Saturday, April 26, 2008

Evidence of Strategic Voting in Political Futures Market

The Big Question

The big question in the political realm today is who between Hillary Clinton and Barack Obama has a better chance of beating John McCain come November. Both candidates have been quoting different statistics from pollster.com, some support Hillary while others support Obama. But while polls only show a small advantage for Obama, the political futures market* on intrade.com points strongly toward Obama as being a better candidate for beating McCain.

The question of who has a better chance of beating McCain in November can be answered uses futures prices in a more reliable way then in a national poll. For one thing, people can lie in national polls. They are anonymous and don't stand to loose anything if their opinion is proven worng. On the other hand, people betting in the futures market stand to loose cash and have no incentive to move a national poll on way or the other.

The answer to who can beat McCain rests in how the demand for McCain futures changes with increases in demand for Obama or Clinton futures. If McCain futures increase in value Clinton futures increase in value then investors believe that McCain has a better chance of winning if Hillary wins the Democratic Nomination. If no relationship exists between McCain futures price and Clinton's then investors do not perceive there to be any difference between Obama and Clinton in terms of their edge over McCain.

The data points strongly in favor of Obama at being better able to beat McCain in a general election. When Obama's future's price goes up by $1 (10% better chance) then McCain's futures price for winning the general election fall by $0.13 (1.3% worse chance). An opposite and slightly larger effect is true of Clinton futures where a $1 increase for her futures price leads to a McCain's futures price for winning the general election going up by $0.16 (a 1.6% better chance of winning the general election).

Note to Statisticians: These results are significant at the 10% level and come from data based on the month of April. The data is available at intrade.com. To replicate these finding use a simple regression of McCain closing futures price on the "anyone in the GOP to win the general election" as a control, and then use Obama's and Clinton's closing futures price for winning the nomination in two separate models to avoid multicollinearity problems. No other control variables are needed to see these results.

The evidence of strategic voting comes from the huge difference between the futures price estimates and national polls on who is better able to beat McCain in November. The average poll on pollster.com for the month of April shows Hillary with a 1.1% smaller chance of beating and Obama with a 0.2% better chance of beaitng McCain with some individual polls supporting wither candidate. The estimated difference based on the futures market where voters stand to loose cash for being wrong has Obama with a 24.8% margin to beat McCain while Hillary stands to loose at -13.2% margin. This difference is huge. Anonymous polls are notoriously bad at giving unbiased estimates compared to markets that require voters to “put their money where their mouth is”. The reason why the estimates from a futures market would differ so strongly from those of a poll rest in who stands to gain from Hillary winning the nomination. The Republican party seems is the only party who could benefit from over estimating Hillary Clinton's chance of beating McCain besides supporters of Senator Clinton.

Further note to statisticians: The margin difference calculation can be replicated using the coefficients from the ordinary least squares analysis. Hillary Clinton futures as of April 25th is trading a $1.75. If she wins the nomination, her futures price will need to raise from 1.75 to 10 which, given our analysis, will raise McCain’s futures from 40% to 53.2% leaving Hillary with only 47.6% to win in November. A similar approach can be done for Barack Obama leaving him with 62.4% to beat McCain's 37.6%.

If you have any questions reguarding this analysis or need help getting the same results feel free to email me, Joseph Eugene McPhail, at joseph.e.mcphail@gmail.com. Thank you for your interest.

*What is a futures market? The political futures market allows anyone with a credit card to bet on which candidate they think will win the nomination and/or general election. This form of market is more often associated with the trading of buying a selling of financial assets, but can also be used for events such as who will win the presidential election. Anyone can go to intrade.com and buy a futures for Obama, Clinton or McCain for a price of less then $10. If your candidate wins you receive $10. If they loose you get nothing. Before the election is over you have the option to sell your futures back to someone else. If the chance of a candidate like McCain goes up for winning the election then people from around the globe rush to buy his futures until the price rises to better reflect his chance of winning. For example, a futures for a candidate like McCain winning the general election selling for $39.2 can be interpreted as McCain having a 39.2% chance of winning.

2 comments:

Sir Gorpster said...

The participants at intrade are not a random sample whereas the polls attempt to get a random sample. The intrade might be more accurate because $$ at risk but difference does not have to be explained by "strategic voting". It could just be the random sample influenced by other factor - such as (my favorite) the big advantage CLINTON name has in "name recognition" over OBAMA. The polls still show most people at random don't know much about him and what they know is often inaccurate.

This is why there is always such an OBAMA surge after attention turns to a state primary such as happened in OH, TX, and PA.

Unknown said...

Hmm, an interesting argument Joe! I can appreciate the math behind your intrade analysis. I believe Gene's argument is similar to the one I made that night at your apartment, about the sample not being random from the population. However, I think you explained that while they may not be random, they should represent a more informed, or at least more interested, population sample, and therefore their "votes", with a material risk, should be a more accurate representation of what will actually happen.

I still don't know if I believe that ;) But regardless, looking at the degree of correlation is something I can understand, and is a cool way to come to this conclusion!

I wonder what other world events, weather changes, etc... could be correlated against intrade values? Perhaps a scan of top news stories every time the Obama or Clinton value drops or rises past a significant threshold would show something interesting too?

You're a genius man :)