Representation Matters, But How? What We Learned from Studying California School Boards
Anya Dalal
Last Thursday, we presented a poster at the American Political Science Association's 122nd Annual Meeting in Boston. Our research asks a narrow question: Does the demographic composition of a California school board affect downstream outcomes that school boards control? While the project may appear like a traditional political science paper presented at a political science conference, its origin story actually lies with the California Association of Youth Commissions (CAYC).
When we founded the CAYC, one of our main goals was to help youth across California establish youth commissions in their own communities. With so many towns lacking youth commissions, we needed a way to determine where our efforts might make the greatest impact. We decided early on to prioritize municipalities where the gap between the adults governing education and the students actually sitting in the classrooms was widest. We reasoned that where a board least reflects the community it serves, perspectives that differ from those already represented, including those of young people, may be less likely to have a seat at the table.
To rank towns by that gap, we first had to find a way to measure it. Over two years, we assembled a database of California school boards and paired it with district-level demographics. The demographic side was difficult because the Census does not publish estimates around school district geography. Therefore, we developed techniques to draw boundaries around California school districts and apportioned American Community Survey block-group estimates into them, weighting them according to the share of each block group that falls inside a given district’s boundaries. For board members themselves (where no demographic data was available at all), we estimated race and gender probabilistically from names and counties of residence, averaging across three separate prediction packages: wru, predictrace, and rethnicity.
What started as a spreadsheet for deciding which city council to reach out to had become a district-by-year panel of California’s school boards. Once we had assembled the data set, we realized it could answer questions we had not originally thought about. For example, if you know the composition of every board over years and if you can link that to decision outcomes such as teacher compensation decisions, you can explore whether board composition has a relationship with specific outcomes.
To answer this question, we used two main methods. First, we used linear mixed-effects models estimated on the district-by-year panel, with outcomes measured from four years before an election through four years after. The pre-election years function as “placebo” tests: if effects appear before a board member is elected, something other than that election is driving them. The second method used was a close-election regression discontinuity design. In close elections where the last candidate to win and the first candidate to lose differ in race or gender, and the margin between them is very small, the one who gets ultimately elected could be considered random, which gets us closer to a causal estimate.
Across both approaches, most of our estimates cluster near zero. The close-election design, by construction, disregards every race that was not competitive, and the resulting loss of statistical power means our confidence intervals are wide enough that we cannot rule out effects that would matter a great deal in practice. What we can say is that we found no consistent, broad relationship between a board member's race or gender considered on its own and either teacher salary or student achievement, across grades, subjects, and demographic groups.
There were, however, some interesting patterns. In our panel models, Hispanic representation was associated with higher teacher compensation two and four years after an election. When we examined the interaction between race and gender, that signal was concentrated specifically in districts that elected Hispanic women. This finding should be taken with a grain of salt; we tested a lot of outcomes, increasing the possibility that some statistically significant findings arose by chance. Still, the pattern raises an interesting possibility, that if representation shapes what school boards do, its effects may not operate uniformly across demographic categories or policy areas. Instead, they may emerge at the intersection of particular identities and particular policy domains.
Earlier work in California, using similar data and a similar design, reached a more optimistic conclusion. Kogan et al. (2021) found that electing minority board members improved achievement among non-white students, with gains accumulating roughly six years after the election. Their elections ran from 1998 through 2014; ours ran from 2010 through 2024. We do not reproduce their result in the more recent period. Our differences could stem from different ways of measurement, or an actual change in what school boards accomplish over the years, but regardless, it is a caution against treating representation effects as a fixed constant of local governance rather than something that heavily depends on context and timing.
It would be easy to read a page of estimates clustered around zero as a definitive conclusion. However, it’s important to remember that a null research result is a statement about what we can detect, not about what exists. Therefore, none of this changes what we set out to do nor how we do it. Our commitment to helping California towns that want youth commissions to build one continues, and we will continue prioritizing the amplification of youth voices. Since our CAYC founding three years ago, I am proud of the fact that we have helped 15 youth commissions get going, and we are actively helping more!
Image credit: 74 Million