The Stubborn Unresponsiveness of Youth Voter Turnout to Civic Education: Quasi-Experimental Evidence From State-Mandated Civics Tests

36,627 respondent-wave observations (approx. 5,000 respondents per wave across seven waves)

US citizens aged 18-22 in the Current Population Survey Voting and Registration Supplement, presidential elections 1996-2020.

Difference-in-differences plus event study exploiting the staggered adoption of the Civics Education Initiative (CEI) - a state requirement that high-school students pass a test modelled on the USCIS naturalisation civics test - by 18 states from 2015 onward. State and year fixed effects, individual and state-level time-varying covariates, state-specific linear trends, placebo checks on 28-32 year olds, triple-difference models comparing 18-22 with 28-32 year olds, and separate 'strong' (graduation conditional on passing) versus 'weak' requirement arms. Parallel pre-trends verified visually and by event study. GRADED C, NOT B, FOR TWO REASONS: the outcome is SELF-REPORTED CPS turnout, not validated voter records, and the design is a staggered-adoption two-way fixed effects DiD, which is exactly the setting where TWFE estimators are known to be fragile. READ DEPTH CAVEAT: read in full as EdWorkingPaper 22-668 (November 2022); the version of record is Educational Evaluation and Policy Analysis 46(4):764-787 (2024), which was not re-read.

State-mandated civics tests do not move youth voter turnout. The largest DiD estimate is a statistically insignificant 1.5 percentage points, and the preferred specification gives -0.012. Event studies put zero inside the 95% interval in every post-period. Results are unchanged whether the requirement is 'strong' (passing conditions graduation) or 'weak', and hold under placebo checks, triple differences and comparisons against states with historically low prior civics requirements. One uncomfortable subgroup result: Black youth in CEI states show a 9.7 percentage point DECREASE in the probability of voting (p<0.10), which the authors flag with appropriate caution about power, and interpret against a literature warning that test mandates narrow civics teaching from discussion and activity toward fact memorisation. CEI did not affect high-school graduation rates either way. The authors' framing is the one this cluster needs: applied political learning beats memorising facts about government, and the gap between civic knowledge and civic engagement is persistent.

Medium. Identification is off state policy adoption, not individual choice, so individual heritable disposition is not the threat; the threats are state-level confounding policy shocks and the known instability of staggered TWFE. Since the finding is a NULL, confounding would have to be masking a real effect, which is the less likely direction.

Converges with Weinschenk and Dawes 2022 (civic coursework, family fixed effects, null on turnout) and with Manning and Edwards 2014 (systematic review: little evidence of effect on voting or registration). Together these are three independent designs on three data sources agreeing that civics instruction requirements do not move youth turnout. It also sits directly on top of Campbell and Niemi 2016, which found the same policy instrument DOES move knowledge - the cleanest available demonstration that the knowledge channel and the behaviour channel come apart.