Sondheimer and Green: Using Experiments to Estimate the Effects of Education on Voter Turnout

Sondheimer, Rachel Milstein; Green, Donald P. 2010. Using Experiments to Estimate the Effects of Education on Voter Turnout. American Journal of Political Science 54 (January): 174-189.

Suppose you're in a room full of people and you want to know which of them are most likely to be active voters, but you're not allowed to ask them about their political activity. The best question you can ask them: How many years of schooling they have. We've known for many years that education is among the best predictors of voting (Wolfinger and Rosenstone 1980).

But what hasn't been clear until now is whether education caused voting, or whether it was merely correlated with voting. After all, education is caused by family background (parents' education level, family wealth) and personal characteristics (intelligence). Does education cause voting, or do the things that cause education also cause voting? A major knock against the "education as cause" theory came when Brody (1978) pointed out that education levels have risen dramatically since the 1960s, but turnout has not.

So how can we figure out whether education causes turnout? Well, shoot, what if we did an experiment that randomly caused one group of kids to get more education than a control group of their peers? Then we could just see whether those who were randomly induced to get more education also ended up voting at higher rates.

Genius. In the current issue of AJPS, Sondheimer and Green have an article that does exactly that.

Experiments and Results

Sondheimer and Green dig up three old studies from the education literature. All three studies used randomized experiments to see whether certain treatments would increase high school graduation rates.

• The Perry Preschool Experiment took a randomly selected group of disadvantaged kids and gave them an intensive preschool experience in the 1960s. The treatment group's graduation rate was 46% higher than the control group's. And its turnout rate in 2000/2002 was 44% higher than the control group's.
• The "I Have a Dream" Foundation of Boulder, Colorado, identified a bunch of fifth-graders in 1992 and offered them tutoring, extracurricular activities, mentoring, and other assistance from the time of their selection through high school. Once again, a treatment group participated while a control group did not. The treatment group's graduation rate was 28% higher than the control's; the treatment group's turnout rate through 2004 was 26% higher than the control's.
• The Tennessee STAR experiment randomly assigned kids entering kindergarten in 1985 to have regular class sizes (22-25 students) or small class sizes (13-17 students) from kindergarten through third grade. Those in the treatment group graduated at a rate 6% higher than the control; they voted at rates 10-11% higher than the control.

You can look up Sondheimer and Green's article to see the advanced statistical analysis, but the percentages above tell the whole story.

What we Learn

Education does, indeed, have a robust causal effect on voter turnout. This finding is all the more striking because the authors did not expect it. Both authors had previously argued that education's correlation with turnout was probably spurious. But after conducting this analysis, they change their minds. As they put it, "The data presented here have led to a reversal of this assessment."

So we know that education causes turnout. It's not just correlated. We still don't know why. Maybe education gives kids the skills they need to figure out how to vote. Maybe it promotes interest in politics. Maybe it expands kids' social networks. Maybe it increases their confidence, or "efficacy." Maybe it increases their later affluence, hence their political interests. We still don't know. All we know is that education does have a genuine, strong, and robust causal effect on turnout.