Association for Public Policy Analysis & Management

Throughout time, many criticisms have been leveled at school boards, one of which is that they do not serve an important function in a nation where citizens engage in their election at very low rates and where there is sparse evidence that citizens participate in ways that make sense given the logic of having elected boards in the first place. This study seeks to contribute to this literature by answering four questions: First, to what extent do school quality perceptions match up with actual achievement data?  Second, to what extent do perceived or actual student achievement trends influence voter turnout? Third, to what extent do these same trends influence incumbent fortunes in school board elections? Fourth, do different types of information influence these outcomes to different degrees? School board elections, being less partisan focused than other levels of the political process and being tied directly to a government service, allow for important theoretical advances in the area of citizen engagement in politics and the degree to which previous performance is tied to electoral outcomes.  If one is to find evidence of retrospective voting in any level of electoral politics, it is likely to find this evidence in school board elections.

This project uses panel data from California school board elections, California public opinion surveys from the Public Policy Institute of California, the American Community Survey demographic statistics, and local school performance information, Achievement data was pulled from both the California STAR system, which has information on actual standardized test scores, as well as the API system, which has data on scores that are used in California’s school accountability policies. The public opinion surveys asked a representative sample several questions including an evaluation of the performance of one’s local schools.  California school board election data is able to provide rich data on voter turnout and outcomes of elections.  All of this data is available from 2000 to 2013.  Using panel data allows us to account for many of the confounding characteristics of districts as well as election cycles that have made this kind of research difficult to interpret in the past.

Negative binomial regression was used for voter turnout estimates, while least squares regression was used for incumbent fortunes models.  All models include district fixed effects and year fixed effects.

Preliminary results find that the largest significant associations between school performance and voter behavior were between lagged STAR scores and total voter turnout. We find little evidence that API scores influence voter turnout, nor was significant evidence found for gain scores. In addition, little evidence was found for any influences on incumbent fortunes. Significant results were found for all subgroups when looking at STAR data, but these results were only found for the association between lagged scores and voter turnout. This work provides some evidence that citizens do exhibit retrospective voting behavior in local school board elections, but that the evidence of retrospective voting patterns is limited.  These findings have important implications for retrospective voting theory and for electoral accountability.