The development of the election movement in the last few weeks meant that the opinion polls received a lot of attention. The distance between Tidöblocket and Magdalena Andersson's government background shrank rapidly. At the same time, the Liberals were approaching the barrier, against all odds.
The rapid and large shifts in the electorate contributed to the fact that measurements from various institutes spread from day to day. In retrospect, it can still be stated that the accuracy of the last measurements before election day ended up very close to the historical average, i.e. a deviation per party of 1.1 percentage points.
Ipsos, which DN collaborates with, came closest of all those who did measurements before the election, with a deviation of 0.7 percentage points on average. Indicator, which takes measurements for Echo, came next. For the blocks, Infostat scored the most correctly, with Ipsos in second place.
- We are the only institute that has all the parties and blocs within the margins of error, says Nicklas KĂ€llebring, opinion analyst at Ipsos.
A blemish on the record is that Ipsos and many others scored the Moderates low in relation to the election results. It is one of the questions that Nicklas KĂ€llebring will look at when the methods are to be tuned before the next election.
- The moderates are a critical issue. Another is that we have overestimated the Sweden Democrats two elections in a row.
Nicklas KĂ€llebring points out that opinion polls are not forecasts but a summary of the situation during the measurement period. This means that the focus of DN/Ipso's last survey was not on the Friday before election Sunday, but rather in the middle of the week when about half of the interviews were done. In other words, the election was then three to four days away.
It matters according to Nicklas KÀllebring, because there were clear trends in the final stage in favor of the Tidö parties. The Moderates probably had a strong momentum up until election day - which continued thereafter. In DN/Ipso's first voter barometer after the election, M received 22 percent, the highest figure in eight years.
That M and SD have become communicating vessels is a factor that can make it more difficult to estimate the parties' support.
- We do not know exactly which currents have taken place between them in the last few days.
Another factor is first-time voters, where the Moderates are strong. The group is particularly difficult to measure because you cannot weight the result against how you voted last time. It is otherwise one of the methods used in opinion polls as compensation for a group being underrepresented.
- There is a job to be done here. The challenge is to understand and distinguish late voter movements from random variations and systematic errors, says Nicklas KĂ€llebring.
Voter behavior changes over time. Which parties have been difficult to measure has therefore also varied historically.
- Once upon a time it was the Social Democrats. Later, everyone underestimated the Sweden Democrats. And now it's the Moderates, says Nicklas KĂ€llebring.
In the case of SD, part of the solution was to develop the methods so that voters had the opportunity to answer in different ways. For a long time, many people were reluctant to say outright in a telephone interview that they sympathized with SD, because the party was so badly regarded in broad circles. Then digital surveys became a way to get answers that otherwise would not have been received.
- The Sweden Democrats are a typical example that we who do measurements have managed to deal with the systematic errors that have arisen. But it may take a couple of elections to overcome it, says Nicklas KĂ€llebring.
