Episode notes
The Uncertain World of “Likely Voters” and Polling Error: Methodological Insights
Pre-election polls face a web of uncertainties at every step, from identifying who will vote to accounting for biases and error. The widely reported “margin of error” typically addresses just one factor – random sampling noise – while leaving many other sources of inaccuracy unmeasured. In lower-turnout contests like primary elections, these uncertainties are amplified, and seemingly small errors can shift outcomes. In extreme cases, a combination of misjudged turnout (“likely voter” errors) and polling biases can even produce surprise results – for example, both top finishers in a top-two primary coming from the same party, contrary to initial poll expectations. Below, we detail why a “likely voter” isn’t necessarily a “probable” (actual) voter, and examine how sampling error, design effects, nonresponse, weighting, likely-voter screens, and pollster-specific “house effects” contribute to polling uncertainty. We then show how these factors compound in primary elections and under top-two systems. [nuundigital.com], [biologyinsights.com]