Back in the summer of 2013, we first commented on what we called "Phantom Markets" - displayed quotes and prices, in not only equities, FX and commodities but increasingly in government bonds, without any underlying liquidity. The problem, which we first addressed in 2012, had gotten so bad, even the all important Treasury Borrowing Advisory Committee to the US Treasury had just sounded an alarm on the topic.
Since then we have sat back and watched as our prediction was borne out, as bond market liquidity slowly devolved then sharply and dramatically collapsed recently to a level that is so unprecedented, not even we though possible, leading first to the October 15 bond flash crash and countless "VaR shock" events ever since.
And while we urge those few carbon-based life forms who still trade for a living to catch up on our numerous posts on market "liquidity" and lack thereof, here is a quick and dirty primer on just why there is virtually no bond market left, courtesy of the man who, weeks ahead of the Lehman collapse when nobody had any idea what is going on, laid out precisely what happens in 2008 and onward in his seminal note "Are the Brokers Broken?", Citigroup's Matt King.
Here is the gist of his recent note on the liquidity paradox which is a must read for everyone who trades anything and certainly bonds, while for the TL/DR crowd here is the 5 word summary: blame central bankers and HFTs.
* * *
The more liquidity central banks add, the less there is in markets
- Water, water, everywhere — On many metrics, liquidity across markets seems abundant. Bid-offers are tight, if not always back to pre-crisis levels. Notional traded volumes in credit and rates have reached all-time highs. The rise of e-trading is helping to match buyers and sellers of securities more efficiently than ever before.
- Nor any drop to drink — And yet almost every institutional investor, in almost every market, seems worried about liquidity. Even if it’s here today, they fear it will be gone tomorrow. They say that e-trading contributes much volume, but little depth for those who need to trade in size. The growing frequency of “flash crashes” and “air pockets” – often without obvious cause – adds weight to their fears.
- Yes, street regulation has played a role — The most frequently cited explanation is that increased regulation has driven up the cost of balance sheet and reduced the street’s appetite for risk, and hence ability to act as a warehouser between buyers and sellers.
- But so too have the central banks — And yet this fails to explain why even markets like FX and equities, which do not consume dealers’ balance sheets, have been subject to problems. We argue that in addition to regulations, central banks’ distortion of markets has reduced the heterogeneity of the investor base, forcing them to be the “same way round” over the past four years to a greater extent than ever previously. This creates markets which trend strongly, but are then prone to sudden corrections. It also leaves investors more focused on central banks than ever before – and is liable to make it impossible for the central banks to make a smooth exit.
How Bad Is Liquidity Reall?
From the BIS to BlackRock, and Jamie Dimon to Jose Vinals, everyone seems to be talking about market liquidity. Chiefly they seem to be fretting about a lack of it. Primary markets might be wide open, thanks in large part to the largesse of central banks, but the very same investors who are buying today seem deeply concerned about their ability to get out tomorrow.
Liquidity as a concept is notoriously difficult to pin down. It has a reputation for being very much in evidence when not required, and then disappearing without trace the moment you need it. For strategists – and regulators – this represents a challenge: conventional metrics like bid-offer and traded volume can go only so far towards capturing what investors mean by liquidity. And yet because investors’ concept of liquidity tends very much to be focused on tail events, by definition, data to help monitor it are scarce.
This paper tries to assess the evidence across markets, and evaluate what is driving it.
Some observers have argued that even if liquidity is disappearing from some markets, it is being maintained – or even concentrated – in others. We argue in contrast that the risk of illiquidity is spreading from markets where it is a traditionally a problem, like credit, to traditionally more liquid ones like rates, equities, and FX. There is a bifurcation – but it is between decent liquidity much of the time and then sudden vacuums when it is really required, not across markets.
We likewise take issue with the widespread notion that the problem is solely due to regulators having raised the cost of dealer balance sheet, and could be ameliorated if only there were greater investment in e-trading or a rise in non-dealer-to-non dealer activity. To be sure, we see the growth in regulation – leverage ratio and net stable funding ratio (NSFR) in particular – as one of the main reasons why rates markets are now starting to be afflicted, and indeed we expect further declines in repo volumes to add to such pressures. But illiquidity is a growing concern even in markets like equities and FX, which use barely any balance sheet at all, and where e-trading is the already the norm rather than the exception.
Instead, we argue that in addition to bank regulations, there is a broad-based problem insofar as the investor base across markets has developed a greater tendency to crowd into the same trades, to be the same way round at the same time. This “herding” effect leads to markets which trend strongly, often with low day-to-day volatility, but are prone to air pockets, and ultimately to abrupt corrections. Etrading if anything reinforces this tendency, by creating the illusion of lliquidity which
evaporates under stress.
Such herding implies a reduction in the heterogeneity of the investor base. One potential cause is the way steadily more investor types having become subject to procyclical accounting and capital requirements in recent years, most obviously for insurance companies and pension funds.
But the tendency towards illiquidity pockets even in markets where insurance and pension money is not dominant suggests a deeper cause. We think the most likely candidate is central banks’ increasing hold over markets. Over the past four years, it is expectations of central bank liquidity, not economic or corporate fundamentals, which have become the main driver of everything from €/$ to credit spreads to BTP yields.
While central banks have always been significant market participants, their role has obviously grown since 2008. Most obviously, their global asset purchases have drastically reduced the net supply of securities available to be bought by investors. At the same time, we have seen a breakdown in a number of fundamental relationships which had previously correlated well with markets – and their replacement with metrics directly linked to central bank QE. Because the herding is not directly backed by leverage, it is unlikely to be reduced by macroprudential regulation.
To date, the air pockets and flash crashes represent little more than a curiosity, having mostly been resolved very quickly, and having had little or no obvious feedthrough to longer-term market dynamics, never mind to the real economy.
But we think ignoring them would be a mistake. Each has occurred against a largely benign economic backdrop, with little by way of a fundamental driver. And yet with each one, investors’ nervousness about the risk of illiquidity is likely to have been reinforced. When the time comes that investors do see a fundamental reason all to sell – most obviously because they start to doubt the extent of central banks’ support – their desire to be first through the exit is liable to be even greater.
There when it’s not needed
First, let us consider the evidence that all the fuss about liquidity is much ado about nothing. In many markets, under normal conditions, it is now possible to trade on more platforms, with more counterparties, and with tighter bid-offer, than ever before. While some market developments may perhaps point to the potential for problems, day-to-day liquidity remains remarkably good.
Plenty of notional volume, with tight bid-offer
The most obvious data point in this respect is that notional traded volumes in many markets are at or close to all-time highs.
US credit, for example, saw nearly $27 trillion in secondary trading last year; HY volumes have doubled since the crisis (Figure 1). In government bonds, the increase was typically pre-crisis, but volumes have typically remained flat even in the face of a growing proportion of bonds being absorbed by the central banks (Figure 2). German Bunds are a notable exception. Notional volumes in equities have fallen from precrisis peaks, but remain close to them in the US and Japan (Figure 3).
Likewise, data on bid-offer across markets mostly paints a healthy picture. The BIS shows that bid-offer in govies has largely fallen back towards pre-crisis levels. Equity bid-offer is likewise close to pre-crisis lows, and seems to suggest that even quite large portfolios could be transacted at tight spread levels – provided volumes were split quite broadly across a large number of stocks (Figure 4). Bid-offer in credit remains wide to pre-crisis levels, as far as we can tell4, but is the tightest it has been since then (Figure 5). For small investors not correlated with the broader herd, these gains in notional volume and tight bid-offers will represent a very real ability to transact.
Turnover in decline
Admittedly, the positive message in these notional data is considerably mitigated when the growth in many markets – and in many investors’ portfolios, especially in rates and credit – is taken into consideration. Once transaction volumes are adjusted to show the difficulty an investor is likely to have moving a given percentage of the market, a very different picture emerges – and one which is much more consistent across markets.
In credit, rates, and equities alike, turnover relative to market outstandings has fallen considerably. Corporate turnover has almost halved since the crisis (Figure 6). The decline in government bond turnover has been more protracted, but is just as drastic, especially in US Treasuries (Figure 7). Equity turnover is more obviously influenced by the rise and fall in outstandings with market movements, but has also suffered a post-crisis decline (Figure 8).
For most investors, turnover is probably a more useful metric than straight traded volume. Monthly mutual fund inflows and outflows have generally grown as fund sizes have increased; for high-yield bond funds, net outflows seem to have been growing even in percentage terms, never mind in notional ones (Figure 9). Given that their share of the market has also been growing, the capacity to move a given percentage of the market is a much better guide than the ability to move a given notional volume. The same tendency is not quite as pronounced in other long-term fund types, but seems also to be true for money market funds (Figure 10).
And yet despite declining turnover, the fact remains that outflows to date have not led to obvious liquidity problems.
ETFs, futures and indices – can they be more liquid than their underlying?
If the mutual funds’ role in contributing to market liquidity risks has been exaggerated, then we think ETFs have been more maligned still. They may well be responsible for some of the reason “day-to-day” liquidity has held up so well, but we struggle to see that they can be blamed for any increase in its tendency to disappear under stress.
The rise in ETFs has certainly been striking; in equities, in particular, net ETF sales have outstripped regular mutual fund sales in recent years (Figure 16), though in fixed income their rise has been slower. More striking still is that ETF trading now constitutes just under 30% of all US equity dollar volume. The rise of ETFs is probably one reason why, also in Europe, a growing proportion of daily volume is shifting away from intraday trades and towards end-of-day auctions (Figure 17).
There may well be a positive feedback loop in which other large trades are increasingly executed at end-of-day auctions, precisely because investors know that liquidity is becoming concentrated there, and to avoid being seen for reporting purposes to have dealt away from the daily closing price.
But what is striking in Figure 17 is that non-auction equity volumes have also been rising in recent years. Rather than ETFs subtracting from cash traded volume, they seem to have added to it. The same might be said for the rise of traded volume in dark pools.
In this respect, we see ETFs as very much akin to the CDS indices in credit, or futures in equities and rates. It is not that there is a fixed quantity of trading to be done, and that the arrival of a new instrument necessarily removes trades which would have been done elsewhere. If anything, we see the opposite phenomenon. Liquidity begets liquidity: when investors have new instruments with which to express views and refine their positions, it tends to encourage still more trading.
In addition, in contrast with open-ended mutual funds, ETFs – like closed-end investment trusts – offer the possibility for prices to diverge from net asset values. This affords an additional cushion, especially during periods of market stress. The fact that ETF volumes have a tendency to increase relative to cash volumes during periods of stress – even though they then are likely to be trading at a discount – suggests that this cushion works well, and may even act to boost liquidity during stressed conditions. We have much sympathy for the way one ETF provider put it: “Everyone complains when they see our prices deviating from the underlying. But what they fail to realize is that at least people can and do actually trade at those prices: in troubled times, the prices shown on broker screens for underlying instruments are often a fiction.”
So while we would agree with the statement that “no investment vehicle should promise greater liquidity than is afforded by its underlying assets”5, we think there is something for an exception for vehicles which are themselves tradable, and can trade with a basis relative to their underlyings. ETFs, CDS indices and futures can and do provide much more liquidity than their underlyings. This, though, is not so much because they “promise” more liquidity, as that they facilitate additional activity, much of which is crossed at the index level, and only a fraction of which is transmitted through to the underlying. While investors are periodically surprised and frustrated by these vehicles’ failure to perfectly track movements in their underlyings, this failure is in some ways the secret of their liquidity. Rather than stealing liquidity from the underlyings, we think their growth has added to it.
Missing when required
But if neither ETFs nor mutual funds can be held responsible for perceptions of reduced liquidity across markets, what can? The finger is most often pointed at the street. Many buysiders feel that dealers’ willingness to act as a liquidity provider even during good times has waned; might such reduced willingness also help explain an increased tendency of liquidity to evaporate altogether? We think this is much closer to the truth – but even so, we doubt it is the full story.
Are the brokers broken?
Such arguments have been voiced most persuasively in Jamie Dimon’s recent letter to JPM shareholders. He cited three factors: higher cost of balance sheet, explicit constraints for US banks as a result of the Volcker Rule, and in extremis the fact that the rescues of the likes of Bear Stearns, WaMu and Countrywide/Merrill have led not to gratitude but to heavy fines for JPMorgan and Bank of America, in at least one case following the abnegation of ex ante assurances otherwise.
The effects of the latter two factors are hard to observe, and would probably become visible only in a proper crisis. Even then, they might in principle be offset by “Rainy Day” funds. These are pools of money being set up by asset managers precisely so as to take advantage of liquidity-related distortions. But it seems doubtful that these will ever reach the scale required to take 2008-style rescues, at least without large amounts of leverage.
The effects of reduced dealer balance sheet, however, are visible already, and are contributing directly to the perception of illiquidity in fixed income. Unlike equities or FX, fixed income trading is fragmented across an extremely large number of outstanding securities. Only rarely can a willing buyer and seller of the same security be found at the same instant. As such, liquidity, particularly for larger trades, relies heavily on dealers’ ability to act as a warehouse, temporarily hedging a long position in one security with a short position in another. This requires both the availability of balance sheet, and the ability to borrow securities freely through repo. Both of these are now under threat from regulation.
Why e-trading is no panacea
E-trading works extremely well as an efficient means of uniting buyers and sellers of a given security – provided those buyers and sellers exist in the first place.
That is, a buyer can probably find a seller faster now than they could a few years ago, when they had to rely solely on email and phone calls. In indices, in equities and in on-the-run govies, where many investors are ready and willing to trade the same security on either side of the market multiple times in an hour, this can bring about significant improvements (Figure 24).
But in much of fixed income (and especially credit), liquidity is intrinsically fragmented. End investors’ willingness to buy and sell a large notional volume of a given security is simply not there in the first place (Figure 25). Even if investors could be persuaded to concentrate their positions in a given issuer on a smaller number of outstanding “benchmark” securities, as BlackRock has called for, issuers could never be persuaded to do so, since it would add to their refinancing risks. Issuers like the diversification that a multitude of outstanding securities brings.
As such, it is no use efficiently putting together buyers and sellers when those buyers and sellers do not exist in the first place. This is why many traders report that – even when they have on occasions been able to offer ‘choice’ markets with zero bid-offer for a while, these have not necessarily resulted in any trading. End user liquidity remains fundamentally dependent on a counterparty’s willingness to act as a warehouse: to buy the security the seller wants to sell, to offer the security the buyer wants to buy, find a hedge of some sort and then move to unwind the position later. This is also why so many bond market participants – buyside and sellside – are opposed to efforts to copy-paste equity-inspired regulations into a fixed income framework in the interest of increased transparency.
Nor is it obvious that there is an easy alternative to the existing broker-dealer model when it comes to warehouses. Bid-offers are tight enough that the market-making model relies upon leverage in order to generate a reasonable return on equity. Even if, say, asset managers or hedge funds are prepared to act as warehouses, for them to make money on the operation they will want to operate with leverage. But – as the clearing houses are now starting to find out, and banks discovered some time ago – such leverage represents precisely the sort of systemic risk that regulators are now keen to limit.
As such, e-trading to date has done a great deal to boost what the IMF calls “flow”, or day-to-day liquidity, for small-sized trades. But when it comes to larger transactions, they can seldom be cleared (Figure 26).
‘Phantom liquidity’ – is the problem getting worse?
There is an argument that the liquidity provided by e-trading seems to be more fleeting than that stemming from voice trades and personal relationships. When markets become volatile, e-trading operators tend to pull the plug – or, at best, reduce the size they are willing to trade. A recent IMF analysis concluded that it was precisely such a reduction in the depth of order books which seems to have led to the ‘flash rally’ in US Treasuries on October 15th 2014. A high dependence on electronic trading also seems to have contributed to the flash crash in equities on May 6th 2010.
It is this phenomenon that is known as “phantom liquidity”, or the “liquidity illusion” – a tendency to evaporate when really needed. It does seem possible that e-trading may have added to such a tendency, by improving the appearance of liquidity under normal conditions, and then withdrawing it in periods of stress. This could help explain why some of the most obvious instances of recent illiquidity have occurred in markets which already have high proportions of trades conducted electronically.
And yet beyond the anecdotal, quantitative data demonstrating an increased tendency towards “phantom liquidity” is extremely hard to pin down. After all, it makes predictions not about day-to-day circumstances but about tail events, which by definition are few and far between.
The best evidence is probably some increase in bifurcation in day-to-day trading conditions, visible in daily volatility levels across markets. Rather than there being a steady stream of moderately volatile days (and liquidity conditions), volatility seems to be becoming more clustered than it used to be: there are many days with tight ranges and good liquidity, and then occasional days of extreme intraday volatility and reportedly poor liquidity – even though (as in the flash crash and flash rally) volumes on such days can actually remain quite high.
Often, the volatility is thought to occur intraday, and therefore may not be captured by the net daily changes implicit in volatility metrics. Looking at intraday high-low ranges in markets paints a slightly clearer picture – but not decisively so. It does in general seem to us that intraday ranges have become more bifurcated since around 2005, most obviously with a period of low volatility prior to the financial crisis being followed by the extremely high ranges during the crisis itself, but also with post-crisis daily trading ranges being more bifurcated than prior to 2005.
Low day-to-day volatility, punctuated by occasional sharp corrections, are exactly what we might anticipate if markets were becoming less liquid. In credit, for example, they are one of the features which distinguishes the cash market relative to the continual bouncing around of the CDS indices. And we have argued previously that markets seem to be becoming more subject to positive feedback loops, which see them trending steadily upward only to fall back suddenly and often unexpectedly.
And yet the limited number of observations, and the variations across markets, make it hard to make confident statistical statements about any change in the shape of distributions. We are left with the unsatisfying conclusion that evaporating liquidity is as much a feeling voiced by many market participants as to what might happen under stress, illustrated by a few idiosyncratic examples, as it is a statistically demonstrable phenomenon.
The evidence of increased herding
A recipe for a perfectly liquid market would be one with a small number of homogeneous securities being traded by a much larger number of heterogeneous participants. This would do a great deal to improve the likelihood of a buyer and seller both wanting to transact in the same security at the same time, and hence being able to agree upon an appropriate price. Indeed, it has even been suggested that we might quantify markets’ potential for liquidity along such lines, taking the number of distinct market participants and dividing by the number of securities traded.15 This is one reason why liquidity in indices and futures is often so good, as they concentrate a large amount of activity in a small place.
One potential explanation for growing illiquidity is that markets have been evolving in an exactly opposite direction. Not only has the number of securities traded been growing (in credit in particular), but the heterogeneity of market participants seems to have been reducing as well. Here too, the regulations are partly to blame, with more and more investors being forced both to mark to market and to hold capital or cover pension deficits on the basis of such mark to market calculations. In addition, though, it feels to us as though market participants are increasingly looking at the same factors when they make their investment decisions.
For the last few years, valuations in more and more markets seem to have stopped following traditional relationships and instead followed global QE. Likewise in meetings with investors, we have been struck by how little time anyone spends discussing fundamentals these days, and how much revolves around central banks. Record-high proportions of investors think fixed income is expensive and think equities are expensive. A growing number of property market participants seem to think real estate is expensive. And yet almost all have had to remain long, as each of these markets has rallied. Could it be that central bank liquidity has forced investors to be the same way round more so than previously, and that this is making markets prone to sudden corrections.
While it is hard to demonstrate conclusively, a growing weight of evidence would seem to point in such a direction. CFTC data on net speculative positioning in futures and options markets has become more extreme in recent years, and abrupt falls in net positioning have often coincided with sharp market movements. Net shorts in Treasuries reached record levels immediately prior to the flash rally, for example (Figure 29); net longs in commodities contracts preceded the sharp fall in commodities indices in the second half of last year, and record net euro shorts (Figure 30) and dollar longs are being squeezed at the moment. However, it could be argued that growth in notional contracts outstanding is a normal part of financial market deepening; normalizing by the net open interest would (at least in some cases) suggest recent positioning is not too far out of line with history.
But other data also point to an increase in investor crowding. Our own credit survey shows that investors’ positions in credit since the crisis have not only been longer on average than ever previously, but also less mean reverting, and exhibiting less dispersion and less mean reversion (Figure 31). Fully 83% of those surveyed were long credit in December last year – a sizeable imbalance for any market. Similarly, the BAML global investor survey shows that investors have been long in equities for a longer period than would historically have been normal.
Research specifically designed to detect investor herding has reached the same conclusion. An IMF analysis of the correlation between individual securities transactions by US mutual funds, using the vast CRSP database, shows a clear pickup in herding with the crisis, and then another one in late 2011. The herding seems to have occurred consistently across markets, but was more intense in credit and especially EM than in equities. It also occurred both among retail and among institutional investors. The IMF were unable to test for herding in government bonds and FX because of the much more limited number of securities.
What we find striking about the herding numbers is the way they correlate with the metrics we use to track the scale of central bank interventions: rolling global asset purchases by DM central banks, and global net issuance of securities once central bank purchases (and, in this case, also LTROs) have been subtracted out. Over the past four years, we have had to use these metrics to help explain market movements when traditional fundamental relationships have broken down.
Investors likewise agree on the dominance of central banks: in a survey of global credit derivatives investors we conducted in January this year, fully two thirds thought “central bank actions” would be the main driver of spreads this year, well ahead of “credit fundamentals”, “global growth trends” or “geopolitical risks”. Even if the central banks are only having to intervene because the systemic risks they are confronting have become bigger, the effect remains the same.
This, then, would seem to be the final piece of the puzzle as to what is making markets prone to pockets of illiquidity. Central bank distortions have forced investors into positions they would not have held otherwise, and forced them to be the ‘same way round’ to a much greater extent than previously. The post-crisis increase in correlations, which has been visible both within credit and equities and across asset classes (Figure 35), stems directly from the fact that investors now increasingly find themselves focused on the same thing: central bank liquidity. Every so often, when they start to doubt their convictions, they find that the clearing price for risk as they try to reverse positions is nowhere near where they’d expected.
This explains why the air pockets have not just been in markets where the street acts as a warehouser of risk. It explains why they have occurred not only in the form of sell-offs which could have caused multiple market participants to suffer from procyclical capital squeezes. It also explains why the catalysts have often, while often trivially small, have nevertheless been macro in nature, since they have boosted expectations of a change in central banks’ support for markets.
Unfortunately, it leads to a rather ominous conclusion. The bouts of illiquidity will continue until central banks stop distorting markets. If anything, they seem likely to intensify: unless fundamentals move so as to justify current valuations, when central banks move towards the exit, investors will too.
Rather than dismissing recent episodes as relatively harmless, then, we are supposed to worry how much larger a move could occur in response to a more obvious stimulus. While financial sector leverage has fallen, debt across the nonfinancial sectors of almost every economy remains close to record highs, meaning that the potential for negative wealth effects in the real economy is very much there.
In principle, markets could gap to a point where they went from being absurdly expensive to being absurdly cheap, and then – as investors stepped in again – gap tighter, perhaps even without very much trading. But the existence of the feedback loop to the real economy means that the fundamentals tend also to be affected by extreme market moves: “cheap” may be a moving target. This in turn could force central banks to step back in again.
To sum up, we are left with a paradox. Markets are liquid when they work both ways. Market participants, though, find themselves increasingly needing to move the same way. This is not only because of procyclical regulation; it is also because central banks have become a far larger driver of markets than was true in the past. The more liquidity the central banks add, the more they disrupt the natural heterogeneity of the market. On the way in, it has mostly proved possible to accommodate this, as investors have moved gradually, and their purchases have been offset by new issuance. The way out may not prove so easy; indeed, we are not sure there is any way out at all.
* * *
To which all we can add is: Good luck with the "exit"