Estimating Youth Turnout in Primaries
Since CIRCLE’s founding in 2001, our staff has worked to provide independent, accessible data focused on opportunities to increase, diversify, and sustain youth civic and political engagement. Over the past two decades, we have had regular discussions about maintaining rigor while doing timely and influential work. One critical discussion, especially during election years, is about the sources of youth voter turnout data. Measuring election participation is difficult to do in an exact way.
Starting in 2014, CIRCLE began to use and evaluate how aggregated national voter files compare to the other methods of estimating youth voter turnout (i.e., the Census Current Population Survey [CPS] and CIRCLE’s day-after early-indicator estimate). We’ve done due diligence to understand the differences between these sources in midterms and presidential elections, and we continue to track these differences. Our analysis suggests that exit polls improved in 2018 vis-a-vis voter file estimates, and we will continue to use exit polls for day-after, preliminary estimates that can serve as early indicators of major trends. We update these estimates as soon as voter file data (or CPS, when voter files cannot answer some questions) becomes available.
Participation in primaries, however, is recorded in only two ways that allow us to estimate turnout: through a national aggregated voter file and through exit/entrance polls. We used exit poll data to calculate CIRCLE’s day-after, preliminary estimates of youth voter turnout in the 2016 primaries, and we examined both data sources by comparing turnout in 19 states where data from both sources was available. In most but not all cases, the day-after, exit poll-based, preliminary estimate was higher by more than 2 percentage points. To some extent, this is to be expected for a number of reasons related to each method, but in the primary data we did not see consistent differences across all of the states. In some states, the estimates between the exit polls and voter file were remarkably similar. We spent substantial time looking for patterns and testing hypotheses that might explain these differences. In addition to sources of error inherent to each method, we are not privy to specific information on factors involved in entrance/exit polling methodological decisions.
As a result, in 2020, we’ll continue to provide these day-after preliminary estimates, and will continue our extensive analysis of data sources in order to provide the most rigorous analysis.
