This is the third post in a sequential four-part series on the Federal Reserve’s foreign repo pool. The previous Parts, I & II (see summaries at the very end of each post), noticed the increasing size of the pool, analyzed the user base and how the pool fits into the users’ reserve allocation process, but left questions concerning the cause of the reallocation unexamined.
This post starts the inquiry into the multifaceted ‘Why’ question by exploring the most important factor: the foreign repo pool rate (FRPR) and subsequently how the rate is set (with a simple model to calculate a high(er) frequency proxy) and why Basel III regulations and US extensions thereof seem to lead to a structurally higher FRPR, attracting inflows into the pool.
III. The Foreign Repo Pool Rate
Transactions in a free-market environment should, according to textbook definitions, lead to wealth and/or utility gains for all parties involved. Applying this principle to FOIs lending to the Fed via its foreign repo pool requires an evaluation of the risks and returns this activity entails. Bluntly stated, rationally behaving FOIs should only lend cash to the Fed if relative risks & returns offered by the pool exceed alternatives elsewhere.
From a risk perspective, the evaluation is simple: there is no safer counterparty than the central bank endowed with the authority to create money by crediting its own account. In addition, since the pool is at the end of the day structured as repos, cash lenders receive high quality collateral from the Fed’s SOMA portfolio, i.e. either Treasuries or Agency MBS. If MBS is involved at all, its role is likely small since pre-GFC, the Fed didn’t own any and usually CBs prefer to limit their liquidity operations to the ‘purest’ safe assets, i.e. Treasuries.
An allocation to the foreign repo pool, in consequence, is at least as safe as owing Treasuries. The only distinction, besides some minor operational features, is the interest rate paid to FOIs.
An initial problem, even before starting the analysis, is posed by the general lack of transparency surrounding the pool which unfortunately also extends to the foreign repo pool rate (FRPR).The regular daily/weekly/monthly releases do not contain any useful information. Thanks to Zoltan Pozsar (who over the past decade or so has done a significant amount of research, shining a light on US money & repo markets)’s sharp eyes, it is now known where to search for the FRPR: in the unaudited Federal Reserve Banks Combined Quarterly Financial Reports, where the so far accumulated interest expense during the year related to the Fed’s reverse repo operations is listed. The report is only issued three times a year which is why Q4 averages are not available.
The data points that can be gathered this way show a slowly rising FRPR even before the eventual liftoff in December 2015 and a roughly 25 bps jump thereafter, similar to the increase of the fed funds rate and other short-term interest rates.
Prior to thinking about how the FRPR is set and why it drifted upwards during 2015, three separate issues should be touched on briefly:
- the structure of US repo markets
- a variety of short-term interest & repo rates
- and finally, how the foreign repo pool is managed from an operational perspective, based on the little information released over the years and some subjective dot connecting
[Figure 26], (click to enlarge), is a stylized scheme of the US repo market and serves as a reference to the sections that follow. There are much better/more comprehensive versions of this chart elsewhere; for the current purpose this simplified one should suffice. Accordingly, the details of lesser importance to this study will not be discussed. From a high-level perspective, the raison d’être of the (Tri-party) repo market is the intermediation between liquid cash pools & households (on the left) and borrowers with pledgeable securities (on the right). Dealers/brokers and especially primary dealers act as intermediaries in this process. A large share of these transactions takes place via the Tri-party repo system, where recurring trades against general collateral (GC) are handled in a highly efficient manner by outsourcing back-office tasks to the clearing banks, Bank of New York Mellon (BNYM) and JPMorgan Chase (JPMC). Due to the involvement of a third party, tracking transactions and the interest rate on them is much easier than in the more obscure bilateral repo market, where counterparties themselves are responsible to ensure proper settlement, usually by structuring trades as ‘Delivery versus Payment (DvP)’ repos. Finally, the importance of the Fed (blue box) has increased since the crisis due to the existence of excess reserves on banks’ balance sheets.
The short-term interest rate space has also changed a bit and is shaped by the effects of new regulations and CBs’ large-scale asset purchases (LSAPs).
The Effective Federal Funds rate, the Fed’s interbank, unsecured target rate before the crisis has declined in importance. Only lenders, mostly FHLBs, effectively excluded from the new interest rate setting environment, lend funds to banks which can place these more profitably with the Fed as excess reserves earning Interest on Excess Reserves (IOER). IOER is intended to incentivize banks to pass interest rates changes, enacted by the FOMC, on to other rates by acting as arbitrageurs with the ability to access the highest risk-free rate, IOER. The private full allotment ON RRP, finalized in December 2015, completes the new corridor system by providing a complementary floor to the IOER ceiling.
The other famous unsecured interbank rate, LIBOR, too came under scrutiny due to its survey nature and the resulting LIBOR scandal. The current rise, the result of pending money market fund reforms, is also not supportive of LIBOR’s status as widely used benchmark rate, which is set to end sometime in the yet undetermined future.
Then, there are a whole host of customer-to-bank funding rates, all relatively unchanged from the investors‘ point of view post-crisis though, behind the scenes, affected by new regulations that seek to limit short-term funding by disincentivizing flighty deposits which, as a result, are shunned by banks. Frequently quoted customer-to-bank rates: CDs, financial CP and Eurodollar deposits.
Treasury bills do obviously also still exist, are however affected most strongly by ‘flight to quality flows’ or influenced by the occasional lack sufficient issuance, limiting their use as neutral benchmark.
And finally, returning to the issue at hand, there are repo markets and repo rates. The creation of repo indexes is tricky since there are at least two main risks to consider: the counterparty and the riskiness of the pledged collateral. The risk assessment is then transformed into an interest rate and a haircut, in theory opening the space for multiple equilibria. In addition there are many non-price variables that can and do affect repo rates: operational control during the contract (Tri-party repo, Delivery repo, Hold-in-Custody repo), GC repo or special, direct counterparty exposure or CCP-intermediated, reusability (or not) of collateral, possibility of collateral substitution if term >1 day.
Despite these difficulties, there are two repo indexes, both settled on the books of the tri-party clearing banks, released daily, worth examining due to their later importance in reconstructing the FRPR.
BNY Mellon Tri-Party Repo Indexes: These rates, available for three GC categories (Treasuries, Agency debt, Agency MBS), reflect the activity in the BNYM cleared section of the US tri-party repo market. As displayed by the black lines in [Figure 26], the lending flow is almost exclusively one-way, from cash pools to large dealers. Since the tri-party clearing banks only settle the bilaterally negotiated trades and do not act as CCPs by novating & netting exposures, the interest rates of single trades, even within the same collateral category and executed at the same time of the day, can differ due to the unique credit risk each counterparty exhibits. The three rates are calculated as dollar weighted averages for new, overnight trades entered for clearing with BNYM. One small drawback of the index referencing repos collateralized by Treasuries: The Fed’s ON private RRP program also utilizes the already existing Tri-party infrastructure. The RRP demand curve is flat at a rate of 25 bps as long as there are unencumbered securities in the SOMA portfolio. Unless BNYM excludes these private-to-Fed transactions, and there have been no statements indicating this as far as I know, the private-to-private repo rate should be a bit higher since the ON RRP, special cases aside, represents the lowest rate on the tri-party platform. The size of the ON RRP has declined for some time now, so that the effects shouldn’t be too large, even during quarter ends.
DTCC GCF Repo Index®: General Collateral Financing (GCF) interdealer repos could be, drawing an imperfect analogy to the unsecured bank funding markets, seen as the federal funds equivalent with the BNYM indexes representing the various unsecured customer-to-bank funding rates. Similar to the fed funds market, the access to GCF repos is limited; to dealers and depositary institutions, as opposed to only depositary institutions in fed funds, and some GSEs. Three indexes are published daily referencing the same collateral classes as the BNYM indexes. While interdealer trades are also settled on the books of the tri-party clearing banks (or only BNYM, after JPMC announced the discontinuation of its GCF brokerage operations), no negotiation between counterparties takes place since the market is blind-brokered and involves a CCP, the FICC, which, once counterparties intermediated by a interdealer-broker anonymously agree to a trade, novates these trades and becomes the counterparty to both dealers. Since the FICC always serves as counterparty to GCF trades, all trades within a collateral category executed at the same time, share a common rate.
With this overview of US repo markets in mind, now on to the pieced-together history of the foreign repo pool:
- The foreign repo pool or foreign RP pool or more formally “Reverse Repurchase Agreements with Foreign Official and International Accounts” already exists for a long time, reaching back into the 20th century.
- At some point in history, CBs around the world, most likely in mutual accordance in an attempt not to tip off private sector banks when intervening in currency markets and to absolve each other of time-consuming counterparty assessments, decided to offer each other deposit facilities (DFs).
- What rate should be paid on this risk-free DF? In order not to politicize the issue, it seems reasonable to tie the rate to some observable market rate and not set it discretionarily.
- The Fed seems to have decided to solve this problem by not paying interest on the DF itself, but to sweep the balances from all FOI accounts into one big pool at the end of the day and arrange ON reverse repos, secured by government collateral, against these.
- There are two ways to arrange these repos: send them to the private repo market as customer-related repos, without affecting reserve levels, and let the market handle the interest rate question. This seems to have been done in the early days but was seemingly phased out, probably since FOIs ultimately transacted not with the Fed but with a private counterparty. Even though the transaction was secured by government collateral and the Fed acted as custodian, it is not equivalent to directly transacting with a nominally default-free entity like the Fed. Alternatively, the Fed can execute repos internally against the securities held in the SOMA portfolio.
- In the pre-GFC interest rate setting environment based on reserve tightness, the Fed usually did some ON repos for reserve fine-tuning purposes with primary dealers (red bracketed arrow), and even more once short-term autonomous factors like the foreign repo pool, turned more volatile.
- As an active repo lender, small enough not to exert pressure on repo rates but large enough to get an idea of the environment, in repo markets before 2008, the Fed didn’t need to rely on outside reported repo rates since it knew firsthand what was payed/received on the repos done for the Fed’s own account. According to statements, the rate on the foreign repo pool, was set some basis points lower than the traded repo rate in order not to incentivize outsized inflows into the pool, which viewed from the banking sector’s angle are large deposit outflows. These would have to be sterilized by the Fed and could lead to problems should large amounts of cash be shifted very late in the day, requiring a large & fast Fed response. For the same reason, limits were placed on how much FOIs could vary the size of their deposits at the Fed.
- From an operational perspective, repos between private cash pools and dealers as well as Fed RRPs with MMFs are mostly settled using the tri-party repo infrastructure. After the tri-party repo infrastructure reform, the unwind, i.e. when cash lenders get back their cash, has moved from the early morning (8:30 am ET) to the afternoon (~3:30 pm ET) to limit the duration intraday credit has to be extended by the clearing banks. As a consequence, lenders only have access to cash late in the afternoon, a rather suboptimal procedure for FOIs that may require funds intraday to intervene in FX markets.
- Repos under the foreign repo pool, in contrast, are arranged bilaterally (FOIs, particularly smaller ones don’t usually have access to tri-party repo and switching to tri-party settlement now is unpractical due to the point above) and are unwound early in the morning. Bilateral repos with FOIs can thus be said to be truly ‘only overnight’ as cash is available in the respective deposit accounts at the Fed during the day, ready to be deployed.
- Up to mid-2008 the system likely worked roughly as portrayed above. Then things started rolling (…) and, at first due to various emergency lending programs and later due to the first round of intentionally unsterilized LSAPs, excess reserves began to increase, rendering ON repos by the Fed unnecessary.
- The foreign repo pool didn’t magically disappear in 2008 but rather began to grow. The question how to set the FRPR resurfaces.
- It’s important to remember that the Fed had to address this question in 2008/2009, not now. Today, due to the increased appreciation of the importance of repo markets, the Fed seems, so far mostly internally, to track bilateral repo transactions more closely via a new dataset based on dealer/broker responses. Back then, there were only two rates somewhat ready to be considered as new reference rate:
- 1. Tri-party GC rates; where MMFs & cash pools lend to primary/large dealers; black arrows in the diagram
- 2. Interdealer GCF rates; where (mostly) dealers/brokers lend to each other; grey arrows
For most of the post-crisis period, there was very little to no information available indicating how the Fed sets the FRPR. This changed in February 2016, when Simon Potter, the Head of the NY Fed’s Markets Group, gave a speech at Columbia University. The entire speech is definitely worth reading and, relevant to this post, also includes a brief section on the foreign repo pool. Potter argues that the Fed still pays a market-based rate, even though the rate setting mechanism was adjusted, in an attempt to dismiss claims the Fed is utilizing the foreign repo pool as policy instrument by paying a discretionarily set high rate of interest.
In this context, the following slide (last page in the accompanying deck), is provided. While it doesn’t explicitly show how the FRPR is set, it offers enough information for some very educated guesses.
During the already concluded quarters (Q1 – Q4 2015), it can be seen that the FRPR is very similar to the tri-party rate in the fourth column. The tri-party rate is a small touch higher, from 0.9 bps during Q2 2015 to 1.4 bps during the last quarter of 2015 and the correlation is almost 1. By implication, the tri-party rate presented in the fourth column seems to be the rate the FRPR is benchmarked to.
As discussed above, the lack of possible repo benchmark indexes left the Fed with only two realistic options: Tri-party GC rates or interdealer GCF rates. According to the underlined notes in [Figure 27], the tri-party rate, and thus also the FRPR, consists of a mixture of the two, both collateralized by Treasury securities.
This information enables the reverse-engineering of the FRPR, by figuring out the relative shares of the two rates and in the end will explain why the FRPR increased even pre-liftoff, moving several basis points above comparable risk-free rates.
Of the three data providers mentioned supplying inputs for the calculation of the tri-party rate in column four, two, BYNM and the DTCC, release daily rates on tri-party GC and GCF repos on their websites. With theses and some simple math, the relative weights can be determined.
Starting with the slide from the presentation, by subtracting column four from column two, the average quarterly ‘Target’ Tri-party rate [for Q1 2015: (0.07-(-0.012)) = 0.082] is calculated. The GCF & GC rates in the next columns in the table below are the average rates of the respective repo indexes during the referenced quarters.
The ‘Target’ Tri-party rate is then attained by solving for the relative share of GCF & GC.
The relative shares are quite stable with an average GCF:GC split of ~30:70. Despite the apparent stability in the composition of the ‘Target’ Tri-party rate, one year alone is a timeframe too short to declare with a fair amount of certainty that the FRPR is truly based on this target rate; more data is required to dispel in-sample concerns.
To do this, the FRPR proxy, using the 29:71 split, will be calculated daily as well as the quarterly averages. This is possible going back to Q4 2012, before which BNYM GC rates were not released publicly. The actual FRPR from 2014-2016 is taken from the aforementioned Federal Reserve Banks Combined Quarterly Financial Reports. Data before 2014, not available in this report, is taken from the NY Fed’s “Domestic Open Market Operations During 2015” report, which includes the FRPR going back to 2011.
Apparently, neither the formula nor the weights shifted noticeably during the portrayed timeframe; the 29:71 modelled tri-party rate correlates well with the actual FRPR. Interestingly, the target rate (blue) is always 1-2 bps above the actual rate, possibly a remnant of the pre-GFC period, when a similar sized buffer disincentivized inflows into the pool by deliberately lowering the FRPR below comparable private sector repo rates.
Correcting for this buffer…
…yields the final equation of the in volume terms second most important Fed-set interest rate, the FRPR, …
.. which can now also be quoted at higher than quarterly frequencies.
Equipped with a simple yet effective and more importantly realistic model of how the FRPR is set, the question why the rate increased more than other risk-free rates since 2014 can be addressed by analyzing the two repo rates entering the calculation of the FRPR.
Up to mid-2014, GCF & GC rates, both secured by Treasuries, were almost indistinguishable. From then on however, interdealer GCF rates slowly began do diverge to the upside. The GCF-GC-spread, below 5 bps and as low as 2 bps before mid-2014 (in this post-crisis sample), moved up to an average value of 6 bps during H2/2014, 12 bps during 2015 and 17 bps this year through August.
There are a number of short-term factors that temporarily, but in a fairly predictable rhythm, can affect GCF & GC rates differently, leading to higher spreads (e.g. Tsy & Class A settlement, GSE P&I payments, financial reporting dates).
The enduring drift higher in the GCF-GC-spread since 2014 is not attributable to such factors and is not a temporary phenomenon but the result of new regulations that slowly but surely produce the (un?)-intended effects.
The global regulatory paradigm just before the crisis, Basel II for most and somewhere between I and II for the rest, relied to a large extent on risk-weighted assets, which in hindsight didn’t exactly work out as planned. The crisis and subsequent quick abandoning of regulations solely based on Basel II ideas, gave rise to the behemoth that is Basel III.
Basel III bestowed on the world many new acronyms and a more holistic & complex approach of regulating financial institutions. Among the new rules, as a complement to more subjective risk-weighted measures, is the leverage ratio (LR), a supposedly simpler across-the-board metric to evaluate appropriate balance sheet size by comparing capital to all on-balance sheet assets (+ some off-balance sheet/contingent exposures).
Due to its calculation, the leverage ratio effectively ignores the varying amounts of risk embedded in different assets, by requiring a constant amount of capital for all assets.
The US implementation of Basel III was released in 2013 and a final expanded version in 2014. The US version, the Supplementary Leverage Ratio (SLR) is based on the 3% LR proposed by the Basel Committee, but adds additional capital buffers for US based G-SIBs. Under the ‘enhanced Supplementary Leverage Ratio’, these have to maintain a >5% SLR at the holding level and >6% SLR at insured depository subsidiaries. The eSLR will only become binding from 2018 but is required to be disclosed starting 2015, pushing banks towards early compliance.
The effect of the eSLR on US G-SIBS, which are/were not by chance also the largest dealers on the street, is substantial. The running of matched repo books, characterized formerly by large gross exposures combined with low spreads, has become much more costly as capital has to be held against repo positions, even when dealers only act as intermediaries in overcollateralized lending agreements, passing cash to borrowers and high quality collateral to lenders. Netting could help, but necessary requirements for netting down exposures are strict and usually not helpful for most transactions where dealers intermediate between pure lenders and pure borrowers (i.e. where net equals gross positions).
To illustrate the effects specifically, consider a primary dealer in [Figure 26]. For simplification purposes, let’s introduce the unrealistic assumption (neglecting bilateral repos) that this dealer only trades on the tri-party platform, borrowing at GC from cash pools and lending at GCF to other smaller dealers that, due to their size, cannot borrow directly from cash pools, both secured by Treasuries. The spread between the rates equals the profit the primary dealer makes by running a matched repo book. This activity is quasi-risk free and could be run with almost no capital due to favorable netting rules and no/ less stringent leverage ratios before the crisis and also post-crisis before Basel III/the eSLR were finalized. With the eSLR introduced and most banks already (over)-complying, G-SIB affiliated dealers have to hold at least 5% capital against matched book repos, rendering single digit GCF-GC-spreads economically unviable.
Large banks then face the decision of either phasing out their repo desks or can attempt to increase the spread their business runs on, in order to push returns closer to ROE hurdle rates, by either
- lowering the borrowing rate
- or increasing the lending rate.
The first option is technically unfeasible; the fixed-rate ON full allotment private sector RRP program guarantees that tri-party GC rates won’t decline below 25 bps for long since lenders will switch from private counterpartys to Fed RRPs.
As a result, the only option for primary dealers to run their matched repo books somewhat profitably is to increase the lending rate, leading to the observed GCF-GC-spread divergence. This process of GCF-GC-spread widening, due to regulatory pressure on the old model, then filters through to the FRPR which contains a ~30% GCF component. The elevated level of the FRPR relative to yields on comparable safe assets generates a pull effect incentivizing lower yielding safe assets to be transformed into reverse repo lending to the Fed via the foreign repo pool.
One final summary, before merging the information attained in Parts I-III in a final post in order to analyze how quickly the pool may grow under various scenarios, why this is likely should the required set of conditions arise and the effect this could have on the Fed’s balance sheet normalization process and broader fixed income markets.
The FRPR, from all evidence available, seems to be a combination of Tri-party GC repo rates (where large dealers borrow from cash pools) and Interdealer GCF rates (where the same large dealers lend to smaller dealers at a spread over GC). Basel III & the eSLR significantly reduced the profitability of dealers running matched-repo books with the result of lower repo volumes overall and also higher lending rates to, at least partly, remunerate the players left for the higher capital requirements. The increase in the lending rate filters through to the FRPR, which is now higher than rates on other safe assets, establishing the pool as the most attractive location for FOI short-term risk-free dollar-denominated claims.