My post-election Intrade newsletter
I thought I had posted this already. I submitted this on Thanksgiving morning, November 23rd.
How well did Intrade's contracts really do?
That the Democrats would retake control of the House of Representatives was not surprising to most people. That the Democrats would win enough seats for a Senate majority was a victory that surprised many political experts. It also surprised a lot of Intrade traders: the Senate.GOP.2006 contract (and similar ones provided by other prediction markets) incorrectly predicted that Republicans would retain control of the Senate -- or did it? It's important to remember that such contracts never predict absolute outcome, but rather the pure probability of the event occurring. A probability is not just a percentage chance, but a prediction that if the election is performed an infinite number of times, if the die is rolled an infinite number of times, etc., then a particular outcome will occur that percent of the time.
If an event has (or is predicted to have) a 90% probability of occurring, it may be extremely probable that it will, but keep in mind that it is still not absolute. A hundred-sided die has a chance of coming up with any one desired side as opposed to the other 99: on any particular roll, any side can come up. Then we delve into conditions of much lower probability, such as the high 50s that the Senate.GOP.2006 contract saw in its last hours: those odds are not significantly better than a coin toss, because the chances of the event not happening are still in the 40s and quite good. With a five-sided die, there's quite a good chance that you'll get sides 1 or 2, as opposed to the other three. As a real-life example, I recently half-joked with a friend that it was a 60% probability (a number I pulled out of the air) that she'd be laid off on the day she expected to be. It didn't happen, prompting her to ask why she still had a job. I replied, "Well, that was the 40%." The same applies to the Senate elections: this election was part of the 40-ish percent probability that the Democrats would retake the Senate.
As the election neared, I was extremely careful in "giving" the 2006 Senate races to either party. Going into the final day, most races had a contract in the 70s or above, which historically (at least in Intrade's contracts) is a strong prediction of the outcome. Voter surveys showed the elections were close in Maryland, Montana, Ohio and Tennessee, but those weren't battlegrounds according to those states' Senate contracts, which had one party's contract in the 80s or higher. TN.Senate06.GOP was moving to the GOP in the final days, reaching 86.7 on the night before the election. I was tentatively giving Virginia to the GOP through the end of October, although I included a caveat on the 31st: "Some races, like Virginia and maybe Maryland, could go against what the trading markets currently predict." In the next week, VA.Senate06.Dem started climbing and closed at 60.9 on the night before the election, apparently because George Allen's newest tactics backfired. Also, MO.Senate06.GOP had gone from the 50s to the 40s while MO.Senate06.Dem had gone from the 40s to the 50s, making it and Virginia the two races that would determine Senate control.
In the end, the general Senate.GOP.2006 contract was incorrect (though as I said, it was statistically not much more sure than a toss-up), but each of the 33 individual Intrade Senate contracts correctly predicted the winner. That in itself is astounding success, and another feather in prediction markets' cap. With that knowledge, an entrepreneurial trader could succeed. It would not suffice that he be Schumpeterian in calculating the risk: he must be also be Kirznerian in noticing (in fact, being able to notice) the opportunity for profit. Once cognizant of what the individual Senate contracts were predicting, he could bet against the favorite in Senate.GOP.2006. That also occurred in the parlay contracts combining the overall House and Senate races, not judging by the contracts' prices, but by the volume.
I noted in our Election Day edition that over 1900 contracts of GOP.House.GOP.Senate had been sold, with nearly 1800 of Dem.House.GOP.Senate, a very surprising 750 of GOP.House.GOP.Senate, and over 2500 of Dem.House.Dem.Senate. The favorite to expire at 100 was Dem.House.GOP.Senate, which closed at 57 on the night before, but as we've already discussed, that's not too strong a chance. So like the Senate.GOP.2006 contract, these parlay contracts also failed to predict that the Democrats would win back control of the Senate. However, consider that over 30% more contracts of Dem.House.Dem.Senate were sold than either GOP.House.GOP.Senate and Dem.House.GOP.Senate. I dismissed it at the time as the contract merely being "interesting" to traders, but hindsight (perfect as always) lets us conclude that someone bet on the general contracts based on the individual contracts.
We no longer have the sales data available to download, but my personal conclusion is that some people correctly expected Dem.House.Dem.Senate to be the contract to expire at 100, and whether they were buying or selling contracts, the price stayed down. Presumably others were applying the general contract to the specific ones, rather than the specific to the general. So the successful traders exploited (which is not inherently a bad word) others' expectations that later proved erroneous, in a perfect example of asymmetrical information, "the market for lemons," "buyers as suckers," whatever you'd like to call it.
Perhaps this is too easy an explanation, and it's true that this is only one general contract out of six. Nonetheless, this is why I like to look at volume as well as price, particularly after some news that might affect a contract. Even if prices stay about the same, higher volumes indicate that someone knows (or at least suspects) something. In the movie "Trading Places" when the Duke brothers continually bought thought the price kept going up, one trader noticed and said to his associate, "Let's get in on it!" The wonderful thing about this market system is that no one is ever coerced into buying or selling; all participants are fully willing, and the winners are those who can compete best. Brute strength does not factor into the competition, and merely acquiring better information does not guarantee success: one must be able to analyze it astutely.
How well did Intrade's contracts really do?
That the Democrats would retake control of the House of Representatives was not surprising to most people. That the Democrats would win enough seats for a Senate majority was a victory that surprised many political experts. It also surprised a lot of Intrade traders: the Senate.GOP.2006 contract (and similar ones provided by other prediction markets) incorrectly predicted that Republicans would retain control of the Senate -- or did it? It's important to remember that such contracts never predict absolute outcome, but rather the pure probability of the event occurring. A probability is not just a percentage chance, but a prediction that if the election is performed an infinite number of times, if the die is rolled an infinite number of times, etc., then a particular outcome will occur that percent of the time.
If an event has (or is predicted to have) a 90% probability of occurring, it may be extremely probable that it will, but keep in mind that it is still not absolute. A hundred-sided die has a chance of coming up with any one desired side as opposed to the other 99: on any particular roll, any side can come up. Then we delve into conditions of much lower probability, such as the high 50s that the Senate.GOP.2006 contract saw in its last hours: those odds are not significantly better than a coin toss, because the chances of the event not happening are still in the 40s and quite good. With a five-sided die, there's quite a good chance that you'll get sides 1 or 2, as opposed to the other three. As a real-life example, I recently half-joked with a friend that it was a 60% probability (a number I pulled out of the air) that she'd be laid off on the day she expected to be. It didn't happen, prompting her to ask why she still had a job. I replied, "Well, that was the 40%." The same applies to the Senate elections: this election was part of the 40-ish percent probability that the Democrats would retake the Senate.
As the election neared, I was extremely careful in "giving" the 2006 Senate races to either party. Going into the final day, most races had a contract in the 70s or above, which historically (at least in Intrade's contracts) is a strong prediction of the outcome. Voter surveys showed the elections were close in Maryland, Montana, Ohio and Tennessee, but those weren't battlegrounds according to those states' Senate contracts, which had one party's contract in the 80s or higher. TN.Senate06.GOP was moving to the GOP in the final days, reaching 86.7 on the night before the election. I was tentatively giving Virginia to the GOP through the end of October, although I included a caveat on the 31st: "Some races, like Virginia and maybe Maryland, could go against what the trading markets currently predict." In the next week, VA.Senate06.Dem started climbing and closed at 60.9 on the night before the election, apparently because George Allen's newest tactics backfired. Also, MO.Senate06.GOP had gone from the 50s to the 40s while MO.Senate06.Dem had gone from the 40s to the 50s, making it and Virginia the two races that would determine Senate control.
In the end, the general Senate.GOP.2006 contract was incorrect (though as I said, it was statistically not much more sure than a toss-up), but each of the 33 individual Intrade Senate contracts correctly predicted the winner. That in itself is astounding success, and another feather in prediction markets' cap. With that knowledge, an entrepreneurial trader could succeed. It would not suffice that he be Schumpeterian in calculating the risk: he must be also be Kirznerian in noticing (in fact, being able to notice) the opportunity for profit. Once cognizant of what the individual Senate contracts were predicting, he could bet against the favorite in Senate.GOP.2006. That also occurred in the parlay contracts combining the overall House and Senate races, not judging by the contracts' prices, but by the volume.
I noted in our Election Day edition that over 1900 contracts of GOP.House.GOP.Senate had been sold, with nearly 1800 of Dem.House.GOP.Senate, a very surprising 750 of GOP.House.GOP.Senate, and over 2500 of Dem.House.Dem.Senate. The favorite to expire at 100 was Dem.House.GOP.Senate, which closed at 57 on the night before, but as we've already discussed, that's not too strong a chance. So like the Senate.GOP.2006 contract, these parlay contracts also failed to predict that the Democrats would win back control of the Senate. However, consider that over 30% more contracts of Dem.House.Dem.Senate were sold than either GOP.House.GOP.Senate and Dem.House.GOP.Senate. I dismissed it at the time as the contract merely being "interesting" to traders, but hindsight (perfect as always) lets us conclude that someone bet on the general contracts based on the individual contracts.
We no longer have the sales data available to download, but my personal conclusion is that some people correctly expected Dem.House.Dem.Senate to be the contract to expire at 100, and whether they were buying or selling contracts, the price stayed down. Presumably others were applying the general contract to the specific ones, rather than the specific to the general. So the successful traders exploited (which is not inherently a bad word) others' expectations that later proved erroneous, in a perfect example of asymmetrical information, "the market for lemons," "buyers as suckers," whatever you'd like to call it.
Perhaps this is too easy an explanation, and it's true that this is only one general contract out of six. Nonetheless, this is why I like to look at volume as well as price, particularly after some news that might affect a contract. Even if prices stay about the same, higher volumes indicate that someone knows (or at least suspects) something. In the movie "Trading Places" when the Duke brothers continually bought thought the price kept going up, one trader noticed and said to his associate, "Let's get in on it!" The wonderful thing about this market system is that no one is ever coerced into buying or selling; all participants are fully willing, and the winners are those who can compete best. Brute strength does not factor into the competition, and merely acquiring better information does not guarantee success: one must be able to analyze it astutely.
posted by Perry Eidelbus at
2:16 PM
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home