Prediction Market Efficiency - US Presidential Prediction Markets

US Presedential Election Prediction Markets

The question of how efficient prediction markets are, has long haunted economists; well, at least those that are not blinded by their rather heavy prediction market baggage. How quickly do prediction markets incorporate all of the available information?

Gilder and Lerman have presented a system for predicting price fluctuations in Prediction Markets, such as TradeSports and the Iowa Electronic Markets, utilizing both market history and public news flow, published before the beginning of each trading each day, to produce a set of recommended investment actions.

The authors hypothesise that past/present events can potentially assist in predicting future prices in prediction markets. They empirically show that prediction markets are surprisingly predictable, even by purely market-historical techniques.

Evaluating their system on eight political markets from 2004, the authors show that they could have made effective investment decisions based on their system’s predictions, with profits greatly exceeding those generated by a baseline system.

The authors conclude that;

In summary, we have presented a system that predicts price fluctuations in prediction markets well enough to make a significant profit. We observe that while using market-historical features is enough to make a good profit, adding a linguistic analysis of news boosts returns significantly. We further show that standard approaches cannot capture news information and that our more sophisticated NLP methods yield improved accuracy."

In a nutshell, the research suggests that the prediction market prices studied do not reflect all available information. Accordingly, in terms of said prediction markets, the efficient market hypothesis does not hold. A likely explanation for this finding, is that the prediction markets that were studied are simply not very liquid.

Taking hold of the baton from Gilder and Lerman, Panos Ipeirotis and George Tziralis have developed some techniques for extracting news flow signals to see whether they can indeed be utilised to predict the future performance of markets on the InTrade prediction exchange.

As part of a larger study, the results of which they will present in a paper at a later date, they provide a preliminary result for the prediction market on whether Hillary Clinton will be the Democratic Presidential Nominee in 2008;

"Our sentiment index (in maroon) is close to 1 when we predict that the market will move higher, and it is close to 0 when we predict that the market will move down. Typically, it works pretty well for predicting long periods of price increases and declines. To put our money where our mouth is, the signal for the last few days shows that Hillary's market price will edge lower in the new few days/weeks."

In the tables below we see the Intrade prediction market and the Betfair market (top) on whether Hillary Clinton will be the Democratic Presidential Nominee in 2008 as of 10.45 GMT on December 3 2007. Whilst the Intrade market suggests that Clinton's probability of victory is 67%, the Betfair market gives a current reading of 69%. (For the purposes of this study we assume that there is no over-round in the Betfair market).

In the following tables we see the Intrade prediction market and the Betfair market (top) on whether Hillary Clinton will be the Democratic Presidential Nominee in 2008 as of 08.45 GMT on December 7 2007. Whilst the Intrade market had suggested on December 3 that Clinton's probability of victory was 67%, it was now suggesting that her probability of victory was 64.5%. The Betfair market which had given a reading of 69% on December 3 as regards her probability of winning the democratic nomination, was now suggesting that her probability of victory was 67%.

In the following tables we see the Intrade prediction market and the Betfair market (top) on whether Hillary Clinton will be the Democratic Presidential Nominee in 2008 as of 18.20 GMT on December 10 2007. Whilst the Intrade market had suggested on December 7 that Clinton's probability of victory was 64.5%, it was now suggesting that her probability of victory was 61.5%. The Betfair market which had given a reading of 67% on December 7 as regards her probability of winning the democratic nomination, was now suggesting that her probability of victory was 64%.

In the following tables we see the Intrade prediction market and the Betfair market (top) on whether Hillary Clinton will be the Democratic Presidential Nominee in 2008 as of 20.45 GMT on December 17 2007; two weeks after we first looked at the markets. On 12 December a WMUR/CNN poll had put Clinton on 31%, down five points since last month, and Obama on 30%, up eight points, while a poll by Suffolk University in Boston put her on 33%, down one point, to Obama's 26%, up four, since last month. During the past week, a poll released by the Davenport, Iowa, Quad City Times, suggested that Obama had surged to a 9-point lead over Clinton in the race for the Iowa presidential caucuses. Whilst the Intrade market had suggested on December 3 that Clinton's probability of victory was 67%, two weeks later that probability had fallen to 56.2. The Betfair market which had given a probability reading of 69% on December 3, was now suggesting that Clinton's probability of victory was 57%.

It is quite clear that during the two week period under observation, the markets were indeed responding to stale news. Those that had sold Clinton on Betfair at 1.44 on December 3, on the back of Panos Ipeirotis and George Tziralis' advice, are now sitting on a healthy profit; albeit one must factor in the lack of liquidity in both sets of markets.
The race to become Democratic nominee for the 2008 Presidential election has also seriously called into question the efficiency of prediction markets. Ahead of the New Hampshire Caucas, the market on Intrade was suggesting that Barak Obama had a 75.9% probability of victory, with the market suggesting that Clinton had a probability of victory of 22.5%. Following Clinton's victory in New Hampshire, the market was suggesting that Clinton's probability of victory had increased to 58.5%, whilst Obama's probability had fallen back to 40%. In an efficient market, one would not expect to see such wild variations in price over such a small period of time.....

Looking at the latest Intrade market on the Democratic nominee race (January 30 2008), we can see that the sum of the probabilities is greater than 1! This fact alone, raises quite significant efficiency concerns rergarding this market and its probalistic potentialities.

The following table, showing the odds 24hrs before Super Tuesday, reveals that the gap between Obama and Clinton is narrowing in the betting market, once again.

Mid-way through Super Tuesday and Obama is back in the lead, in response to him having performed well in the States that had called early. Throughout the evening, the lead was to once again pass back to Clinton.

After all the results are in from Super Tuesday and Obama has extended his lead over Clinton in the market.

Following on from Super Tuesday Barack Obama recorded four weekend primary victories in Nebraska, Louisiana, Washington state and Maine, representing a considerable geographical and political range. In Maine, with 99 percent of precincts reporting, Obama had won 59 percent of the delegates to the state party convention, compared with 40 per cent for Hillary Clinton. Looking ahead, both Maryland and Washington D.C. have sizable African American populations, which, it is anticipated will line up behind Obama, and he is also averaging a 17 point lead over Clinton in Virginia polls. Clinton, on the other hand, may not see another victory until until the March 4 votes in Ohio, Texas, Vermont and Rhode Island.

On Tuesday February 12 Obama racked up further victories in Maryland, Virginia and the District of Columbia, beating Clinton by a margin of nearly 2-1 in Maryland and Virginia and 3-1 in the nation's capital. He also narrowly pulled ahead among delegates for the first time in the contest. Slowy but surely it would appear that he is eating into Clinton's core white working class voters. Following these victories, the betting market on Intrade was betting that he now had a 73.5% probability of being the democratic presidential nominee in 2008.

On Tuesday February 19 Obama notched up his ninth straight primary victory since Super Tuesday, winning the Wisconsin primary by 15 percentage points. Obama also took a big lead in the Hawaii caucus, where, with 7.5 percent of precincts reporting, he gained 77 percent of the vote, compared with 23 percent for Clinton. Following on from these victories, the betting market on Intrade was now betting that he had a 78% probability of being the democratic presidential nominee in 2008

On the eve of the vote in Texas in Ohio, the betting market on Intrade was now betting that he had a 80%% probability of being the democratic presidential nominee in 2008. (Just before voting opened, this had actually climbed to 83.7%, as the markets and polls seemed to suggest that Obama would win Texas).

Hillary Clinton threw the Democratic race wide open today when she bounced back to win the two mega-states of Texas and Ohio, in the process ending Barack Obama's long string of caucas victories. Obama, who had been trading on the Intrade democratic presidential nominee betting market with a probability of 83.7% on the morning of the vote, ended the day back at 73% - with a number of his more bullish supporters having had their wings clipped a little. The momentum is now likely to build behind Clinton's campaign, and it would be no surprise were Obama's probability rating to fall back even further over the next few weeks. Once again, when it came to the vote in Texas the prediction and betting markets called the result wrong.

Play Online Poker

Copyright © 1997 - 2007 Bettingmarket.Com et/ou entreprises liées. Tous droits réservés. Terms and Conditions Apply .