The first known example of opinion polling was in 1824 in Pennsylvania. It correctly predicted the local election. Was this a sign of dedicated research or was it just luck? We will never know. What we do know is that as the years went by, opinion polling gradually became more scientific and popular as it spread throughout the United States. As time went by it was exported to the ‘old continent’ Europe.
Though it is a relatively old phenomenon, we see that last year(s) the quality and influence of opinion polls are still heavily debated. Did it poorly predict (for example) the US elections and the Brexit and if that is the case, what was the magnitude of influence? Some countries, including France, Canada and Italy, even forbid the use of opinion polls during election campaigns. Moreover, in Germany there is a gentleman’s agreement to not publish any.
There are three types of methodological risks when it comes to polling. Firstly, there is a coverage bias, which simply means that it is too difficult to reach everyone and answer a survey. Historically there have been problems with reaching people without a phone. Nowadays it is harder to reach people that barely go online.
Secondly there is a non-response-bias. Some people simply don’t fill in surveys, even if they have reached them. For Example, people that are not interested in politics or dispute the political system.
Thirdly there is a response bias. This means that people sometimes fill in a survey in a way they feel is socially acceptable, instead of giving their real, maybe more extreme opinion.
by Maxim le Clercq