A Broken Eurodollar System Will Lead to Another Monetary Event
From Friday, August 11, through Friday, September 1, this year, the equivalent yield on the 4-week US Treasury bill was calculated to be less than 1% in each trading day during that time. There is, of course, no single 4-week Treasury bill, and they don’t pay any interest. You buy them at a discount and redeem them at par on maturity. The difference in the discounted price and par is the equivalent yield, blended among the several bill CUSIPs that make up the 4-week constant maturity proxy.
There really isn’t any reason for bill yields to trade less than 1%, however. The Federal Reserve has set its reverse repo rate (RRP) at 1%. The RRP is intended as a floor for all money rates. Treasury bills while being ostensibly debt instruments are considered money equivalents, particularly at the shortest maturities - such as one month.
Therefore the RRP “floor” should be applied equally to T-bills as federal funds. After all, consider the choices: lend at 4 weeks to the federal government at less than 1% (per annum), or lend continuously overnight to the Federal Reserve collateralized by US Treasury securities at 1%. If the equivalent 4-week yield is persistently less than RRP, then we know for sure that some number of money market participants is choosing the former option.
On Monday, September 5, the calculated yield for this bill, which had been 96 bps the Friday before, and as low as 93 bps (8/22) in the weeks leading up, suddenly spiked beyond all expectation. The Treasury Department tells us that the calculated yield for that day was 130 bps, higher than the so-called money market ceiling set by IOER (which began its life as the floor, poignantly enough).
The reason for that rejection was never made plain. Some said it was the debt ceiling, where the possible lack of agreement between the White House and the people’s House represented a very real danger for technical default. But it was only the one day; the very next day, September 6, the 4-week equivalent yield was back down to 104 bps. The day after that, the bill was below RRP again where it had remained each additional day until this past Tuesday.
If money market agents are paying a premium for 4-week federal paper, it is surely because of collateral concerns in the repo market more broadly. This point is inarguable, as we have seen enough repo problems over the last ten years to establish empirically this fact. For much of those years, the Treasury bill instruments were left out of it in favor of the benchmark 10-year note (and other similar notes toward the belly of the yield curve).
At and near the zero lower bound, there isn’t enough spread to make bills a prime collateral candidate for securities lenders and borrowers despite their alluring liquidity characteristics. Getting away from the zero lower bound has had the effect of restoring bill action to the forefront of repo hierarchy.
On September 7, two days after the surge in 4-week equivalent yields, the repo market was struck by enormous imbalance, indicated by the 10-year GC rate. The general collateral (GC) rate is supposed to track closely other money market rates. On this particular day, however, the GC rate for the 10s fell below what is meant to be yet another floor, this time the “dynamic fails penalty” imposed in May 2009 upon UST repo.
The fails penalty is charged for any repo transaction that ends with, obviously, a fail. A “fail” is a repo market situation where for various reasons securities that were pledged to be resold aren’t. A repo transaction is really four parts since it is technically still a repurchase agreement; that is, a contract to buy/sell securities on one day and then reverse the transaction at an agreed upon future date. The GC part comes into it as to what is being sold and then rebought (or, from the cash-holder counterparty perspective, bought and then resold).
There is and has always been some unspoken leeway in how this collateral is treated. As ICMA puts it in explaining general collateral, “the repo market as a whole is indifferent between general collateral securities.” I shouldn’t care if I “sell” my 4-week bill to you at the GC rate, receiving your cash in consideration, and then “buy” back a 10-year UST at maturity, returning your cash plus interest (the GC rate), because that is what you are reselling. From my perspective, it’s all interchangeable UST collateral of sufficient liquidity.
The trades are typically described from the cash perspective, which history has taught us is a dangerous assumption. Sometimes I do care which UST is resold back to me. When that happens, it is called “specialness” and historically it was described as a function of shorting bonds. Unlike repo, a short needs to replace the exact security in which was borrowed.
It’s not always (or over the last decade, ever) shorts who define specialness and fails. Why would they? If you really want to short UST’s because you believe interest rates have nowhere to go but up (for the fourth time) there is a much more liquid futures market that provides a more flexible range of such opportunities. Shorting in the cash market isn’t a cause of fails, if for no other reason than that there are no fails problems in futures.
But on September 7, the market really did care about the 10s; so much so that cash was available to be borrowed at an insanely negative rate in order to procure those particular notes. The dynamic fails penalty, 300 bps below the RRP, is charged to create an economic cost associated with failing to return repo collateral to keep orderly function. Why the desperation over 10s?
We can’t know for sure, of course, because much of what takes place in repo is only part of the overall whole. Even fails are a partial snapshot, those captured by FRBNY reported to that branch of the US central bank by the primary dealers.
On the 7th, I speculated that it was the unexpected repudiation of the 4-week bill that did it:
“What do matter are haircuts. As a cash lender/collateral holder your only concern apart from the interest rate or spread is security of that collateral. The haircut is that security, cushioning your position against any violent move in the price of the collateral instrument. Even an overnight repo is subject occasionally to such risk (repo traders, I can assure you, do not fondly recall 2007 MBS, before then the perfect example of placid trading).”
The setting of the haircut is a key element in repo/collateral protection. That’s what caused so many systemic problems in 2008, starting in 2007 with MBS. Haircuts were set far too low because everyone assumed that placidity in securities trading, even subprime mortgage structures, was a permanent feature. You didn’t need a big haircut, you thought, because prices never moved all that much on a single day; until they did.
That seems to have been the case with the 4-week bill and September 5. All of a sudden the price of those instruments changed, by a huge amount, defying all haircuts that were set prior on the belief it wasn’t likely to happen. If you had used a 4-week bill in repo on September 5, you may have been faced with a collateral margin call of sorts on the 6th; your repo counterparty was happy to lend you cash again at the same or nearly the same interest rate, but you had to put up more collateral to do so if all you were offering was still a 4-week bill.
From that situation, you can more easily understand the sudden scramble for 10s over the next few days. If you can’t any longer post 4-week bills on the same easy terms, then you have to find whatever else is next on the list. The benchmark 10s approach bills in terms of liquidity characteristics repo counterparties are after, so it stands to reason they would be in such great demand as a potential replacement.
But this was all confined to just three days in early September. Even though we don’t really know what happened to the 4-week bill on that particular Monday, though I have some theories that have nothing to do with the debt ceiling, in some ways it doesn’t really matter. What does is that it started a chain reaction that may have only just now been resolved this week.
Late on Friday, September 15, FRBNY reported what I had been expecting. Repo fails during the week in question had surged to $435 billion (combined “to receive” plus “to deliver”), up sharply from an eight-week average of $242 billion. That average pace was already elevated, significantly more than what used to constitute something like normal in UST repo. At the last week in May 2014, for example, just prior to the “rising dollar”, the eight-week average for UST repo fails was all of $99.8 billion.
Though we expected a heavy dose of fails for the week of September 6, perhaps less predictably it continued on the week after that; and then the week after that. FRBNY tells us that repo fails in UST’s submitted by the primary dealers were $325 billion the week immediately following these events, and then $359 billion the week of September 20 (the latest estimates available). We can surmise that fails might have been severe again the week of September 27 given the T-bill rate remained below the RRP floor (but not so much below it that it makes for an easy prediction).
Three consecutive weeks at that level of fails is a warning; it has been that in the past, including the recent past. As I recalled upon the release last Friday of that $359 billion:
“The last time that [three weeks in a row of at least $250 billion fails] happened was for five straight weeks beginning last November during the big bond selloff, and then earlier in 2016 in March (Japan problems). Before those there were four weeks in December 2015 just prior to the second wave of global liquidations, and then for three weeks in June 2014 kicking off the whole ‘rising dollar.’”
It is unusual even in these turbulent times to find such volume in fails; it’s even more so given what little started it.
The real question we need to be asking is why a one-day shift in the 4-week bill has cascaded into a systemic repo warning (again). If things were working as they should, there may have been at most a few days of outlier rates and numbers and not anything more. Securities lenders would have entered the market to take advantage of the insane cash rates being offered (if they wouldn’t have entered the market at the appearance of favorable cash rates before they became insane), unlocking additional collateral available to whomever may be on the wrong side of the 4-week bill’s torment.
Clearly that didn’t happen. Instead, what followed was at least three weeks of persistent unavailability of alternative collateral supplies – including that of the RRP itself! One of the main features of the reverse repo program offered by the Fed was that it opened a channel for the trillions in UST securities sitting otherwise idle in SOMA. But for all the theoretical benefit, it has never worked in that fashion. I wrote to start October 2014 that the RRP had already been proved a “fairy tale” as far as that theory was concerned. The language I used should seem far too familiar
“They can try to convince themselves that repo fails are strictly limited to the size of the short position in UST, but that is largely irrelevant – it matters not why there is demand for collateral, the only focus should be on why the “market” cannot meet said demand. If an imprint of short selling in UST, expecting rates to rise (which I don’t buy as the primary problem in repo right now, at all!), can rattle repo as it has, then that is very a concerning sign for when there is much more demand for collateral under real, true strain; a condition that is inevitable.”
Two weeks after I wrote all that the “buying panic” of October 15 happened, nothing more than an enormous collateral call on the whole global market. Things were never the same thereafter; indeed, the warnings and monetary insufficiencies only grew worse from that point forward.
Why can’t the repo market meet demand for collateral?
Balance sheet capacity is the answer.
Many will have you believe that increased regulation is the reason for the lack of it, but that, too, is demonstrably false. We are told by these people that guidelines like Basel 3 make it harder and more expensive for dealers to hold inventories of UST’s and the like, therefore diminishing their capacity to respond in these kinds of situations.
But FRBNY in its other statistics on primary dealers shows quite the opposite. If anything, dealers are holding now (2017) more inventory than at any point in decades; especially as it comes to bills. When reporting these positions, the dealers are more net long.
Long or short in terms of dealers does not mean long or short as you may immediately understand. Before August 2007, dealers were quite net short in bills as well as coupons and not because they believed Alan Greenspan was going to overcome his “conundrum.” Instead, they were short UST’s as an indication of how they were (fluidly and cheaply) renting out securities in securities lending.
As soon as crisis hit and MBS collateral more and more rejected on useful terms, dealers began to hoard UST’s as a safeguard for their own liquidity concerns. Thus, the inventory of dealer holdings rose (relatively more net long) and fell (relatively more net short) over the last ten years in close relationship with each “reflation” and then the inevitable re-emergence of liquidity crisis; first 2011, and then again in 2014.
Following the “rising dollar” and the “reflation” that gripped markets in the second half of 2016, dealers unlike previous times chose instead to continue hoarding UST’s almost as if nothing had changed. As such, they have been holding even larger net long positions throughout most of 2017, especially after February. That’s not regulations, but good sense; risk perceptions that have been proven once again more accurate than all mainstream forecasts.
If dealers are no answer to desperate collateral demand, and the RRP nothing but a joke, then it would make sense that one little blip in bill trading unleashed several weeks at least of huge and concerning monetary tightness. From August 2007 to March 2008 and the failure of Bear Stearns to September 2008 and Lehman, collateral tightness is every bit as important as if there was a currency shortage. In the modern wholesale system, it can be, and has been, more valuable than cash.
That is what bill holders are paying for in their premium prices over the past six months. Collateral conditions are not and have not been good as a baseline. The repo system remains vulnerable no matter what Janet Yellen says; after all, the Fed believes in interest rates not money supply, discontinuing an M3 stock measure that included at least a small part of the repo picture more than ten years ago.
Thus, the current problem is twofold: the lack of response to even small disturbances, which suggests very low collateral elasticity. A well-functioning, highly elastic system for collateral would be as I described above, one where dealers respond easily and readily to profit opportunities that outlier circumstances provide.
We have heard constantly over especially the last ten years that dealers can’t make money without volatility, and that the Fed’s policies smoothed out too much volatility. That’s another pro-central bank myth easily disproved, especially by the “rising dollar.” And here it is contradicted yet again, for dealers have been absent in September 2017 given often massive profit opportunities to step in.
And that leads us to the second current concern: why are dealers still so reluctant? If they are still hoarding inventory, and they are, what are they worried about? It may be as simple as this redundancy issue, meaning that dealers learned (the hard way) that when the collateral system is stressed they are as vulnerable as the little guys out there funding on much the same terms. It is, in other words, a self-reinforcing problem, a positive feedback loop where collateral problems instead of triggering a favorable dealer response that solves them lead instead to greater dealer reluctance that amplifies them (October 15, 2014).
What’s missing is what is always missing; balance sheet capacity. It’s why I consider balance sheet capacity more “money” than cash is. It dictates all these complex facets of global wholesale funding and far more. Dealers who are unconcerned and more unrestrained in their own balance sheet capacities are those who would break the positive feedback loop by acting like money dealers; especially when they might be reasonably assured that their own capacity is shared by the rest of the dealer network. After all, there is almost a hive mind characteristic to it, where what one dealer does the rest does also.
It’s a major structural flaw of this eurodollar system, one inherent contradiction among several. There is no redundancy because the central bank is largely irrelevant, the RRP providing the next in a long line of proving arguments, and the dealers themselves are essentially one common thing rather than several dispersed individual entities standing on their own idiosyncrasies and merits. The system is incestuous particularly when it comes to risk capacity; therefore what happens to one is immediately shared by all.
This is why the eurodollar system is irreparably broken, and also why I continue to expect these warnings are leading us toward another of these monetary events. The next one, should it happen, will be the fourth such. Three times was sufficient only to instill some doubt in Economists and central bankers, who through the second one (2011) and its aftermath remained naïvely steadfast. Even if three weeks of fails was actually related to bond market shorts, illustrating a less troublesome affliction, it would still deserve widespread attention and caution.
