I must admit to having somewhat mixed feelings about the Ma-Rao-Sears article. As one of the very few academics to believe that coordinated circuit breakers might even be worth discussing, let alone enacting, I was pleased to note the message the three authors drew from their study of what happens in commodity futures markets before and after daily price limits are hit. They conclude that following such limit moves, prices tend to stabilize or reverse; and that volatility also tends to drop off substantially when trading resumes. They thus appear to find, as some of the more enthusiastic supporters of circuit breakers, like the Brady Commission, have argued, that commodity price limits act like the beryllium rods in nuclear reactors. They cool off an overheating market but without permanently damaging market liquidity, since the three authors find also that volume does not drop off substantially once trading resumes.
Provocative as these new empirical findings may seem at first glance, their real probative value in the debates over circuit breakers is far from clear. For one thing, commodity price limits, though they do bring trading to a halt, differ in other important respects from the coordinated circuit breakers proposed after the crash and subsequently installed by the New York Stock Exchange and the Chicago Mercantile Exchange. F'urthermore, even if price limits really were the appropriate analog, the attempt to draw 'valid inferences from the post-resumption activity faces difficult and possibly insuperable obstacles because of selection bias.
Price limit moves, after all, can occur for several days in a row, leaving an investigator of post-limit behavior with several unsatisfactory choices. The most deceptively simple option, of course, is just to throw the multiple-move cases out of the sample. One-day limit moves are already rare events; multiple-day moves are even rarer. Can there be much harm in removing so few points from the sample? The answer, alas, is yes. Trimming out the multiple-move days leaves a sample not just of one-day limit moves, but of limit moves where, by selection, the next day's move could not have been an equally large or larger move in the same direction! Not surprisingly, prices, returns, volatility, and volume would all seem to be going back to normal levels on the days following these selected one-day-only events.
Nor does it help much to put the multiple-move days back in, treating the second move as both the follow-on of a one-day move and the start of a new sequence. When the second limit move in the sequence is considered a post-limit day relative to the first move, the