The article discusses how election polls can be inaccurate due to biases and excess variability. The researchers used a hidden Markov model to analyze polling errors in US Presidential and Senate elections from 1992 to 2020. They found that previous estimates of bias in Presidential elections were too extreme by 10%, bias in Senatorial elections was too extreme by 25%, and excess variability estimates were also too large. This new approach is less sensitive to the timing of polls and provides more accurate and interpretable results.