Paul Krugman is right: prediction markets are looking pretty pathetic this morning. On Monday I noted that Obama had become the Democratic heir presumptive more or less overnight, saying that "the speed with which Clinton and Obama have traded places is nothing short of astonishing"; today, we’re back at the status quo ante. The InTrade contract which expires at 100 if Obama gets the Democratic nomination collapsed from above 75 to below 40 in the space of a few hours last night: the times are Irish, since that’s where InTrade is based.
A few lessons can be learned here. Firstly, the strategy of "buy any favorite trading around the 70 level", on the grounds that such trades nearly always win, is very dangerous. Secondly, prediction markets are pretty bad at working out the margin of error on polls. And thirdly, if you’re Justin Wolfers, it’s probably smart not to make unhedged statements saying that Barack Obama has "better than a nine-in-ten chance of winning" the New Hampshire primary.
Dan Gross says that the lesson of New Hampshire is that "these are less futures markets than immediate-past markets" – by which I think he means that they give too much credence to very recent polls.
Taking the other side of the debate, Chris Masse makes a couple of very good points: firstly that prediction markets, suffer from GIGO as much as anything else (garbage in, garbage out); and secondly that they should be judged not on their absolute reliability but rather on their relative reliability, against polls.
On the other hand, it might just be that Hillary really did save her political life last night, and that that was indeed a low-probability event. Stranger things have happened in politics.