Sunday, November 2, 2008

Prices in Senate and Presidential Markets

Intrade provides markets in several states where there will be elections for both President and for Senate. Polling data are available for both sets of contests, and analysis reveals a few anomalies where candidates with similar poll performance have significantly different Intrade prices.

Either Intraders have information that is not in the polls or there are profit opportunities available (disclosure: I do not hold a position in any of these markets).

Above is a table (made Sunday afternoon, November 2) of the last 13 states whose Senate races Real Clear Politics (RCP) put on their “Battle for the Senate” tab that have recent polling data page (Louisiana and Nebraska had insufficient polling data to generate an RCP average and were excluded). Aside from Alaska (obviously a special case since the Republican candidate for Senator was convicted and the prior polling data may have factored in a possibility that Sen. Stevens would be not guilty), there is a rough correspondence between the margin in the polls and the Intrade prices. But a few interesting points emerge that I offer for the purpose of debate, knowing of many possible problems (such as the lack of liquidity in certain contracts) that might also explain these findings.

Comparisons among Senate Prices

Even though the Republican candidate Norm Coleman leads in the RCP average, Intraders believe that Al Franken is roughly an even chance. But the race is clearly very close and the polling data are fairly skimpy, with the most recent poll, from the Star-Tribune, has Franken significantly ahead. The interesting observation for Intraders might be the comparison with Georgia, where Saxby Chambliss has a lead comparable to Coleman’s but trades at a significantly higher price.

The other interesting Senate observation is in the spread between the prices of the Democrats who are 5 point RCP favorites compared with the prices of a Republican with a similar edge. Challengers Kay Hagan in North Carolina and Jeff Merkley in Oregon both have leads roughly comparable to incumbent Republican Mitch McConnell in Kentucky. Yet both Democratic contracts trade a bit higher than the Republican contract.

Finally, Intraders believe that the polls for Roger Wicker in Mississippi seem to overstate his chances. Jeanne Shaheen in New Hampshire has a slightly smaller edge in the polls (against an incumbent) and trades five cents higher.

Comparisons among Presidential Prices

The states that stand out are North Carolina, Colorado and Virginia; in North Carolina the polling is virtually even (the last four polls were Obama +2, tied, Obama +2 and McCain +3), yet the Obama contract trades over 60 cents while in Colorado and Virginia Obama’s 5 point lead corresponds to prices in the mid 80s. In New Hampshire, Obama has a larger lead and a lower price. To put these prices into context, candidates with 5 point leads in the Senate (in Kentucky, North Carolina and Oregon) all sell below the Virginia Presidential contract.

Senate-Presidential Price Comparisons

Anomalies: In Colorado and New Mexico, Obama runs well behind Mark and Tom Udall in the polls, yet trades at a slightly higher price. In Mississippi, McCain and Wicker have similar polling results and presumably most voters for McCain will vote for Wicker. Yet McCain’s price remains significantly higher than Wicker’s.

Tuesday, October 28, 2008

Intrade Election Market: No Bradley Effect

Analysts of the 2008 election have repeatedly cited the Bradley effect as a reason to discount polling data in elections pitting white candidates against black candidates, arguing that some voters tell pollsters they would vote for a black candidate when they intend to vote for the white candidate. Yet analyzing the data, it appears that Obama shares are performing as well as (or perhaps outperforming) his RealClearPolitics (RCP) polling data, indicating that Intrade investors believe that the RCP polls understate Sen. Obama's margin of victory.

The Bradley effect refers to the California gubernatorial election of 1982, in which the black mayor of Los Angeles, Tom Bradley, lost an election he had been expected to win. In a number of other elections featuring black candidates versus white candidates, the black candidate appeared to fare worse on Election Day than expected (see here for anecdotes from elections from the 1980s through 2006). While there is still active debate about whether the Bradley effect may have disappeared as American social attitudes changed in recent years, the evidence from Intrade markets is clear: Intrade market participants completely discount the Bradley effect.

If Intrade investors believed in the Bradley effect we would expect to see the price of Obama shares below 50 when the polls were even, only rising above 50 when Sen. Obama had a lead in the polls greater than the expected Bradley effect. For example, if 2% of the voters told pollsters they would vote for Sen. Obama when in fact they were supporters of Sen. McCain, then Obama shares would trade at 50 only when he had a lead greater than 2% in the polls.

Since Sen. Obama has had a large lead in the RCP polls in recent weeks, it is difficult to know whether the price of Obama shares accurately reflects the polling data. But for a brief period in early September the polls were very close. On September 2, after Sen. McCain cancelled the first day of the Republican Convention, Sen. Obama led in the RCP polls by 6.4% and his Intrade shares were trading around 60 cents. Then, after the surprise announcement of Gov. Sarah Palin as VP and a reasonably successful convention, Sen. McCain started to rise in the polls.

Sen. McCain actually led in the RCP polls for a short period; he had a small lead from September 7 until Sen. Obama came back to even in the polls on September 17 (and Obama has maintained a lead ever since).

Sen. Obama's closing Intrade price was 57 on September 7 and his shares remained above 50 through September 10, despite Sen. McCain's small and growing lead in the RCP poll. Only when Sen. McCain's lead reached 2.5 points on September 11 did Obama prices drop below 50. When the RCP polls were tied on September 17, Obama shares closed at 48.9, but they rallied to 51.8 on September 19 as Obama gained a 1.9 point in the RCP poll.

In other words, Intrade traders clearly believe that the Bradley effect will not be a factor in the 2008 Presidential election. Prices of Obama contracts remained above 50 at a time when he was behind in the polls, indicating that Intrade traders believed that the polls actually underestimated Obama's likely election performance. When Sen. McCain's lead in the polls grew, the price of Obama shares briefly dropped below 50 but they quickly recovered with the evidence of improvement in the polling data.

Political scientists will spend the next decade analyzing the 2008 polling data and election returns for evidence of a Bradley effect but Intrade investors have already cast their vote.