.
.--.
Print this
:.--:
-
|select-------
-------------
-
The World According to Mark Rubinstein

Mark Rubinstein is the Paul Stephens Professor of Applied Investment Analysis at the Haas School of Business at the University of California at Berkeley. He is renowned for his work on the binomial options pricing model (also known as the Cox-Ross-Rubinstein model), and was a founding partner of Leland O’Brien Rubinstein Associates, a firm that developed portfolio insurance strategies for institutional investors. His many publications include Options Markets as well as more than 50 publications in leading finance and economic journals. His most recent effort is Derivatives: A PowerPlus Picture Book, an electronic textbook available at www.in-the-money.com. He spoke with editor Joe Kolman in May.

Derivatives Strategy: Did the Long-Term Capital Management affair remind you of anything in your experiences with portfolio insurance at Leland O’Brien Rubinstein during the 1987 crash?

Mark Rubinstein: It sure did. In fact, the resemblance was uncanny. In both cases, the potential difficulties for the economy were enormous; and in both cases, the Federal Reserve felt it necessary to call investment bankers or corporations together to save the day.

Another similarity was the origin of both situations. Both were rare events. We used to warn our clients that the only thing that could do us in would be a stock market crash—that our strategies wouldn’t work if the market made a sudden large move either up or down.

It was the same thing with LTCM. LTCM thought that all of its individual investments made sense. They were hedges, for the most part, and they were diversified internationally across different kinds of instruments. It was difficult for the firm to see how all its investments could go down together. And yet that’s more or less what happened. The fund didn’t have the diversification it thought it had. Neither did portfolio insurers in 1987—they were implicitly relying on the diversification provided by a widely dispersed portfolio of equities.

Still another similarity was that we were blamed for the very events that undid us. That was the same for LTCM. The hedge fund had become too large relative to its markets, and the investments turned out to be much more correlated than was expected. They weren’t fundamentally correlated, but the investments became correlated because of the activities of other traders in the market. So when things began to get bad, other investors began to sell out of the same strategies that LTCM was in—and that caused the markets to move in the same direction, even though they weren’t fundamentally tied together. LTCM’s positions were so large that they could not be liquidated. Similarly, the portfolios that were being “insured” in 1987 were so large that it was impossible to sell enough futures in such a short period of time to keep the portfolio insurance strategy on track.

The ironic thing in our case was that our very strategy may have been the vehicle that helped create a new situation in which an extreme event could occur. For me, as a simple professor at the University of California at Berkeley, it was difficult to believe that I could have been involved in a small company that had such an incredible effect on the markets.

DS: What exactly was your strategy at the time?

MR: Portfolio insurance requires that we gradually sell out of the market for our customers as the market falls and buy in as the market rises. By selling out, for example, we would place a floor on how much a client could lose if the market went down. On the upside, the client would get a good portion of the increase in the market if it went up. So naturally, when the market began to fall, we were forced to sell.

At first glance, nothing seemed to have caused the market crash. People had trouble figuring out what had happened since no news at the time seemed able to justify what had occurred. But when the dust settled and the U.S. government commission finished its study, the commission argued that the very strategy we were advocating and following was largely responsible for the crash. The theory was that massive selling by portfolio insurers was responsible for most of the decline on October 19, 1987.

DS: How big was your selling? Do you believe it was the spark that moved the whole market down?

MR: It’s difficult to say. Some highly respected people believe it was. But in my own opinion, that is probably wrong. There’s no doubt that portfolio insurers sold large quantities of stock and index futures on October 19. To put the matter simply, anyone who sold on net that day was partially responsible. My estimate is that portfolio insurers sold about 20 percent of the overall stock and index futures that were sold. But others sold the remaining 80 percent. Portfolio insurance was certainly one of the causes of the crash, but perhaps not the most significant one.

DS: How much money were you managing at that point?

MR: We managed about $5 billion or $6 billion directly, in our own name. But the software we had licensed to others was probably controlling another $55 billion. We estimated that another $35 billion of competitors’ money was following a similar strategy as ours, since our success attracted other people to do the same thing. In a sense, we were ultimately undone by our own success.

DS: What do you think is the most common misconception people have about this whole event?

MR: Some people felt that the crash discredited the whole strategy of portfolio insurance. I definitely don’t believe that. The key feature of portfolio insurance is that we can say to our clients, If the stock market behaves this way, then this will be your return. In other words, if the stock market goes down, you’ll break even; but if the stock market goes up, you’ll get 90 percent of the increase. It’s a risk-control strategy that has general appeal.

At the time, the short maturity of exchange-traded options and their position limits severely restricted their use. Because we had extremely large portfolios and our customers wanted long-term insurance, we had to implement the strategies dynamically—by buying as the market went up and selling as it went down. We managed portfolios of index futures contracts dynamically.

“Some people felt that the crash of 1987 discredited the whole strategy of portfolio insurance. I definitely don’t believe that.”

I don’t think the stock market crash demonstrated that dynamic hedging or portfolio insurance is useless. If you believe that jumps in market levels will make dynamic trading difficult, then you’ll find portfolio insurance implemented through dynamic hedging less useful. To the extent that investors took this lesson away from the market crash, this should have reduced the use of these strategies. But it certainly shouldn’t have killed them.

DS: Portfolio insurance isn’t really dead, is it?

MR: I see traders today who don’t use the phrase portfolio insurance, but who do other things that are roughly equivalent.

DS: They call it dynamic trading.

MR: It often takes the form of a dynamic trading strategy that effectively tries to insure a portfolio by selling out when the market goes down. Traders have, presumably, done that for hundreds of years. So I think it’s still alive.

DS: Here’s what strikes me as particularly strange about the LTCM fiasco. Doesn’t everybody know that correlations fall apart during periods of market stress? And wouldn’t LTCM have known that everybody was doing similar carry trades or similar spread trades, and that the market would be under stress if these things fell apart?

MR: I think it’s fair to say that, yes, it’s something we both should have known about. In our case, we knew that if the market moved that way, the strategies wouldn’t work. We simply didn’t think it would happen. The same thing was true for LTCM. It was a conceivable scenario, but it was quite unlikely.

When we did our simulations all the way back to the Great Depression, there was never a situation that could have posed a really serious problem. We weren’t expecting a move like that. We used to wonder what could cause such an event. It would either be World War III, a Martian landing or something else really incredible. We didn’t think it could be anything that was not clearly fundamental—or, even worse, our own self-defeating strategy.

DS: How were you using stress tests to assess your risks?

MR: We would go back historically, through the Great Depression, and run our strategies over those 60 years. We found that our strategies worked the way they were supposed to almost all the time. Once in a great while, there would be a significant deviation in the return we were supposed to get, but nothing similar to what we saw in the 1987 stock market crash.

There wasn’t an event like that crash in the historical record, so naturally it didn’t show up in our analysis. I think that’s a problem with stress testing in general. How do you form your opinion about what a nasty event will be? Usually, you look at the past and assume it can’t get any worse than it has been in the last 50 years. Then you say, “OK, that’s my worst outcome.” But the fact of the matter is, the world changes.

DS: LTCM claimed that nothing in the data could have helped it predict what happened.

MR: I think LTCM is saying essentially the same thing that we thought. We thought—if we ignored our own influence—that there was nothing in the data in the last 50 years that should have caused us to be worried.

DS: In both cases, there was a failure to appreciate the dangers of potential illiquidity.

MR: I think it was certainly something we failed at, although we simply assumed liquidity wouldn’t be any worse than it had been over the last 50 years. I think modeling liquidity is difficult. You have to realize that markets are changing and that an illiquidity event that’s quite different from anything on record in the past could nonetheless happen in the future.

DS: When LTCM modeled its risk, the fund claimed that the chance of a 50 percent drop in its net asset value was one in 10 to the 9th power.

MR: That’s right. I did a similar study of the stock market crash. Let’s suppose you believe the standard finance model, which states that the returns of the stock market are lognormally distributed. And suppose you believe the 70-year historically observed standard deviation, which is about 20 percent per year. In that case, a one-day decline of 29 percent, which was the decline of the Standard & Poor’s 500 futures price, wouldn’t happen in the life of our universe, which is 20 billion years. Indeed, it wouldn’t even happen if you were to live through 20 billion of those universes.

“How do you form your opinion about what a nasty event will be? Usually, you look at the past and assume it can’t get any worse than it has been in the last 50 years.”

If you’re using those kinds of models, you are going to discount completely the possibility of such a big crash in a single day. But we can look back historically and realize that there are larger negative and positive events than would be expected from a lognormal market model. Still, we’ve never seen anything as large as what we observed on October 19th.

DS: But if your model showed that you would go bust only once in every 20 times the life of the universe, wouldn’t you begin to wonder if you were doing something wrong?

MR: Yes, but people at LTCM, for instance, could have said to themselves, Let’s say we’re wrong by 10 orders of magnitude—we’re still fine. Or they could have guessed that the chances were one in a million. LTCM was on the right track. It was making enough money to make a good risk/return tradeoff. But if it thought the chances of catastrophe were one in 50, then it would have begun to be less sure.

DS: It seems like lighting struck twice.

MR: That’s how I look at it. It certainly struck twice in my lifetime. I’ll mention another analogy that helps with the LTCM case, and a little less with the case of portfolio insurance. LTCM claimed it had found markets that were inefficient, in which the return per unit of risk was higher than it should have been. That may be true, and LTCM may yet prove it to us. But you can also take the view that the profits the fund was making were similar to what would happen if you went to Las Vegas and played roulette.

Let’s suppose, for the sake of this discussion, that the roulette wheel only has numbers one through 36 on it, and that it doesn’t have a zero or a double zero. If you place one chip on each of the numbers from one through 35, you will win a small amount most of the time. Now suppose you borrowed most of the money to finance the 35 chips you bet. You’d find that most of the time you’d earn a high rate of return on your capital, but that, on average, one out of 36 times you would lose it all.

That’s the way you can think about LTCM, if you wanted to make its strategies consistent with an efficient market. LTCM was drawn into situations in which there was little apparent mispricing. The fund would hedge to take advantage of that slight mispricing, and would then leverage itself. But if you consider the fact that you can land on 36, there isn’t any mispricing. I don’t know enough about LTCM’s markets; in our case, when people looked at portfolio insurance they thought it would work reasonably well, but they ignored the possibility of landing on 36. Once they realized that, the interest in portfolio insurance lessened.

“You have to look at the strategy and ask, Can I think of a stress test in which everything goes wrong? Then you have to go further and be creative in thinking about the future.”

The general problem with one–35 and 36 is one of evaluating investment strategies. A lesson we can draw is that when we look at the performance of an investment manager, we now have to ask ourselves, Is this a strategy that wins 35 out of 36 times, but really blows it the 36th time? Because of this problem, it’s easy to be deceived by a manager’s track record.

DS: What measures would you use to avoid that problem? Standard methodologies such as stress testing and Sharpe ratios?

MR: The problem is that the managers we’re talking about would have sailed right through those tests because of their reliance on past data. I don’t have a silver bullet for this. In some cases, you can catch it because it is obvious—for example, all a manager may be doing is selling out-of-the-money put options.

You have to look at the strategy and ask, Can I think of a stress test in which everything goes wrong? Then you have to go further and be creative in thinking about the future. What kinds of things can happen? Are there situations in which liquidity will dry up and I won’t be able to get out of my positions? And then you have to be even more creative than that. Because the next event may not be a liquidity event. It may be something else.

Was this information valuable?
Subscribe to Derivatives Strategy by clicking here!

--