Political attitudes
How Poles’ attitudes have changed since the end of communism. Each trend pools every comparable survey reading I could find, from the early 1990s to the latest polls, into one monthly estimate, and can be broken down by party preference, age, gender, education and place of residence. The first attitude is support for Poland’s membership of the European Union; more will follow. No single survey has asked the same question about the European Union from 1990 to today. CBOS has asked whether people support Poland’s integration with (since 2004, membership of) the EU since 1994; Eurobarometer, the European Social Survey, the Polish National Election Study, the Polish General Social Survey, the European Election Study and commercial pollsters have asked other questions at other times. The trend combines them all. Every reading is turned into the share of respondents taking a side who give the pro-EU answer (those who say “don’t know”, or choose a neutral middle option, are set aside). A Bayesian measurement model (Stan) then treats each question as a noisy measure of one underlying level of support that changes from month to month. Each question has its own level and its own sensitivity to that underlying support, estimated from the periods in which it overlaps with other questions, and its own allowance for error beyond sampling error, which absorbs differences of wording, fieldwork organisation and mode. The underlying support follows a random walk, so it is smooth when little happens but can move quickly when many surveys show a change. The line is expressed on the scale of CBOS’s question in its face-to-face surveys (up to April 2020): it reads as the share of supporters among those with a view who would have been recorded by that question. CBOS’s surveys since May 2020, which mix face-to-face, telephone and online interviews, are treated as a separate question, because the change of mode may have shifted answers. Dots show each reading placed on the same scale, after allowing for its question’s level and sensitivity. Bands are 80% and 95% credible intervals. For the breakdowns, each group’s line is the national trend plus a deviation that changes slowly over time, estimated from every survey that reports that group. Where surveys report categories that do not match exactly (for example, Eurobarometer’s “small or middle-sized town”), the reading is modelled as a mix of the groups it covers. Between the surveys that report a breakdown, a group’s line keeps its last known distance from the national trend and its interval widens. Party preference is the party respondents intend to vote for, or voted for at the last Sejm election, followed across elections by party lineage; a party’s line is shown while it existed and its voters were surveyed. The data come from survey microdata wherever they are available, and published results fill in the rest. Survey readings are estimates too, and the early 1990s rest on few of them, so the intervals there are wide. Because CBOS changed the way it interviews in 2020, the line since then sits a few points below CBOS’s own published figures: it is read on the scale of the face-to-face question, and the other surveys suggest the mixed-mode question reads somewhat higher. Sources:How this works