Parliamentary elections will take place on November 22th in The Netherlands. In the period before the elections several organizations attempt to predict the outcomes of this elections.
In fact, the Dutch public is more or less a washed by opinion polls. Each weekend, at least three polls are published. And of course, their predictions differ.
Politicians consider these polls very important. One sometimes wonders whether they are more interest in poll results than in what the general public really wants. It seems that politics follows the polls instead of polls following politics.
The opinion polls get a lot of media exposure. But these media (radio, tv, newspaper) never ask the question how reliable these polls are. A good polls is based on a random sample (probability sample) from the population, were every member has the same, non-zero probability of being selected. This is not the case for the current opinion polls. They collect their data using the Internet. It means that people without an Internet connection (primarily elderly, low educated, and ethnic minorities) have a zero probability of being selected. Moreover, no random is selected. Instead these polls rely on self-selection of respondents. It means that selection probabilities are unequal, unknown, and sometimes zero. Consequently, it is completely unclear how the sample resembles the population. And also no measures accuracy can be computed.
The website PolitiekePeilingen.nl, which I built with my dad, keeps track of the results of three major opinion polls. By combining them in one graph it becomes easy to compare results. The site also contains methodological commentary on poll results being published by the media, in the news section. This new viewpoint tries to place news items the proper perspective.