Saudi Arabia’s Cash Reserves & Equity selling

A significant amount of news was created last month by the release of disaggregated Treasury holdings for the Middle East and especially for Saudi Arabia. The recently returned ‘Don of Capital Flow Analysis’, Brad Setser, had some good posts on what new information is contained in the release  and equally important, pointing out the considerable amount of  grey area that still exists (here and here).

What follows, without too many repetitions from the links above, are some supplementary notes, mostly on equities and the currency breakdown & location of the sizeable cash holdings of the Saudi Arabian Monetary Agency (SAMA).

Equities & Agencies

The release of the ‘Annual Cross-U.S. Border Portfolio Holdings ‘ by the Treasury last week went almost unnoticed even though it contained the second half of disaggregated data from BBG’s FOIA request. Again, as with the initial Treasury Holdings release around mid-month, there wasn’t too much to be learned from the total holdings, if only because the data is almost a year old by now.

S.A. total holdings.PNG

[click to enlarge]

In addition to the ~$100 bn Treasuries held directly, Saudi Arabia’s direct equity holdings in June 2015 amounted to $52 bn, long-term Corporate and Agency bonds to another $ 20 bn .

The annual positions going back to 1994, with >1 year time intervals in the early years, were also provided. Two interesting points:

1. As Agency spreads, along with everything else less liquid than Treasuries, widened during the crisis, Saudi entities shifted a non-negligible part of their reserves into Agencies, collecting the wider spreads.

S.A. Agency holdings

Even though S.A.’s private sector held ~$45 bn of debt securities globally in 2008, Agencies are typically more frequently used by Central Banks (CBs) in search of Treasury substitutes. Moreover the rapid rise and decline is rather indicative of a top-down decision by one institution than many private sector investors deciding in the turmoil of 2008 to acquire Agencies. Consequently, the spike is likely attributable to SAMA.

2. S.A. has probably been selling US equities even before the peak value in mid-2014. The blue line in the chart below shows the equity holdings taken from the newly released annual dataset. Since Saudi entities most likely hold a broad range of companies from various sectors to minimize idiosyncratic risk, it seems reasonable to divide the period-to-period changes into valuation gains/losses and the actual underlying flows.

S.A. equity holdings.png

The red line shows the expected value of the portfolio by taking the previous value of [a] and multiplying it by the realized performance of the S&P 500. The cyan bars display the implied flow into and out of US equites by S.A. .

From 2002 – 2012, S.A. constantly added to their holdings. Importantly, even during the GFC, there was no selling. Unfortunately, quantifying the amounts held by SAMA and the private sector respectively, is difficult. Although it is conceivable to attribute the total holdings to SAMA, the Saudi Arabian private sector has accumulated a $100 bn  position in global equites. Thus, it seems likely that the reported $52 bn are owned by a mix of private and official accounts. Despite the unknown split, it seems unlikely that SAMA sold much of their direct equity holdings in the US during the last time of low(er) oil prices during 2008.

Taken together with the move into the slightly less liquid Agencies, SAMA’s behavior appears to have been neutral with some hints of liquidity provision as asset prices fell, thereby dampening the decline in risk assets during the GFC. Their role at the moment is rather different.


Saudi Arabia’s high allocation towards currency and deposits is relatively uncommon and not seen in comparable countries. Unfortunately SAMA does not provide a breakdown by currency. The consensus, as pointed out in one of the links above, is that most of these deposits are kept in US dollars as SAR is ultimately pegged to the USD.

While that makes intuitive sense, it is also conceivable for the Euro allocation to be sizeable, as AAA sovereign short term rates are quite negative, hoovering a bit below the deposit facility rate of -0.4%. There may be single cases where banks have passed on negative rates to their customers; broad data from the ECB however shows that while front-book deposit rates are approaching 0, they are not yet negative, incentivizing switching from  Schatz/Bobl & Co to deposits.

Fortunately, banking is a heavily regulated industry where a lot of activity is being tracked and oftentimes released to the public by the regulators and/or CBs. This factor opens up some indirect ways around SAMA’s nondisclosure.

As always, to every asset (cash held by S.A.) there exists a balancing liability, in this case on global banks’ balance sheets. The BIS publishes, in their Locational Banking Statistics (LBS), a currency breakdown for the total liabilities on banks’ balance sheets for 220 countries and a few regional aggregates. The currency breakdown provided by LBS is based on the residence principle, i.e.  it’s not possible to track SAMA’s position exclusively, but only that of all Saudi entities; official (SAMA) + private (NFCs + HHs + banks (ex-SAMA)).

As an initial check whether it’s sensible to proceed, due to potential data issues caused by fallacious nationality assignments, the asset and liability positions, equivalent in a perfect world, are compared. This is done by summing up the cash position SAMA provides in their monthly reserve release and the foreign currency (FC) held by the private sector reported in S.A.’s IIP and charting it against the total liabilities reported by banks around the globe for which the nationality of the holder has been identified as Saudi Arabian.

deposits asset liability comparison

The lines are fairly comparable, both level- & delta-wise; it appears safe to proceed. Stepping back for moment, it is quite an achievement by the data collection departments  of central banks contributing to the LBS, to be able to assign the nationality of the deposit holders in their respective location to this extent so that textbook logic (assets = liabilities) can, to a large degree, be replicated in practice.

As mentioned previously, the BIS doesn’t release a currency breakdown by subcategory (banks; central banks; non-banks) but only for the composite (blue line above). It is therefore necessary to ensure that SAMA is responsible for a major part of the total holdings and not overshadowed by the private sector. If for instance, SAMA accounted for only say 10% of cash holdings, it could be possible for their currency breakdown to differ quite strongly from the aggregate, without affecting the composite breakdown too much.

At the close of Q3 2015, SAMA’s contribution to the total cash held by Saudi Arabian entities in banks abroad was 67%; enough to draw conclusions from the composite currency breakdown.

With these formalities taken care of, finally, the LBS data for Saudi Arabia:

S.A. LBS.png

The US dollar share of total Saudi Arabian deposits in global banks, taking into account the ‘unallocated category’, is 85% [182/(221-8)].  Euros and Pounds account for 6% and 5% respectively.

Conclusion: The consensus is correct, SAMA’s deposits are almost solely kept in US dollars.


Now, that the currency breakdown has been established, the real fun begins: locating where SAMA keeps the deposits. While all US dollars are eventually borrowed back into US territory even when they are not required to fund banks’ assets, in which case they end up at the Fed to earn IOER, there is no reason to believe SAMA keeps all their deposits directly in US banks. In fact, any bank above the mere local level will generally accept foreign currency deposits and will act as middleman, on-lending the funds back to the US.  As will be seen shortly, SAMA is a heavy user of Eurodollar deposits.

The statistics sections of the major CBs offer some insight with regard to the location of SAMA’s cash deposits. Some CBs publish their ‘local’ LBS results, which are later summarized by the BIS; some publish numbers on the external business of banks operating in their territory more generally and some decide to report nothing, at least not on their international websites.

In order to obtain a quantitative result, a few assumptions are necessary. No CB issues numbers by specific institution, so SAMA’s position has to be abstracted from the overall cash holdings of Saudi Arabian entities. As established above, SAMA accounts for two-thirds of the overall Saudi deposit base abroad.

In many cases, S.A. isn’t broken out separately but reported in a Middle East aggregate. As not every CB releases the country breakdown of this category, the countries in the ‘Asian Oil Exporter’ category used by the US Treasury will serve as guideline. Some CBs also add Africa to the Middle East category. To extract the amount of deposits belonging to Saudi Arabian entities from regional composites, the LBS reported Saudi numbers are compared to the regional LBS aggregates (Middle East or Middle East + Africa), resulting in shares of 45% and 34% respectively.

SAMA deposit location

For instance, the BoE provides a number for S.A. itself, i.e. the (x) column remains blank and (y) is in this case not the result of a calculation, but a directly imported value. 97.2 is then multiplied by (a), to obtain the implied SAMA deposit holdings in the UK of $64.15 bn.

Jersey on the other hand reports only a Middle East aggregate, which is entered into column (x). (y) is calculated by multiplying 28.73 by (b). From then on, the same procedure as in the UK example.

What can be learned from these back-of-the-envelope calculations?

  1. The UK is by far the most important location for Saudi deposits;
  2. Deposits kept in the US are relatively modest;
  3. Offshore/Tax-havens account for ~20% of the identified $136 bn ;
  4. The sum explains two-thirds of SAMA’s ~ $200 bn  cash holdings.

Where are the other $65 bn? Partly in the countries not in the table above, either because they don’t report their contribution to LBS on their website or don’t take part in LBS at all. Luxembourg, Belgium, Ireland, Hong Kong and China are  some of the more promising names. Another possibility, with a similar effect as holding securities in custodian banks in other countries, is the acquisition of MMFs outside the US. Finally, data issues (missing nationality data; depositing cash via firms with a different nationality…) can likely be blamed for some billions as well.


What did S.A. sell & the way ahead

What is (sort of) known:

  • Direct treasury holdings have not declined since reserves peaked (see Brad’s blog for charts)
  • Treasury holdings in custodial centers haven’t declined either
  • The deposit portion of reserves is also flat from mid-2014 – now, but initially declined by $20 bn from mid 2014 – mid 2015
  • S.A. (official+private) sold US equities worth $30 bn  from mid-2014- mid 2015

Points 1-3 are fairly straightforward: SAMA probably didn’t sell much of their safest, most money-like securities.

With regard to equites, it seems reasonable to assume that the selling reported in the Annual TIC data was mostly from the official side for two reasons:

  1. The global equity holdings of the S.A. private sector didn’t decline during the same time frame.
  2. The incentives for the private sector to sell foreign equities are not really in place at present. If the general expectation of a depreciating local currency is correct, it makes no sense to switch back from a supposedly stronger US currency denominated asset to a weaker SAR based one. At some point, it’s possible that repatriation flows will set in, but with a top heavy wealth distribution and no deep recession despite the oil decline, this is rather unlikely for now.

By separating the decline into two phases, from mid-2014 to mid-2015 and from mid-15 to now, and some assumptions about equity holdings outside the US, a still murky but somewhat clearer picture can be painted.

As pointed out at the beginning of this post, it’s difficult to figure out the share of SAMA’s holdings in TIC data. But even without this knowledge, some relative estimates about the allocation to equities outside the US can be made.

The US share in MSCI World is typically somewhere between 50-60%. If SAMA regards this as a guideline, the unknown US allocation times ~[1/0.55] should reflect their total equity holdings.

It’s contestable if SAMA holds a lot of EM equities, as the EM risk vector is somewhat common to both the health of the Saudi Arabian economy and EM equites, and could introduce unwanted wrong-way risk. If that’s the case, the 50-60% share of US equities assumed above, could be higher. On the other hand if SAMA’S regional equity split resembles that of SWFs from the region, it seems likely that 50-60% is an overstatement, as SWFs appear more tilted towards Europe & UK vs the US.

Unable to get a better grip on the real number, let’s call it a wash and stay with the initial midpoint of 55%.

Based on the assumption that all selling from mid-2014 to mid-2015 came from the official side and the selling was conducted without altering the regional percentage allocations, the selling in total amounts to $55 bn [30 bn (US equities) * 1/0.55].

In addition, during the same time cash holdings declined by $20 bn, putting the total implied reserve drawdown at 75bn US. The actual reserve decline during this period at $67 bn isn’t that different, so with the risk of still not knowing the detailed allocation of a substantial part of the reserve portfolio, the guesswork seems fairly good.

The second time period, from mid-2015 – now is trickier as the information about equity holdings, used in the prior period, is not available yet. Nevertheless it appears as if SAMA further derisked the portfolio as points [1-3] from above still hold. Flat direct positions in Treasuries + in custodial centers + an increase in the cash position back to mid-2014 levels despite an overall decline in reserves points to further selling of the riskier assets still left in the portfolio.

With this much selling likely to have already occurred, further equity selling by SAMA is a risk that shouldn’t plague investors too much going forward simply due to the fact that there is likely not much left to be sold and it’s kind of unlikely for the net position to go negative.

SWFs in neighboring countries, with typically much higher equity allocations, appear more likely of hitting the bid should oil prices turn down again. At this time, should further pressure arise, SAMA could meet these for some time without selling assets by simply running down the – until now – untouched cash holdings. If instead, SAMA begins to sell fixed income securities despite the large cash position, this would be rather indicative of how they view the direction of interest rates going forward than them being forced by the market to liquidate assets.

Saudi Arabia’s Cash Reserves & Equity selling

China’s ‘hidden’ current account deficit; invoicing decisions and liquidity effects on EMs


  • China’s current account is responsible for a substantial portion of the reserve decline
  • caused by unfavourable changes in export & import conversion ratios and invoicing decisions
  • current account contribution to reserves even more volatile than the capital account
  • the acceptance of CNY/CNH as an invoicing currency in Asia is at risk as trade partners require US $ payments
  • further pressure on EM vendor financing mechanism


As another month has come to a close, it’s almost time again to recycle the competitive devaluation and capital flight stories published over the last six months. Although the declining reserve position last year was reported widely, the mechanism of the various flows and their consequences have often been left unexplored.

The “State Administration of Foreign Exchange” (SAFE) provides a relatively, at least by Chinese standards, extensive dataset which presents the FX buying and selling by Chinese Banks on the one hand  and all other non-financial actors on the other hand. Banks usually only operate as pass-through institutions and rarely alter their FX position by much; they occasionally absorb flows as during August last year when they (willingly?) sold dollars to non-financial actors to accommodate outflows.

[Chart 1]

[Chart 1] shows the cumulative dollar demand and supply by Chinese non-financial actors in 2015. All values are in bn CNY equivalents even though the reference unit is foreign currency i.e. mostly US dollars. The single categories aren’t netted yet so that gross quantities are observable.

As an example, exports of goods contributed 7354bn CNY or ~1160 bn US $ to China’s reserve position. In other words, Chinese corporations sold goods to other countries and received an amount X as payment in foreign currency (FC)1 and decided to exchange all/something/nothing, say Y, of their FC received for local currency (LC) at the PBoC. During 2015, Y equaled said 7354bn CNY. A similar amount of FC, 7612bn CNY, was demanded by Chinese corporations to pay for imports.

The “Net” category is the resultant flow effect on the PBoC’s foreign exchange position. The drawdown of 3248 bn CNY during 2015 divided by the average USD/CNY rate (~6.25), yields a reserve drawdown close to the frequently cited number of 500 bn $.

The vast bars in the export & import of goods section demonstrate the importance of gross trade flows. They outshine capital account items with ease. The netted  numbers, see [Chart 2], by category for 2015, therefore somewhat understate the importance of current account transactions on foreign exchange reserves as small percentage changes in FC demand or supply in these can easily overwhelm other categories.

[Chart 2]

By only looking at the above data, the solution seems simple: cut the services deficit and clamp down on the residual outflow category2. The problem with this is that even though the two categories worsened in 2015 vs 2014, both were already quite negative before.


The principal difference in 2015 was the sharp decline in FC brought in via the net exports of goods category while during the same period the current account surplus, a bit counterintuitively,  widened.

Export & Import conversion

The current account is primarily intended to measure the difference between goods & services purchased and sold. It does however not offer insight into the invoicing currency. Thus, the amount of goods/services bought and sold alone is not sufficient to describe the effect on reserves. The missing piece is the share of exports receipts that is translated into LC and the share of imports that can be paid in LC; i.e export & import conversion ratios.

[Chart 3]

In the chart above the blue line shows what percentage of FC export proceeds are swapped into LC. During Q4/2015, 0.52 $ were exchanged for LC for every dollar of exports. Similarly (in red), during the same time frame 75% of imports had to be/were paid in FC.

The significance of the up move in the import conversion ratio and the down move in the export conversion ratio cannot be overstated. Despite the increased current account surplus during 2015, reserves were used up to enable China’s trade operations. The swings in the conversion ratios are the primary driver of reserve accumulation. Despite a relatively steady current account surplus post GFC, the below data displays the volatile contribution current account transactions had on the reserve position. Even during 2015 when the attention was squarely focused on capital account transactions, the current account was at least as much a culprit as the capital account.

[Chart 4]

Quite rationally, as [Chart 4] shows,  the direction of conversion ratios varies with phases of CNY appreciation and depreciation expectations, very similar to “pure” capital account flows. Given the swings in the current account contribution to reserves, it may be appropriate to mentally classify them somewhere in the middle of the spectrum with *regular* current account transactions (solely conducted in the seller’s currency) on one end and regular capital account transactions on the other. The effect is equivalent to the latter even though it is rather the passive allocation of profit attained in the global real economy vs mostly financial profit expectations which drive the capital account.

In appreciation mode, Chinese actors want to swap FC into LC to benefit from (expected) future FX gains, while foreign countries which export to China are willing to accept CNY as payment. Even if they don’t believe in the appreciation themselves, the fervent appreciation belief of others provides ample liquidity to exit later and thus accept payment in CNY in the present. At the moment, given depreciation pressures, the above processes occur in reverse order. Exporters keep more of their FC receipts and other countries demand payment in FC as the acceptance of CNY payment in the contemporary skewed environment  would be akin to selling a put option without demanding an adequate premium.

Since these conversion processes aren’t affected by how the proceeds are eventually used, they shouldn’t be expected to stop naturally with the passage of time. This is relevant insofar as the supposed assumption that the export conversion ratio will turn up again when proceeds were used to pay down FC debt is hardly conclusive. Of course, changes in sentiment due to debt repayment could push upward pressure on the export conversion ratio but there is no scientific law that will become operative once debt has been paid down.

While the accumulation of FC has been flagged on some occasions, the change in the import conversion ratio, not only due to a larger percentage move, seems more important in an international liquidity context. The decision by exporters to retain FC receipts is entirely up to them. On the contrary, the decision of the invoicing currency whenever Chinese corporations import goods lies mostly with foreign countries, as they take on the FX risk and might entail losses if further CNY weakness materializes.

The move higher in the import conversion ratio can consequently be interpreted as other countries getting more cautious accepting CNY.


Why is this important? As is well known, many EMs borrowed in dollars due to interest rate differentials & expectations of local currency appreciation and now face pressure due to a stronger dollar and rising US interest rates. The problem, well covered by the BIS and the FTAV crew, is that this credit was oftentimes funded by reserve accumulating EMs (vendor financing) with very favorable terms, was frequently geopolitically motivated and thus came with low/too low? risk premiums attached. Due to the declining reserves in many EMs, the Eurodollar market for the emerging world started to experience pressures as the dollars either go home (via imports)  or are transferred to private actors to delever or diversify their assets in a world without EM currency upside pressures.

China is at the center of the vendor financing mechanism as its companies got access to cheap dollar credit and it used its reserves and the-always-appreciating-CNY to provide funding to other, more risky, places (Latin America & Silk Road).

All this wouldn’t be too concerning if the RMB already was a seasoned currency with solid official and private backing. The SDR inclusion may be the first step for further official flows but private institutions don’t seem too eager to push money into China at the moment.

If the elevated level in the import conversion ratio persists, the prospect of the RMB as a serious alternative invoicing currency to the dollar in Asia declines. In this case, EM Asia first lost easy Eurodollar liquidity access and subsequently decided not to revert to the second best alternative, CNY, worsening liquidity conditions further.

Should the notion of a vendor financing mechanism be accurate, the insistence by other countries to receive $ instead of CNY accelerates the drawdown of reserves and amplifies the pressure on the system in its current form.


The above paragraphs may paint an overly pessimistic picture and are more of a worst case scenario than a forecast. The situation is unendingly reflexive and riddled with path dependencies, complicating the formulation of timely directional market calls. As shown, a big part of the pressure on the Chinese currency was exerted by corporates and outside countries’ varying future expectation of the CNY path. These corporate flows could in theory be reversed or neutralized more easily, when depreciation expectations were dampened, compared to one-way-flows like embezzled funds fleeing due to the anti-corruption campaign. In case the conversion ratios normalized and the PBoC & Co showed improved communication skills, short-sellers may decide to cover and a medium term equilibrium could be reached.

For the moment though markets seem locked in depreciation mode and the question arises what the next moves the  Chinese authorities could take, if further depreciation pressure mount, would be. Some more or less realistic, fragmentary thoughts:

a) Increase the export conversion ratio back to its 2010-2013 average of 70%. Combine the Chinese leadership’s far-reaching power with some nationalist “profits-must-be-repatriated-to-Middle-Kingdom” and increasing the export conversion ratio by 20% would result in 533bn $ extra inflows assuming unchanged export numbers. Regulations of this type wouldn’t be new for China. Before the current framework was introduced, the system moved from a full FC surrender regime at first to a quota based system later before full discretion was granted.

b) Lower the import conversion ratio. Rather more difficult than a). Rhetoric rather useless. Could use quasi-status as monopsonist in some goods to force partial acceptance of CNY. No long-term solution but temporary method to break positive feedback loop.

c) Travel is the current account item showing the biggest change over the last few years following the CNY-REER appreciation. Unfortunately, the conversion ratio of this category is likely close to one as hotels in Tokyo or shops in Seoul aren’t accidentally waiting to accumulate CNH.  Levying charges on or even restricting outbound tourism wouldn’t boost national sentiment and are thus likely lower on the to-do list. More realistically, elevated scrutiny of larger/trophy purchases abroad could dampen the zeal of wealthy citizens travelling abroad.

d) Suspend ODI to neutralize the fall in inbound FDI.


1. This simplification isn’t entirely correct as some payments to China by other countries for imports  is now settled in CNY/CNH, enabled by the growth from 2010-2014 of offshore CNH trading hubs. As CNH trade settlement is rather small and relatively static over shorter periods, the disregard shouldn’t weaken the main arguments discussed below.  »


2. Calculated by subtracting FDI and financial asset flows from the overall Capital Account.   »
China’s ‘hidden’ current account deficit; invoicing decisions and liquidity effects on EMs

Foreign currency credit in China, Pure offshore transactions and attempts to reconcile BIS and PBOC data

While the author believes to have a decent understanding of G7 markets, he is not an EM specialist. Familiarity with the datasets used is thus tenuous at times and results, particularly sections towards the end, should for the moment rather be seen as hypotheses than affirmative statements. Criticism & feedback is welcome.

(Rough) Outline:

  • Compilation of foreign currency assets and liabilities by type and source
  • Pure offshore transactions
  • World ex-China exposure to Chinese non-financial corporations
  • Developments after the August mini-devaluation


  • BIS: Local banking statistics (LBS); Consolidated banking statistics (CBS); Debt securities statistics
  • PBOC: Sources & Uses of Credit Funds of Financial Institutions (in FC)
  • MAS: External assets and liabilities
  • HKMA: External claims and liabilities


Foreign currency balance sheet of Chinese non-financial companies

Note: The following line of thought deems Chinese (CN) banks in aggregate unable to ‘carry trade’ due to regulatory constraints and more comprehensive oversight compared to non-financial corporations (NFCs). The net position of single banks surely fluctuates intraday to accomdate customer flows; quarter end on-balance & off-balance exposure is remarkably stable though, oscillating between -1% and +1% of total assets.

[Chart 1] is regularly presented in discussions about the size of foreign currency (FC) lending to CN companies.

[Chart 1]

While trends are accurately represented by this time series, quite a few adjustments would have to be made, as it includes superfluous categories and omits necessary ones, to attain an accurate picture of total FC lending to CN NFCs. The non-participation of CN in BIS statistic panels (more on this later) doesn’t facilitate the analysis, consequently it appears best to construct the FC balance sheet of NFCs in CN line by line.

As the BIS notes, BIS data alone is not inevitably sufficient and local data, from the People’s Bank of China (PBOC), is necessary to complete the picture.


 In order to calculate the gross/net FC position of CN NFCs, the following FC balance sheet items will be examined and additional information is provided, if available.

3Subsequently the obtained numbers will be added to the above table.

I. Non-CN banks & CN NFCs

Data source: quarterly, BIS CBS/LBS

[Chart 2]
[Chart 3]

The data is sourced from the BIS Consolidated banking statistics (CBS) and Locational banking statistics (LBS) panels with 31 and 44 respective participating countries. The construction methodology behind both is quite lengthy (see here), with certain characteristics often only contained in one of them. Small differences aside, the overlap is deemed considerable so that rough conclusions can be drawn from both.

Currency breakdown:

A breakdown by currency is not available for claims on CN NFCs, but there is one for total claims of LBS reporting banks, i.e. on CN NFCs + CN banks, which subsequently on-lend to maintain a net positions close to zero.

[Chart 4]
Nationality of banks lending to CN:

LBS don’t offer insight into the nationality of the lending banks, CBS do however. CBS put the total claims of reporting countries on China at 1.249 bn $ (total shaded area [Chart 5]). 263bn $ (blue area; 1.249 – 986) should be subtracted, since they represent claims denominated local currency, without currency risk for CN NFCs. This leaves 986 bn $ for which the nationality of the lending countries is relevant.

[Chart 5]
As can be seen in the chart on the left [Chart 5], the bank nationality breakdown is based on foreign claims (it includes loans from abroad denominated in CNY), thus results are once more only indicative.

[Chart 6]

Interestingly, the share of clearly defined countries is relatively small with the ‘Big Four’ accounting for less than 40%. The ‘Other banks’ category is a bit more opaque. Adding Australia, Chinese Taipei, Korea and Switzerland, the 50% mark is breached, leaving 13% of the total or 50% of the ‘Other banks’ category unaccounted for. These have to be, by definition, in CBS reporting countries and it seems likely that Singapore and Hong Kong are responsible. Both financial centers seem to be included in CBS; however their data is not released by the BIS for unobvious reasons.

The final category, ‘Outside area Banks’, is even more opaque. The BIS explains this category in their Q2 2015 statistical release (p.3 here) as follows:

“Includes branches or subsidiaries of Chinese banks located in BIS reporting countries, as well as banks whose activities are not consolidated by a controlling parent institution in another reporting country (eg the banking subsidiary of an insurance group).”

FC loans to China extended by offshore branches of CN headquartered banks are obviously denominated in FC but it should be noted that the foreign exchange and credit risk is on CN banks’ balance sheet. In case severe losses are incurred, the allocation thereof depends on the funding structure of the CN offshore subsidiary and possible guarantees by the main institutions.

These loans contradict the casual notion that FC credit automatically increases risk exposure by World ex-CN institutions.

Maturity structure

[Chart 7]

Again, information is only published for the aggregate (banks + NFCs). Nevertheless, it seems fair to conclude most of the credit, due to the short maturity, is used for carry trades.

II. CN banks & CN NFCs

Data source: monthly, PBOC, Sources & Uses of Funds of Financial Institutions (in Foreign Currency)

[Chart 8]
[Chart 9]

As seen in a previous section, BIS reporting banks lend not only to CN NFCs but also to CN banks. Since CN banks maintain balanced FC books, they  on lend the obtained funds to NFCs in CN. This activity is not included in BIS statistics since CN does not participate in their panels. PBOC data captures this activity.

The next sections try to establish the factors that affect [Chart 8/9], reasons they correlate and scenarios that lead to divergence. The balance sheets in the illustrations are not those of single institutions but of consolidated sectors.

[Example 1] International lending – outflow/inflow (terminology adopted from p.2 here)

[Example 1]

A bank outside CN loans FC to a bank in CN (blue). The CN bank records the loan as a liability and at the same time adds a deposit to its assets. It then lends its deposit to a CN NFC (red) and records this loan as an asset (orange). The NFC adds the loan as liability and the deposit as an asset. The deposit is again recorded on the banks’ balance sheet as liability (green) and asset (purple). To complete the cycle CN banks lend the FC deposit back to World ex-CN banks.

The red and green transactions are reported to the PBOC. The blue and black transactions are reported to the BIS by World ex-CN banks.

In this example FC assets and FC liabilities increase/decrease in lockstep, depending on the volume of new credit creation and repayment of existing loans. Also note that blue =red = green = black in this particular example. There wouldn’t be any need to analyze PBOC statistics if the world was as simple as in [Experiment 1]. The FC loans by CN banks are equivalent to the BIS derived lending of World ex-CN to CN banks.

[Example 2]: Pure offshore transaction (POT)

[Example 2]

All transactions in Example [1] are considered to have already occurred except the last one, the back channeling of deposits overseas (black). The CN bank balance sheet is funded via a loan from abroad and a deposit and holds a claim against CN NFCs and a FC deposit, that is not yet lend back offshore. Assuming no regulatory constraints, CN banks can again extend credit (as in [Experiment 1]) to NFCs (red and green). Each time a new loan and a new deposit is added to CN banks’ balance sheet, NFCs add a loan to their liabilities and a deposit to their assets.

The transactions, as recroded in BIS stats, considered in Experiment [1] are represented by the blue circle.

Each round of pure offshore credit creation by CN banks, only one in [Experiment 1], can be thought of as adding another red circle. These activities are however not reflected in BIS statistics as the multiplication process takes place within CN borders. The final back channeling of funds by CN banks to World ex-CN banks is unaffected.

Attempts will later be made to quantify the size of pure offshore transactions.

Example [3]: Alternative sources of funds

The examples above considered a loan by a bank outside CN to CN banks as first step. Alternatively the initial funds could be  retained export proceeds or FC loans directly obtained by CN NFCs. CN banks can then either follow the steps considered in [Experiment 2] or back channel the funds directly. In the extreme case, FC credit could be extended within CN (via export proceeds) with BIS statistics showing no FC credit whatsoever.

Example [4] PBOC intervention and carry trades

[Examples 1-3] all made the implicit assumption that CN NFCs in aggregate are willing to hold FC deposits. To profit from interest rate differentials and potential local currency appreciation, CN companies have to exchange their FC deposits into local currency. Since CN still exhibits a positive current account balance and few RMB hubs, with previously accumaulated RMB deposits outside CN exist, the only real exchange mechanism is via the PBOC (via the banking system which intermediates the exchange).

Regulators’ attempts to complicate the direct exchange of FC into LC are rarely effective as over-/under- invoicing structures combined with multiple transactions with subsidiaries abroad enable NFCs to effectively exchange FC to LC.

Introducing the PBOC to the graphical representation finally enables CN NFCs to put on carry trades.

[Example 4]

The blue, red and orange arrows are taken from [Experiment 1]. This time, the deposit on CN banks’ balance sheet must be crossed out a as NFCs swap their FC deposit for LC at the PBOC (green and black).

CN banks still run a balanced FC position (liability: loan from World ex-CN bank; asset: loan to NFCs), BIS statistics however paint them as net debtors. This is due to the back-channeling now occurring indirectly via the PBOC and not via CN Banks. The actual currency mismatch happens on CN NFCs’ balance sheet and can only be examined with PBOC data.

Each time a POT takes place a locally funded FC loan and FC deposit are created. While the net FC indebtedness doesn’t rise, an unfavorable sectoral distribution can still lead to increased risk.

The whole process is similar to the difference between corporate bank vs corporate bond based financing. While loans are created endogenously, bonds usually are issued as an exchange of upfront cash vs a promise to repay cash + interest later. In such a system, a deposit is required first, before it can be lend via bond issuance. As the initial deposit spins through the economy there are theoretically no limits to how often it can be lend via new bond issuance. Similarly, the initial FC deposit can’t be created in CN but has to be imported, either financially or via real exports, before it can be used for lending. How often it is relend in CN is totally up to CN institutions & regulations.

III. Bond issuance by CN NFCs

Source: quarterly, BIS Debt securities statistics

[Chart 10]

90+ % are dollar denominated, long-term and have a fixed interest rate. There is no information on holdings of FC bonds or other FC financial assets by CN NFCs.


All data obtained can now be entered into the table introduced at the beginning.


Above table presented graphically; from 2005 – Q2/2015


[Chart 11]

[Chart 12]

[Chart 13]


Gross FC liabilities shy of 1.2trn$; lending mostly intermediated by CN banks; FC bond issuance increasingly important and rather longterm vs short term bank loans.

FC asset growth, contrary to popular opinion, almost kept up with liabilities.

Net FC position -600bn $; flat since Q1/2014 after steep rise (300bn$) starting Q4/2012 as CN/US interest rates diverged the previous year and the dollar was trading close to local lows.


‘True’ exposure of World ex-CN institutions

Even though the aggregate net position is -607 bn $, in a worst case scenario 1172bn $ could be at risk in case of an unfavorable distribution of FC assets/liabilities across individual companies.

As shown previously it’s hasty to equate gross FC liabilities of CN NFCs with claims World ex-CN holds on CN as (1) a substantial share of the funds required by CN banks to on-lend in CN are extended by subsidiaries abroad and (2) pure offshore transactions in CN do not increase the risk faced by World ex-CN institutions.


The calculation of the size of pure offshore transactions seems fairly straight forward. As shown in the [Experiment 1-4], the various lending activities by CN banks are recorded in BIS (the initial funding) and PBOC (the on-lending) statistics. As long as they mirror each other, no POT has taken place. In cases where PBOC numbers > BIS numbers, POTs have occurred.

Unfortunately, the BIS data shows CN banks only as aggregate, which is unfortunate as direct comparisons with the previously developed CN NFC data may provide an false image since CN households have also acquired FC deposits and CN banks have lend overseas in increasing amounts recently. Therefore total loans and total deposits on CN banks’ balance sheet will be used. The share of POTs in total loans can then be applied to the NFC subset.

In an idealized framework, in which all excess deposits CN banks hold are lend back overseas to collect interest, the implied pure offshore transactions can be estimated via both the difference in loans and the difference in deposits.


[Chart 14]

The results obviously don’t correlate nicely, as they should in theory. They are however, aside from the small negative print in 2007, not negative as that would be impossible according the definition, but you never know what happens when you combine two independently collected sets of data. Furthermore, despite the different trajectories, the endpoints are relatively close, compared to their history.

Which of the two methods is a better representation of POTs is open for debate; I tend towards Method 1, as I’m not familiar accounting wise if all FC deposits are finally back channeled to World ex-CN banks given ZIRP. Criticism & feedback on this point is particularly welcome.

The size of POTs is 308bn $ or 256bn $ according to Method 1/2, which translates to 33% or 28% of total FC loans extended by CN banks.


With the adjustments above, the gross exposure by World ex-CN institutions to CN NFCs is cut by roughly a third. As mentioned, overseas lending by CN banks (Silk Road?) is increasing steadily and World ex-CN banks most likely supply the initial funds. Since these claims are ultimately backed by the creditworthiness of CN banks and not the overseas lending projects, which are small when compared to the total (FC+LC) CN bank balance sheet, the exposure other countries have on CN indirectly could be a bit higher.

After the devaluation

So far all statistics presented contained data through Q2/2015, as that’s the latest BIS data available.

Indications of FC deleveraging since August can be approximated with monthly data provided by the PBOC, Honk Kong Monetary Authority (HKMA) and the Monetary Authority of Singapore (MAS), which provide timely and geographically relevant data.

The PBOC data series is the same as used in [Chart 8/9] and shows monthly data through October.

[Chart 15]

The data taken from the MAS shows Singaporean banks’ exposure to East Asia and is also through to October.

[Chart 16]

The data taken from the HKMA shows HK banks’ exposure to Mainland CN and ends with August.

[Chart 17]

The FC deleveraging has obviously begun. PBOC and MAS data seems benign so far; FC dealings in HK show clear signs of a reversal with FC credit, primarily to CN banks, down sizably and FC deposits spiking.

While there is no public high frequency data available on FC bond issuance of CN NFCs, according to street research and newspapers, issuance has been subdued since the start of the year as bond issuance shifted to the local market.

BIS statistics will be released later this month and will finally provide the complete view.


This has gotten much longer than initially assumed. If you find inconstancies or think the approach is flawed, feel free to comment or contact me directly.

Finally, as this piece has only dealt with aggregates, in case someone has broken down or seen a detailed breakdown by industry, the comment section welcomes you.


twtr: @ stwill1      or      stwill1 at zoho dot com

Foreign currency credit in China, Pure offshore transactions and attempts to reconcile BIS and PBOC data