Investigating Industry Bias in Swedish Polling from 2010 to 2022

Abstract: Industry bias is the bias associated with polling estimates, and that is caused by systematic errors. We aim to test for its presence in Swedish political polling from 2010 to 2022 and simulate industry bias as a consequence of nonresponse bias. We estimate industry bias per party as the mean error of the last polls per polling house before an election and calculate the probability of observing the number of polling underestimates and overestimates per party and election year. Our results indicate that industry bias is present to varying degrees in each Swedish parliamentary election year from 2010 to 2022. Moreover, our results indicate that polls with and without industry bias as a consequence of nonrespondent bias can be replicated.