This is the first post in a series of four examining the Federal Reserve’s foreign repo pool. After an introduction, it attempts to figure out who is responsible for the rise in the pool since 2014.
Subsequent posts in this series:
The growth of foreign exchange reserves over the past decades gave rise to a new investor class: a class predominantly focused on safety, excessively concerned about the liquidity terms of their investments and a class trying to avoid private counterparty risk at almost all costs.
In a sense, FX reserves managers epitomize the first half of the old saying about the return of capital vs. the return on capital.
In order to fulfil their mandate, FX reserve managers usually have outsized allocations to G10 fixed income instruments with a heavy tilt towards public sector securities. Occasionally there is some equity exposure, though most of it is, in cases where deployment is certain to be in the distant future, outsourced to SWFs. While G10 government bonds are very liquid, most FX reserve managers also hold some cash (usually bank deposits) or cash substitutes (government bills/MMFs/short-term repos) as a first line of defense, with even less duration & credit risk than longer-term bonds.
This liquidity tranche and in particular a rapidly growing place in the US of almost absolute safety is the topic of this series of posts.
In a previous article, the cash holdings of a particular FX reserve manager, SAMA, were analyzed.
Even though almost all of its cash holdings are US dollar-denominated, the share actually located in US territory is surprisingly small (~20%) as most of SAMA’s cash holdings are deposited in the Eurodollar market either in London or in offshore centers. Underlying the analysis at that time was the assumption that bank deposits are exclusively ‘private sector bank’ deposits. This assumption, although historically more or less accurate, is not anymore generally descriptive of how FX reserve managers allocate their cash, especially in the US.
[Figure 1], created from data contained in the monthly TIC release, illustrates the level and breakdown of how Foreign Official Institutions (FOIs) allocate their short-term claims on the United States.
The usual TIC disclaimers (custodial bias etc.) apply, still some general trends are observable:
- Short-term Treasuries dominated historically & especially during the GFC when FOIs bought $400 bn.
- Other short-term negotiable securities rose from after the dot-com bubble until 2008 and then fell rapidly. Solely based on this behavior, this category seems largely composed of commercial paper.
- Other liabilities grew steadily until 2008, stagnated from then until 2014, with very rapid growth thereafter. Is responsible for total short-term claims approaching 2008 highs.
The growth of the “Other Liabilities” category is a bit tricky to appreciate in the stacked chart, the unstacked version based on the same data is more effective at highlighting the growth since 2014.
After coming close in 2008, “Other Liabilities” is rapidly approaching Treasury bills and if recent growth persists, will overtake them in the next few months. Given this trajectory the question arises what this seemingly residual category consists of. The answer is not too difficult to find since there exists a very similar time series published in the weekly H.4.1 release: “Reverse Repurchases Agreements with Foreign Official and International Accounts”.
Along with another time series provided in the TIC data, which references all reverse repos by FOIs, “Other Liabilities” can be split into three categories.
Before the crisis, FOIs engaged in reverse repos with private counterparties to the tune of up to $120 bn. While the counterparty in these arrangements is a private entity, it should be noted that the collateral can be both, a security issued by the private sector or the government & agencies.
Starting in September 2008, FOIs began to cut their repos with private counterparties to ~$30 bn and increased their transactions with the Fed. Overall balances still declined, probably funding some of the $400 bn short-term Treasury buying.
The recent rise in “Other Liabilities” is thus solely due to the increase in reverse repos with the Federal Reserve, as repo lending to private counterparties never recovered from its post-GFC decline, still hovering around $30 bn.
Given the increased importance of FOI reverse repos with the Fed and the generally high level of transparency of the Federal Reserve System, one would expect a lengthy discussion of this facility somewhere, along with regular releases on its use over time. Reality is much bleaker tough, with only the aggregate level (blue in [Figure 3]) being released regularly in an easily accessible manner.
The NY Fed’s website has, among other things, this to say on the program, which closely mirrors the usual response on requests for comment on the program:
[…] This investment service has been a standard provision of the New York Fed to foreign public sector account holders for many years and is separate from monetary policy operations, including the overnight and term reverse repo operations. […]
Nothing to worry about then? Maybe, but the lack of causal explanations for the $150 bn rise and discussions about what this means for monetary policy, the users and financial markets more broadly, asks to be analyzed a bit.
The following sections gather what can be extracted from various sources and paint a picture of a facility which is anything but pedestrian with potentially interesting & considerable implications. Four broad questions guide the process:
- Which countries use the Fed’s foreign repo facility and more specifically who is responsible for the rise since 2014? [Today]
I. The Users
The Federal Reserve unfortunately only releases an aggregate time series on Fed repo transactions with FOIs, as the release of single country data could provide insight into the reserve allocation of these countries and similarly to IMF reserve releases, too much insight into particular holdings is deliberately avoided.
There is a country breakdown in the TIC banking data but it isn’t helpful at first for two reasons:
- as always, data on single countries in the TIC data system, are reported only in aggregate (private + official) form and not separately, which is particularly problematic in this case, as the private sectors’ “Other liabilities” is far greater (about 5-to-1)
- during the 2014-2016 timeframe, i.e. when FOI repos with the Fed increased, the foreign private sectors’ “Other Liabilities” declined by about the same amount as the official category rose, impairing the drawing of conclusions from the net change per country
These limitations combined complicate figuring out which countries transact with the Fed. TIC banking data by country is therefore more a confirmatory signal than an initial one.
A possible way around this, similar to what was done in the Saudi Arabia post, makes use of the comparatively easy traceability of deposits from both, the asset side of the owner and the liability side of the borrowing bank.
For now, and this obviously simplifies matters a bit but isn’t too far off as will be discussed later, it’s best to think of FOIs engaging in reverse repos with the Fed as them placing cash in a Deposit Facility (DF) provided by the Fed exclusively to them with the expedient feature of not carrying any counterparty risk.
In a stylized world with ‘n’ countries, i.e. ‘n’ Central Banks (CBs), where every CB provides a DF to FOIs in all other countries, the following tautological equations holds:
In words, the sum of all DFs is by definition equal to the sum of cash deposited there by FOIs.
As a simple example, let’s assume the ECB intervenes in the foreign exchange market and places the received US dollars with the Fed. Somewhere in the ECB’s balance sheet there now exists an entry, roughly titled “Cash held in other Central Banks”. The Fed, the recipient of the deposit, now has a new deposit liability in their official DF. The asset view from the ECB and the liability view from the Fed consequently are equivalent in size.
The equation in its current form is a required first step but doesn’t provide insight into who places deposits at the Fed. Some simplifying assumptions are necessary to get there eventually.
The largest part of FX reserves is denominated in four currencies: USD, EUR, GBP, JPY; so that the liability view of the equation can be rewritten as the summation of DFs of the CBs issuing the named currencies.
With regard to the right hand side, fortunately it isn’t necessary to collect the amounts held by every single FOI, a rather impossible task since many FOIs don’t report their cash levels, but rely on the by far largest subset of the FOI universe: foreign CBs with FX reserves (also see the box above). It isn’t necessary to locate the data on all local CB websites as the IMF provides quite some details in addition to the headline reserve numbers. A,(b),(i) in the SDDS data template titled “total currency and deposits with: other national central banks, BIS and IMF” is precisely what’s necessary to proceed.
The countries provided by the IMF are narrowed down to 47 by excluding any country which over the last 12 months doesn’t at any point hold at least $1 bn of deposits with other CBs.
The simplified formula greatly reduces the amount of work required but still doesn’t explain who deposits cash at the Fed. Most of the time, with movements in all of the accounts of the four CBs, the nationality of the FOIs varying their allocation would be close to untraceable. The period during which the Fed’s repos with FOIs increased, from late 2014 to early 2016, is special in a highly convenient way: The ECB’s, BoE’s and the BoJ’s DFs seem to have all moved relatively little or if they did move, the relevant FOIs can be easily traced and excluded. The Fed therefore seems to be the only CB with changes in its DF.
The equation, from October 2014 onwards, then looks as follows:
It’s difficult to prove the equation directly since while the Fed and the ECB release explicit data on their foreign RRP program in the former case and the Deposit Facility in the latter, the BoE and BoJ are more opaque.
For both CBs, regular reserves held by banks, resultant from the respective QE programs, account for the largest part of their liabilities.
The BoJ releases some other liability categories, none however called foreign deposit facility or similarly, but none growing by much during the reference period.
The BoE, in transparency terms, is even worse as the high frequency publication of “other assets/liabilities” was discontinued in 2014, i.e. no direct statement about their DF can be made. Indirect effects of possible increases/decreases should be somewhat traceable though: During the 2014-2016 period, no QE program was conducted, so reserves should be roughly unchanged. If a foreign CB held bank deposits in the private banking system and decided to switch this portion over to the BoE, reserve levels should have decreased. Similarly, if a FOI sold/let mature and not roll over UK government bonds (or other UK assets), some other entity must have bought the bond and reserves would have decreased again, assuming a negative to neutral fiscal balance which doesn’t withdraw reserves via tax payments exceeding government outlays. Reserves didn’t decline so most likely there wasn’t much movement in the BoE’s DF which is confirmed by the BoE’s annual reports which break down the balance sheet by geographical origins.
The ultimate test for the approximate equation and the considerations above is the extent to which the Fed’s foreign repo pool is correlated with the sum of “Cash held in other Central Banks” (ChioCBs) over the reference period.
The black line represents the ChioCBs by 47 CBs taken from the IMF’s reserve template. It’s important to note that the period-to-period change of this line isn’t only affected by the rise and fall of cash positions of the existing members, but also by the addition of new countries reporting in SDDS format. The big double spikes during 2007 for instance are due to the addition of India and the United States. [Figure 5] contains some more information about the sample members.
The result is quite decent: from October 2014 – April 2016, ChioCBs rose ~ $130 bn, very close to the actual $140 bn increase in the Fed’s repo transactions with FOIs.
While the start-to-end change confirms the assumptions already, the trajectory is a bit off at times, requiring some refinement:
- The US cash held in other DFs, although there are almost no changes, is excluded since the objective is to model the cash deposited with the Fed by foreign FOIs and not the Fed’s foreign currency holdings deposited at other CBs.
- Switzerland was responsible for the spike during early 2015 and is excluded for two reasons: 1. The spike reversed quickly and 2. The rise in deposits didn’t occur in the US, but were mostly deposited in the ECB’s DF, as [Figure 6] shows.
- The number of reporting countries drops of quickly after Feb 2016, as smaller CBs aren’t as fast reporting their results to the IMF. The February numbers contain 41 countries of a maximum of 47 at the top. The six countries without numbers provided through at least
February are excluded to avoid a
drop-off in the level. The dark blue spots in the table show which CBs are excluded; as said, mostly small countries without movements large enough to affect overall results.
With these adjustments, the result looks very similar to the actual trajectory, confirming the assumptions made about DFs in other areas.
Since the asset view, i.e. the Cash deposited in other CBs, is simply the sum of all cash held by foreign CBs in the sample, it’s now possible to look at the changes by country during the reference period, in order to identify those which increasingly engage in reverse repos with the Fed.
There is one clear standout, Japan with an increase of over $100 bn, and three minor contributors, Brazil, Thailand and Turkey.
Summary: Section I
FOIs increasingly engage in reverse repos with the Fed, which is similar to depositing cash in a default-free account at the Fed. There is no official information besides the aggregate size of the foreign repo pool. With some assumptions and a bit of luck (no large movement in other CBs’ DFs), the liability view (FOI repos on the Fed’s balance sheet) can be replicated from the asset side (ChioCBs) and the depositing countries can be traced out.