Sunday, May 9, 2010

The day the computers panicked PART 1

A real instance of science fiction

By now you probably know something about what happened Thursday. In case you don’t, let me make a citation simply because it states the obvious. The headline is “Computers, Not Human Error, Likely Caused Market Meltdown.” It happened simultaneously in multiple markets: commodities (e.g., gold and oil), bonds (e.g., one could easily track in real time in Treasuries, but other debt markets also), currencies, equities, and derivatives. The plunge went furthest on purely automated exchanges. Do you think it might have been computer driven?

But for the sake of thoroughness, let’s include a more extensive quote:
“Computerized sell programs triggered by global events-rather than trader error or "fat finger"-appear to have caused Thursday's unprecedented market swing, according to market pros who are reconstructing the nearly 1,000-point stock sell off.
These experts think the intensely accelerated electronic trading was sparked by the Greek debt crisis and other events and not a trader who typed a "b" for billion instead of "m" for million in executing a trade on Thursday.”

If the link’s still good, you can grab the article at:
But really, do you need it? Something, could have been “fat fingers” or Greece, set off a panic. It would be ironic if it was Greece because that would make the discussion of sovereign risk and the rush to post on debt on Wednesday seem prescient. However, for what follows, it is irrelevant whether it was “fat fingers” or Greece.

Let’s see who panicked. It wasn’t people. It happened too fast, in too many places, and in too many markets. Put bluntly, a computer or multiple computers panicked. In case you think computers don’t panic, remember this isn’t the first time. If one follows markets, one has seen a few computer panics. Think about October 1987 for a dramatic example, but it’s just one. Yes, computers can be, and have been, programmed to panic. There is probably someone programming a computer to panic right now. The people who do it are called “quants.”

This blog has pointed out the destructive impact of quantitative trading numerous times. Perhaps too subtly, but it was the topic of “On Quants” on March 9. On February 28 in “Is the Volker Plan shadow boxing or can it help?” the posting stated: “Quantitative trading did cause some of the contagion even if it wasn’t the root cause. But, hedge funds seem to have been the more important vehicle, not banks. In fact, hedge funds were probably instrumental in transmitting the downturn from debt markets to equity markets where they had an impact on a broader set of individuals through their 401k’s and IRAs.” And, again the thought scenario in a posting entitled “Beware the risk-free return” explained why quant funds blow up. Perhaps the posting should have been more explicit about the collateral damage they cause.

So, here is the science fiction. On Thursday a bunch of computers panic; after a minute or two and into Friday the panic spreads to people. Sound familiar? We know people can sense panic. Casual observation, history, behavioral economics, and the social sciences all have confirmed it. It seems computers sense each other’s panic, too. Unlike organisms, which, according to scientists, seem to have multiple methods of sensing fear, computers undoubtedly sense it from behavior. It also seems people can sense computer panic. Witness the reporting and reactions of people during Thursday’s meltdown.

Your response might be: What’s the big deal? Panics happen.” But, we remember them and record them for history because they have consequences. In the case of Thursday, the volatility of every asset class has changed. It has changed in terms of how it is measured for investment decisions, which means it will be reflected in prices, and thus it will have an impact on capital allocation. In the last step in this chain, capital allocation will influence future economic growth.

That’s damage done. The quants update their data to incorporate the revised volatility measurement. Actually the updates are preprogrammed and have often occurred already. Consequently, post-event trading reflects this dog-chasing-its-tail computer intelligence.

But, now let’s allow people into the picture. Over the weekend they’ve also had a chance to update their perceptions. Their reactions cover a much broader range of behaviors than just the multiple markets that the programmed trades influence. Most people probably aren’t feeling as secure as they did Thursday morning. If they hadn’t already thought about the consequences of a computer panicking, they should feel less secure. Realizing computers panic is quite different from panicking and warrants a very different response.

The response from a policy perspective will be PART 2, and the response from an investment perspective will be PART 3.

No comments:

Post a Comment