Treasury Liquidity Freakout: Searching for a Market-Maker

As someone old enough to have done finance in the Paleolithic pre-personal computer era (yes, I did financial analysis using a calculator and green accountant’s ledger paper as a newbie associate at Goldman), investor expectations that market liquidity should ever and always be there seem bizarre, as well as ahistorical. Yet over the past month or two, there has been an unseemly amount of hand-wringing about liquidity in the bond market, both corporate bonds, and today, in a Financial Times story we’ll use as a point of departure, Treasuries.

These concerns appear to be prompted by worries about what happens if (as in when) bond investors get freaked out by the Fed finally signaling it is really, no really, now serious about tightening and many rush for the exits at once. The taper tantrum of summer 2013 was a not-pretty early warning and the central bank quickly lost nerve. The worry is that there might be other complicating events, like geopolitical concerns, that will impede the Fed’s efforts at soothing rattled nerves, or worse, that the bond market will gap down before the Fed can intercede (as if investors have a right to orderly price moves!).

Let’s provide some context to make sense of these pleas for ever-on liquidity.

Believe it or not, capital markets did a perfectly fine job of providing funding for companies without having perma-liquidity. For instance, it was generally understood in the stone ages of my youth that equity markets were receptive to dodgy wares like technology IPOs only two or three years out of every five, and even then, the issuer better have six quarters of earnings. Yet that period of not very good IPO access was one of the most stellar periods for the establishment of computer hardware and software giants: Lotus, Microsoft, Apple, Cisco, Sun, Oracle. In other words, entrepreneurs may have been BETTER motivated to build lasting enterprises by focusing on the business of the business and not fixating on The Great IPO Exit, which would eventually be there for them as long as they kept their venture growing and profitable.

Similarly, back in the days of stone knives and bearskins of the 1980s, individual corporate bond issues were understood not to be all that liquid, but buyers didn’t seem terribly bothered. If you really did need liquidity, you’d stick with the issues that were well-traded, such as AT&T or big utility bonds. But corporate bond buyers could get adequate pricing even if their particular bond wasn’t heavily traded because corporate bond prices were easy to grid: dealers and investors could extrapolate pretty well from the liquid issues, adjusting for the feature of each deal (ratings, maturity, coupon, sinking fund, call protection. For lower rated issues, you might have to look into the covenants). And remember, an investor expected to get a little yield premium for taking an issue that was less liquid. After all, everything in finance can be solved by price.

In the intervening years, we’ve had persistent efforts both by regulators and from industry participants themselves for more liquidity. And it’s not hard to make a case that the push for extra liquidity has gone well beyond the point of optimal value, at least as far as having capital markets provide benefits for society is concerned. Various studies have concluded that high frequency trading in stocks is detrimental, in that it adds liquidity when no one needs it, as in when liquidity is already ample, and drains it when it is most needed, when markets are roiled. And this means equity market structures, despite the appearance of greater ease of transactions. As former derivatives trader Craig Heimark and we wrote in April:

Perversely, much of the regulation of the last twenty years has been nominally in the interest of “market efficiency” but has come at the expense of market integrity. Far too many of the arguments and studies saying the promotion of competition among exchanges (and dark pools) has led to greater efficiency look at the efficiency as measured by the bid ask spread (plus fees) only of trading in the top stocks (because if they are trade weighted so that is where all the volume is). But this greater efficiency comes at the expense of no reciprocal liquidity obligation (witness the flash crash) as well as reduced liquidity in less frequently traded stocks.

The societal benefit of trading is to reduce cost to raise capital for actual companies. Does anyone really think that narrowing the spread on Google by a penny or two makes any difference to its weighted average cost of capital? In contrast, incidents like the flash crash and the feeling the market is rigged keep many small investors away from the market. The penalty for reduced liquidity in small stocks may actually be material to small company capital formation.

And these small investors are right to be concerned. The old exchange system was a hub and spoke model, which was a stable system architecture. The internet was an outgrowth of a DARPA project to make a communication system so decentralized that it could not be taken out by a nuclear strike. Hub and spoke models are stable, but subject to an outage, say by a nuclear bomb or electrical failure. What chaos theorists have found is that highly decentralized networks are stable, as are single node networks (like exchanges), but that slightly decentralized networks are fragile. And that is what we have now thanks to the SEC’s misguided efforts to “modernize” the stock market via Regulation NMS.

So regulators have left investors with the worse possible market structure. We no longer have liquidity obligations to make orderly markets as we had with the old model. Our current system is more complex due to some decentralization, but it is not so decentralized that it is robust (in technology-speak, a synchronized mesh network). The complexity of keeping the slightly decentralized model synchronized is what makes the system unpredictable and more fragile. This is not just an academic network construct. It is why we saw some exchange crashes recently (like Nasdaq) that were due to code changes in the linkages and feeds between exchanges.

Back to the current post. For mere mortals, “reciprocal liquidity obligation” is a fancy way of saying, “no one is particularly obligated to make a market”. Recall that in the traditional stock exchange system, that fell to specialists. For bonds, that responsibility once belonged to the lead manager of a new bond issue (in fact, the money-making opportunities for them were much greater in the secondary market, since it was the lead manager who would know who held which bonds. The more you knew who owned what, the more you could suggest trades that might fit a particular investor’s portfolio needs while allowing you the trading firm to make something on the shuffling. A dominant position produced great network effects, which was the foundation of bond king Salomon Brothers’ success in its heyday).

Consistent with Heimark’s views, regulators and investors themselves have peculiarly acted as if technology, as in order-matching, could somehow substitute for market-making. It isn’t hard to imagine that there are times when lots of investors want to dump holdings for reasons good or bad (such as news, rumors, or a wave of automatic trade execution resulting from too many people using models that are too similar. The first time we saw that movie was in the 1987 stock market crash). Economist Paul Davidson observed:

Davison discusses as some length the LPT view versus the EMT view, which lies at the foundation of the fondness for (among other things) the push for more automated trading. Davidson points out, but frankly could have said it more boldly, that EMT has a “continuous markets” assumption, that liquidity is ALWAYS there. And why is that assumption deemed to be reasonable? Because prices are assumed to reflect fundamental values, and those values can be estimated because risks can be modeled. To put this in slightly oversimplified layperson terms, people can agree on prices because there is a presumed rational basis for pricing:

Davidson points out that the future does not confirm to nice, neat, statistically “normal” (as in “tractable” or Gaussian) risk models and forecasts. Uncertainty is far wilder and efforts to model the extremes break down. This is the critique made by Nassim Nicholas Taleb, most famously in his book The Black Swan, following the work of mathematician Benoit Mandelbrot. Davidson refers to those types of risks as “nonergodic” as in not having a propensity to self-correct. And look where understanding the true nature of market risks takes us:

What we’ve had since the 1980s, in addition to a belief that more liquidity was ever and always better, was a policy preference in the US for moving financing for business out of banks and more and more into capital markets, out of the view that bank capital was too costly and capital markets were more efficient. And we’ve seen, stealthily and overtly, that the central bank indeed has used its liquidity muscle to backstop markets. Greenspan was widely heralded for jawboning (and more) in the 1987 crash; we’ve seen how the Fed ran to support all sorts of credit markets during the crisis. So Davidson is simply making a clear-cut statement of where we are.

But are all these markets deserving of support? And per Heimark, have we made market structures so much worse as to make them even more dependent on central banks than they need be? We have evidence of this in the Financial Times story. The focus is a mini Treasury melt-up on October 15, when the yield on the 10 year bond dropped from 2.38% to 1.86% in a bit more than an hour. The article perversely treats this as a “meltdown” when this is in fact a “gap up,” as in lower bond yields = higher prices. But only when shorts are getting killed are price increases depicted as a bad thing (click to enlarge):

I have to confess I also find it hard to get worked up about big price swings in a trading day. Investors have become complacent about market risk thanks to the tender ministrations of central banks since the crisis; October 15 was a rude wake-up call.

The article goes on about “seven standard deviation moves” as proof that Something Went Horribly Amiss. It might be more accurate to say that these metrics illustrate the fallacy of using “normal” distributions to measure trading risk. Let us remind you how badly these models anticipated the market action during the financial crisis. As Paul De Grauwe, Leonardo Iania, and Pablo Rovira Kaltwasser pointed out in “How Abnormal Was the Stock Market in October 2008?“:

We selected the six largest daily percentage changes in the Dow Jones Industrial Average during October, and asked the question of how frequent these changes occur assuming that, as is commonly done in finance models, these events are normally distributed. The results are truly astonishing. There were two daily changes of more than 10% during the month. With a standard deviation of daily changes of 1.032% (computed over the period 1971-2008) movements of such a magnitude can occur only once every 73 to 603 trillion billion years. Since our universe, according to most physicists, exists a mere 20 billion years we, finance theorists, would have had to wait for another trillion universes before one such change could be observed. Yet it happened twice during the same month. A truly miraculous event. The other four changes during the same month of October have a somewhat higher frequency, but surely we did not expect these to happen in our lifetimes.

But sadly, that type of risk modeling been institutionalized with the widespread use of VaR, as in Value at Risk models, not just for banks’ own convenience, but for regulatory purposes.

But what is mind-boggling, given the foregoing discussion, is that supposedly savvy investors were shocked. From the Financial Times article:

One hedge fund manager recalls being bewildered by subsequent events: “What on earth was charging through the market to want volume at such a price and why, in response to that catalyst, did the electronic marketplace just take any and all liquidity away?”

Without going through the details in the article, the explanation is really simple: a huge order imbalance, in this case, more buyers than sellers, driven by a combination of “flight to safety” purchase plus technical buyers, as in shorts who’d bet that the end of QE would lead to higher interest rates (lower bond prices) not being able to take any more pain, and closing out their positions. And there was a dearth of opportunists on the other side because so much trading is now electronic. Prices moves so quickly as to outpace normal human reactions to unexpected events. This was similarly a prime driver of the 1987 crash, when order execution moved at a snail’s pace compared to today:

The head of trading at a major dealer-bank says: “Once volatility shows up, you don’t want to make a mistake in a fast market and so you always see dealers pull back from providing prices.”…

The two main electronic trading venues for US Treasuries are run by Nasdaq’s eSpeed and Icap’s BrokerTec. In recent years these platforms have opened up to a range of broker-dealers and high-frequency traders. These firms do not underwrite US Treasury debt sales and are often viewed as opportunistic – providing prices when they spot a quick profit and then retreating when trading turns tricky.

Customer orders are now transacted and almost instantaneously hedged, or offset, by computer systems – a type of automation that works well when trading is orderly but rapidly breaks down when the situation changes. At such moments, turning off the machines becomes a necessity. This contributed to the downdraft in liquidity on October 15.

“Dealer-banks don’t really position in bonds,” said one head trader at a large US bank. “They basically act as a pass-through to places like BrokerTec and eSpeed or match off their client flow. The market-makers in this new market are not obligated to be there when everyone’s selling.”

Now while this sort of thing also happened in the stone ages of my youth (the Treasury market did come close to not trading in the worst of the 1987 crash; the Fed called the Bank of Japan and basically told the BoJ to hop to it. The BoJ told Japanese banks to start buying, and they did). But generally speaking, in bad markets, rather than prices gapping down, you were simply prevented from dealing. Your friendly bond dealer wouldn’t pick up the phone, unless it was Salomon or Bear, that staked its reputation on making a price even in really bad markets.

And that despite the now-widely-accepted view that investors have a right to liquidity, dealers not answering the phone during a panic actually is salutary. It’s an unofficial circuit breaker, exactly the sort of device that operates formally in stock markets when market moves during a day are deemed to be so large as to reflect panic. The circuit breaker allows the more speculative-minded buyers to assess what the hell is going on that led to the meltdown or up, and step in when markets re-open. It also gives the panicked seller time to cool off and determine if dumping investments at already stressed levels really is warranted.

At least this article does place primary blame on the increased role of high-speed trading and electronic order-matching. It also argues that regulations play a part, in that banks are no longer allowed to engage in proprietary trading. Without belaboring that issue, this is one of the few areas where using VaR as a measure of trading risk would be helpful, and allowing banks to take outsided positions on one side of a market or other to facilitate trading that they’d be required to “flatten” in a reasonable period of time (to be determined with further analysis, but the idea would be more like days, rather than much longer periods that many prop positions were built and then liquidated. As we’ve written, this was a serious flaw in Volcker Rule implementation and does warrant being revisited).

Where does this leave us? We intend to pursue this topic in future posts, but one take-away is that dealers are already petitioning the authorities informally for an intervention because they anticipate a rerun of a classic bond bear market, where investors who are long dump positions in order to stem losses (bond markets are so large that there is not remotely enough credible hedging capacity for a major player to go market neutral, let alone short. Recall Goldman which did try to get out of the way of the mortgage meltdown early, was short only in certain mortgage portfolios; it was still net long mortgage backed securities overall. And remember AIG was effectively bankrupted merely over providing hedges to one product, high grade CDOs, plus making bad subprime bets in its securities lending operation).

An exit is almost certain to be more violent if the Fed ever does tighten because at super low bond yields, a 1% interest rate increase results in vastly greater losses in principal value than if prevailing ten year bond yields were at, say, 4% or 5%. We think the outcome is more likely to be that the Fed is unlikely to make a serious exit of its artificially low yield posture, unless that process is extremely attenuated. But it’s not impossible, as in 1997 Japan, that the central bank comes to believe its own PR about the state of the economy and starts down the removal of low interest rate punchbowl path anyhow. And so the market nervous nellies may be proven correct, that the Fed will unleash a deluge. Whether it is just a taper tantrum redux or something more serious remains to be seen.