Eric Burroughs's Profile
Warning signs on market liquidity risks
If the great commodity selloff of 2011 shows nothing else, it is that markets are undergoing serious structural changes that need to be followed closely. Our commodities analyst John Kemp has compared the oil plunge with the May 2010 flash crash in U.S. shares, and rightfully so. Four standard deviation moves in oil futures are not normal, even if Gaussian distributions underestimate the chance of such a move. The rise of high-speed electronic trading appears to be creating imbalances between buyers and sellers in nanoseconds that lead to outsized moves. It would probably be manageable if these problems were tied to smaller markets not so correlated with the rest of the world. But in this age of highly correlated global markets, these changes matter and need to be better understood — both by market participants and regulators.
One curious outcome about the rise of algo-driven trading is the volume is not leading to better liquidity, especially in these flash crashes. Liquidity — defined as the ease with which trades can take place without causing a major price impact (and not referring here to overall bank liquidity/funding risk) — appears to suddenly vanish in some of these big market moves, leading to massive swings.
Liquidity is a fundamental part of market dynamics — how to think about where to buy and sell — and is almost always at the root of market stress. On a day to day basis, the dynamic is fairly natural. Traders closely follow chart levels, such as Fibonacci retracements, in part because stop-loss orders can accumulate around those levels, leading to sudden lurches — a relatively illiquid imbalance in trading as the burst of selling pushes bid/ask spreads wider. Such moves can define the trend of the day, change the near-term trend or even the long-term trend when they strike at a crucial moment. There’s no doubt that old-fashioned leverage, stop-losses, option-related gamma selling and the sort played a role, as our Reuters special report fleshed out. Yet those factors still do not explain some of the severe intra-minute and intra-hour volatility. There is nothing normal about these new kind of market moves. As much as bubbles are normal, and silver looked bubbly, there was no reason why the silver selloff should have had a bigger impact on oil than gold — and expose weaknesses in the trade of WTI and Brent futures, even with the leverage that had built up in those markets. Leverage alone does not explain this new brand of chaotic market swings.
There’s a few ways to guess what happened here. One is that algo-driven selling dominated to the point that it overwhelmed buyers and potentially even led other players to pull bids. Computers created the two liquidity crises the SEC described in its report on why the Wall Street flash crash unfolded the way it did, with the hectic HFT trading prompting bids to be pulled. As the SEC described it: “This sudden decline in both price and liquidity may be symptomatic of the notion that prices were moving so fast, fundamental buyers and cross-market arbitrageurs were either unable or unwilling to supply enough buy-side liquidity.” A perfect recipe for a price vacuum and severe downdraft.
What it suggests is that the computers are not providing liquidity, and in fact may be pulling liquidity in nanoseconds at just the time it is most needed. Air pockets seem to develop, leading to a flurry of offers and thinning bids — and that’s when the moves are the most severe. In the past year, flash crashes have now happened in benchmark U.S. equity indexes, oil futures and even dollar/yen (when the plunge to record lows happened on heavy structured product hedging in the least liquid hours of global FX trade). What happened to market makers? That’s a question that needs to be answered. It appears that computers are reading computers, knowing when others are pulling back and prompting other funds/banks to do the same (not to mention the occasional runaway algo).
Why so? This may be the impact of liquidaty-adjusted value-at-risk gauges, otherwise known as LVaR. LVaR makes a lot of sense: you want your trading desk and portfolio risk to be judged not just on standard deviations of volatility, but to also factor in market liquidity on the price at which you might be able to sell since VaR is all about minimizing potential losses. On a basic level, LVaR can incorporate unusual changes in bid/ask quotes — using standard deviations on mean bid/ask spreads — to give risk signals. The further the
mean bid/ask spread from the mean, the more dangerous the market conditions.
A 2007 paper published by New York University’s Stern School of Business explains the potential negative feedback loop from market participants using LVaR gauges.
“Subjecting traders to an LVaR gives rise to a multiplier effect. Tighter risk management leads to more restricted positions, hence longer expected selling times, implying higher risk over the expected selling period, which further tightens the risk management, and so on. This feedback between liquidity and risk management can help explain why liquidity can suddenly drop. We show that this ‘snowballing’ illiquidity can arise if volatility rises, or if more agents face reduced risk-bearing capacity— for instance, because of investor redemptions, losses, or increased risk aversion.”
This is a recipe not just for more severe market volatility, but also raising the risk of a market crash or shock with no economic origin which still inflicts economic damage. We’re going to have to look more into what the bid/ask data of the big selloff days show.
The other question is the role of ETFs. Some of the most volatile swings in silver and oil happened in less-liquid Asian hours. Asian trading is a time when any big buyer or seller would almost certainly avoid stepping in, knowing full well that liquidity is not what it is during New York and London hours. Computers obviously have fewer such compunctions. Metals traders in Asia blamed ETFs for the sharp silver selloff. To be honest, I don’t know how ETFs trade into market moves in a 24-hour market cycle, given the need to track index performance. It’s something I’ll be following up on. But are ETFs pro-cyclical traders in illiquid market moves? Are humans or computers making the trading decisions? Are the algos able to make split-second liquidity trading decisions? ETF prospectuses don’t give much clarity on how they trade, other than warning about liquidity as one market risk. But the rise of ETFs, especially leveraged ETFs, only appears to be adding to the liquidity risks now gripping markets. The Financial Stability Board recently noted the risks from ETFs, especially those that provide on-demand redemption and liquidity but deal in less-than-liquid markets (think silver): “The expectation of on-demand liquidity may create the conditions for acute redemption pressures on certain types of ETFs in situations of market stress, which could in turn affect the liquidity of the large asset managers and banks active in this market.” And of course flash crashes in commodity markets don’t just affect professional investors, thanks to ETFs.
A lot of smart firms have invested a lot of money into liquidity risk management. But these massive intraday market swings must be keeping a few risk managers up at night. Just as computer-driven portfolio insurance sparked the uneconomic 1987 crash, we are now starting to see clear signs that modern algos and ETFs are likely causing uneconomic market dislocations. A better understanding of the role of their impact on market liquidity is urgent.
One last note on silver. The clues for what happened that sparked the severe silver volatility, April 25 appears to be one of the days to focus on. That’s when volumes in the iShares Silver Trust (SLV) ETF suddenly surged and trading became more erratic. On that day, trading activity soared: the ratio of daily volume to the 30-day average shot up to more than 4 times, the second highest ever on data going back five years, and it would have been the highest if not for an odd one-day volume spike on Nov. 9 last year. Volume only swelled from there as the market peaked and turned. A pure guess in reading the charts is that the seeds for the ensuing chaos were sowed on April 25.