned from EU assets by the big five US banks, and represents a measure of conservatism in our figures.

This breakdown of IBAs can be compared to the comparable figures for the UK banks (both in billions of dollars):

UK and US exposure breakdown $bns

The numbers are broadly similar for the UK and the US top five, suggesting that the assumptions we used were reasonable. This is the result expressed in terms of % of total exposure:

UK and US exposure breakdown %

The key line is loans to customers, and here our assumptions result in us estimating that these represent a lower percentage of the top five US banks total EU exposure than they do for UK banks. However, if we strip out investment bank focused Goldman Sachs and Morgan Stanley, who have a very low percentage of their balance sheets devoted to loans to customers, the two aggregate balance sheets are very similar:

UK and US ex MS and GS exposure breakdown %

As mentioned, the figures for other exposure for Morgan Stanley and Citigroup includes exposure to NBFIs, as did Lloyds Bank’s. Since not all banks break NBFI exposure out, I have chosen to leave it in loans to customers to maintain consistency. However, these exposures are different from loans to non-financial customers in terms of their relevance to the Brexit debate: as mentioned above, insurers like Allianz or fund managers like Belgium’s Degroof Petercam can borrow from UK located lenders without a passport, since they are financial companies themselves. Here is the EU NBFI exposure gleaned from the three banks which disclose it (Lloyds’ figures have been converted to dollars):


To put this into context, ~$61bn, or more than 10% of the total top 5 US banks’ EU exposure of $545bn, is represented by the disclosed NBFI exposure of only two banks. We can subtract the disclosed NBFI figure in order to calculate the amount of loans to non-financial customers. However, the NBFI figures disclosed by Citigroup and Morgan Stanley are simple exposure figures. They do not distinguish exposure through loans and exposure through securities such as bonds. For the sake of simplicity, we will assume that the whole disclosed $61bn is in the form of loans, but also that the disclosed exposure is equal to the total exposure, in other words that the banks do not have NBFI exposure in the smaller EU countries they didn’t disclose. It also assumes the other three banks do not have meaningful NBFI exposure. The assumptions offset each other, and effectively we end up assuming that any NBFI exposure in the form of securities is offset by loans to NBFIs in Citigroup and Morgan Stanley’s smaller countries and by loans to NBFIs by JP Morgan, Bank of America and Goldman Sachs.

US bank adjusted loans to customers

Having estimated that the top five US banks held $85bn of loans to non-financial customers at the end of 2015, we can assume that those assets generated the same 3.13% return that was generated in aggregate by the UK banks in order to estimate the following retail and business banking income at risk from Brexit for those banks:

US income estimate

With these assumptions, the income at risk would be $2.6bn or £1.8bn. Adding this number to our estimate for the UK gives you a total for the nine banks of £6.2bn, less than half the total £14bn revenue estimate derived from OW.


Using BIS data to calculate EU exposure for the rest of the UK located banks

We now need to examine the exposure of UK located banks to EU entities not captured by the big five US investment banks and the four UK headquartered banks. This represents $469bn in exposure (the numbers in this chart are different from those in the chart above because that chart showed claims numbers and this one shows exposure numbers):

Other countries exposure

That exposure will be divided up among:

  • Swiss banks;
  • Japanese banks;
  • Canadian banks;
  • US banks banks with lower EU exposure than the big five;
  • EU banks such as Deutsche Bank which might book claims on EU entities in their London branches for a variety of reasons;
  • Asian banks, Middle Eastern banks and others.

To assess the impact of this remaining exposure, we need to organise the BIS data to make it comparable to the information from the top five US banks and the UK headquartered banks. As discussed above, the A.6 classification of claims on each country (into debt and other securities, loans to banks and loans to non-banks etc.) is similar to that employed by the banks in their individual reports, and should, on the face of it, enable us to figure out the composition of the $469bn of exposure held by the other banks.

Although the BIS does not provide this precise breakdown for claims of banks located in individual countries, it offers a breakdown of those claims into loans and other claims and publishes claims on non-banks, including loans to non-banks, as a memo item, in its A.6-2S tables. Here are the data for UK located claims on entities in Slovakia, for example:

A6.2 S screenshot

Total loans and advances ($633m for Slovakia) minus loans and advances to non-banks ($539m for Slovakia) is equal to loans to banks ($94m). Total claims ($550m) minus total loans ($633m) equals the total of debt securities, other securities and unallocated. As you can see with Slovakia, there is a negative unallocated line which represents a consolidation adjustment of the loans figure. Such discrepancies only occur in small markets like Slovakia. The BIS, as can be seen from this screenshot, does not publish the breakdown of loans to non-banks into loans to NBFIs and loans to non-financial customers for the different locations of the banking offices making the loan. However, we can use the breakdown of loans to non-banks into loans to NBFIs and loans to non-financial customers for the country as a whole, provided in the main A.6 report, to estimate that breakdown for banks located in the UK. For example, loans to non-banks in Belgium at end December 2015 were $60,592m, of which $48,631 or 80.3% were to non-financial customers and the $11,961m (19.7%) balance to NBFIs. The A.6-2S reports that loans to non-financial customers in Belgium by UK located banks was $11,082m at the end of December 2015. We can use the percentages for Belgium as a whole to estimate that this was broken down into $8,899m loans to non-financial customers (80.3% x $11,082m) and the balance of $2,183m (19.7% x $11,082m) to NBFIs.

Here are the raw figures for loans by UK located banks to non-bank entities in each EU country:

UK nonbanks


Tax driven lending

These figures are not estimated or adjusted figures but copied verbatim from the relevant A.6-2S reports. One thing jumps out at you from them. The top three countries for UK located bank loans to customers claims are the Netherlands ($103bn), Luxembourg ($100bn) and Ireland ($75bn), which together represent over half of the claims on non-bank customers in the form of loans. This is rather odd. Luxembourg is bigger than France and Germany combined and the Netherlands is almost equal to all the remaining countries put together. With no disrespect to Dutch corporate innovation and enterprise, it is pretty clear that not all of these loans are to Dutch, Luxembourgeois or Irish industrial companies or individuals. What all three countries have in common is either a low corporate tax rate, in the case of Ireland and Luxembourg, or the offer of attractive tax structures, such as the legendary Dutch B.V.

The BIS makes clear in its guide to its locational banking statistics (i.e. those which include the A.6 report we are studying) that the counterparty country of a loan is the country where the claim is located, not the country in which the parent company and ultimate counterparty of the claim is located (p 30):

foreign subsidiary definition

What this means is that the BIS stats are, to a large extent, picking up loans to tax driven structures in these EU countries. The ultimate borrower in such cases may not even be an EU entity at all. These loans could be to Amazon Eurofinance plc, Dublin or Nestlé B.V., Amsterdam or Proctor & Gamble sarl, Luxembourg. This means that the BIS figures may include exposures to non-EU entities booked in EU tax structures; such exposure will not be recorded as European exposure in the company accounts, which will look through the tax structure to the underlying borrower who determines the credit risk. It also means that a significant proportion of the borrowers in those three important countries are large, financially sophisticated companies, which has an important impact on the calculation of what percentage of loans to EU entities from banks located in the UK will be affected by the passport.


BIS exposure calculation for banks located in the UK

The figures for each individual country can be combined to produce the following breakdown of UK located claims on all EU entities, which can in turn be translated into exposure figures using the assumptions outlined above:

claims exposure total

Our $209bn in loans to EU customers by the UK banks and of $145bn by the US banks include NBFIs. They therefore have to be compared to the BIS exposure figure for total loans to non-banks of $477.6bn, calculated by deducting credit provisions of 1% from the BIS disclosed claims figure of $482bn. This would imply $123bn in remaining loans to EU customers by the other banks ($477.6bn minus $209bn minus $145bn).

Note that the NBFI loan claims figure of $215bn is almost the same as that for loan claims on non-financial companies of $267bn. This figure can also be compared to Morgan Stanley and Citigroup disclosures, which together, at $61bn of exposure, are equal to ~30% of the entire NBFI loan exposure estimated from the claims data recorded by the BIS. This probably explains why they are given separate disclosure by those two banks.

Here are the BIS numbers as percentages, compared to the UK banks exposure to the EU and the US banks total exposure, excluding Morgan Stanley and Goldman given their investment banking focus:

BIS breakdown vs UK EU and US ex inv bnk

The BIS locational data data shows debt securities as a percentage of total IBAs for the BIS universe as a whole is less than a third of debt securities as a percentage of total IBAs for the big four UK and three big US banks’ (the big five minus Morgan Stanley and Goldman) EU exposure. This is a significant divergence.


Securities: location, location, location

If we look at the figures in absolute numbers, we can see that the total figure for claims in the form of securities of the BIS at $239bn is barely more than the figure for the UK banks alone of $233bn and $254bn less than the UK’s $233bn and the US top five banks’ estimated $270bn combined. We did make estimates for HSBC’s exposure and for the five US banks, which, one could argue, are responsible for this inconsistency. But if we just add Lloyds, RBS and Barclays to JP Morgan and Bank of America, all of which disclose their European securities exposure, you get the following picture:

specified securities exposure

The explicitly disclosed debt securities exposure of these five banks is $32bn greater than the figure for all UK located banks recorded by the BIS. And clearly HSBC, Morgan Stanley, Goldman Sachs and Citigroup will have some EU debt securities holdings. This demonstrates that there is a different country allocation principle at work between the BIS and the individual banks’ reports. We saw above that the BIS consolidated, headquarter data, seemed to be coherent with the data for UK headquartered banks as collected by the BIS in the B.4 report. The B.4 report for UK headquartered banks records securities issued by EU counterparties in all those banks’ offices. Again, it is a consolidated report, so it consolidates all the securities issued in the EU and held across the world by banks headquartered in the UK.

The UK components of the locational A.6 reports, however, will only record securities issued in the EU and held in banking offices in the UK. The BIS uses the term “banking office” as a collective term which can consist of either a bank branch or a separately incorporated subsidiary. The way that securities holdings are reported by any banking office will depend on whether it is a branch or a subsidiary. If it is a subsidiary, then it is very likely that it will report all of the securities it holds. This will be a clearly established entity with its own accounts and its own solvency reporting. Therefore, extracting data for the BIS will be relatively straightforward (though I’m sure the people who have to do this would roll their eyes if they heard me say so). By contrast, branches will typically not hold securities at all. They are just there to take deposits and make loans. The deposits they take in may be invested in securities, but this will typically be done by the central treasury arm or other part of the bank.

That being said, even if the BIS can record the EU issued securities a foreign bank holds in its UK located subsidiary, that doesn’t mean that it holds all or even any of its EU issued securities in the UK. On the contrary, holdings of EU securities by the majority of foreign banks with UK offices will not be held in their subsidiaries in the UK but in their headquarters, by a central treasury arm or other department. This means that the branches will not report them to the BIS as UK located exposure to EU securities. Although they will appear in company reports as exposure to EU securities, they will not appear at all in the BIS locational statistics for the UK, as they will be held elsewhere.

For UK headquartered banks that will not have much impact; most of their securities holdings will be booked in offices in the UK, where their central treasury arm and other securities management divisions will be located. However, any securities held by their subsidiaries in Europe, such as HSBC France or Trinkaus, will not be recorded by the BIS in their A.6 reports on claims held by UK located banks – because those claims are held by subsidiaries located outside the UK. In the case of HSBC, its two subsidiaries HSBC France and Trinkaus hold $63bn in debt securities, so out of HSBC’s ~$89bn of EU debt securities exposure, only $26bn or just over a third was potentially held at the central level in the UK. All $89bn of that securities exposure will be recorded as headquartered in the UK in the B.4 report, but only up to $26bn will be recorded as located in the UK in the A.6 report (HSBC France’s $60bn and Trinkaus’ $3bn will be recorded, respectively, in the France and Germany A.6 reports). To put this in perspective, the ~$63bn of debt securities holdings from these two HSBC subsidiaries alone, which won’t be included in the BIS A.6 report, is equal to 26.4% of the ~$239bn EU debt securities exposure estimated from the ~$342bn claims number recorded by the BIS, which gives a clear indication of why securities holdings are much lower for UK banks in the locational A.6 reports than they are in the consolidated B.4 reports.

The three other large UK banks will also, to a less significant extent, book some of their securities holdings in subsidiaries located outside of the UK, but we cannot quantify this in the absence of silo accounts (Lloyds’ report to the EBA gives some information on this, but the exposure is calculated in a different way from the reports and accounts).

For the US banks the picture is likely to be very different than it is for the UK banks. Any US bank’s subsidiaries located in the UK do require a certain amount of liquidity, part of which will be held in the form of securities. But the US parent companies are gigantic entities with big, diversified securities portfolios. Therefore, a large proportion of those $270bn estimated holdings of EU securities by US banks are likely to be held at the parent company level, and therefore not in the UK. The low value of UK located bank holdings of EU debt securities recorded by the A.6 data, relative to the value of EU debt securities on US and UK consolidated balance sheets, strongly suggests that the subsidiaries of the other foreign banks located in the UK with exposure to the EU will, like the big five US banks, not have significant holdings of EU securities booked in their UK offices.

Unfortunately, there isn’t enough data to calculate a likely breakdown of UK located EU securities holdings between the banks included in the BIS data. Since banks do not require any authorisation to hold securities issued in any country in the world where there are no capital controls, which – for the time being – includes all countries in the EU, the exact value and breakdown of EU issued securities held by UK located banks will not impact our forecast. Nevertheless, for the sake of clarity, the following table presents a likely, rough breakdown of this exposure reconciling the consolidated securities exposure from the individual banks reports and accounts with the BIS data:

securities adjusted

The $313bn adjustment to the figure derived from the UK and US banks’ individual accounts includes the ~$63bn observed EU issued securities holdings by HSBC in subsidiaries outside the UK, as well as an estimated $250bn (representing the majority of the total $270bn) of EU issued securities held by the five US banks we have analysed in their headquarters or in subsidiaries outside the UK. This implies a small holding of securities ($49bn) by the smaller banks outside of of the nine UK and US banks we have analysed, and which probably consists of regulatory capital for any subsidiaries maintained by them.


HSBC France and Trinkaus loans to customers not located in the UK

We know that HSBC France’s ~$41bn and Trinkaus’ ~$8.5bn loans to customers are not booked in UK locations but at the subsidiary level. They will therefore not show up in the BIS A.6 report. The other UK banks, by contrast, do not have important European subsidiaries. RBS’ ~$1bn loans to customers in the Netherlands were in all likelihood booked in their Dutch subsidiary (inherited from their acquisition of ABN Amro) and mostly funded at the subsidiary level. The fact that Lloyds plans to turn their Berlin branch into a subsidiary implies that they didn’t have a subsidiary in the EU. Barclays’ main European subsidiary was Spain, which it sold. The French activities were only established as a subsidiary, Barclays SA, when preparing Barclays France for sale, and so did not constitute a subsidiary in 2015. Barclays’ Dublin subsidiary has a banking license, but was not fully capitalised and so did not hold significant IBAs of any sort in 2015 (this will change when Barclays starts to use Dublin as its EU subsidiary post-Brexit).

As for the US banks, we can assume all their loans to EU customers are located in the UK, either in the form of direct loans or in the form of intragroup loans to any EU subsidiaries they have. Other foreign banks with UK offices may also make some of their loans to EU entities through their EU subsidiaries. However, as with the US banks, and as we shall see below, these will, to a significant extent, be funded by intragroup loans from the London office to the EU subsidiary. Therefore, even if made via an EU located subsidiary, these will be picked up in the A.6 tables’ records of UK located claims on and exposure to EU customers, but as intragroup loans to banks rather than as direct loans to customers. Here, accordingly, are the UK consolidated EU loans to EU customer numbers, reconciled with the A.6 BIS numbers:

loans to customers adjusted

In terms of Brexit risk, the loans booked in HSBC France and Trinkaus and in RBS’ Amsterdam subsidiary are already held in the name of a regulated EU entity, and therefore not at risk from any loss of EU financial services passport. However, given the BIS $468bn total of loans to non-bank customers hasn’t changed, this ~$51bn reduction in loans at risk from UK headquartered banks merely means that there is a corresponding increase in loans to non-bank customers written by banks outside the UK headquartered banks and the US top five, which now stand at an estimated $175bn.


Intragroup loans: loans to customers in disguise?

We now turn to the intragroup loans with which, as just discussed, banks transfer funds from one office (branch or subsidiary) to another. By way of background, we can observe that the percentage of IBAs consisting of loans to customers in the BIS numbers as a whole is higher than it is for the nine UK and US banks we analysed, but that is purely due to the lack of securities exposure. If we focus on the ratio of loans to customers to loans to banks, we see that, roughly speaking, loans to customers are ~1.8x loans to banks for the BIS data as a whole, compared to ~2x for UK banks and over ~3x for the US banks ex Morgan Stanley and Goldman. In other words, the BIS sample’s loan exposure as a whole seems more exposed to loans to banks than the US and UK banks whose figures we analysed above.

This picture does not entirely reflect the underlying reality. As highlighted above on many occasions, the numbers in the BIS A.6 tables are not consolidated. They will therefore contain loans from one bank’s offices located in the UK to one of its offices in an EU country. The low percentage of loans to customers and high percentage of loans to banks in the BIS numbers, relative to the numbers in the individual bank balance sheets we have looked at, is likely due in large part to this. The BIS gives us a figure for the loans from UK located banks to EU banks which are intragroup loans to one of their banking office in the EU. In some cases, these loans will be used by that EU banking office to make loans to EU customers. The net effect in that case would be that a loan has been made to a third party EU customer from an office in the UK, but that loan has first passed through a banking office in the EU. On the bank’s consolidated balance sheet, the intragroup loan will be eliminated, and what the BIS records as an intragroup loan to a bank in the cross-border statistics will appear on the bank’s own balance sheet as a loan to a customer. This means that if we were to reclassify those intragroup loans from one office to another office of the same banking group, which the recipient office uses to make a loan to a customer, loans to customers would represent a higher proportion of the BIS claims figure and loans to banks a correspondingly lower one.

We therefore need to make assumptions about what percentage of the intragroup loans by UK located banks to their EU offices, recorded by the BIS, are eventually used by those EU offices to fund loans to EU customers. This will impact on our estimate of what proportion of the residual $469bn exposure of UK located banks to EU entities, outside of the UK headquartered and big five US banks, consists of customer loans versus loans to banks, which impacts in turn on the total loans to customers figure for UK located banks we finally end up with. In other words, to the extent that you assume that a certain amount of the $839bn claims recorded as loans to banks by the BIS were in fact intragroup loans which ended up as loans to non-banks on the bank’s consolidated balance sheet, you will increase the figure for total exposure in the form of loans to non-banks above the figure of $478bn originally derived from the BIS, as illustrated in the following diagram:

Remaining claims and inter company loans


Using BIS data to quantify UK intragroup claims

BIS in the A.6 reports offers a breakdown of claims on any particular country by broad sector of exposure: banks and non-banks, with non-banks being further split into NBFI and non-financial. This is very similar to the breakdown we have analysed before, except it is for all instruments, in other words it shows all claims on banks including loans, debt securities and other securities. Here is the table for France in 2015:

BIS France intragroup

$749,425bn, the second figure underlined in red, is the figure for loans to banks we saw in the table above. The first figure underlined in red of $1,008,846, represents total cross-border claims on banks in France, of which $749bn, the majority, was held in the form of loans, and the balance of $260bn was held in the form of debt and other securities. Underneath the figure for total claims on banks is a figure for the portion of those claims which are “intragroup.” That figure for France in at end December 2015 was $477bn, or 47% of total claims on banks. Since most of the claims on banks are in the form of loans, the ratio of total intragroup claims on banks to total claims on banks is a reasonable indicator of the percentage of intragroup loans to banks as a percentage of total loans to banks. Here are the BIS figures for total claims on the banking sector, where you can see the French figure of $1,008bn:

Exposure to banking sector inc securities

Here are the figures for intragroup claims, where you can see the French figure of $477bn:


Here is the intragroup claims as percentage of total claims on banks, where you can see the French figure of 47%:

BIS Intragroup % total

50% of total claims on EU banks across all instruments are intragroup claims.

We can then multiply these percentages by each country’s UK located loans to banks (from the A.62-S tables) to calculate a figure for UK located intragroup loans to banks. For France, we multiply 47.3% by the UK located loans to French banks from the A.6-2S reports of $282,322bn.

UK intragroup to EU

It’s worth noting that the estimated country breakdown of intragroup lending from UK located offices to their EU offices (which replicates the disclosed country breakdown of loans to EU banks from UK located banking offices) is more in line with the size of the countries concerned, with France and Germany being the biggest counterparty countries, and Luxembourg being small. UK intragroup loan claims on entities in the Netherlands at $73bn represents 17% of total intragroup lending by UK located offices, compared to the Netherlands share of 10% of total EU claims in the form of loans to banks. So the Netherlands continues to “punch above its weight” in terms of the UK’s intragroup loan claims. Ireland at $27bn represents 6% of total UK intragroup loan claims versus its 3% share of total EU loans to banks; the small overweight in the UK is easily explained by the UK’s proximity to Ireland. Apart from the Netherlands, therefore, the counterparty country breakdown of the UK’s intragroup lending, unlike its loans to non-banking customers, roughly follows the size of the counterparty country’s economy. It is not dominated by tax driven lending to the same extent as the loans to non-banking customers, except, perhaps to a small extent, in the case of the Netherlands.


How to reconcile consolidated with unconsolidated data

Careful readers may have noticed a potential methodological flaw in the argument so far. The ~$1.6 trillion of total BIS locational (A.6) claims include intragroup loans which are excluded from the bottom up company data figures we are analysing, in other words excluded from those companies’ balance sheets. If we subtract consolidated company data from unconsolidated BIS data, then, you might argue, the residual will be inflated by intragroup loans from banks whose balance sheets we have included in our bottom up calculations. In other words, Citigroup’s intragroup loans from London to their Madrid office would be in included by the BIS in the A.6 data but not in their reports and accounts. If you subtract Citigroup’s EU reports and accounts data from BIS A.6 data the residual would include Citigroup’s intragroup loans from London to their Madrid office. But the residual is only supposed to capture the contribution of the claims of the other banks which were not included in the bottom up calculations: it’s there to capture loans by Japanese or Canadian banks, not intragroup loans from UK or US banks. How do we know what the true picture is? This would be the confusing table that might result from an attempt to make sense of this situation:

A6 vs B4 confusion

To resolve this issue, it will be useful to show how an intragroup loan from the UK to an EU customer might appear, in parallel, in consolidated accounts and in the BIS statistics. Let’s imagine a $100m loan by a UK located bank to an EU customer made via an intragroup loan to an office in the EU. So far, we have exclusively looked at the BIS’ cross-border statistics. To capture the full transaction, as presented in the company accounts and by the BIS, we need to include the BIS’ domestic claims data as well as its cross-border claims data:

BIS versus company business flow presentation

So the BIS locational, unconsolidated cross-border data has a $100m intragroup loan which will fund a loan to a customer which also appears as a $100m loan in the BIS’ domestic claims data. Therefore, although the intragroup cross-border loan will be included in the BIS cross-border locational A.6 data, the loan to the EU customer which appears in the bank’s balance sheet will not appear in the cross-border locational data, only in the domestic data. In other words, the $641bn of cross-border claims on EU entities by UK headquartered banks ($539bn exposure) we have calculated using the UK company accounts and reconciled with the UK consolidated BIS statistics (in the B.4 tables) will include loans to EU customers which do not appear as loans to customers in the cross-border BIS A.6 data for UK located banks, but only as intragroup loans to banks.

In order to present an accurate picture of the breakdown of UK located claims on EU entities we therefore need to convert intragroup loans into loans to customers and the other types of claims for which that intragroup funding is ultimately used. We are effectively reclassifying what BIS records as intragroup claims in its cross-border data and as claims on third parties in its domestic data into one cross-border claim on a third party. In so doing, we are making the BIS data comparable to the company accounts data:

BIS versus company adjusted to be comparable

Here is the corresponding, rough reconciliation of the company accounts data with the BIS locational data, some of the detail of which was displayed above:

total claims adjusted

This would offer the following, more coherent overall picture:

comparable BIS and company accounts


Analysis of the other banks

The important question is, what proportion of these $435bn in intragroup loans to banks are used to fund loans to customers, rather than used to fund loans to banks, or even securities or other assets? In the table above we marked this ambiguity by putting a question mark next to loans to customers and loans to banks under the $435bn domestic exposure. As we saw above, the higher that proportion is, the higher our final estimate of loans to EU customers by UK located banks will be.

Before estimating this proportion, it will be useful to make a stab at firming up our bottom up estimate of the loans to customers of the banks outside the nine UK and US banks we have analysed in detail. That will give us a rough idea of the total loans to customers we should end up with as a result of reclassifying some of the intragroup loans to banks into loans to customers. Our starting point for estimating the likely remaining loans to customers outside the UK banks and the five big US banks will be to look at a broad selection of banks with London operations and significant EU exposure. Our aim will be to estimate the value of those banks’ EU IBAs and what proportion of them consist of loans to customers.


The Swiss

One of the most high profile Brexit-related job relocation announcements came from UBS, when its chairman, Axel Weber, told the BBC on 18 January: “We have roughly five thousand people in London, real passporting business is probably down to about a thousand of those employees in London, and for them we need to look at what the ultimate deal will be mapped out with.” The media was, of course, less interested when UBS revised that estimate to as low as 250. UBS’ 2015 annual report‘s risk section provides the following disclosure (p 226):


“Banking products” mainly consist of loans, and the “other” category does not include governments or banks. The “other banking products” therefore correspond, broadly, to loans and advances to customers. That number, however, includes unfunded guarantees and commitments which will not be recorded as claims by the BIS. UBS helpfully identifies these in the “unfunded” column for banking products as a whole. We can therefore apply the percentage of banking products which are unfunded to the “other banking products” to estimate “funded other banking products” which will be, for all intents and purposes, identical with funded loans to customers. UBS does not break down the CHF105bn of banking products exposure to “other countries” into sovereigns, local government, banks and other, so I assume that “other” as a percentage of “other countries'” banking products exposure is equal to the average percentage for all the individually disclosed countries (for whom this breakdown is available). That percentage is 85.2%. Note that, surprisingly, Germany is not reported separately and is seemingly included in the “other countries” line. Here are the results:

UBS exposure analysed

Of the CHF5.9bn of banking products, CHF2.1bn or 35.6% in aggregate were unfunded, leaving CHF3.8bn of funded banking products, in other words CHF3.8bn of loans. Of these CHF5.9bn of banking products, just over CHF5bn were to “other” customers, i.e. not to banks or government entities. If we assume 35.6% of this CHF5.05bn was unfunded, that leaves CHF3,237m funded loans to customers on UBS’ balance sheet. This is only for the Eurozone. Non-Eurozone EU countries in the BIS A.6 tables above represent just under 11% of Eurozone exposure. We can therefore take 11% of the CHF3.2bn Eurozone exposure or CHF356m as an estimate of UBS’ non-Eurozone exposure. This is generous, as all banks have particular strengths in particular regions, so the aggregate BIS data, which includes data for all banks active across all regions of Europe, will not be representative of UBS’s  more narrow exposure. However the estimated non-Eurozone figure is not large. This leads to an estimate of CHF3,594m of loans to European customers on the part of UBS, which would equate almost exactly to $3,594bn dollars given the exchange rate at end 2015 was nearly even.

Before accepting that number at face value we need to compare it to another disclosure, namely UBS 2016 Pillar 3 capital adequacy disclosures under the Basel III regime. Unfortunately, this particular report is not available for 2015, but, as we saw with other data sets and disclosures, loans to European entities have tended to stay in a similar ball park from 2015 to 2016, so this 2016 data has some value as a reference point (p 16):

Pillar 3 exposure

The figure underlined gives the value of loans on the balance sheet to Europe ex Switzerland and therefore including the UK, excluding loans to banks which are on separate lines of the exposure table. The figure given for this exposure here is just under CHF30bn, compared to the figure of just over CHF3.5bn derived from the 2015 annual report’s exposure report. The larger Pillar 3 report figure however includes the UK, and it is therefore entirely possible most of the CHF26.5bn difference between the CHF30bn Europe ex Switzerland (including the UK) Pillar III figure and the CHF3.5bn Europe (excluding the UK) annual report figure represents loans to the UK, where UBS has a strong franchise in financial services lending, particularly securities related. Citigroup and Morgan Stanley report exposure of $53bn and $58bn, respectively, to UK NBFIs, in comparison with which UBS’ implied UK exposure of CHF26.5bn ($26.8bn at end 2016 FX rates) is not unreasonable. It therefore seems plausible that UBS should have loan exposure to EU countries of roughly $3.5bn, which is not a large amount.

Note that the exposure statement in the annual report does not give a breakdown of securities exposure to Europe or Eurozone customers. This means that, although we know that banking products exposure splits roughly into 85% loans to customers and 15% loans to banks, we don’t know how much either line is as percentage of total European IBAs because we can’t calculate total European IBAs without a figure for European securities. However, the table in the Pillar 3 report does allow us to perform this calculation for Europe including the UK. Here is the result:


This shows that the percentage of UBS’ other EU (including UK) IBAs consisting of loans to customers is 33.7%, close to that of the four UK headquartered banks.

CSFB presents its exposure data in a similar format to UBS (2015 annual report and accounts, p 178). I have shown the total figure rather than the individual countries which takes up a whole page. However, it should be noted that CSFB’s table includes non-Eurozone countries like Croatia, and therefore represents its European, not just its Eurozone, exposure:


Unfortunately, this exposure report only includes countries rated below AA or equivalent by at least one ratings agency, and therefore does not include Germany, France, the Netherlands, Belgium, Finland and other countries in Europe with a strong rating. No disclosure is made regarding these countries. CSFB’s Pillar 3 capital adequacy disclosures under the Basel III regime provide the following information (p 12):

pillar 3.PNG

The figure of $8bn for other Europe corporate exposure is roughly consistent with €8.9bn figure for “corporates and other” in CSFB’s risk report. But the $8bn doesn’t include individuals and is calculated using the AIRB or “advanced internal risk basis” for exposure, not the IFRS method for accounting for loans on the balance sheet, while the €8.9bn doesn’t include most of Europe’s largest economies. However, from the commentary in both reports, it doesn’t seem as if CSFB has a particularly developed balance sheet activity in those countries. Although there are exclusions, the €8.9bn will include exposure to securities as well as loans. So we can tentatively assume the countries excluded are offset by the securities included and retain the ~$9.7bn figure (at a closing 2015 $/€ rate of 1.088) from the annual report for exposure to loans to EU corporates.

This is a significant figure. Although CSFB’s exposure to corporates and other is roughly 50% of total credit exposure disclosed in the Pillar III disclosure, we can’t arrive at a figure for what percentage of CSFB’s total IBAs is constituted by corporate and retail loans, as we could for UBS or the UK and US banks, as we don’t have information on the breakdown of that exposure into loans and securities. It is important to note, both for UBS and CSFB, that the loans to European entities we have estimated could be booked in their Swiss or other European branches, therefore not all of them will be booked in London. Estimating this breakdown however would be pure guess work.


The Japanese

The Bank of England wrote in the 80s that “Japanese banks are London’s largest foreign bank group, as measured by the size of balance sheet assets” and they still have large amounts of assets in London today.

A number of Japanese banks’ plans to set up a base in Europe post-Brexit were reported in the press: MUFG, Nomura and Daiwa. Japan is a country with an excess of savings, and Japanese banks have been active in lending those savings internationally. They can therefore be expected to provide useful clues about and represent a significant chunk of the remaining recalculated $837bn EU exposure of UK located banks outside the nine UK and US banks we have analysed above in detail.

Unfortunately, reading their annual reports reminded me of my discussions with my colleagues after their meetings with Japanese banks, particularly one wag who suspected that he always left the meeting with less data than he entered it, because the Japanese investor relations officers were using specially developed technological devices to erase his memory. MUFG’s 2016 form 20F (20F is the SEC’s equivalent of the 10k for foreign issuers) only provides high level data on total assets in Europe as a whole, including the UK, on page 110:

MUFG Europe assets.PNG

At March 2016 ¥/$ exchange rates of 111, this implies assets in Europe of $238bn. Unfortunately, MUFG won’t tell you how much of this is loans to banks or customers and how much is securities, or how much is to the UK and how much to the EU. MUFG does disclose that only ¥50.65tr of their total ¥292.6tr foreign assets are loans, implying a massive c¥240tr in securities. To put this into perspective, ¥240tr is $2.1 trillion, which is half a trillion dollars more than the entire BIS claims of UK located banks on the EU. This is what years of quantitative easing, zero interest rates and anaemic domestic economic growth does to a bank.

MUFG foreign loan porfolio

Taken together, these two disclosures show the following breakdown of MUFG’s foreign assets:

MUFG foreign asset breakdown %

MUFG’s 12.3% of non-financial loans to total foreign IBAs is much lower than the ratio for loans to customers as a percentage of total claims we have observed in the aggregate BIS data and most UK and US banks we analysed.

In order to estimate the size of MUFG’s exposure that could be at risk from the loss of an EU passport, the best we can do is apply the 12.3% percent of total foreign assets which consist of commercial, industrial and other loans to the ¥26.2tr of MUFG European assets:

MUFG EU consumer loans.PNG

This results in an estimate of ¥3.2bn, or $29.2bn, which is significant, in fact a little below Lloyds’ exposure to EU customer loans. This will, however, include UK exposure, and so the EU passport related business will be even less. Unfortunately, there is no obvious means to estimate this split. If we look at the big five US banks’ other exposure (in the case of Citi, Morgan Stanley and Goldman) and loans (in the case of Citigroup and Bank of America) to European entities and compare that with the number disclosed for those metrics for UK exposure, we find the UK varies from 16.5% of UK and EU combined exposure for Goldman to 57% for Bank of America, with the aggregate ratio sitting at 42%. 42% is nearly half. Using that number will considerably lower our estimate of MUFG’s EU exposure. It is possible that the numbers observed for the five US banks in our sample are excessively skewed to the UK. The UK’s own A6 data shows that at the end of 2015 there were ~$3 trillion of cross-border claims on entities in the UK, compared to ~$8 trillion on entities in the EU. According to these BIS numbers, the UK represents 31% of claims on European including UK entities. To be prudent, we can apply the 31% rather than the 42% to the MUFG number in order to estimate the UK share of what MUFG classifies as Europe, which would imply $20bn of MUFG’s European loan exposure was to the EU and the remaining $9bn to the UK. MUFG is Japan’s largest bank, so an exposure of this size, just above Bank of America’s, is reasonable.

In conclusion, MUFG provides us with one example outside the UK banks and the US top five with significant volume of loans to EU customers, although loans to customers are only 12% of their foreign exposure.

Nomura’s 2016 SEC 20F contains the following disclosures. But beware. These are in millions of Yen, not billions or trillions like MUFG. Page F-132:

Nomura long lived asset

Page 57:

Nomura leveraged finance 2016.PNG

Leveraged finance is provided to non-banking corporate customers. Interestingly, this asset class represents just under a third of Nomura’s total European balance sheet, more in line with the UK and US banks. But this amount represents less than a billion dollars. Nomura is more of an investment bank and does not have a large balance sheet in Europe.

Finally we have Daiwa, but Daiwa isn’t really a bank at all but a securities broker and investment bank. It clearly has exposure to Europe (as detailed in the Companies House format filing of its Daiwa Capital Markets Europe subsidiary) but no loans there.

In addition to the banks announcing potential new European headquarters, we can look at Sumitomo Mitsui and Mizuho, both of which announced plans to expand lending in Europe when the European sovereign debt crisis reduced the ability of European banks to offer credit. Here are the relevant extracts from Sumitomo’s 2016 20-F (p F-106):

Sumitomo loans.PNG

Europe here includes the UK, whose exposure is detailed on page A-10:

Sumitomo UK

We can assume that “Others” for UK exposure includes securities as well as loans. Note the ¥4,982,442m figure for European loans is gross loan exposure. In order to make that figure comparable to the UK  “Others” figure we will estimate what European net loans are, using the ratio of  total “Allowance for loan losses” and “Unearned income etc.” to total gross loans (1.04%) and apply it to European loans. Accordingly, we can perform the following calculation (in ¥m except the last line) :

Sumitomo EU loans

Sumitomo clearly is a Japanese bank with significant EU exposure, according to this estimate. To put the estimated $36bn into perspective, it is larger than Lloyds’ $30bn and just below JP Morgan’s $39bn. This estimate is supported to some extent by Sumitomo’s disclosure of its exposure to Portugal, Ireland, Italy and Spain (p 101):


Sumitomo’s exposure to non-financial corporations in these five countries alone (which the text specifies is mostly in the form of loans, not securities, but which includes undrawn facilities), is $8.5bn, a little under CSFB’s total estimated corporate exposure to Europe. Sumitomo is also an example of a foreign bank for whom a large percentage of their total IBAs (53%) are loans to customers.

However we can refine these estimates further by using the silo report of the Queen Victoria Street headquartered Sumitomo Mitsui Banking Corporation Europe Limited (SMBCE), available from Companies House (company number 4684034). In addition to useful information, such as staff numbers, the accounts provide details of the assets held in London by SMBCE on behalf of its parent company. Here is the geographic breakdwon of SMBCE’s exposure (p 33):

SBME geographic

The figure of $16,961m for the UK in the silo accounts is almost identical to that estimated using the ¥/$ exchange rate of 110:

Sumitomo UK SMBCE vs parent

We can therefore, satisfyingly, assume consistency between the parent company and silo account exposure numbers. In the silo accounts, the total EU ex UK exposure amounts to $16.9bn. This is does not tally with the total in the parent company accounts: Europe ex-UK loans alone were estimated above at $32bn. This implies that some European assets, including loans, are held either at the parent company level or elsewhere outside SMBCE:

EU assets held outside the UK

Given the identity between UK exposure as presented by SMBCE and the parent company, it would seem that all UK exposure is booked in the UK in SMBCE, meaning that the European assets booked in Japan, outside SMBCE, will specifically be EU assets. Why would Sumitomo hold loans to EU customers outside SMBCE? That’s hard to say. There is almost no hard data on this. Such details are very much recorded at the management accounts level, and little discussed with outsiders such as investors or journalists, as the topic is not material to the banks’ financial prospects. The most likely explanation is that these are large European companies with some sort of business relationship with a Japanese entity (think Renault with its shareholding in Nissan) through which a banking relationship was established with Sumitomo in Japan. We will discuss similar loans in connection with UK located emerging markets banks below, as well as BIS data showing that loans to EU counterparties by Japanese located banks have grown faster than those of UK located banks in the last decade.

There are a few reference points which we can use to understand why the choice of booking these loans in Japan rather than abroad in SMBCE is an easier one for a Japanese bank like Sumitomo to make than it is for banks in certain other countries. Japan is a very open country in terms of financial flows. It is very easy for Japanese banks to shift money out of the country. It is also stable and easy to navigate in terms of tax and regulation. This makes Japan different from emerging markets countries whose banks also lend to companies in Europe thanks to the borrowing companies’ relationships with an emerging markets company which is a client of the emerging market bank. Moving that money to a London office, for the emerging markets bank, is a good excuse for keeping it away from the local regulator and tax man. It is also no longer subject to any capital controls that may be imposed. Because these advantages do not obtain for Japanese banks, they are more free to choose to book their loans in subsidiaries like SMBCE or elsewhere based on other factors. Loans to a Japanese company’s tax structure in Luxembourg, the Netherlands or Ireland will, likely, be booked in Tokyo without passing through subsidiaries like SMBCE. Other loans to the subsidiaries of large corporations driven mainly by FX management will also probably be booked in Tokyo. Loans to domestic European customers and to the operational subsidiaries of Japanese companies (operating factories, warehouses etc.) will probably be booked in subsidiaries like SMBCE.

SMBCE’s risk report doesn’t break its Europe ex-UK exposure down into loans to customers and other assets (it is, in that respect, less detailed than the parent company accounts), but the breakdown of total IBAs for all geographies (of which Europe ex-UK represents 43% and Europe including the UK represents 86%) allows us to estimate what portion of SMBCE’s Europe ex-UK exposure consists of loans to customers. In the table below, I have used the 39.4% of total IBAs constituted by loans to customers for SMBCE as a whole as a guideline for the percentage of its European exposure constituted by loans to customers. However, I have used the higher figure of 50% (highlighted in gray in the table below) to estimate the percentage of its European exposure constituted by loans to customers, rather than the observed figure of 39% for SMBCE as a whole. In other words, I have assumed that SMBCE’s European exposure has a higher percentage of loans to customers than SMBCE as a whole. SMBCE’s exposure is 43% Europe ex-UK and 43% UK, and its IBAs consist of 39% loans to customers and 49% deposits at central banks. Therefore, saying that SMBCE’s Europe ex-UK exposure has a higher proportion of loans to customers as a percentage of total IBAs implies, mechanically, that its UK exposure has a higher proportion of deposits at central banks as a percentage of total IBAs. This makes perfect sense, as SMBCE is an English company regulated by the Bank of England, and which holds its liquidity reserve in the UK. Here are the results:

SMBCE loans to EU ex UK customers

Sumitomo, according to this more precise estimate, had $8.5bn of loans to EU customers located in London, which would put SMBCE in the same league as CSFB’s total exposure to EU loans to customers (not all of which, as we noted above, will be booked in London). SMBCE’s European loan exposure is almost exactly equivalent to its parent company’s overall exposure to PIIGS countries. Now, SMBCE’s other European exposure was only $5.3bn. Therefore a substantial amount of Sumitomo’s PIIGS exposure must be held at the parent company level. This implies that it is Sumitomo’s central treasury department, not its London office, which is responsible for its still significant exposure to high risk EU countries!

The wider lesson to be drawn from the examination of SMBCE’s silo accounts is that a significant proportion of the EU exposure analysed in MUFG’s accounts, above, and Mizuho’s, below, are likely to be held at parent company level and therefore not to appear in the BIS A.6 report. We have no reliable parameters with which to estimate this exposure. We noted above that we were similarly unable to specify how much of the Swiss banks’ estimated European loan exposure was booked in the UK or elsewhere. To the extent that we could estimate this and that we removed it from our estimates for MUFG and Mizuho, we would increase our residual estimate of the exposure of other UK located banks to loans to EU customers, which is analysed in more detail below.

As for Mizuho, the relevant information can be found in its annual report and accounts for March 2016 (p 290). The level of detail here is not bad:

Mizuho credit exposure

The figures in the first column include loan commitments, loans to banks and the Western European figure includes exposure to the UK. Here is an estimate of the on balance sheet exposure to loans to EU customers, using Mizuho’s ratio of total loans to banks to total loans at the consolidated level to estimate loans to EU banks as a percentage of total EU loans and the same assumptions as for other Japanese banks to estimate commitments and UK exposure:

Mizuho loans to EU customers

Again, this is sizeable exposure, just below Lloyds in size. Mizuho’s loans to customers represent a full 41% of its IBAs at consolidated level.


The Canadians

Canada’s top five banks are large, internationally active and, helpfully, provide data on their European exposure which splits that exposure out into securities, loans to banks and loans to customers, and reporting the UK and Switzerland separately. How very Canadian. Here are the results:


All of these Canadian banks have important London subsidiaries, and it can be expected that all of this loan exposure is booked in London.


Wells Fargo

Wells Fargo, as one of the big US banks which does not have a significant investment banking arm and is therefore more of a banking pure play, is also worth investigating due to its sheer size: its 2015 annual report reveals $905bn of net loans to customers, more than any of the top five US banks we surveyed. Its country exposure is presented as follows (p 68):

Wells Fargo

Its total Eurozone exposure ex sovereigns is $8.75bn, but this includes loans to banks, money market placements and unfunded commitments. Assuming around 20% of this unfunded and 20% of the remainder is loans to banks, you would have exposure of just under $5.6bn to loans to customers, which is close in size to Goldman, TD and RBC. Despite its great size overall, Wells Fargo is a small player in the European customer loans market.


Summary of exposure of large banks apart from UK headquartered and US top five

If we take the information we have for the eleven banks outside the nine UK and US banks we initially analysed we find the following:

other countries individual banks

We have therefore estimated $89.5bn of loans to customers in total by those eleven banks, or just under half of the amount identified in the big four UK banks and just over half of the amount identified for the US big five. This compares with $165bn which we estimated above for all banks with EU exposure outside of the big four UK and big five US, based on the BIS numbers without reclassifying any BIS intragroup loans (but eliminating the $51bn UK headquartered banks’ loans booked in subsidiaries outside the UK which we know will not be included in the BIS A.6 report for UK located banks).

Although this isn’t a comprehensive list of all banks making customer loans in Europe, it does capture the largest international banks in the five countries with the largest cross-border exposure to European loans and with significant operations in London. Here is a chart of the loans to customers we estimated for the twenty banks we have analysed so far:

pareto chart all loan to customers in decreasing size

As you can see, there is a sharp drop from HSBC and Barclays to a group of four banks with around $40bn in loans to customers, followed by another drop to the $20-30bn range with four banks, and a tail of ten banks with less than $10bn. This distribution once again supports the hypothesis that cross-border loans to customers will be concentrated among the larger players, with a long tail of smaller exposures, and is strongly consistent with a Pareto style distribution of cross-border loans to EU customers.


Some examples of smaller banks

We can test our assumption of a drop off in the size of the banks in the tail by looking at the EU balance sheet exposures of a few banks with smaller EU exposure than the twenty large banks in the chart above.

Bank of New York Mellon (BNY) is a large bank with international operations, although these are principally in custody and fund administration, which are off-balance sheet activities serving the fund management industry.  BNY reports exposure to France and Germany of around $4-4.5bn and of around $3bn to the Netherlands in the cross-border risk section of its 2016 SEC 10k report (p 29):


As you can see, the French and German exposure which wasn’t to banks or to governments was less than 10% of total French and German exposure in every year.

In addition, BNY’s note 18 (p 197) shows a scattering of smaller claims amounting to just under $6bn, in other countries like Spain, Italy, Belgium and Ireland, all of which are in the form of floating rate notes and government bonds.

BNY fair value.PNG

In other words, the composition of BNY’s small exposure is heavily weighted toward banks and government bonds in a way which is consistent with the BIS data. Note that, at $14bn odd, BNY’s EU exposure is 2% of the $837bn exposure to EU entities of UK located banks outside the UK and US top five banks, consistent with the picture above of a concentration of exposure in the hands of a few players and the remainder a tail of much smaller exposures.

State Street bank is BNY’s leading competitor in custody and fund administration, and a significant player in asset management with a diverse international reach which is particularly strong in Europe. State Street’s 2015 10k reveals that its exposures to entities in countries outside the US, including Europe, contain no loans to customers at all:

State Street

Total outstandings are equal to securities and other assets plus derivatives and securities on loan. There are no loans to customers here.

Bank of London and the Middle East (BLME) provides a good example of a Middle Eastern bank with international lending ambitions pursued through its London headquarters. Its 2016 annual report shows it is small with only £1bn in credit exposure. All of its European exposure is to financial institutions (this can be confirmed by combining information from pages 4 and 98). BLME is a good test case of a small bank which has no exposure to loans to customers outside its core region.

Similarly, Europe Arab Bank, according to its Companies House filings for 2016 (p 46), has around €1.37bn in loans and advances to customers, including the UK, in 2015, so its EU exposure will be even lower. For all their petrodollar wealth, the Arab banks are not big players in the European loans to customers market.

We also saw earlier that UK banks outside the top four in terms of EU exposure, like Co-op or Standard Chartered, had only nominal EU exposure, as did Nomura and Daiwa.

Looking at these other banks with smaller exposure, we find that their exposure to EU loans to customers is less than a billion dollars. There is no bottom up evidence of any substantial loan to customer exposure outside of the twenty banks we have investigated in detail. This is entirely consistent with the loan distribution we observed above.


Implications of company level data for BIS intragroup data

If none of the intragroup loans we reclassify, as described above, in order to make the BIS data homologous with the company level data, are reclassified as loans to customers, then the loan exposure of the remaining UK located banks outside the “top twenty” (UK headquartered plus US top five plus eleven other banks analysed above) to EU customers would only be $85.5bn (“other banks” outside the UK and US top five of $175bn minus $89.5bn calculated bottom up for the eleven Swiss, Japanese and Canadian banks and Wells Fargo). However, if, at the other extreme, we assumed that all the BIS intragroup UK located loans to EU banks were eventually used to provide loans to EU customers, then we would need to add a whopping extra $435bn to the total value of UK located loans to EU customers. That would mean around $520bn loans to EU customers from UK located banks would have to be accounted for by other, smaller banks (the initial remaining ~$85bn plus the $435bn reclassified intragroup).

But our analysis of the EU exposure of banks outside the “top twenty” shows that the vast majority of it is in the form of loans to banks and securities, not loans to customers, and that loans to customers exposure of any individual bank we examined struggles to make $1bn. Therefore, if all of the $435bn intragroup loan amount is reclassified as loans to customers, this would imply a tail of over 500 banks, each with small, sub-economic loan exposure to EU clients managed from the UK. However TheCityUK, a high profile campaign group purportedly seeking to promote the City’s interests, in its Practitioner’s Guide to Brexit, estimates that only “250 foreign banks operate in London” (p 5). It is therefore very likely that not all of the intragroup loans to banks are used to make customer loans. To understand why that is, and to arrive at a more precise estimate of how the intragroup funding from UK located offices to EU offices is used, we need to analyse intragroup funding in more detail.


Intragroup: branches versus subsidiaries

We need to start with a better understand what the BIS includes in intragroup lending. The BIS’ guide to its locational banking statistics, in which the intragroup data appears, doesn’t use the word “intragroup” but its close synonyms “intrabank” and “inter-office.” It offers the following definition on page 25:

definition of intragroup

On pages 22 and 28, respectively, it clarifies:

definition of affiliates

definition of own offices

In short, intragroup cross-border lending can be lending both to an individually incorporated subsidiary company in another jurisdiction and to a branch in another jurisdiction which could be part of the holding company or of one of its subsidiaries.

The idea of an intragroup loan to a subsidiary is easy to understand, as the subsidiary is an independent legal entity. Many subsidiaries of larger banking groups have either an excess of deposits versus loans (positive funding gap) or of loans versus deposits (negative funding gap). It makes sense for the subsidiary with the positive funding gap to lend it to the subsidiary with the negative funding gap, rather than to build up excessive holdings of loans to banks or of securities within that subsidiary. Sometimes, the subsidiary with the positive funding gap will lend it to the parent company which then lends some of those funds to subsidiaries with negative funding gaps.

Because the subsidiaries are independently established and regulated legal entities, these intragroup claims must be carried out on a basis which does not jeopardise the financial soundness of either subsidiary or parent, and therefore have to be carried out with properly documented terms and conditions. They must also be recorded on the audited balance sheets of both entities. The intragroup loan numbers recorded by the BIS, in connection with loans to or from subsidiaries, are therefore reflected in a legal loan document and an audited accounting entry published by a recognised and regulated legal entity. At the consolidated level of course, these intragroup claims are eliminated. But the terms of their existence, both economic and legal, are clearly established. If the parent company or one of the subsidiaries were ever to become insolvent, the entity within the group which had made an intragroup loan to them would be able to establish itself as a creditor in the insolvency proceedings on the basis of the terms of that loan.

The situation with a loan between branches is quite different. The branch does not exist as a legal entity in the same way as a subsidiary. There is no loan documentation between the branch and another branch or the parent company. In insolvency proceedings, no branch can be the subject or the pursuer of a claim against either the parent or another branch. Unlike intragroup loans involving subsidiaries, any interest rate charged, nominally, on an intragroup loan to a branch is not economically meaningful, only the interest rate earned by the loan the branch makes with that money compared with the bank’s cost of funding in that country and the credit risk of the loan. The branch is simply a conduit for the money it puts to work. The loan will appear in the bank’s management information as booked under the branch, but there will be no audited accounting entry published by the branch as a separate legal entity. The intragroup loan numbers recorded by the BIS, in connection with loans to or from a branch, are therefore not reflected in any legal document or any audited accounting entry published by a recognised and regulated legal entity.

Because the subsidiary is independently regulated and constituted, it has to hold its own liquidity. It will therefore not only hold loans to customers but loans to banks and securities too, as will be demonstrated when we look at individual bank subsidiary or “silo” balance sheets. Any intragroup loan to a subsidiary will, accordingly, not ultimately end up being used exclusively for loans to customers, but will be treated by that subsidiary as one of its sources of funding for the many assets on its balance sheet.

To understand intragroup loans to branches, you need to understand why they might be made. Say a Rotterdam branch of a US bank with its European head office in London makes a loan to a Dutch SME. The branch is a business lending branch and doesn’t have many depositors. The US bank has a central funding group which manages the savings of its branches across the world and the proceeds of its securities issuance and inter-bank borrowing. It therefore provides the funding for the loan booked by its Rotterdam branch. As we shall see, it will typically do so using an intragroup loan from its European head office in London. But it doesn’t necessarily have to do things this way. That loan could instead be booked directly between the bank’s European HQ in London and the Dutch SME. But the Dutch SME will, typically, have a current account with the Dutch branch through which it will service the loan. If payments to service the loan are made by the SME to the European HQ in London, because the London HQ lent directly to the SME without using an intragroup loan to the Rotterdam branch, then the money would have to pass through international payment structures, which likely would carry some unnecessary costs. Moreover, the local relationship with the branch is a competitive advantage and represents a form of goodwill. Using an intragroup loan to the Rotterdam branch, rather than making the loan direct to the SME from another office, makes the loan seem local to the Dutch SME, which might support that local goodwill, whereas presenting it as from a big remote corporate center might do the opposite. Therefore, it might make more sense for the money to be transferred from central funding in London via an intragroup loan to the Rotterdam branch, where it is then loaned to the Dutch SME.

Having checked with the BIS’ helpful staff member, it seems that once the loan is disbursed by the Rotterdam branch to the Dutch SME, it isn’t re-classified by the BIS as a cross-border loan to the Dutch SME from the London office. It continues, instead, to be recorded by the BIS as a domestic loan to the Dutch SME from the Amsterdam office, which continues to be funded by a cross-border intragroup loan from the London office to the Amsterdam office. Intragroup loans to branches are therefore not temporary in nature, according to the BIS’ treatment. Had they been temporary, you could have expected that only a small fraction of the intragroup loans recorded by the BIS was to branches. That’s because those loans would only exist as intragroup loans for a few days while the necessary steps were taken to disburse the funds to the ultimate borrower. Instead, a loan to a customer arranged in this way, i.e. via an intragroup loan to another branch, will stay as an intragroup loan for the lifetime of that loan. That means a larger proportion of the intragroup loans can be expected to be loans to branches than if they were temporary in the way described.

This is important because branches, unlike subsidiaries, have no need to maintain liquidity buffers. They only exist to process loans and deposits. All the liquidity is typically held at subsidiary or parent level. Therefore, we can assume that virtually all the intragroup loans to branches will eventually be turned into loans to customers, not into loans to banks or securities holdings.


What percentage of the intragroup loans to EU countries from the UK are loans to branches and what percentage are loans to subsidiaries?

It is well known that a theoretical advantage of holding an EU financial services passport is that it allows you to start operating in a country with a branch, rather than having to set up a separately capitalised and incorporated subsidiary. This received idea might lead you to expect that most intragroup loans recorded by the BIS would consist of loans to branches. However, although the European financial services passport should, in theory, enable direct transfer of funds from a bank’s London “hub” to an EU branch or “spoke,” without going through any EU subsidiary on the way, the practice (as is so often the case) is different.

The EU financial services passport only got going relatively recently, and, both long before and during its existence, even before the introduction of the Euro, European banks, including UK, and even banks outside the EU, were attempting to gain scale by making cross-border acquisitions in the EU. When the consolidator banks acquired these EU banks they inherited a series of well established payment mechanisms. Even though, in theory, the subsidiaries which lay at the center of those payment mechanisms might be no longer required after the introduction of the European financial services passport, in practise, reorganising or reprogramming any payment or transfer or intragroup loan arrangement carries expenses and significant risks or potential bugs which banks are exceptionally keen to avoid. Any change requires communication with customers and staff. Communication of any such change, however well executed, can raise questions or even just distract staff and customers, leading to lost business. All it takes is for one detail to spoil or delay a transaction and the bank will suffer loss of income or extra expenses or reputational damage. The banks’ attitude is therefore, rightly, “if it ain’t broke, don’t fix it.” That is one of the reasons why you get the use of an often complicated network of subsidiaries persisting after the introduction of the European financial services passport, as we shall show below with specific examples.

Another – also theoretical – argument in favor of branches and against subsidiaries is that banks could make their balance sheets leaner with the “hub and spoke” model, by centralising all the liquidity in one EU subsidiary (hub) and only operating in other EU markets through branches (spokes), which didn’t require any liquidity. However there is a very significant advantage to operating through a subsidiary: if the subsidiary collapses the parent company can allow it to fail without endangering the other banks it owns in other countries. It is therefore less risky. The surplus liquidity held in the subsidiaries is a relatively low cost firewall protecting the rest of the group from any credit issues in individual subsidiaries. Small businessmen will use the same principle when they open a string of shops as separately owned trading partnerships or limited companies, even though they could – theoretically – save on Companies House filing and accounting costs: if one of the shops struggles and can’t pay the rent, the landlord can’t force the profitable shops to make the payments.

Franklin Allen and Xian Gu discuss this point in a lively 2011 paper on the use of intragroup lending “Corporate governance and intra-group transactions in European bank holding companies during the crisis.” They write (pps 10-11):

Wharton subsidiary advantages.PNG

In addition to the “firewall” argument above, the authors identify governance, deposit insurance and other regulatory fees, brand equity and tax optimisation as other advantages to using subsidiaries. To test their hypothesis, they painstakingly collected the following data on numbers of subsidiaries and branches in the EU in 2001 and 2009, to see whether the EU financial services passport resulted in any change in those numbers. I have reproduced their figures below (from p 52 of their paper), adding the ratio of branches per sub in 2009 implied by those figures. From 2003 to 2009 the number of subsidiaries has indeed decreased by just under 100 or ~14%, while branch numbers increased marginally by 21 or ~3%. One would have expected a reduction in branches in some markets, ceteris paribus, due to increasing use of internet banking and the resulting cost cutting. Therefore, the small rise in branch numbers indicates that internet related branch rationalisation may have been offset by some banks expanding in new markets with branches rather than subsidiaries. However, the numbers involved are very small. Despite the introduction of the EU passport, the numbers of foreign branches per subsidiary in what the authors describe as the “old market segment” (OMS), in other words the mature Western European markets, ranges from 0.4 in Luxembourg to 5.2 in Italy, with the aggregate standing at 1.2. What this implies is that there are very few branches operating without any subsidiary, and therefore that, the financial passport notwithstanding, the minor rationalisation of the networks over the period of the study has not made a significant change to its structure, which remains dominated by subsidiaries:

Subsidiaries versus branches #


HSBC and Mizuho: complex networks of subsidiary holdings

The existence of a cross-border network in the EU dominated by subsidiary relationships continues to be evident today. The organisational structures of two groups, HSBC and Mizuho, illustrate this well. Every set of report and accounts or equivalent regulatory disclosure contains a list of subsidiaries. These are rarely read in any detail, which is a pity, as they often contain very useful information, some of which might have helped those expatiating on the impact of Brexit on the City not make so many ill-informed statements. Here are the relevant extracts from Mizuho’s 2016 annual report page 135:

Sumitomo subsidiaries.PNG

Mizuho has a UK  (“Mizuho Capital Markets UK Limited”) and a Dutch subsidiary (“Mizuho Bank Nederland N.V.”). Interestingly, its branches in Spain and Belgium are branches of its Dutch subsidiary, but it also has branches in Paris and London which are branches of the parent company. It may well be that most of the lending work goes on in London, but any intragroup loans to at least two of the branches will go through the Dutch subsidiary, and possibly even those to the Paris branch may go via that route too. Bottom line, in this example, most intragroup loans from London are likely to take the form of a loan to a subsidiary, not a direct loan to a branch. Mizuho offers one concrete example of a bank that adopted the “if it ain’t broke don’t fix it” approach and, accordingly, kept transacting its business through several established subsidiaries rather than going for the theoretically more efficient “hub and spoke” model of a single subsidiary with many branches. It is banks like Mizuho which contribute to the statistics in Allen and Gu’s report showing the continued extensive use of subsidiaries in EU intragroup funding, despite the theoretical benefits of the “hub and spoke” model enabled by the EU financial services passport.

HSBC is a particularly rich source of information on how such a set-up operates because of its French and German subsidiaries’silo accounts. In addition, one of HSBC France’s real estate lending subsidiaries, HSBC SCH, also provides independent reports and accounts.

An interesting detail can be observed in HSBC France’s list of subsidiaries (n 33 p 284):

HSBC France Trinkaus.PNG

HSBC France held a 90% share in a Trinkaus subsidiary which was liquidated in 2016. While not economically meaningful, this shows a complex holding company network in which the parent company invests in a subsidiary via a subsidiary, the opposite of the direct “hub and spoke” model the EU financial services passport was supposed to have enabled.

HSBC’s holding in Trinkaus and Burckhardt AG, whose balance sheet we have analysed, is in fact carried out via another holding company (Trinkaus annual report, p 76):

Trinkaus holding via Holding AG


Intragroup asset recycling

What HSBC’s holding company and silo accounts also demonstrate is that there is a significant two way flow of funds between the parent company and its subsidiaries or from one subsidiary to another. What that means is that a significant proportion of intragroup loans from the UK to the EU will be offset within the same group by intragroup loans to other subsidiaries, including loans from the EU to the UK. These recycled intragroup loans are by definition not used to make loans to customers.

We can first examine the parent company balance sheet for HSBC holdings, the holding company which owns all of the subsidiaries consolidated by HSBC. The main purpose of the holding company is to raise wholesale funding and allocate that funding to the group’s subsidiaries and branches (p 80):

Holdings definition

Here are its assets and liabilities (p 191):

HSBC parent balance sheet.PNG

In 2015 HSBC parent company loaned $44.4bn to its various subsidiaries across the globe and in return received $2.1bn from other subsidiaries with excess liquidity. Although this is not directly observable, some of the $2.1bn deposits may have been made by subsidiaries which were also recipients of a portion of the $44.4bn loans.

J. P. Morgan, on page 271 of its 2016 annual report, offers a more dramatic example of this dynamic when it reveals that its parent company went from being a significant net lender to its subsidiaries to a small net borrower from them from 2015 to 2016:

JP Morgan intragroup.PNG

Clearly, not all of JP Morgan’s intragroup lending consists of funding for long term customer loans, but instead is part of a liquidity management strategy. In 2016, the parent company’s principal function was, effectively, to invest its subsidiaries positive funding gap in a variety of debt securities.

HSBC explicitly says that it raises wholesale finance in order to maintain a presence in local funding markets. Since it is primarily the holding company which raises wholesale finance and primarily its subsidiaries which are active in local wholesale markets, the loans and advances it provides its subsidiaries are partly destined to fund loans to banks rather than loans to customers (p 76):

Local wholesale.PNG

HSBC France’s DDR for 2014 provides an interesting insight into the way HSBC Holdings provides capital to its subsidiaries, recording an issue of subordinated debt by HSBC France subscribed entirely by HSBC plc (note 28, p 186):

HSBC France subordinated debt.PNG

As the note makes clear, this bond with a long term maturity of 15 years is eligible as Tier 2 capital and therefore bolsters HSBC France’s solvency ratios. In other words, this is an example of intragroup funding being used to support the solvency of a subsidiary as a whole.

We can turn to the silo accounts of both HSBC France and HSBC Trinkaus to see the flow of capital between HSBC’s subsidiaries and the parent company, whose balance sheet we discussed above, and directly between each other.

HSBC France

HSBC France records both assets and liabilities toward other HSBC entities (n 31 p 283):

HSBC France intragroup exposures.PNG

(“Actifs” means “assets” and “Passifs” means “liabilities”). This could be both with HSBC parent and with other subsidiaries like Trinkaus. Note the loans to other HSBC banks of €809m are counterbalanced by deposits from other HSBC banks of €566m, which, although not significant in size, show that intragroup funding is not kept to a minimum but taken in loaned out in a way that allows excess liquidity to build at the subsidiary level.


HSBC SFH is a fully owned subsidiary of HSBC France dedicated to real estate lending. SFH is able to issue covered real estate bonds which provide an attractive source of funding. In its 2016 silo accounts, SFH makes the following statement (p 4):

HSCB SCH will help finance HSBC France

“HSBC SFH (France) will continue to participate in [i.e. to contribute to] the medium and long term funding of HSBC France.” In other words, SFH will use its ability to issue covered bonds to help its parent company, HSBC France, with its funding. But if we look at SFH’s accounts (p 52) we notice something interesting: SFH’s loans to banks (“créances envers les établissments de crédit”) were €5,287m. This excludes loans to central banks.


HSBC France, HSBC SFH’s parent company, held loans to banks, excluding central banks, of only €4,660m – less than its subsidiary! What this means is that HSBC France has deposits of €627m with HSBC SFH, which appear on SFH’s silo balance sheet but which are eliminated as an intragroup balance in the consolidated accounts of HSBC France. Again, these are small amounts, but what they demonstrate is that SFH issues covered bonds to raise funds for HSBC France, part of the proceeds of which HSBC France then deposits back with SFH.

HSBC Trinkaus

We can also see this tendency to take intragroup deposits at the same time as making intragroup loans in Trinkaus’s balance sheet (n 69 p 170):

Trinkaus intragroup loans

Again, far from an efficient, capital lean, “hub and spoke” model, HSBC’s subsidiaries are sitting on excess liquidity provided by the parent company and other subsidiaries.

(Note that, in theory, we should have reduced the $51bn HSBC and RBS loans held in subsidiaries outside the UK – and therefore excluded from the A.6 data – in the table above, by the amount of these intragroup loans from HSBC plc to HSBC France and Trinkaus. These would have been picked up by the A.6 tables as crossborder intragroup loans to EU entities, which partly funded the French or German located assets we excluded. In practice, the amounts were too small to matter).

These HSBC specific observations are confirmed by data compiled in Allen and Gu (2011). Using Bancsope data, annual report and accounts and subsidiary or silo reports and accounts they compiled the following picture of intragroup loans and deposits for a selection of European banks from 2007 to 2009 (p 59):

intragroup wharton.PNG

As you can see, loans and advances made from a parent to a subsidiary, or a subsidiary to a parent, are often accompanied by loans and advances made in the other direction. Taking the larger flow as the denominator and the smaller as the numerator, we can perform the following analysis of the data:

intragroup wharton %

In each year for each entity we calculate the percentage of the larger loan which is loaned back by the recipient (whether from subsidiary to parent or vice versa). The total figure is the aggregate of all the smaller loans divided by the aggregate of the bigger loans. GE Poland is the only subsidiary which made deposits to its parent company without receiving any funding from that parent. This case study supports the hypothesis that a significant percentage of the intragroup loans made by UK located banks to EU located bank branches will be loaned back to other subsidiaries in the group or the parent company, rather than loaned to third party customers.

A final piece of data which supports the hypothesis that a significant proportion of intragroup loans from the UK to the EU are loaned back to other intragroup entities, including entities in the UK, can be found by looking at intragroup loans to the UK by EU entities, as recorded in the UK’s A.6 report. The figures in that report allow us to estimate the percentage of claims on UK entities constituted by intragroup loans to banks in the UK:

UK intragroup loans to banks %

The A6.2-S country table for the UK breaks claims down into loans to banks and non-banks, for each EU country apart from Germany. We can therefore apply the 46.4% intragroup ratio to the bank loan figure disclosed for all countries except Germany, and the 24.2% intragroup loans to banks ratio for the total claims figure disclosed for Germany.

UK intragroup loans

Bank offices in the UK are the recipients of an estimated $313bn cross-border intragroup loans, three quarters of the cross-border intragroup loans provided by UK located offices to offices in the EU. We can pair these two figures up for each country, showing the estimated intragroup loans to banks to the UK from that country and from the UK to that country:

UK vs EU intragroup by country

With the notable exception of Italy, each country on which we can estimate the UK had significant intragroup claims also has significant and sometimes larger intragroup claims on the UK (notably in Luxembourg and Belgium). This macro-level picture is consistent with the thesis that a portion of UK intragroup loans to EU subsidiaries finding their way back to the UK entity making that loan, which is supported by HSBC’s silo accounts and the data in Allen and Gu (2011).


Basel requirements

The holding, some might say hoarding, of liquidity at the subsidiary level is in fact encouraged by international banking regulation: the BIS encourages self-sufficiency in funding at the subsidiary level, as the brilliant Julien Noizet wrote in 2014:

Globally, Basel 3 regulations now require each subsidiary of international banking groups to hold high levels of liquid assets and comply with a Net Stable Funding Ratio. By itself, this means that subsidiaries have a limited power to transfer liquidity intragroup even if they don’t need it at a given moment. Only liquidity/funding in excess of those (already high) limits could be transferred.

Noizet argues that this is an inefficient way for international banks to manage their balance sheets.  Not only does it trap money unnecessarily in the subsidiary, but it reduces the ability for parent and subsidiaries to come to each others’ aid in times of difficulty, making the parts safer at the expense of making the whole riskier :

When a truly large crisis strikes, healthy banks won’t be able to support their struggling sister banks, which can potentially even endanger their own existence through indirect contagion.

It won’t have escaped the careful reader that the “hub and spoke” model enabled by the EU financial services passport provides banks with the perfect tool to stymie the efforts of both the BIS and EU national bank regulators to ensure the banking system in a particular country is sufficiently capitalised. By using branches rather than subsidiaries, banks can avoid the need to inject solvency capital in any given EU market in which they operate. This is precisely the agency or governance issue flagged by Franklin and Gu (2011), as quoted above. Such regulatory contradictions are a common feature of EU policy. The BIS rules, in this case, effectively make the maintenance of the “firewall” described above more onerous.

Whatever the unintended consequences of these regulatory contradictions, the implications of Noizet’s perceptive analysis for our own is that regulatory pressure forces subsidiaries receiving loans from other subsidiaries or from their parent to allocate a smaller proportion to their core activity of lending to customers and more of it to liquidity reserves.


Putting numbers on the proportion of intragroup loans used to fund loans to customers

The purpose of this analysis of the arcane subject of intragroup loans was to estimate how much of the $435bn cross-border intragroup loans provided by UK located offices to offices in the EU are used by those EU offices to make loans to their domestic customers and how much is used for other purposes (loans to banks, securities, other forms of liquidity). Note that there we are not analysing the impact of intragroup securities holdings in this analysis. By definition, the securities cannot be issued by a branch and therefore will be issued by subsidiaries. The reasons not to include these intragroup securities are twofold. First, we don’t know what proportion of those securities are debt and what proportion are equity. For example, that figure would have included Unicredit’s 32.8% equity stake in Poland’s Bank Pekao, which Unicredit sold in two transactions in 2016. Such equity holdings are not interest bearing and therefore do not directly contribute to revenues at risk from the loss of passport. Second, although both bonds and equity issued by EU subsidiaries of UK located banks help to finance those banks’ loans to EU customers, there is no requirement for the UK located bank to hold an EU financial services passport in order to own those bonds.

The first assumption that needs to be made is what percentage of intragroup loans are loans to branches and what percentage are loans to subsidiaries. The analysis above, namely the historic expansion of London located banks into Europe via acquisition, the work of Allen and Gu and the examples of HSBC and Mizuho all suggest that banks involved in such intragroup lending to the EU from their offices in London and which have subsidiaries in the EU will do most of that business via their EU subsidiaries. There will however be some, mainly smaller, banks which only have a subsidiary in London and do their intragroup lending in the EU via their EU branches. However, such banks constitute a minority of the overall volume we have analysed so far. Our work therefore suggests that the majority of intragroup lending from UK offices to the EU is to subsidiaries, rather than branches. However we will, prudently assume that only two thirds of it is.

As discussed above, intragroup loans to subsidiaries will fund that subsidiaries’ whole balance sheet, not one particular part of it. Moreover, the analysis above supports the hypothesis that significant liquidity is held at subsidiaries which are recipients of intragroup funding, with subsidiaries both lending to and borrowing from parent or holding companies or other susidiaries. In terms of loans to customers, it is reasonable to assume that the proportion of any intragroup loan which is used to make a loan to a customer will be the same as the proportion of the subsidiaries’ balance sheet used to make loans to customers.  To estimate that proportion, I have used loans to customers as a percentage of total IBAs for the nine US and UK banks analysed in detail, excluding Goldman Sachs and Morgan Stanley as their investment banking focus will not be typical of most banking subsidiaries in Europe. This results in an estimate of 27.6% of subsidiary assets being invested in loans to customers:

intragroup subsidiary IBAs to total assets.PNG

This results in the following estimate of intragroup loans converted to loans to customers:

Intragroup calc

This estimated $231bn of loan claims on EU customers ($229bn exposure after deducting 1% credit provision) funded by UK office intragroup loans to their EU offices (53% of total intragroup loans) leads to the following estimate of adjusted loans to EU customers and loans to EU banks by UK located banks:

Remaining claims and inter company loans adjusted

This in turn allows us to present the following comprehensive picture of UK located claims on and exposure to EU entities:


A few points to note:

  1. For the sake of simplicity and because it doesn’t impact the estimate, I have assumed all intragroup loans not used to fund loans to customers are used to fund loans to banks. As we saw, intragroup loans to subsidiaries are used to fund securities as well, but estimating this would have further complicated our table without improving the accuracy of our estimate.
  2. I assume that all intragroup loans which ultimately fund loans to customers are all to non-financial customers. That is because NBFIs as relatively sophisticated financial institutions themselves might be expected to be able to borrow directly from the London office. In the absence of any data I prefer to adopt this assumption which, conservatively, maximises the estimated Brexit impact.
  3. The total exposure is $1,554bn rather than $1,556bn previously, because provisions have increased by $2bn, representing the provisions on the intragroup loans reclassified as loans to customers (on which no provisions had been assumed when they were treated as loans to banks).

This leads to the following estimate of loans to customers, broken down into loans to non-financial customers and loans to NBFIs. The “Eleven other banks” are the Swiss, Japanese and Canadian banks and Wells Fargo, analysed above:

Total exposure income

Note that the number for NBFI exposure from the BIS has allowed us to refine our estimate for the UK. We assume that the ratio of loans to NBFIs to total loans to non-banks is the same for the “Eleven other banks” and the “Smaller banks” as it is for the adjusted BIS claims data as a whole ($215bn out of $714bn). We retain the $61bn of NBFI exposure from Citigroup and Morgan Stanley for the US top five. The UK NBFI exposure of $31bn is equal to the total $213bn NBFI exposure estimated from the BIS data minus the NBFI exposure of the US top five ($61bn), the eleven larger banks ($27bn) and the smaller banks ($95bn).

Note too that the loans to EU customers by EU subsidiaries in receipt of intragroup loans from UK located banks are relatively easy to protect from any loss of EU financial services passport. As long as the EU subsidiary is sufficiently capitalised to take deposits, the UK located bank can use that subsidiary as a convenient way of continuing to service EU customers.

Our estimate of the retail and business banking income at risk from the passport, including the twenty banks analysed above and the smaller banks captured by the BIS reports, is $15.4bn or £10.4bn. This is just over two thirds of the OW £14bn estimate. By removing income from assets not affected by the passport we have taken just under £5bn or one third off of our estimate. If we add the £1bn from private banking we get £11.4bn of Banking Revenue at risk, less than half of OW’s original £25bn estimate.

exposure reduction


Long tail analysis

Our analysis above presents us with two seemingly contradictory facts. On one hand, the analysis of individual banks shows a steep drop off in the size of the book of loans to EU customers outside the ten banks with loan books above $19bn, and that it is hard to find any bank outside the “top twenty” with loans over $1bn. On the other, the BIS total figure for loans to customers, increased by assuming that some crossborder intragroup loans ultimately fund loans to customers recorded as domestic loans by the BIS, together with the loans to EU customers observed in the “top twenty” banks (duly adjusted by removing loans booked in the EU by some of their subsidiaries), implies that these smaller banks have, in aggregate, loaned $315bn to customers in the EU, around double the UK top four and over double the US top five.

TheCityUK’s estimate of 250 foreign banks operating in London implies a total of 257 banks have contributed to the numbers we are analysing (250 foreign banks, the four big UK banks and the three smaller UK banks briefly mentioned above). We can, accordingly, derive the following picture of the distribution of exposure to EU customers among UK banks:

whole market distribution

This would imply that 237 of those banks had extended $315bn of loans to customers, or an average of $1.3bn each. That is entirely consistent with the analysis above showing that twelve large banks had loans to EU customers of $8bn and above, six had loans of $3-8bn, and the last two, BMO and CIBC, were around $1bn. So it seems the other 237 banks on average have a loan similar in size to the two smallest banks of the 20 we analysed in detail. This is also consistent with the observations we made of smaller banks in terms of EU exposure, like BNY or British Arab Bank.

However, this indicates a longer tail than one would have expected, for example under a Pareto distribution. 80% of the total loans to customers of $707bn is equal to $565bn. The “top twenty” banks we analysed in detail account for $392n of those, meaning that $173bn must be accounted for by other, smaller banks. These smaller banks will be the largest of the 237 smaller banks. If those banks have slightly larger loan books than the $1.3bn average, say around $3bn, like UBS, that would imply another ~58 smaller banks included in the 80% of total assets held by the largest banks, implying a total of 78 (58 plus the “top twenty”) held 80% of loans to EU customers by UK located banks. 78 is ~30% of the 257 total number of banks. If this is correct, then the distribution of loans is significantly different from a Pareto distribution, where 80% of the loans would be expected to be written by 20% of the banks. A Pareto distribution would have implied that 80% of loans were held by 51 banks (20% x 257), so the actual number of banks estimated here is 27 – around 60% – higher than the Pareto number. What that means is that the loans are concentrated among the larger banks, as expected, but the tail is longer and more significant than might have been expected using Pareto distribution as a prior assumption. The fundamental question is, how can 237 banks be profitable in the EU with loan books averaging $1.3bn? At 3.13% return, they would be making an average of ~$45m per annum. RBS said it would require an extra tens of millions of pounds to fully staff its Amsterdam subsidiary post-Brexit, giving some idea of the fixed costs which might need to be absorbed by that ~$45m average income. Therefore, if any of these banks have significant overheads in London they will be loss-making.

There are two explanations for this contradiction. The first is that it would seem, from some of the announcements made by banks after the Brexit vote, that some European headquartered banks have been booking some of their European loans and other IBAs in their UK offices. Here is a report of Deutsche Bank’s John Cryan’s comments to the press on his post-Brexit plans:

Deutsche Bank

What is noteworthy here is the reference to moving the “balance sheet.” We shouldn’t be distracted by the journalist’s inept use of “markets” (nor indeed his omission of the apostrophe). Trading of securities on the market is for the most part off-balance sheet. Even Deutsche’s on balance sheet proprietary trading of securities is in securities which can be and are freely traded by investors all over the world. It doesn’t matter where you book them. The only balance sheet assets which might have to move because of Brexit would be loans to customers. In all likelihood therefore, it is to these that Cryan is referring.

We saw above that a large portion of UK located lending to EU entities was made to entities in the Netherlands, Luxembourg and Ireland. It would make perfect sense if, for example, Deutsche Bank lent to Beiersdorf B.V. Amsterdam from its London office rather than from its Hamburg office. Many European banks may also choose to book certain loans to European customers in London in order to help their other London desks to sell ancillary products such as interest rate or FX swaps to the borrower. In terms of Brexit impact, a company to whom such products can be sold will typically be large enough to buy ancillary products from banks outside the EU. Therefore, although there will be many such loans booked in London by European banks like Deutsche, the loans that will actually have to move will likely only be those to medium sized customers too small for banks outside the EU to be authorised to sell it complex financial products out of London, but large enough to have any requirement for such products in the first place. What is true of Deutsche Bank will be true of other European banks with significant investment banking arms, such as Société Général, BNP Paribas, Commerzbank (which acquired Dresdner Bank), ING and Unicredito.

The second explanation for this long tail is that many international banks, particularly from the emerging markets, accompany their clients abroad. An Indian firm opening a business in Europe will usually try to finance that business in Europe, mainly to benefit from lower European interest rates relative to Indian rates (entirely hedged by European revenues), but also to reduce currency exposure and sometimes for tax optimisation. The simplest way for them to do so is to get the money from the bank they work with in India. There are many such companies across the world outside of the EU which have EU subsidiaries of one kind or another. Indian companies in particular, because they face domestic capital controls, have to finance any EU subsidiaries on an autonomous basis. Rutgers Belgium, a coal tar manufacturer headquartered in Gent, for example, has been majority owned, with a separate listing, by India’s Rain Industries since 2012. Rutgers almost certainly had its own banking relationships before it was acquired by Rain, but companies like Rain will frequently obtain funding for their newer, smaller EU subsidiaries from a local Indian bank which knows the parent company well and, in Rain’s case, can obtain security for the loan against Rain industry and its subsidiary. From time to time, even a larger company like Rutgers might obtain funding from an Indian bank’s London office at more attractive rates than its existing lenders due to the parent company pledge. In any case, the debt component of the acquisition consideration paid by Rain will likely be sourced from an Indian bank but serviced by Rutgers Belgium’s cash flows. The majority of these emerging markets banks will have established an office in London as a convenient way to manage such loans and often to locate other activities and assets away from their domestic regulator. Examples might include Bank of Baroda and Bank of India, as well as new banks such as Yes Bank which was reported in the press as applying to set up a branch in London explicitly to support their Indian clients abroad:

Yes Bank

Such a setup would represent a typical example of how emerging markets banks operate out of London. Other examples might include China CITIC, Thailand’s Bangkok Bank or Indonesia’s Bank Negara‘s London offices:

Bank Negara

In terms of our analysis of the “tail” of banks with small amounts of loans to EU customers, such foreign banks might well have small exposure of around a billion to a few companies from their respective domestic market seeking to establish operations in the EU.

Crucially, the EU companies to which these emerging markets banks’ London offices are lending will not be EU companies at all but foreign companies. And, unless the subsidiary is a sizeable, independent company like Rutgers, the bank which lends to that EU subsidiary will present the exposure represented by that loan in its reports and accounts as domestic exposure to an Indian company, not EU exposure, so this is exposure which will only appear as EU exposure in the BIS reports, not in the bank’s own risk exposure. By not using the BIS data, OW was unable to capture this dynamic.

The reason that these banks’ operations are sustainable with small loan volumes and without the overheads and network normally necessary to be active in the EU retail and business loans market is that their business in Europe is an extension of their business in their home country and gained through their network there. In fact, they are not really active in the market for retail and business loans to domestic EU customers at all. The tail of loans written by these emerging markets banks are based on different market shares, entry costs and economies of scale than those in the domestic EU market, and therefore would not be expected to conform to or be included in any Pareto type distribution obtaining there.

The average loan of the smaller banks of $1.3bn will mask considerable variety. Some banks like the smaller UK banks will have less than a billion. Some banks may have no EU exposure at all in some years. Some Asian banks will grow year to year as their clients expand, while others contract as theirs repay their loans. European banks will hold large but fluctuating loans to various EU customers booked in London.

In summary, the remaining $315bn exposure outside the “big twenty” will consist of:

  • Loans by European banks like Deutsche Bank to European clients but booked in London;
  • Loans to European subsidiaries of emerging markets companies by emerging markets banks with a relationship to that company in the emerging market concerned, booked in their London office;
  • Loans by smaller banks such as Arab Bank or BNY;
  • Probably in that order.

The important point about these smaller loans outside the “top twenty” is that only a small portion of the London booked European banks’ loans to European clients need to move, since most of them are to large companies, and none of the loans to emerging markets subsidiaries need to move because of Brexit … since they’re not European companies in the first place. Unfortunately, trying to quantify the relative weight of the different components really is guess work.


Size matters

Although we have now shrunk from a £23-27bn to an £11.4bn estimate (including the £1bn private banking revenue figure), a further reduction needs to be be made. As we explained above, the loans to to non-financial customers figure we estimated included loans to large companies like Germany’s Reckitt Benkiser or Belgium’s Inbev, for which UK located banks do not require an EU financial services passport (to accept their deposits or offer them financial advice). We therefore need to estimate what percentage of those loans are to large companies and reduce the estimate of income at risk from Brexit accordingly. The BIS does not break loans down by size of borrower, but a few national banks provide those statistics. Incredibly, the Bundesbank, in the land of the Mittelstand, does not provide a breakdown of credit by size of business:


Our data is from the largest European banks with good quality data records, the Bank of England and Banque de France (see here for personal credit statistics). This data includes both personal (mainly mortgage) and corporate loans. To use this data to estimate what percentage of cross-border loans to EU customers from UK located banks are to large companies, we need to determine what proportion of these cross-border loans are to individuals and what proportion to corporates. We will estimate loans to EU individuals  by UK located banks separately, mainly relying on individual bank disclosures, in order to estimate the residual amount of loans to EU corporate by UK located banks. We will estimate the percentage of corporate loans in the UK and France which are to large companies, and which we will use in our Brexit impact calculations, based on the Bank of England and Banque de France figures for corporate loans.

The BIS do not disclose the percentage of cross-border loans to individuals as a separate statistic, and the fact that they break loans down into “non-banks,” “non-bank financial” and “non-financial” strongly suggests that their cross-border statistics are dominated by loans to corporates. However, RBS, Lloyds and HSBC all record lending to individuals in Europe. Of the US banks, only Citigroup mentions that its “Other” exposure is to “corporate and households.” There is no disclosure from the other banks we have analysed which would allow us to attempt to estimate whether they have any exposure to individuals or households at all. We do know, however, that loans to individuals, mainly mortgages, are typically smaller and more numerous than business loans. Therefore the scale and network arguments which we used to rationalise the fact that most loans to customers were concentrated in the hands of a few banks apply even more to loans to individual customers. Such loan books typically need a significant number of customer service employees to manage the thousands of mortgage customers. The only exception would be an internet only operation such as that offered by ING direct, but none of these has grown to any scale.

Looking at the Japanese, Swiss and Canadians, there is no evidence that any except the Swiss are involved in the individual loans market in Europe (apart from any loans to private banking customers by the Swiss), and, as we saw, their EU loan books are small and any individual component may not be booked in the UK. As for the emerging markets banks we can rule out any cross-border individual lending in Europe, given their business model and network. It would make no sense for the European banks to book any of their mortgage loans or other individual loans in London. Therefore we can expect that banks outside the four UK banks and big five US banks will have no individual exposure.

The UK banks which have individual exposure do so for different reasons. Of RBS’ and Lloyds’ respective £14bn and £9bn personal loan exposure, £13.5bn (96%) of RBS’ was in Ireland and 97% of Lloyds was roughly evenly split between Ireland and the Netherlands. In other words it is highly concentrated in two countries in which both those banks made considerable investments. Lloyds exposure to loans to individuals in the Netherlands is mainly due to the fact that it was the first bank to offer online mortgages there in 1999. Now that online mortgages are standard any competitive advantage Lloyds had in the Netherlands is gone. RBS’ exposure to Ireland is due to its ownership of Ulster Bank which it acquired when it acquired NatWest bank. Both banks’ exposure to Ireland is due to a combination of Ireland’s proximity to the UK and the now notorious Irish property boom. Lloyds, as mentioned above, is exiting the Irish market and the book of mortgage loans of which its exposure to individuals in Ireland is composed is in run-off. Therefore, the circumstances leading to this exposure is idiosyncratic and cannot be used as a basis to assume that other UK banks or the top five US banks had any individual loan exposure at all. Here, for the record, are RBS’ and Lloyds’ ratio of individual loans to total loans to customers (in £bns):

Lloyds RBS individual

These legacy mortgage books are just under 50% of both banks EU loan exposure.

The picture for HSBC is entirely different. Here are its gross loans to customers already reproduced above (from p 104), with the detail on the type of counterparty given by HSBC:

HSBC EU personal

The first two columns represent HSBC’s European individual loan exposure. Most of it is booked in its HSBC France and Trinkaus subsidiaries, again demonstrating the importance of scale and franchise for writing such loans. However, its “Other” European operations have ~$5bn of mortgages and other personal loans. This represents a third of its total “Other” European loan book. That is significant exposure. As we shall discuss further below, it seems that HSBC has been more aggressive than most banks in pushing cross border retail banking expansion, and this $5bn individual loan book in its “Other” European countries seems to support this impression.

In terms of read across, we could use this ratio as some sort of guide for Barclays and the US banks. However, none of those have the same extensive European network as HSBC. Nothing we have observed is enough to even rule out the possibility that some of them have no exposure to individual loans in Europe. Taking these observations together, it would seem prudent to assume that Barclays and the top five US banks are less exposed to individual loans than HSBC. Below, on the left hand side of the line, are the figures for individual exposure for the three UK banks which disclose them, in billions of dollars and as a percentage loans ex-NBFIS and excluding any loans booked in European subsidiaries. The UK estimate for NBFIs of $31bn excluding Lloyds $3bn has been allocated pro-rata to the other three UK banks. Using this new denominator, which is consistent with the figures used in our income impact calculations, the percentage for individual loans has increased:

Individual %

Based on HSBC’s ratio of 43% of loans ex NBFIs in its “Other” European countries I conservatively estimate that 35% of Barclays and the US top five exposure ex NBFIs is to individuals. That estimate can be benchmarked against Bank of England and Banque de France data for individual loans as a percentage of total loans recorded by these institutions:

Individual % central banks

The numbers recorded by the two central banks are dominated by domestic lenders, and the proportion of loans to individuals is due largely to mortgages, which represent 87% and 82% of UK and French loans to individuals, respectively. They also represent 72% of HSBC’s and the bulk of Lloyds’ and RBS’ loans to individuals. In the absence of a large book of mortgages, you would expect the percentage of other banks’ loans to individuals in the EU to be around HSBC’s ratio of consumer loans as a percentage of total loans, namely 9% ($1,393m/$15,054m) – if they have any at all. Therefore, the assumed ratio of 35% for Barclays and the US banks allows plenty of headroom for some of those banks to hold a significant book of mortgages in Europe, though not as large as Lloyds, RBS or HSBC. A google search of Barclays and the US banks shows little evidence of any presence in the European mortgage market, which is not surprising.

Having estimated the percentage of loans to EU customers ex individuals, we can turn to the Bank of England and Banque de France data to estimate what percentage of the corporate loan book is to large companies. Here are the two numbers and an average:

percentage large

As you can see, both in France and the UK the weight of large companies in total borrowing is (i) quite close and (ii) around half the total market. Note this is the average for all large companies, including NBFIs. NBFIs are more represented in larger companies than in mid-sized or smaller companies, and some of them, particularly insurance companies, are more highly leveraged than the typical large company. Unfortunately, we have no data to quantify this effect. Given the high exposure of the UK economy to finance and the relatively high exposure of France’s economy to large industrial companies such as Lafarge and St Gobain and relatively small exposure to NBFIs (pace Axa, CNP and reinsurer Scor), the lower French ratio would seem to be appropriate to use in our estimate of the weight of large companies in BIS loans to customers ex NBFIs.

We can apply this 41.3% ratio to the non-financial loans, with one important exception: the Netherlands, Luxembourg and Ireland. For those three markets, as discussed above, it is likely that large companies represent a much larger percentage of all borrowers. If you’re big enough to set up a tax structure in Luxembourg you’re probably big enough to borrow internationally without your lender requiring a passport. Here is how I translate the claims figures provided by the BIS into an estimate of non-financial exposure for these countries:

Netherlands etc non financial exposure

Since we reclassified $231bn of the $435bn intragroup loans (53%) as loans to customers, the Netherlands, Luxembourg and Ireland need to take a share of that. I calculate that amount based on the intragroup loans estimated specifically for each country multiplied by the 53% of all intragroup loans assumed to be used to fund loans to customers. I assume that there will be the same proportion of NBFIs in this set of countries as there is for the BIS population as a whole. Even though all three countries have large NBFI industries there are insufficient tangible grounds to assume otherwise.

Next we have to estimate what percentage of the non-financial customer loans in each area is to subsidiaries of large companies from other countries which have been established in those countries for tax optimisation. Here, the BIS can help us again. In its B.4 reports, it discloses “Claims on an immediate counterparty basis” and “Claims on an ultimate risk basis” for individual countries. “Immediate counterparty basis” means where the borrowing entity is located, whereas “ultimate risk basis” means where the parent company and ultimate source of credit risk has its headquarters. The difference between the two, expressed as a percentage of total claims on an immediate counterparty basis, represents the proportion of loans to entities in that country which are subsidiaries of parent companies located elsewhere. Here, by way of illustration, is the report for Luxembourg:

Luxembourgh immediate vs ultimate

As you can see, there are $598.66bn claims on entities located in Luxembourg, but only $477.88bn of those are claims on companies whose parent company is headquartered in Luxembourg, implying $120.8bn (or 20% of total claims on an immediate counterparty basis) of those claims are on subsidiaries established in Luxembourg by companies headquartered elsewhere.

By contrast, Germany has ~$10bn more claims on an ultimate than immediate basis, meaning that some of their companies will borrow through entities in other locations, like Luxembourg:

Germany immediate vs ultimate

We can use the difference between claims on an immediate and ultimate risk basis, expressed as a percentage of claims on an immediate risk basis, for the Netherlands, Luxembourg and Ireland, as a proxy for the percentage of UK claims on entities in those countries which are on subsidiaries of large companies from other countries which have been established in those countries for tax optimisation and other reasons. Here are the numbers:

tax driven country percentages

Although there is a big tax driven business in the Netherlands, the country does also have a significant industrial and commercial base. Ireland is a smaller economy and is therefore likely to be more tax driven than the Netherlands. However, because of the UK’s proximity to Ireland, as well as the long standing presence of some of its banks in the market, it is very likely that the over-representation of Irish loans to non-financial customers in the books of UK located banks, relative to the market as a whole, is due in part to loans to Irish industrial or commercial companies and not exclusively to tax structures. Luxembourg, finally, is a tiny economy and the lending will be heavily driven by tax structures.

The numbers above suggest a much lower influence of tax driven structures than one would expect on this basis, with Luxembourg’s ratio almost equal to that of the Netherlands. Surprisingly, Luxembourg borrows almost as much as the Netherlands on an ultimate risk basis, despite its tiny economy. This is very likely due to a significant amount of lending to Luxembourg banks and other financial institutions. However, trying to quantify this is subjective, so we will, conservatively, stick with the somewhat counterintuitive, lower numbers from the BIS. Before performing the final calculation we need to separate individual and corporate lending. We could use the average calculated above of 17.8% of total UK located loans to EU customers being to individuals. However, the disclosures from RBS and Lloyds tell us that both companies had $26bn of loan exposure to individuals in Ireland, almost half the $62bn estimated exposure to non-financial customers in Ireland calculated above. By contrast, Lloyds disclosed $7bn exposure to the Netherlands is only 7.3% of the estimated exposure to non-financial customers in the Netherlands calculated above. So for Ireland we have to take account of the higher disclosed exposure to individual credit. For all other markets, we have to use the percentage of individual loans to non-financial customers, ex Ireland.

Ireland rest of Europe individual

This allows us to calculate the following break-up of loans between large company and other (personal and small or medium sized business lending) for the three tax driven countries:

Netherlands etc large cap

Note the BIS figures for cross-border claims on an immediate and ultimate basis that we are using using to calculate the percentage of tax driven loans in each of these countries will include both individual and corporate loans. The percentage of loans to tax driven subsidiaries would be higher if we were to use cross-border claims on an immediate and ultimate basis for corporate loans only, but we have no reliable way of doing so as BIS does not provide a breakdown of loans into individual and corporate. Therefore our estimate will understate the volume of tax driven loans in these countries to some extent.

We can subtract the tax driven loans (which will be to large companies) and non-tax driven loans to large companies to estimate the loans at risk from loss of passport. This leads to the following estimate of loans to large non-financial companies and revised income impact estimate:

size effect

Our estimate of Retail and business banking income has shrunk by another £3.9bn or 38% to £6.5bn, well under half of the £14bn mid-point estimate by OW for that figure. Including the £1bn from private banking, OW’s initial £25bn estimate of “European banking” revenues at risk from a loss of passport has shrunk to £7.5bn, well under one third of the original.

final reduction

This estimate may be generous. The $493bn loans to EU non-financial customers from UK located banks we calculated almost certainly includes loans to foreign company subsidiaries. Bank Negara doesn’t need a passport to lend to an Indonesian company’s subsidiary! (Bangkok Bank’s and Bank Negara’s 2016 annual reports mention Brexit as impacting market sentiment and certain economic indicators, but not as an impediment to their ability to lend). I have not adjusted my estimate for this because there is considerable conjecture in the size of the foreign subsidiary component, and because some of that exposure will be included in the larger company and tax driven structures already eliminated, above, from our impact estimate. If we were to assume that one third of the $220bn of loans to non-financial EU customers outside the “top twenty” is to emerging markets subsidiaries, and half of that had not already been included in tax driven and larger company exposure, you would reduce the income loss forecast for all banking activities (including private banking) by £0.8bn to around £6.7bn, well under half OW’s £15bn mid-point estimate.


What was the point?

This has been a laborious exercise. In theory, we could have saved ourselves the trouble by just using the BIS data. It was that which gave us the final figure for loans to EU customers with which we tried to reconcile the company specific data. Moreover, it provided the fascinating insight regarding the location of the counterparties to those loans, with the disproportionate share of loans to the Netherlands, Luxembourg and Ireland strongly suggesting that a more significant proportion of the loans to EU counterparties from banks located in London were tax driven and made to subsidiaries of large companies which could borrow from banks outside the EU. OW really missed a trick by not examining this data.

However, we did need the company reports to estimate the asset margins on loans to customers. Moreover, it was only the company specific analysis of balance sheets and in particular silo accounts which gave us data points on which we could base our $231bn estimate of how much to add to the BIS’ stated loans to customers figure due to the reclassification of intragroup loans. Moreover, it was the company data which allowed us to estimate the $88bn quantum of cross-border loans to individuals and observe their concentration in Ireland. It was this which allowed us to estimate exposure to loans to corporate customers for different groups of banks and, in turn, apply central bank derived percentages of loans to large customers to those estimated corporate loan exposures to estimate the size impact with any precision.

The loans observed in HSBC’s silo accounts also led to a ~$50bn increase in the estimate of the volume of loans to EU customers written by banks outside the “top twenty.” It was the specific data on the size of individual bank exposures to Europe which allowed us to understand the distribution of the total loans to customers figure recorded by the BIS, adjusted for intragroup loans, and which therefore confronted us with the question of why there was such an unexpectedly large tail. Understanding that large tail led us to the investigation of the role of European banks booking loans to European customers in London and of the London branches of emerging markets banks. The comparison of the securities figures from individual bank balance sheets and the aggregate number provided by the BIS helped us understand the fact that EU issued securities held by foreign banks with London branches are primarily held at the parent company level, not in the London subsidiary. The analysis of Sumitomo’s silo accounts demonstrated that this was true to a lesser extent of loans to customers. In short, being able to reconcile the BIS data with the company bottom up data of over twenty banks gives us increased confidence that our estimates are robust and helped understand the detailed composition of UK located loan exposure to EU customers.

I can see why OW didn’t want to look at the data. As we saw, it is often confusing. It forces us to dig deep into abstruse areas like silo and parent company accounts, and to try to reconcile heterogenous data. It is much easier just to pick up the reports and accounts of the big banks you work for, call around and get a number handed to you. The SMBCE silo accounts’ revelation that some of Sumitomo’s loans to EU customers are booked in Tokyo and others in London is one of the most salient examples of the contradictory nature of some of the data we examined. It was particularly hard to reconcile with the assumption that emerging markets banks like Negara or CIBC, contrary to Sumitomo, book virtually all their loans to EU subsidiaries of their domestic customers, from countries like Indonesia or Canada, in their London office. When you perform analysis like this, it would be comforting to be able to apply one assumption in a uniform manner and not have to deal with such contradictions. Management consultants like OW are particularly attracted to such clarity; it helps make their reports seem more convincing. But reality is not neat and tidy like this. Only by looking at the widest range of data, and trying to think through the inconsistencies that the data throw up, can we understand the intricacies of the international banking market and how they apply to the risk Brexit poses for banks lending to the EU from the UK (or lack thereof).


Sanity test of the overall numbers

Although we have made a number of assumptions, both at the high level and in some of the detail, what really counts are those assets which require an EU financial services passport, and those are the loans to customers, ex NBFI and large company customers. It was useful to calculate the other forms of exposure insofar as it allowed us to make sure that this estimate formed part of a coherent and rational picture of all the available data.

Our figure for total UK located claims on EU entities is based on hard numbers from the A.6 reports. The UK headquartered bank customer loans figure of $209bn is based on explicit disclosures in their annual reports. We also have a hard consolidated numbers from the BIS for the UK headquartered banks’ total claims on EU entities (which is very close to the figure for total exposure). So the residual amount of claims on EU customers, ex the UK banks, is based on two hard numbers: the total BIS number for claims of branches located in the UK ($1.6 trillion) minus the total claims of UK headquartered banks ($640bn), leading to a number for claims of foreign banks located in the UK of ~$1 trillion.  The BIS also gave us hard numbers in the A.62-S tables for the breakdown of that $1.6 trillion into loans to customers, loans to banks and securities. We were therefore able to make a minor (~1%) adjustment to the loans to customers numbers for credit provisions and to the securities numbers for an estimated $102bn of equity securities (based on BIS data for equities as a percentage of cross-border securities claims on the EU as a whole).  We also adjusted the UK headquartered figure for HSBC EU debt securities and loans booked outside the UK ($113bn), based on explicit disclosures in the annual report, and assumed that $250bn of EU debt securities exposure reported by US banks were held in their US central funding departments and not in their UK offices. This allowed the bottom up debt securities data to reconcile with the BIS A.6 data.

This left us with around $1 trillion in claims on EU entities by UK located customers outside of the UK headquartered banks. What we have done, in essence, is calculate that a certain percentage of that remaining ~$1 trillion was loans to customers. For that estimate, we were able to rely on both the hard figures from the BIS for loans to EU non-bank customers from UK located banks by country and data from five big US banks and eleven other banks. The only estimates which really impacted the final number were (i) the estimate of the percentage of intragroup loans which were made to subsidiaries and the residual percentage to branches and (ii) what percentage of intragroup loans to subsidiaries were used to make loans to customers. That had a significant impact on the increase of our estimate of loans to customers over the figure disclosed by the BIS A.62-S reports. Regarding (i), Franklin and Gu’s data on the number of subsidiaries relative branches, the fact that lending via a subsidiary is less risky for banks and evidence that the biggest UK located banks which are longest established in Europe have subsidiaries in Europe through which they channel their loans to EU customers all support a higher number than the two thirds that we assumed. Regarding (ii), our assumption had a lesser impact than (i) and was supported by detailed analysis of intragroup balance sheet data and ultimately reflects a fairly typical bank balance sheet structure.

It is worth reflecting though on what would have happened to our estimates if we had used assumptions which increased our estimate of the intragroup loans which were ultimately converted into loans to customers. That would have, effectively, only increased the tail of loans which, as we analysed above, is largely due either to European banks booking some of their loans to EU customers in the UK or to emerging markets banks following their customers into Europe. Since the former are mostly loans to large companies, and the latter are not loans to EU companies, an increase in the size of this tail would actually not have had much of an impact on the amount of revenue at risk from Brexit. Again, it would have been impossible to understand this without looking at both individual company reports and aggregate BIS data.

Finally, our estimate of how much of the EU loans to non-financial customers were to large companies had a significant impact, but that is based on (i) the percentage of loans to individuals as a percentage of total loans to EU non-financial customers, for which we used a combination of disclosed figures and conservative assumptions based on individual company data which showed little evidence of exposure to individual loans outside a handful of companies, (ii) the percentage for large companies as a share of corporate loans from the Banque de France, which is lower than the one in the Bank of England data, as well as (iii) a small percentage for tax driven lending based on BIS data which if anything seemed to understate the amount of tax driven lending in the Netherland, Luxembourg and Ireland.


Why did OW get it so wrong?

This is a hard one to guess. However, we can imagine that the “top twenty” banks on our list would be OW’s first port of call. They are big enough to be able to employ OW and matter to them as a customer. If we take the total European exposure of the UK headquartered banks and the US top five, you get a total exposure, including securities and loans to banks as well as loans to customers, of $766bn. Add the loans to customers from the other eleven banks of $89bn and you get $855bn. Use the 3.13% rate of return we have been using in our calculations and you get income of $27bn or £18bn at the 2015 exchange rates on which OW’s estimate was likely based. That’s not far off from the top of the £13-17bn range of EU related retail and business banking revenue implicit in OW’s figures. In other words, it seems as though OW has simply taken all the retail and business banking revenue earned from EU counterparties by the banks that they have contacts with.

The key phrase which contributed to OW’s downfall, I suspect, is the vague “related to the EU” (p 6). How related? Is revenue earned by Barclays from holding a Bund “related to the EU”? Is income earned by Citigroup’s London office from an interbank loan to Rabobank in Rotterdam? Is fee income earned by Sumitomo from arranging an FX swap for the Intel Corporation’s Dublin subsidiary? Quite possibly. But none of these income streams are actually at risk from Brexit. OW’s sin was not to make the effort to distinguish between securities and loans; loans to banks and loans to customers; loans to large customers and loans to small customers; loans to NBFIs and loans to non-financial customers; loans to European customers and loans to EU subsidiaries of foreign customers; intragroup loans to subsidiaries and intragroup loans to branches etc. If you take a broad brush approach like that, it is very, very easy to come up with such big, eye catching numbers.


Setting up subsidiaries

Loans are intangible. You can move money around at the touch of a button. As we discussed above, one of the reasons banks need to be authorised in the EU to lend to certain EU customers is so they can accept their deposits. To overcome this hurdle, there is no need for UK located banks to move any of their lending infrastructure per se. All they need to do is to set up a fully capitalised subsidiary in the EU that can accept customer deposits.


Equity impact

The implication of UK located banks setting up a subsidiary in the EU for the UK economy is, first, that they have to move some equity capital from the UK to the EU. That equity will back the deposits which enable them to make their loans. The amount of deposits which need to move to the EU in order to continue making the $307bn loans at risk from the passport, on which we based our £6.5bn revenue impact, is not $307bn. The actual amount can be inferred from the offsets that HSBC and Lloyds indicate reduce their loan exposure relative to the loans they book on balance sheet, net of provisions. As discussed above, in the case of loans to customers that offset is a proxy  for deposits made by the borrowing entity. We can use it in reverse to estimate what deposits the UK located banks would need in order to continue to make the loans at risk from Brexit. The figures are 4% for HSBC (p 87) and 1.4% for Lloyds (p 262, reproduced above). In other words, a company borrowing $100m from HSBC would, on average, be expected to have a deposit of $4m with HSBC, reducing HSBC’s exposure to that customer to $96m.

You wouldn’t expect that percentage to be too large, as it is more efficient for companies to use the cash to pay down debt or to expand the business. Nor, however, would you expect it to go too low either, as companies need to hold cash reserves in case they suddenly need it (for an emergency, an acquisition, a fine – whatever). We can assume the higher HSBC figure to be on the safe side. Therefore you might expect up to 4% of the total $307bn of loans to non “mega-cap” non-financial EU customers to constitute deposits which would require an EU subsidiary, which means UK located banks in aggregate would need to hold deposits amounting to $12bn. If those deposits require a 10% tangible equity ratio, then the amount of equity would have to be $1.2bn. Most UK located banks with significant EU operations have an EU subsidiary, though as noted Lloyds did not have any subsidiary and Barclays’ was not fully capitalised. Assuming, conservatively, that three quarters of the subsidiaries are not already fully set-up and capitalised, you would expect an equity capital transfer from the UK to the EU subsidiaries of ~$0.9bn. That transferred capital will in the main be invested in safe government bonds. The UK will lose the return on those free funds, which is likely to be remunerated at around 2%, given current government bond yields. The equity revenue impact is therefore $18m or a derisory £12m per annum.

Even the lower number we arrived at overstates the impact on UK located banks from Brexit. To keep the EU related NII, all the banks need is a subsidiary through which they can accept EU deposits. OW not only overestimated the revenue at risk from Brexit but seemingly did not reflect on how why that revenue was at risk. Understanding that all the banks needed to keep their EU related NII was to be able to accept a small amount of EU deposits enables you to quantify how minor the impact of this is.


Jobs impact

The key worry about all this is, understandably, that high quality “banking jobs” will have to move to Europe. OW forecasts 3,000-4,000 job losses in banking as its low end, high access, estimate and up to 75,000 in its high end, low access scenario. But who actually needs to move? As we have demonstrated elsewhere, most of what the City does involves wholesale transactions with other financial institutions which, by definition, do not require a passport. Yet, as we shall see, banks have announced some job relocations to the EU as a result of Brexit – despite the absence of any impact on their key wholesale business. So there must be some activities which are impacted. The work we have done above on the quantum and structure of the various banks’ exposure to loans to EU customers helps us understand why any jobs need to be lost:

  • One activity which does require EU authorisation is deposit taking from EU customers, usually in support of a lending relationship. In order to meet that requirement, UK located banks will have to set up a subsidiary and staff it with a small amount of personnel.
  • We also identified any people currently in London selling interest rate or currency Swaps or any other financial products to EU non-financial companies outside the large companies – essentially SME customers deemed too small to purchase such sophisticated products from non-EU regulated entities – as needing to relocate. The regulation of sales of Swaps to SME customers became a hot topic after inappropriate Swaps were sold to SMEs in the run up to the financial crisis, as reported here in the FT for example. People selling such services to EU SMEs would have to do so from an EU regulated entity, and so this activity, too, can no longer be carried out in London, post Brexit, without the UK securing a right to EU financial services passports for firms in the country. This is the sort of activity which contributes to the commission income which, we saw, generated income of about 0.37%-0.9% of average customer loans to the three UK banks for whom we calculated the ratio. It is quite likely that some banks sold EU customers a number of complex banking products, such as Swaps, from the UK. Note that EU SMEs who are sold such products from London will typically be relatively large. You wouldn’t expect the local corner shop or garage in Nijmegen to buy interest rate caps from JP Morgan in London.
  • Finally, any sale of financial products to individuals in the EU from London is also financial advice and therefore requires authorisation in the EU for which the person giving the advice has to be in an EU jurisdiction or one benefiting from a financial services passport. Most of this activity is classified as private banking.

If the bank only needs to set up a subsidiary to attract deposits, job creation in Europe and any associated losses in the UK will be small. However, if they need to move sales staff from London, to service SMEs, individuals or both, the impact will be greater. We can therefore investigate the banks we have analysed according to the following criteria:

  • Do they have a subsidiary in the EU and how large is it?
  • How extensive is their existing branch network in the EU?
  • Do they have significant disclosed exposure to individual or SME customers?
  • What Brexit related job relocation announcements have they made?

Some banks have made no announcement at all, implying there is no need for any job losses:

Some banks have indicated they don’t need to make any job cuts in the UK, although they would have to hire small numbers in a new EU subsidiary:

Goldman Sachs announced the creation of around 200 jobs in  Frankfurt, with some transfer of jobs from London, but crucially said there wouldn’t necessarily be any net transfer from London, implying job creation in London might offset any transfers to Frankfurt. This implies minor numbers are involved and contradicts some alarmist forecasts circulated in the media, such as this piece in the Guardian from 21 February 2017, which estimated 3,000 jobs or half of Goldman’s London work force might have to go:

job loss rumor

A number of banks have discussed the creation of jobs in their new subsidiary while citing small or unspecified numbers of job losses in London :

The case of Sumitomo is particularly interesting. Although much has been made in the press of Sumitomo’s vaguely worded press release, SMBCE’s annual report for the year to March 2017, published in July of that year, is altogether phlegmatic about the prospects of Brexit (p 6):


Here are the announcements above, or estimates derived from them (marked in gray) juxtaposed with the staff employed by the bank in London, where known:

job losses

Contrary to the widely quoted figure of “20%,” the announced job loss numbers, even making generous assumptions, nearly all fall well below 10% and mostly below 5% of London staff for the banks making the announcements. One of the figures above 10%, Daiwa, is based on my estimate, while Bank of America’s trading jobs announcement may well prove excessive as UBS’ was. The aggregate figure is derisory, just over one thousand or 3% of the London work forces of the banks concerned.

It seems that all but a few of the job announcements detailed above, such as they were, were required at least in part for the establishment of a subsidiary. Nearly all of the groups concerned had subsidiaries of one kind or another, but only some of them were registered as banks: Barclays (Dublin), RBS and Mizuho (Amsterdam). The rest need to establish a banking subsidiary in Europe. Even those with existing, licensed banking subsidiaries will need to beef up their operations to some extent, RBS seemingly more so than Barclays. Mizuho, as noted above, has not opted to use Mizuho Bank Nederland N.V. Whatever the case, this requires a certain number of corporate and regulatory staff, and it is likely that the bulk of the new hires in the new locations will be the result of this, not because actual revenue generating activities are “moving from London.”

Note however that Sumitomo, Nomura, Daiwa, Citigroup, RBS, TD and others mention the ability to execute “trading” for their customers as one of their reasons for moving jobs or staffing up their new subsidiary. Crédit Agricole’s CEO mentions “some activities”; the bank is headquartered in Paris and therefore has no need to set up a subsidiary, so it is likely also moving trading jobs. As we have written elsewhere, wholesale trading for other financial institutions or on its own account (“prop trading”) constitute the bulk of any bank’s trading activities, and these do not require EU authorisation as they are with wholesale customers or counterparties who can trade internationally in the ordinary course of their business. The one revenue generating activity which needs to move due to Brexit is trading which either involves the sale of financial products to SMEs or with investment advice to individual, private banking type clients, as described above. It is interesting, therefore, that banks have named “trading” as a reason for moving staff and that they have said that the job losses are based on a worst case assumption that UK located firms will lose access to the EU financial services passport post-Brexit. The small UK job losses they have announced are therefore already assuming such trading can no longer be carried out from London. In other words they already factor in a no deal, worst case scenario. What this implies, if confirmed, is that the job loss estimates in the table above include those which are due to the only activities carried out in the EU by the banks concerned which really are affected by Brexit. These numbers therefore represent the total number of staff who will need to move if no financial services passport is granted.

This contradicts what seems to be the received wisdom on the issue in certain parts of the media. The low job loss numbers announced so far have not been widely reported, with media commentary focusing instead on the aggressive and, as demonstrated by this blog, wildly inaccurate forecasts of job losses from various sources (including OW). Any acknowledgement of how low the job losses have been so far is typically qualified with a forecast that the announced figures are only the precursor to larger job loss announcements. EY, author of a stupendously bad analysis of the impact of Brexit on the City, reviewed elsewhere, claims that job loss announcements so far are “the tip of the iceberg.” This comment by Reuters on the results of a survey they conducted on the subject is typical. They acknowledge that the findings of their survey “suggest that the first wave of job losses from Brexit may be at the lower end of estimates by industry lobby groups and firms,” but add that “most respondents said bigger moves could be in store in a decade or more.” This implies that there might be further Brexit impacts, in other words that the small job loss numbers announced so far are just a prelude.

But this statement is misleading. As we have often noted, saying something “could” happen means it has a > 0% chance of happening. Given the pressures under which the financial industry finds itself – due to passive investment, MIFID, QE, automation and any number of other factors – anyone saying there was no chance of any layoffs in their firm would be barking mad. But the fact that there is a >0% chance that larger layoffs may happen in a bank in the future does not mean it is likely that there will be any further layoffs due specifically to Brexit. The article however in a not very subtle rhetorical device uses this anodyne > 0% statement to imply that there may be further Brexit related job losses, when there is no evidence whatsoever to support that expectation.

The article, slyly, goes on to support its suggestion that the (disappointingly, for them) small numbers of job losses announced only represent a first cut of many with the following unattributed quotation:

If it is going to happen it won’t be in one big bang,” said a senior executive at one of Europe’s largest banks, which took part in the survey. “There will be a slow drain of jobs from London over a number of years.”

If what is going to happen? The only way you can make sense of this statement is to interpret it as saying that if London’s wholesale business (which is unaffected by Brexit) really is going to move, due to competitive changes in the industry, such as a repeal of Dodd-Frank in the USA or the growth of Asian centers like Hong Kong or Singapore, then that would indeed manifest itself in a slow erosion of London’s pre-eminence. But there is no reason why that should happen. The loss of business to London or Asia is just a potential event. It’s an “if.” And in the unlikely event that this “if” does happen, it won’t be due to Brexit. Reuters has used another rhetorical device to trigger the narrative fallacy in its reader: by placing this quote, which says any potential manifestation of a shift of financial services business away from London in general will take place over the long term, just after their suggestion that the announced Brexit related job relocations are only a prelude to further job cuts, it makes it seem as though one explains the other. In other words it gives the impression that general and hypothetical decline of London specifically describes the mechanism according to which the actual job loss announcements specifically due to Brexit will unfold. It thereby implies that those announced job losses are not final but only part of a first wave. But this is a metaphorical transposition. It has no connection with the reality of any bank’s specific Brexit job plans. Indeed, so far, the only concrete revision has been UBS and that was a reduction, not an increase.

It is important however to distinguish between vague projections, such as these, on one hand and clear statements and hard numbers, such as those we have analysed, on the other. We already saw that the actual company statements of potential job losses have been well below the numbers floated in the market by various insiders. These are hard numbers, even though they are sometimes revised to the downside, as UBS’ were. And the statements, as described above, clearly say that trading was the reason. We explained that individual or SME trading are the only activities that need to move. Therefore, contrary to what is implied by Reuters and others, these are final, not initial low-ball, numbers.

Probably the difference between the larger job announcements nearer the 200 end and the smaller ones of the “around twenty” variety is that the former include trading jobs whereas the latter only include subsidiary establishment jobs. If these small numbers, maximum 200 people and mostly less than 5% of staff, include both head office and financial product trading, then you can see how small and peripheral such retail trading activities were for the banks concerned.

There are, in addition, five banks with more significant announcements:

  • Deutsche Bank, as quoted above, declared that “some roles” would need to go, later clarifying the number would be around 350, well below the gigantic 4,000 figure forecast by its board member.
  • HSBC, as is well known, has significant staff already in France, where it has a large banking subsidiary. The job impact in London is already quantified at 1,000, with the Guardian, in the quote above, speculating that half are “returning French nationals.” However in October 2017 HSBC’s CEO gave himself potential wiggle room not to make good on the relocation in full when he said HSBC would “wait as a long as possible before shifting jobs out of London as a result of Brexit.”
  • UBS, as noted above, also initially estimated that up to one thousand jobs will be lost in London, out of a workforce of 5,000. That number was significantly revised down in October 2017 to as little as 250.
  • JP Morgan has said that Brexit would result in job losses, but the number of 4,000 attributed to the bank is, in reality, the number of people it employs in Bournemouth. Its CEO, Jamie Dimon, has been happy to allow that forecast to be aired, but never actually endorsed it. In the end, the bank is saying that only around 700 jobs will need to move.
  • Société Générale’s CEO, Fédéric Oudéa, said that 300-400 jobs might have to move from London, while an anonymous insider from BNP Paribas allegedly told Reuters that 300 jobs would have to move, even though its CEO Lars Machenil is on video saying that London remained the group’s financial headquarters and that any adaptations it needed to make were “on the fringe.”

These are big job losses compared to the small numbers suggested or announced by the banks above, and, with the exception of Deutsche and BNP, range from 20-25% of UK staff employed by the banks concerned (the UBS estimate is the larger, initial estimate):

job losses HSBC et al

The 20% of workforce to be lost by HSBC, Société Générale and (initially) UBS is an important number. It has been, surprisingly, extrapolated to the City as a whole, for example in a report by the Bruegel Group report (analysed elsewhere) and by William Wright of the New Financial think tank:

William Wright 20%

Spoiler alert. Having read these reports, including OW’s, there is no evidence in there whatsoever that “20%” is a useful ratio for this discussion. There may be 50+ reports, but they all draw their numbers from the handful of reports we have analysed in detail and found to be wanting. The hundreds of people, in all likelihood, are either all reading the same reports or else have had “discussions” with people who have and are merely repeating what they have heard. This is, in all likelihood, a classic case of Chinese Whispers.

In fact, the 20% number seems to have been plucked from thin air. One of the prompts, or psychological cues, for this number to be adopted in this way, seems to have been these statements from UBS and HSBC. These two banks made their declarations in January and, since then, the 20% ratio seems to have been adopted as a kind of rule of thumb of the likely impact on the City as a whole. This rule of thumb survived the many subsequent announcements of much smaller job losses and the qualification and reduction of their initial announcements by HSBC and UBS, and seems to flap around in the media like a headless chicken, in a textbook example of confirmation bias.

We have seen that these banks are, based on the announcements so far, very much the exception, not the rule. Are there reasons why these five banks might be different from the others? Or are they really, as many seem to blithely assume, a sign of things to come? Again, our analysis of these companies’ loans to EU customers provides a useful clue. HSBC and JP Morgan are, respectively, the second and sixth biggest London located cross-border lenders to EU customers, according to our calculations, but UBS’ exposure to EU customers is among the tail of small exposures. Yet HSBC and JP Morgan initially had near identical job loss forecasts, whereas Barclays and RBS, with the biggest and third biggest exposures, have none. In addition, $50bn of HSBC’s $65bn loans to EU customers were already booked and managed in their French and German subsidiaries. It is therefore unlikely to be a question of the size of the loan book.

Our analysis of HSBC revealed two things which might have had an impact. First, its banking fees as a percentage of loans to customers, at 0.9%, is the highest of all the UK banks. Second, it has a significant book of EU individual consumer loans ($1.3bn), whereas its UK peers’ individual loan books are mostly mortgage related. It may very well be that HSBC’s higher fee to loans to customer percentage is due to the fact that it has been more active in selling financial products to its individual and SME customers. Some of those sales may have been of specialist products which were sold from specialist product centers in London. This would increase the number of sales jobs that needed to move relative to other banks for which such product sales were not so developed, and which therefore did not have such high fee income as a percentage of loans to customers, or which didn’t make those sales to EU customers from London. Moreover, it may be that some of the $1.3bn consumer loans made to individual EU customers were specialist loans sold or serviced by people in London, using telephone or the internet to communicate with the end customer in Europe. This would also increase the number of sales jobs that needed to move relative to other banks which didn’t have a consumer loan book of this size outside their core markets in the EU.

In addition to this, it is highly likely that exposure to private banking will have been a factor for many of these banks. Ordinary individual EU customers will do their trading online. The only individual trading customers that might be affected by Brexit would be high net worth (“HNW”) EU clients, to whom sophisticated investment products were sold from London. Such activity would, typically, fall under private banking and wealth management, and does constitute the kind of activity which can no longer be carried out in London, post Brexit, without the UK securing a right to EU financial services passports for firms in the country, as discussed above. It is very likely that it is these private banking services which are performed by the staff that UBS, for example, plans to move to Frankfurt, given its lack of European loan exposure. This might be surprising, as you would expect HNW individuals to require their adviser to be located in the same country as them; that this would be part of the personal touch they paid for. However, given the increased complexity and specialisation of financial solutions, it is likely that, in addition to their local generalist manager, those HNW clients were also advised on a number of products from specialist product centers, and that these might be located in London. It is likely such product centers which UBS is moving, not the relationship managers who are probably already on the ground in the same country as the client.

The reason given by UBS for the downward revision to its job loss forecast was that back and middle office staff connected with the activities which needed to relocate would not have to move. This is consistent with our hypothesis that these are private banking financial product roles, in which the work is divided between a “front office” sales person and back and middle office workers who process the trades instructed by them on behalf of clients. It also confirms our thesis that only jobs involving the provision of financial advice to retail clients are affected by Brexit, since back and middle office workers do not provide advice to customers.

If the Guardian’s rumor that half of HSBC’s 1,000 relocated staff are returning French citizens this would strongly suggest that, like UBS, it is likely that private banking jobs are a significant contributor, not just the financial product sales to SMEs or consumer loans to individuals described above (HSBC is unusual insofar as no head office jobs need to be transferred there because it has a large, fully regulated and staffed banking subsidiary in the form of HSBC France). HSBC’s private bank is a substantial operation, built up through numerous acquisitions, including Safra, on which the group took a substantial goodwill writedown in 2016. HSBC Private Banking generated $1.75bn in revenue in 2016 from just under $300bn in client assets under administration and with just over eight thousand staff members across the world. The division has seen declining revenue, assets under management, loans and staff over three years, with staff at the close of 2016 more than 8% lower than at the close of 2014, a reduction of over 700. Management comments: “Global Private Banking is much smaller than it was three years ago” (p 7).

The 8,000 people currently employed at HSBC Private Banking could, certainly, include French product center specialists, based in London, selling particular solutions to HSBC France’s significant number of private banking clients (France is one of four countries in Europe and one of twelve globally whose private banking profits are reported separately by HSBC). But it seems very odd that no commentators have asked whether some of the mooted job losses in London by HSBC are not the result of the public, ongoing consolidation in private banking, rather than Brexit. How much of HSBC’s 1,000 job relocation was due to banking product sales such as Swaps and how much to Private Banking is a question which seemingly never occurred to mainstream media commentators. And how much the job reduction which has been ongoing over the last three years and how much Brexit contributed to the Private Banking portion of that job relocation is so far a mystery on which no one in the media has attempted to shed any light.

These insights allow us to venture a number of hypotheses regarding whether there really is read across from the 20% job loss figure from UBS (initially) and HSBC to other banks. Take private banking first. Would we expect all private banks, large and small, to have significant product centers in London? No. You can only make such product centers profitable if you are big enough, and UBS and HSBC are amongst the biggest. There is no read across to smaller private banking operations. Moreover, there is a strong local element to private banking, meaning that many of even the larger private banks will serve their local market with a local product center team. UBS and HSBC’s “hub and spoke” model in this respect is the exception, not the rule.

Next take the jobs potentially lost from HSBC’s banking products team, whose activities we described above. Remember, large companies are sophisticated enough to buy such services globally. And smaller customers are not sophisticated enough to want to enter into complex transactions which require specialist advice. These smaller customers can be sold “vanilla” swap and FX products by their branch or area manager and don’t need the expertise of any central, specialist product center which might be located in London. Therefore the market for such a centralised product center is quite small, and you would need significant SME market share to make one viable. As with private banking, only a very large international bank which, like HSBC, combines investment banking expertise with a significant SME business across many countries is capable of even potentially finding itself in the same situation as HSBC. To actually be in the same position as HSBC, that bank would have to have made the deliberate decision to adopt a hub and spoke model in which product expertise was consolidated in one area, such as London. Other banks with a similar sized book of SME customers and investment banking expertise may, contrary to HSBC, opt for ten Swaps specialists in ten countries rather than a desk of ten Swaps specialists serving ten countries from London. In other words, HSBC’s situation is the exception, not the rule. The same is true of HSBC’s “Other” EU consumer loans division, as we saw above.

We know that JP Morgan was named best global private bank by Euromoney in 2015 and has offices in four EU cities. No data is disclosed on its European private banking client assets however. JP Morgan’s jobs add in the Bournemouth Echo describes J P Morgan’s operations there as “a centre for technology and operational processing with global reach.” In 2014 the Bournemouth Echo said “more than 1,000 technology specialists are based” in JP Morgan’s site there. Operations and technology are the last jobs that require a passport. Otherwise there is nothing to indicate that JP Morgan has significant sales of financial products to SMEs out of London. Based on this limited information, it is very difficult to rationalise any job relocation number remotely approaching 4,000. It seems highly likely that this number was floated by JP Morgan to pressurise the UK into making concessions which enabled it to retain the EU financial services passport. Their 700 relocations are likely due to private banking and, to a lesser extent, financial product sales to SME customers. Neither can be extrapolated to other investment banks or private banks, let alone to the City as a whole.

As for the smaller job losses announced by Deutsche, Société Générale and BNP Paribas, all three have a sizeable book of SME loans and an active financial products team in London, so they may certainly have a number of employees selling financial products to EU SMEs from London. Deutsche bank’s sizeable private banking operation is heavily advertised. SocGen’s private banking business is a big beast, having acquired Hambros Bank, and so it is entirely possible that it sells certain financial products to European private banking customers from London. BNP meanwhile describes itself as the largest private bank in the Eurozone, and was bolstered by the acquisition of Fortis’ private banking equities. It too, therefore, fits the profile of the sort of bank which has sufficient scale sell to EU private banking customers from specialist sales desks in London.

Unlike the banks for which job losses are reported in the tables above, most City firms don’t have EU retail exposure. The ~4,000 job relocations detailed in these tables represent less than 1% of total financial services jobs in London.

The identification of private banking as a potential reason for UBS, HSBC, JP Morgan, SocGen and BNP’s job relocation allows us to test the £1bn in EU related private banking revenue estimate implied by OW’s figures. The total job relocation figure for these banks, assuming 1,000 for JP Morgan and disclosed figures for the rest, is 2,250:

job losses HSBC et al actual

If half of these relocations are due to private banking, that would imply the relocation of 1,125 private banking jobs. If the £1bn revenue estimate is correct, this would imply an average revenue for each relocated worker of £890,000. This can be benchmarked against comparable disclosed figures published by quoted private bank Julius Baer from pages 3 and 78 of its 2016 annual report:

Julius Baer commission per manager

Baer’s commission income per average manager is about £50,000 or 6% above the number for the private banking employees relocated due to Brexit, implied by OW’s revenue numbers and the job losses announced by the UK located banks with significant European private banking operations. That the two numbers are quite similar is in itself satisfying. However, the jobs relocated are not relationship management jobs but product sales jobs. These will, typically, be lower revenue generators and less well paid than the typical private banking relationship manager. Therefore, one would have expected the average revenue for the private banking employees relocated due to Brexit to be even lower, relative to a Julius Baer relationship manager, than the figure implied by OW’s numbers. This would suggest that the £1bn EU related private banking revenue estimate implied by OW’s figures is too high. However, it is not as far above reality as we originally expected, given the domestic nature of private banking business. In fact, the figures we have strongly suggest that there has been a more significant growth in product sales to EU High Net Worth (HNW) individuals than I had anticipated.

In conclusion, assuming, as we did above, that 1,000 jobs are lost at JP Morgan, and a few tens are lost at banks like Barclays and Lloyds who claim they don’t expect any redundancies, the total job losses would amount to just 3,325 (including 250 from UBS rather than 1,000). That is within the range forecast by OW under its high access scenario, despite the job loss forecasts from the banks including trading jobs and therefore assuming no access whatsoever. Given the facts established by our examination of the European banking market, OW’s higher, no access forecasts of 31,000 to 71,000 job losses is several orders of magnitude too high and, basically, completely insane.


Tax impact

The other impact is that some of the banking income currently earned in the UK will be recorded in other tax jurisdictions, and the employees relocated will no longer pay income tax in the UK.

As most of the mooted tax jurisdictions have higher tax rates than the UK, the corporation tax will be paid in the other jurisdiction and lost to the UK. To calculate the corporation tax you need to calculate the profit before tax. Before paying tax the banks will be able to offset labor and other costs. They will be able to charge central costs to their EU subsidiary, so the cost/income ratio of the subsidiary is likely to be similar to that of the parent company, if the banks are able to staff it efficiently. Any inefficiency would raise the cost/income ratio and lower profits earned and tax paid by the subsidiary. I use Lloyds cost/income ratio as its activity is the closest to a pure play European bank. RBS isn’t used because it’s loss making. The following figures are from Lloyds 2016 profit and loss. I have not included insurance premiums income or claims, and stripped volatile trading income out of revenue and regulatory provisions (provisions for claims such as PPI cannot be offset against EU subsidiary profits):

Lloyds Cost Income

We next need to estimate the lost revenues. These include:

  1. Lost income on transferred equity capital (calculated above);
  2. Lost deposit margin;
  3. Lost fee income on banking transactions with individuals or SMEs for whom such transactions involve investment advice;
  4. Lost private banking income.

Here is the income calculation:

taxable income

Even thought the Retail and business banking income requiring a passport is £6.5bn, the loan NII component of that figure can continue to be booked in the UK as long as 4% of the value of the loans is held as deposit in an EU subsidiary. Therefore only a small amount of deposit income and the commission income needs to be booked in another tax jurisdiction:

Lost tax

The amount of tax lost is £270m, which represents less than a derisory 0.5% of the UK’s ~£50bn corporation tax receipts. This is clearly an overestimate as many of the banks will already be providing these fee paying services to clients through local EU based teams, rather than from London, as HSBC or BNP seem to.

If each of the 3,325 employees transferred earned £100,000 basic they would each pay income tax of £35,000 including £4,000 tax on a bonus of £10,000, and NICs of £5,500. If 3,075 employees move that would represent a loss of income tax of £135m, or a paltry 1.5% of total £8.9bn of total income tax.

Combined, corporate and income tax loss would amount to £405m.

So far we have focused on the comparison between our revenue and job loss forecasts and OW’s. OW also gave estimates of lost tax which we can now compare with our own.

OW tax

This is OW’s estimate for the whole industry. It doesn’t attach revenue figures to its tax impacts, but the GVA figure can act as a proxy. GVA is revenue minus non-staff costs. The GVA/revenue OW implies by the figures it gives for the UK retail and business banking sector as a whole (p 4) is 59%. Our estimate of £7.5bn of banking revenue would imply GVA of ~£4.4bn, in other words ~four times OW’s high access scenario GVA. Our jobs estimate, as previously stated, is in the middle of the range of OW’s high access scenario. Based on that low ball GVA and revenue scenario, OW calculates a somewhat higher tax loss than we arrive at, though there may be some rounding involved. OW’s low access scenarios for tax of £8-10bn are therefore completely unreasonable due to the outlandish revenue, GVA and job loss estimates they are based on.


What the BIS data tells us about the financial services passport

The BIS A.6 data goes back to 1977 for most of the European countries surveyed (and those it included later were included in the eighties or early nineties). It therefore offers a perfect test for the impact of European unification and its resulting financial services legislation on financial markets and in particular cross-border lending. If the EU financial services passport is such a boon to the City, we should expect to see an increase, whether absolute or relative, in the claims of London located banks on EU counterparties. Let us therefore examine the A.6 data for total UK located claims on EU counterparties over time.

BIS UK claims on EU $m


Here we can see that the EU 1992 and Financial Services plan of 2005 were both followed by an increase in claims by UK located banks on EU counterparties. However, they were also rising before those events, and have declined since their peak in 2008. I have marked China’s entry into the WTO for reasons that will become apparent in the next chart.

The next chart displays total cross-border claims of all banks in the world, of which UK claims on EU counterparties are a small subset.

BIS total crossborder claims $m

You can see that the two charts are remarkably similar. The key date, in both, is China’s WTO entry, which, as discussed in another post, resulted in an explosion in cross border trade much of which was recycled into holdings of US treasuries and other developed market sovereign debt. This created a wall of money which naturally found its way into the banking system. On an operational level, it also increased the amount of cross-border business as banks were able to lend to subsidiaries of their customers in new countries, and because of an increasing demand for trade finance. The second key date is the bursting of the sub-prime bubble in 2008, which also deflated the bubble in cross-border finance.

It therefore seems that the growth in cross-border lending was driven by the global boom and bust in credit, not any EU legislation. To isolate any impact from EU harmonisation on this series, we would have to look at UK claims on EU counterparties as a percentage of total cross-border claims. If EU harmonisation had any impact, then the former should grow faster than the latter, and increase as a percentage of it. Here is the graph, which appeared at the top of this piece:

BIS UK claims on EU % total


The share of UK located banks is on a downward trend which is only interrupted, briefly and to a small extent, in 1997, when no new EU harmonisation was introduced, and which resumed at the end of 2011 with the European sovereign debt crisis. The share of UK located claims on EU entities has never been lower than now.

Of course, this could just be because the rest of the world, in particular the emerging markets, are growing faster. To test this, we can look first of all at UK claims on EU counterparties as a percentage of claims on developed European counterparties, which represent over 90% of total cross-border claims on European counterparties. This will not be impacted by any underperformance in the growth of cross-border claims on EU counterparties relative to total cross-border claims:

UK to EU % developed Europe

UK located banks have a very significant share of claims on developed EU counterparties, but that share has fallen from over 35% in 1977 to under 25% in 2017. Very clearly, the claims of UK located banks on EU counterparties also fell as a percentage of cross-border claims on developed EU counterparties since 1977, with no discernibly positive impact from any EU financial services harmonisation initiatives.

We can also compare the UK with another large developed market lender to the EU, namely Japan. Japan has a sluggish economy, so its volumes cannot receive a boost from internal growth. If cross-border capital to the EU is growing more slowly than cross-border capital overall because the rest of the world, especially emerging markets, is growing faster, then that effect will be removed when comparing the UK’s EU counterparty claims numbers to Japan’s. To save time I have only tracked Japan’s top eight counterparty markets, which amount to over 85% of total claims. First, here are Japan’s absolute numbers (which, remember, will include items such as Sumitomo loans to EU counterparties booked at parent company level in Tokyo rather than by SMBCE in London):

Japan to EU $m

Japan is a smaller source of cross-border funds to the EU, as you might expect, with claims at the end of the series amounting to around a fifth of those of UK located banks. However, the shape of the line is similar to that of the UK’s, showing again that the growth we are witnessing is due to global factors affecting the UK and Japan alike. Now, here is Japan as a percentage of total claims:

Japan to EU % total claims

To add insult to injury, Japan’s cross-border claims on EU counterparties, although a small percentage of total cross-border claims, were on a growing trend as a percentage of total cross-border claims, while the UK’s were declining on the same basis! This could, conceivably, be in part a reflection of Sumitomo and Mizuho’s well publicised expansion in Europe after the financial crisis. Whatever the case, try as one might, one cannot detect any positive impact from EU harmonisation in these numbers, even adjusting for the fact that claims on the EU may have been declining as a percentage of total global cross-border claims.

The last gambit I considered was to look at the time series of total claims located in the UK minus UK headquartered claims to look at the evolution of claims of foreign owned UK located banks on EU counterparties, to see whether that series shows any influence from EU harmonisation. Unfortunately, the consolidated claims series does not go back far enough to investigate this, and, as we saw above, there are important classification differences between the two series which might affect the residual we would have been observing.


In theory, there’s no difference between theory and practice

There is a common thread between the small impact of Brexit and the potential loss of the EU financial services passport on the City of London we estimated above and the lack of impact of EU financial services harmonisation on the claims of UK located banks on EU counterparties that we just observed in the BIS statistics.

In theory, most banks should have used the financial services passport to reorganise their operations like HSBC’s or UBS’, along a hub and spoke model in which big efficient product centers (hubs) distributed specialised products across their client facing branches (spokes). To get an idea of how deeply influenced the Brexit commentariat is by this theory, you only need to look at how often the “hub and spoke” model is alluded to. OW does not use the hub and spoke trope, but its use of “hub” follows the same logic of exporting financial services to Europe from a hub in the UK:

International and wholesale business is that where a large number of the associated activities can be transacted on a cross-border basis. It is in these activities where [sic] the UK is currently an international hub (n 10 p 6; my emphasis).

Restrictions on EU-related activity from their existing UK hubs may lead banks to establish entities within the EU (p 13; my emphasis).

The UK’s ability to maximise these opportunities could be hindered if its status as a hub is undermined.

However, the hub status of the UK could be undermined in the short term if a suitable transition period was not agreed (p 15; my emphasis).

The phrase is a go-to phrase for the Bruegel group, whose thinly argued 35,000 job loss estimate is reviewed in a separate piece:

  • Charles Goodhart and the Bruegel group’s Dirk Schoenmaker, “The United States dominates global investment banking: does it matter for Europe?LSE Financial Markets Group Paper Series, Special Paper 243 (March 2016) devotes pages 10-13 to the “hub and spoke model” trumpeting their claim that “US investment banks apply the hub and spoke model in their European operations. These banks use the international financial centre of London as their main hub (with 80% of the business) with spokes radiating [!!] out to the other larger and mid-sized European countries” (p 11). The evidence for this claim on pps 12 and 13 is simply a map of Europe colored according to the percentage of investment bank turnover and employees. Not a shred of evidence that this is due to any “hub” nor to any “spokes.”
  • Bruegel’s “Making the Best of Brexit for the EU27 Financial System,” Policy Brief Issue 1 (February 2017) uses the metaphor twice, including in its advice that “ESMA should operate in a hub-and-spoke model” (p 2).

In practice, few banks were large enough to make such a model viable; the appetite for complex products may not have been that great with most SMEs, who were happy with simple interest rate swaps or FX hedges, or with most private clients, who were happy to hold some mutual funds, a few bonds and some gold (or perhaps, these days, cryptocurrency); the local product teams the private banks had worked well and they didn’t want to risk any disruption for the sake of a theoretical economy of scale; again, if it ain’t broke don’t fix it.

Also in theory, we should have witnessed the effects of this harmonisation on a macro level, with a rapid adoption by banks of a hub and spoke model in which they opened branches (the spokes) across Europe through which their central lending office in London (the hub) allocated capital on a homogenous, rational, pan-European basis. In practice, most banks operating in Europe, including those doing so from London, had built up the business decades before, often through acquisitions. They therefore had established subsidiaries in situ with solid market shares and a known brand. Despite the theoretical possibility of opening branches in other countries, it made more business sense, in practice, to use the subsidiaries they had already acquired and developed.

Over the past five years or so, we have observed contradictory developments in this field in the UK. ING Direct, the poster-child for the kind of cross border competition in retail banking which the EU financial services passport should have enabled, was sold to Barclays in 2012 “in line with ING’s strategic objective to sharpen the focus of the bank,” in other words, in an attempt to reduce ING’s participation in markets where it had insufficient scale and franchise. ING’s decision was based on market share and economics, not regulation. And it is one example of the fact that, in practice, the kind of cross-border expansion which the financial services passport should have facilitated is often unprofitable.

On the other hand, we have witnessed rapid growth in the UK by two banks: Metro Bank, which is a copy of Metro Bank USA, mainly in personal banking but also in business banking, and Handelsbanken of Sweden. Metro currently has 50 branches having commenced operations in 2010, while Handelsbanken has 207 having commenced operations in 1982, with most of the growth in the last ten years when it more than quadrupled the 50 branches it had in 2007.

Handelsbanken has been able to operate through branches, without setting up a subsidiary, though its ability to do so predates the passport. It is likely that it could only have operated without a subsidiary at a larger scale because of the passport, judging by its chief executive’s comments following its second quarter 2017 results in July 2017. Although claiming he would prefer to operate without a subsidiary, the prospect of having to open one is not regarded with any concern, and Bloomberg quotes Chief Financial Officer Rolf Marquardt as saying that “switching to a subsidiary structure would cut Swedish resolution fee requirements” (thus confirming one of the advantages of subsidiaries cited by Franklin and Gu, 2011). Metro bank, meanwhile, is a new bank operating exclusively in the UK.

The bottom line is that both banks have grown successfully, despite one benefiting from the passport and the other enjoying no such benefit, while ING struggled, despite enjoying the benefit the passport. Three banks, three different EU financial services passport situations, three outcomes. Why did Metro Bank and Handelsbanken succeed? They both offered services for which there was a gap in the market. Metro Bank was able to secure branches on attractive leases, picking up space vacated by high street retailers for a number of reasons (competition from the internet and, to a lesser degree, out of town shopping centers and subdued consumer spending), to use the new internet platform developed and road tested by Metro Bank USA and to design its operations in an efficient way without any legacy complexities. This enabled it to offer low fees and ease of use, an attractive combination.

Handelsbanken, by contrast, charges high fees and is very selective in its choice of customers, but in return it offers a direct relationship with a bank manager who can take quick decisions and provide small business customers with the finance they need when they need it. This sort of service was lacking from the main banks, which had become very centralised in order to cut costs, and were highly bureaucratic due to accumulated legacy issues. The big banks often lacked the imagination to back good business ideas and even when they did needed too many meetings, authorisations and box ticking exercises to do so with any speed. This is the gap in the market which Handelsbanken successfully filled. The bottom line is that it was business model and market environment which dictated growth with or without the passport for Handelsbanken and Metro, and retreat despite the passport for ING.

The example of the Japanese banks reinforces the message delivered by Handelsbanken. In theory not having a financial services passport should put them at a disadvantage. In practice, they find a way of opening a subsidiary or using their relationship with the customer in Japan or some other means. Banks and indeed any business is confronted with all sorts of regulatory obstacles both abroad and in their domestic markets all the time; if they couldn’t cope with that there would be no business; we would probably still be living in caves. If there is a good business case for opening in a market, most companies will find a way to do so. Only a market that is very corrupt or highly restrictive will put them off. By contrast, if you open in a market just because the regulator made it easy for you, that is unlikely to turn out to be a good business decision.


OW’s exclusion of retail is a strange decision

What is most odd with OW’s report is that they only focus on wholesale activities. OW justify this on page 6, note 10 (cited above):

wholesale n 10

While it is true that wholesale activities are those which are most international, as we have argued elsewhere, they are also the activities which are the least susceptible to regulation by the EU. Wholesale financial transactions are carried out by financial companies, such as banks or funds. Or else they wouldn’t be wholesale. There is a clear and obvious contradiction in OW’s inclusion of activities like retail and business banking and private banking in wholesale activities (the clue is in the name). What this shows is that OW has not properly thought through what activities are really impacted by any potential loss of the financial services passport. In reality, it is only retail activities which are impacted, but it is mostly wholesale activities which run across borders. OW has got itself into a muddle by claiming to look at wholesale as the area where there is most cross-border activity, but including retail activities within this “wholesale” envelope as it is only there that you find companies impacted by Brexit.

This basic confusion is quite extraordinary for such a highly reputed “expert” management consultant. That is, except if you regard it as a rhetorical device. By claiming that the impact they have estimated is on the City’s wholesale activities, they make it seem as though such wholesale activities do in fact require the passport. If they did the City would really be at risk from Brexit – but then the international financial  system as we know it would also not exist. They also make it seem as though any damage from Brexit will be visited on the biggest, most lucrative part of the City and the one in which network effects are greatest. They therefore subconsciously amplify the perceived impact of Brexit on the City for their reader. However, wholesale activities are not affected by Brexit. But OW do need to include activities which are impacted by Brexit to make their claim plausible. Therefore, in order to square this circle they use the verbal subterfuge of saying they are analysing the City’s core wholesale activities while in fact analysing its peripheral retail activities, and slip a few retail activities like private and retail and business banking into their definition of wholesale.

The fact that financial companies are authorised to transact with other financial companies all over the world helps the financial sector of any country or region distribute its products effectively and diversify its sources of funding. That’s why you have banks in South Africa borrowing wholesale from banks in Japan. That’s why it would be futile and insane for the EU to try to control wholesale financial markets. In fact, the ability of wholesale capital markets participants to operate across borders is not only accepted but actively encouraged by the EU:

capital controls.PNG

One of the reasons cited by the EU for its opposition to capital controls is that:

cheap capital

The EU not only doesn’t prevent its financial companies from transacting internationally, it wants them to so they can obtain the cheapest capital possible, which of course enables them to fund investment in the EU. In fact, free movement of capital beyond EU borders is as central to the EU’s vision as free movement of labor within them.

To restrict the UK’s ability to enter into wholesale financial transactions with EU counterparties, the EU would have to abandon Article 63 of its constitution and the commitment to the free movement of capital which underlies it. In fact, it would have to introduce capital controls which prevented non-EU investors from acquiring EU securities or EU pension funds and banks from investing outside the EU, and prevented EU banks funding themselves from non-EU banks or lending to non-EU banks. This would be a monumental catastrophe for the EU. They would no longer be able to fund their government borrowing. Their currencies would collapse. Investment into the block would dry up. Pension funds wouldn’t be able to invest to make provision for an ageing population. Compared to this, the minor impacts of Brexit on the financial system are as nothing.

Our analysis of cross-border lending from the UK to the EU shows the power of such free movement of wholesale capital and how powerless the EU is to stop it. It also shows how tiny and peripheral, in comparison, are the retail activities over which the EU does have some control. The real revelation though is that a prestigious and very highly cited think tank piece on Brexit surrepticiously tried to pass one off as the other in order to make Brexit seem like a menace for the City of London. What does this say about OW’s integrity? Or that of the media who cited the headline impact estimates without even the most cursory of examinations? What does it say about how desperate Brexit’s establishment opponents are? And what does it tell us about how stupid OW and other establishment Remainers think people are? Behind OW’s rhetorical ploy lies a fundamental contradiction between the EU’s commitment to and vital dependence on the free movement of capital – and its misguided pretence that it can punish the UK financial services industry for Brexit.