Can Charter Schools Boost Civic Participation? The Impact of Democracy Prep Public Schools on Voting Behavior

1,060 student lottery applicants (first applicant per family), 2007-08 through 2015-16 lotteries; 5,792 parents

Applicants to Democracy Prep charter schools in New York City (predominantly Black and Latino, low-income Harlem and Bronx families), entering grades 6-11, old enough to vote in the 2016 presidential election.

Oversubscription lottery used as an instrument for enrolment. 35% of applicants were offered admission; 44% of winners vs 19% of losers enrolled, a 25-point first stage, so the LATE is roughly four times the raw ITT. Outcome is VALIDATED registration and turnout from the Catalist national voter database; unmatched records were coded as non-registrants/non-voters, which biases toward zero if match rates differ. Estimation is Bayesian; the authors report a 98% posterior probability that both effects are positive rather than frequentist significance, and explicitly acknowledge 'a considerable degree of statistical uncertainty'. Democracy Prep is the maximal civics treatment in this literature: passing the US Citizenship Test is a graduation requirement, students run voter-registration drives from kindergarten, seniors do a year-long 'Change the World' civic action capstone, and students testify to elected officials. Campbell (2019) calls these results 'probably the upper bound for a school's effect'. READ DEPTH CAVEAT: the APSR letter is paywalled; numbers here come from the Mathematica summary, the Education Next long-form write-up and Campbell 2019, which agree with each other.

Democracy Prep attendance raised voter REGISTRATION by about 16 percentage points and voter TURNOUT by about 12 percentage points in the 2016 presidential election (lottery-adjusted LATE; the raw lottery-winner vs lottery-loser difference was about 6 points on each). This is one of only two randomized results anywhere in the literature showing a schooling treatment moving validated turnout. Two things temper it. First, the uncertainty is large - the authors report a 98% Bayesian posterior probability of a positive effect rather than a conventional confidence interval. Second, the PARENT sample (n=5,792) showed no statistically significant effects, so the trickle-up story that the Kids Voting literature reports descriptively did not replicate under randomization here. Note also that unlike Boston charters, Democracy Prep moved REGISTRATION as well as turnout - consistent with the fact that its programme is literally about registering voters.

Low. Admission was by lottery among applicants, so heritable ability and parental civic disposition are balanced in expectation between winners and losers. Residual risk is differential voter-file match rather than selection.

Cohodes and Feigenbaum (Boston charters, n=9,560) independently reproduce the direction and roughly half the magnitude with a much larger lottery sample. Graded B rather than A despite the lottery design because the sample is small (1,060 applicants), the estimates are imprecise enough that the authors report a Bayesian posterior probability rather than a conventional interval, and the treatment is a SINGLE school network, so the estimate is of Democracy Prep, not of civics education.