Program Trading
back to 1987 crash page

The blame for the crash was initially placed on program trading. Program trading is a computer-assisted method to trade markets, and can be divided into two types. The first - index arbitrage - takes advantage of discrepancies between markets by simultaneously buying in one and offsetting that purchase with a short position of the same size in another closely linked market. Small risk free profits are the result. In the context of the equity markets, an arbitrager would buy an S&P future in Chicago and simultaneously sell all the stocks in the S&P Index in New York when he or she noticed that index futures were undervalued relative to the basket of stocks. Such trades involve hundreds of securities and millions of dollars, and require timing that only computers can accurately achieve.

Portfolio insurance - the sale of stock index futures to protect the value of a falling stock portfolio - is the second strategy referred to when talking of program trading. In the 1980s fund managers bought insurance from portfolio insurers. The job of the insurer was to sell futures on behalf of fund managers whenever the markets headed downwards. Then, as the hedge fund manager's portfolio lost value in a falling market, his or her profitable short position in the futures market would compensate. In theory, a manager was guaranteed that his holdings would never lose more than a certain percentage of their value. Protecting portfolios on the downside with futures was preferable to reducing the size of the portfolio by selling in the stock market because futures markets are very liquid, many stocks aren't, and transaction costs low. It was found that portfolio insurance strategies accounted for some 12% of the selling on October 19

Conventional wisdom has it that the stock market and the futures market had been growing increasingly related, this interaction between the two never so present as the day of the crash. Commentators believed that computers sped up the flow of information and trading and brought futures markets and stocks closer together, these efficiencies both promoting volatility and accelerating markets declines. A potentionally vicious circle was theorized. "If stock futures moved to a discount to stocks, for whatever reason, index arbitragers would purchase futures and sell stocks. If stock prices fell far enough, portfolio insurers would sell futures contracts, creating a new discount in futures markets, leading to further index arbitrage stock sales, and so on into a downward cycle of self-reinforcing selling." U.S. Securities and Exchange Commission, Division of Market Regulation. Report on the Role of Index-Related Trading in the Market Decline on September 11 and 12, 1986. The program trading theory blames the structure of markets, not psychology or events. The circular process it posits explains why markets may have fallen as far and fast as they did, giving responses to two of the important questions any 1987 theory must answer.

Program trading had earned a bad boy's reputation long before the crash. Many people blamed the volatility they perceived in the early 1980's on computers, which had only just started to become more important on Wall Street. Large percentage declines in September 1986 and January 1987 had been attributed to arbitragers and portfolio insurers. Thus, when the market suddenly crashed in October 1987, the world had an easy and ready whipping boy - program trading. It is hard to figure how much of this response was simply the authority's attempt to find a scapegoat as quick as possible, therefore alleviating the public's fear, and how much of the response was an honest attempt to understand what actually happened. Computers and technology always make an easy target, though they are not necessarily a good target. Remember that in 1929, markets collapsed with no computer trading or portfolio insurance.

There are several problems with the program trading theory that render it suspect. First, there is the causal problem. If excess selling by program trading strategies was the cause of the 1987 crash, what caused these programs to initiate sells in the first place? In the SEC's hypothetical cascade scenario I outlined above, stock futures must move to a discount to stocks before arbitragers would sell stocks, initiating a crash scenario. But what caused this initial discount between futures and stocks to emerge? The fallacy of those who blame program trading is their use of a circular argument. They attempt to explain a large price move (the crash) by introducing a process that itself requires a large price move to have already occured. One does not enter into a self-perpetuating process like this by chance, something initiates the entrance into the process. If anything, this is the cause we are looking for, though those who pillory program trading have ignored it. By focusing only on one of the neans by which markets fell, people have been ignoring the larger motivation and getting a far too simplistic notion of the crash.

A second fault in the program trading argument can be found in the order of events. In the context of the equity markets, an arbitrager buys an S&P future in Chicago and simultaneously sells all the stocks in the S&P Index in New York when he or she notices that index futures are undervalued relative to the basket of stocks. A portfolio insurance program sells futures in Chicago as it notes declines in New York. Before the New York Stock Exchange had opened on Monday morning, Chicago traded S&P futures were already down 7% on weekend trading - already a larger decline than any day since 1940. Because the strategy of a program trader is to simultaneously sell futures and buy stocks, they would have been active only when New York and Chicago were both open. The first bout of futures based selling over the weekend and early Monday, the early stage of the crash, occurred only in Chicago - the arbitrager was in the midst of a relaxing weekend. Thus, the crash was already well underway before the program trader arrived on the scene. Some other reason must exist for the first stage of the decline.

In his paper The International Crash of October 1987 Richard Roll criticizes the authorities' almost complete focus on U.S. structural features such as program trading as the cause of the crash, even though there is virtually no evidence to support such a view. "The institutional structure of the U.S market cannot have been the sole culprit, or else the market would have crashed even earlier. There must have been an underlying 'trigger'", writes Roll. "Some have pointed to the U.S. trade deficit, to anticipation about the 1988 elections, to fears of a recession. But no one has been able to substantiate convincingly the true underlying cause of the October decline."

Roll turns to the behaviour of international markets prior to and during the crash in order to better understand the event. He found that the crash was an international phenomenon that caught up markets in Asia, Australia, Europe, and North America without exception. The first point of importance is that significant program trading activity only existed in the United States, not the rest of the world. If program trading drove U.S. markets down 22%, why would markets in other countries without program trading fall by that amount? Those who blame program trading save themselves from this line of reasoning by stating that international markets were simply all watching and reacting to a leading U.S market. Roll shows us this is not a valid argument. Though the biggest daily decline in the U.S that month was on Monday, October 19, Roll found that many European markets split their declines between their 19th and their 20th. Thus, given differences in time zones, parts of Europe started to crash before the U.S. did. Hong Kong, Malaysia, and Singapore also had their major movements on the 19th and 20th, thus preceding the New York crash by more than 12 hours. Hong Kong closed 10% lower even before New York had opened. Given these observations, some underlying factor emerged as the sun rose on the markets on October 19th, sweeping westward and engulfing market after market. If anything, the U.S. market lagged the intital price movement that began that Monday, contradicting the widely-expressed opinion that U.S institutional defects such as program trading pulled down the others on the 19th.

Not only did New York lag temporally, it also lagged in terms of price performance. Out of 23 markets surveyed, the United States had the fifth smallest decline, ie. the fifth best performance. Hong Kong, a market without significant program trading, declined 46%, turning in the the worst peformance. The average decline of five countries in which computer directed trading was prevalent was 6.6 percentage points less than the average decline of the 15 countries where it was not prevalent. Why would the U.S., the market that had been credited as the cause of the crash, have one of the smallest percent changes that day? In the end, Roll found that the prescence of forward trading on the particular exchange did not relate to the extent of the decline. Ironically, computer directed trading was associated with less severe stock market declines, ie. it actually led to less severe crashes. This makes sense - arbitrage tends to lower volatility and fix market imperfections, not create them.

In the end, program trading seems to be an incomplete explanation. Of the five questions a potential hypthosis for the crash must answer, it succeeds at 2 and 3 - the size and speed of the crash. If the event was purely based on the market's faulty structure, then a spectacular market move could have been caused by the system's failure. The theory fails to answer questions 1 and 4, why the crash occurred specifically on October 19 and why it was international in nature, not U.S. specific. Program trading had been in existence since the early 1980', why didn't markets crash in 1985, or 1986? Roll's arguments reveal the theory's failure to explain worldwide decline in markets. Furthermore, the program theory's circularity leaves something to be desired - it requires an initial decline to explain a final decline, focusing on the process by which the final decline might come about but forgetting to give an account for the initial decline.


  JPK and Lope
2002