Right now the United States are deep into presidential primary election. All sorts of pundits are out on TV and media to predict the most likely outcomes of elections. While they do bake in tones of theories and case scenarios into their predictions, most valuable piece of their analysis is based on opinion polls. As the current election season is proving these methods are not the most accurate.
What specifically is wrong with opinion polls:
- Samples. Most samples are extremely small. Can 1,000 people represent the attitudes of a country of more than 300,000? They may, if they are perfectly random. There is no way any pollster can maintain the perfectly random selection, which brings us to the next point...
- Selection. Where do pollsters get the phone numbers? Regular panels (who’s opinions are already known), purchased lists (who’s replies are less than guaranteed), random numbers (that actually may lead to a more random sample, but the process becomes prohibitively complex) .
- Pollsters’ preferences. With average reported (really, who keeps tabs on those) margins of error around 5%, when two front runners have the spread of less than 10%, pollster will feel absolutely no remorse assigning victory to either one of them (and that is actually quite acceptable). In this year’s race where the pendulum swings widely from state to state, the poll numbers are almost entirely in the hands of the poll-takers.
- Timing. It takes time to conduct the poll and compile the results. This process can take weeks. In the meantime, life goes on and races do not stop. The final outcome may be impacted by completely different events than the polls. Mind you, pollsters do adjust the results to reflect the most recent events, but that further diminish the quality of the prediction.
There are even more things that are wrong with polls from the statistical point of view. Rather than diving deeper into that, let me bring up the Existing Alternative to the polls. Political Odds. Yes, pundits have been using them quite a bit as well, but more of the supplement.
We see the odds as by far stronger predictor of the election than the polls, simply because the odds-makers are forced to adjust the odds in order to remain competitive. And the gamblers are betting their own money against bet-takers’ on the outcome. All opinions and political agendas are usually swept aside in these transactions.
In this election we are found OddsChecker.com to be the most reliable compiler of multiple betting houses, which allows us for a fuller market view. Please see the link to the coverage of the 2008 US Presidential Race here. (We are not promoting gambling or were paid by any related service)
One thing to be aware of (outside the terrible side effects of gambling) is the “play money” markets. CNN and other news organization trying to cash in on the political booking boom created these fake “fantasy markets”. The fact that people risk absolutely nothing leads to completely false dynamic that produces extremely unreliable predictions, left alone, entirely unverifiable odds maintenance.
Posts
