Pooling the Poles

What is this blog for?

The purpose of this blog is simple: it pools polls of Poles.

That is, it uses a statistical model pioneered by Simon Jackman (2005) "that tracks changes in voter support over time by pooling the polls, and corrects for variation across polling organisations" to provide "a less biased and more precise estimate of vote intentions than is possible from any one poll alone". I have adapted this model using code kindly provided by Kai Arzheimer.

This model allows me to take all the information contained in the various polls published by CBOS, Kantar, Millward Brown, IBRIS, Pollster, Estymator (and, on occasion, other pollsters) to estimate current levels of support for all relevant parties, trends in support over time, and the likely seat shares of parties both nationally and in each constituency.

A caveat is required: if all the polls are consistently wrong, then the estimated totals will also be wrong. For this reason, poll-pooling models are often better at capturing the shifts of trends than they are at providing snapshot estimates of party strengths at any given point in time. However, at worst, a pooled model will do no worse than individual polls. Generally, it will do better. While Polish polling is volatile - as is the Polish electorate - a model very similar to the one I currently use managed to predict fairly accurately the results of the 2015 election.

Why is this blog needed?

Although reporting of polls in the Polish press has improved significantly over the last few years, there is still a tendency to pay too much attention to individual, eye-catching polls (or clusters of same) without paying attention to longer-term trends. As a result, too much credence is given to noise, and not enough to signal. Since I started pooling the Polish polls in 2014, I have tried to persuade people to pay more attention to trends. At the same time, it is clear that there is also an interest in where parties stand at any one given time, and how that translates into likely shares of seats in the legislature. By pooling the polls and then extracting estimates of present party strengths, I am able to provide a more balanced picture of the current situation than reports on individual polls can. Previously, I published these estimates on Twitter and in an RPubs mini-blog, but have come to believe it would be better to set up a dedicated blog to present these estimates and, where necessary, provide some comment on them.

Why are you doing this?

Because nobody else is.