A market ripe for a fall
Last Thursday the market experienced a sudden air pocket, with the S&P index temporarily dropping by over 100 points. In real-time, it certainly looked as if a crash were underway. At the low point, a record one day intra-day loss was recorded. As is now known, a mixture of errors and glitches contributed to the situation, and it is also known that several HFT (high frequency trading) firms simply stopped trading after the DJIA had fallen by 500 points. Apparently, some of these firms have a rule to withdraw when 'disorderly market conditions' pertain.
Let us be clear beforehand, the stock market was more than ripe for a sell-off. In the weeks leading up to the late April high numerous sentiment indicators went to rarely seen extremes. Especially equity option traders (both large and small) had begun to buy more and more call options for several weeks while refraining from buying protective puts, and various polls (such as Investor's Intelligence) showed an increase in bullish opinion on a par with the bullish consensus seen at major market tops in the past. Lastly, the mutual fund cash-to-assets ratio fell to a new all time low, clocking in at barely above 3% in March. This particular datum is in fact a major medium to long term bearish indicator, and a strong warning that the secular bear market is likely to continue and enter its next phase of liquidation.
The CBOE equity put-call ratio – click on chart for a larger version of the image
The Investor's Intelligence poll – at the end of April, over 55% of advisors were bullish, and less than 20% bearish. This is a very extreme reading – click on chart for a larger version of the image.
The mutual fund cash – to assets ratio – a new all time low of 3,4% as of March 2010 – click on chart for a larger version of the image.
This surfeit of bullish sentiment was certainly aligned with the trend, but that trend had become very stretched. The stock market has been strenuously overvalued for quite some time (more on the stock market valuation issue in a follow-up post). Insiders have been selling hand over fist since the summer of 2009, with nary any buying detectable. In short, the market was in a situation where a period of abrupt and large losses was likely to occur soon. In the trading week prior to the mini-crash, volatility suddenly picked up, a classic pattern of distribution.
The SPX prior to the mini-crash: A pattern of distribution. – click on chart for a larger version of the image.
Bulls made the same erroneous assumption that preceded the 2007-2008 decline: namely that the market's capacity to ignore any and all potentially bearish fundamental news for a while was a sign that these developments 'didn't matter'. A plethora of articles in the financial press suggested for instance that the sovereign debt crisis engulfing the Euro-area could be safely ignored by investors – in some instances it was even argued that the bearish developments in Europe would be bullish for financial assets in the US. This is of course utter nonsense.
Trading systems go haywire
As mentioned above, several HFT firms ceased trading during the market crash in medias res. As the :
|A number of high-frequency firms stopped trading Thursday in the midst of the market plunge, possibly adding to the market’s selloff. Tradebot Systems Inc., a large high-frequency firm based in Kansas City, Mo., closed down its computer trading systems when the Dow Jones Industrial Average had dropped about 500 points, said Dave Cummings, founder and chairman of the firm. Tradeworx Inc., a N.J. firm that operates a high-frequency fund, also stopped trading during the market turmoil, according to a person familiar with the firm. Mr. Cummings said Tradebot’s system is designed to stop trading when the market becomes too volatile, too fast. “That’s what we do for safety,” he said. “If the market’s weird, we don’t want to compound the problem.” Tradebot says it often accounts for about 5% of U.S. stock-market trading volume. The withdrawal of high-frequency firms from the market didn’t necessarily cause the downturn, but could have added to it, some market experts say. A number of high-frequency firms closing down in the midst of a sharp market drop can “widen markets out substantially,” said Jamie Selway, managing director of New York broker White Cap Trading.|
An intra-day chart of the mini crash – each candle represents 3 minutes – the period when the bids were suddenly pulled is clearly discernible – click on chart for a larger verion of the image.
An intra-day chart of venerable Dow stock Procter & Gamble – n.b. this is considered a 'defensive stock' – another crash lasting 3 minutes and reversed just as quickly – click on chart for a larger verion of the image.
According to this Wikipedia article on quantitative trading strategies, HFT recently accounted for more than 70% of all trading volume on US stock exchanges. Does it matter when these firms stop trading and pull all their bids? You bet it does. In discussions with several friends in recent months, I have dubbed this phenomenon the 'fully automated crash' (FAC). My argument centered on several aspects of the market structure. For one thing, the claim made by HFT firms that they have effectively replaced the 'specialist system' and by providing liquidity have actually helped to keep the market stable, never sounded credible to me. These firms use algorithms for trading, designed to arbitrage away what are often tiny price differences. My question was always: What if all of these algorithms want to sell at once, with none of them bidding for shares? Is that an unthinkable situation?
We now know that it can indeed happen. In addition, there are other quantitative trading strategies employed by dedicated hedge funds in the market, most of which use some combination of technical analysis and empirical historical data to determine which trades to execute.
The most salient fact about these quant strategies is that none of them have anything to do with 'investing' in the classical sense of the term. While every system is slightly different from its peers, the main inputs of these systems still tend to increase the herding effect – this means that e.g. favored stocks or commodities will see their trends magnified, adding to the 'normal' herding effect observed in financial markets, and eventually exaggerating the overvaluation of these assets. The other side of the coin is that when a sudden market decline happens, all these systems will sell or attempt to short the market at once.
Once again, apart from systems that use mean reversion as their major input, this will lead to a 'bidless market' and exacerbate declines. The final point to consider are the option markets. Since the crash of 2008, overall trading volume in the stock market has steadily declined. At the same time however, open interest in the option markets has raced from record to record. The increases are by no means small, as you can see when looking at the , the major options exchange in the US.
Consider now how option writers protect themselves against adverse market movements. The most prevalent strategy is referred to as 'dynamic delta hedging'. Delta in options terminology is also fittingly known as the 'hedge ratio' – it compares the price change in the underlying security to the corresponding price change in the derivative. The objective of dynamic delta hedging is to become 'delta neutral' if adverse price movements endanger the position of an option writer. In practice, these strategies are also executed by algorithms, whereby for instance the writer of a naked put will go increasingly short the underlying security if the security falls and threatens to slice through the strike price of the written put.
Conversely, the writer of a naked call will go increasingly long the underlying security if it rises and approaches and eventually breaches the strike price. These strategies are adapted to various parameters such as the premium received by the options writer, but essentially they all aim to offset risk by taking a position opposite to the options position – and they do it dynamically. Naturally, this form of dynamic hedging will then also magnify price movements, especially when these price movements are very large and many strike prices with large open interest are sliced through in a short period of time. Given that trading volume in equities has steadily trended down, while options open interest and trading volume has trended steadily higher, the potential for dynamic delta hedging to exacerbate both sudden and large up and down moves in the market has increased markedly.
Panic of the machines
Considering the prevalence of quantitative trading strategies in today's marketplace, human emotion, which has been the main driver of market movements in the past, has at least in part been superseded by the 'greed and fear of machines'. Naturally, the computers doing the trading do not have emotions – that is precisely the reason why quant strategies have become so popular in the first place. It is held that they remove the human error factor from securities trading, by maintaining cold, pre-programmed objectivity in all situations.
This conviction is a close cousin to the conviction back in the late 1980's that 'portfolio insurance' using S&P futures to protect a basket of stocks dynamically with a likewise fully computerized procedure, could in fact insure stock portfolios against losses. Instead the combination of portfolio insurance and program trading (a form of arbitrage between stock index futures and the underlying baskets of stocks) combined to produce the biggest one day market crash in history in 1987.
Any market in which a specific trader has 'become' an overwhelming percentage of the positions in this market is bound for problems. HFT and other quant strategies are a form of trading that effectively transforms the many different firms engaging in it into a single entity, due to the strong similarities in the strategies employed.
After all, the algorithms used by the black box community have been programmed by humans, all of whom work with the very same set of historical data to come up with effective trading rules that can be back-tested and thus legitimized. Instead of eliminating human error, the computerized strategies are in fact multiplying it, by dint of all of them reacting in virtually the same manner to a given set of inputs. And so the machines are perfectly capable of reciprocating the mindless greed and equally mindless panic that characterize the emotion-laden trading of mere humans, and in fact, it appears that they are capable of exaggerating both.
Without a doubt, quantitative strategies are here to stay. There is growing investor demand for funds employing systematic strategies, and the marketplace has reacted to this growing demand by supplying ever more sophisticated systems that are on the whole highly successful. The panic of last Thursday will likely lead to new regulations and restrictions aiming to blunt the impact these strategies have on the market, but it can be safely predicted that not a whole lot will change in principle – the exchanges would be up in arms if any putative restrictions to electronic trading strategies were deemed too onerous, as they create a good slice of the exchanges' income.
Given that a huge sub-industry has evolved in the hedge fund industry and the CTA (commodity trading advisors) space that revolves around systematic strategies, there are now sizable vested interests that would be threatened by such regulatory action. Besides, we have empirical proof galore that as soon as one potentially market-destabilizing practice is curtailed (such as the curbs on program trading introduced after the 1987 crash), other strategies with similar characteristics will soon take their place. In any case, the market demand for such strategies will only continue to prevail as long as these strategies work as advertised. To wit, portfolio insurance died a sudden and quiet death after the 1987 crash.
There was no need to curb it – the market itself curbed it, by proving spectacularly that it didn't work. So far, most quant strategies seem to work just fine, hence demand for them is bound to continue to be strong. However, investors and traders need to be aware of the potential impact these strategies can have on the market. We may well one day wake up to a market dislocation unlike any other witnessed in market history to date.
A final point I wish to make is that in Thursday's trading another lesson was taught yet again: in a panic situation, when trading volume explodes and prices fall at great speed, the infrastructure that is supposed to handle transactions is prone to failure. It suddenly takes ages to get confirmations of trading executions, and as the many in the meantime reveal, posted bids and offers can fail to reflect reality. For instance, Accenture (AGN) and several other stocks and ETFs at one point traded at plainly absurd prices (AGN went from about $45 to 1 cent in the space of a few minutes).
In other words, if you are exposed to a lot of risk, don't fool yourself into thinking that if anything untoward happens, you will be able to 'get out in time'. As always, market liquidity disappears just when it is needed most, and the backbone of our famed electronic trading spaces is less resilient and reliable than is generally assumed.
Charts by: StockCharts.com, schaeffersresearch.com, sentimentstrader.com.
Thank you for your support!
In case you prefer to donate bitcoins, the address is: 1DRkVzUmkGaz9xAP81us86zzxh5VMEhNke
Follow us on Facebook!