8/15/2012

Presidential prediction (Polling and Betting)

The upcoming Presidential and legislative, and state elections are being predicted by various people and methods. This posting is a collection of links to these various sites. (See: discussion of Polling)

  • FiveThirtyEight’s mission [by Nate Silver] is to help New York Times readers cut through the clutter of this data-rich world. The blog is devoted to rigorous analysis of politics, polling, public affairs, sports, science and culture, largely through statistical means. In addition, FiveThirtyEight provides forecasts of upcoming presidential, Congressional, and gubernatorial elections through the use of its proprietary prediction models. Read more (About) »
  • Poll or set of Polls


Another way to estimate the future is to look at the results of markets. There are a number of market that allow individuals to bet (or pseudo bet) on future outcomes. The "Wisdom of the Crowds" suggests that a large number of individuals will gather a verity of information and if properly managed, isolated, motivated, on average, come up with a more accurate estimate than either they in the field would individually come up with. The following are some of these prediction markets:

Intrade, the world's leading prediction market (Political market)
IEM (Iowa Electronic Markets) is an online futures market where contract payoffs are based on real-world events such as political outcomes, companies' earnings per share (EPS), and stock price returns.

===============================================================

Which presidential polls were most accurate?

================================================================

Polling
 
There are a number of sites that discuss the results of their polls, i.e. their asking people how they will vote. The results are interesting but are not "perfect" for at least the following reasons:

  1. Future -- It is hard to predict the  future: Even if the results were perfect, they represent the results as of the time of polling and the actual vote will be taken from the time the initial absentee ballot is marked till the final election day. Thus the poll results only reflect a snapshot of a distributed process in which about 50% will be actualized in the future on voting day
    .
  2. Sample Bias -- The sample is taken from the people at large. On the other hand since we have only 50% of the population voting, and there will be a different distribution of the voters than the polled population, the pollsters try to adjust by determining the "Likely Voters" and the greater tendency of various groups to vote. Since these are not objective, they can cause polls to be incorrect.
     
  3. Sample Size -- A poll asks people how they "plan" to vote. But the general sample size is often about 1,000 people, and thus has a statistical variation of about 3%,e.g. The results will be  more than 3% off 1/20 times. Thus if the results are within 3%, there is a 5% chance that the true results will be the other way.
Regardless of these  problems Polls are one tool for predicting the future of elections. Perhaps the best way a reader can use the polling results is to looks at results of a number of polls, considering the estimated errors and biases of each, and recognize that though a specific result may be imperfect, the trends and combinations may well be better.

The following are some links to various polls.

No comments: