Testing Civics: State-Level Civic Education Requirements and Political Knowledge

2006 and 2010 NAEP Civics assessments (high-school students) plus a large national survey of 18-24 year olds

US high-school students taking NAEP Civics, and US 18-24 year olds classified by the civics requirements of the state where they graduated.

Cross-state comparison, not a clean shock. The authors test an 'incentive hypothesis' (civics exams matter most where passing is a graduation requirement) and a 'compensation hypothesis' (civics requirements matter most for young people with least political exposure at home - Latinos and immigrants). They defend causal interpretation by arguing the policy is too low-salience for families to sort on, and by controlling for state NAEP scores in maths and reading so the result is not just 'high-achieving states'. That is reasonable but it is not a difference-in-differences with a clean shock, so it is graded C. READ DEPTH CAVEAT: read via abstract, Campbell's own 2019 review (which restates the findings and their limits), and citations in Lindgren et al. 2019 and Jung and Gopalan; effect sizes and standard errors were not extracted.

Modest support for the incentive hypothesis and strong support for the compensation hypothesis - and knowledge only. Introducing a high-stakes civics assessment as a graduation requirement raises NAEP Civics scores for LATINO students, and, among 18-24 year olds, graduating in a high-stakes-civics state predicts higher political knowledge for all Latinos, for first- and second-generation immigrants, and most of all for Latino immigrants. The effect for students who already get political information at home is modest at best. CRITICALLY FOR THIS CLUSTER: the same paper finds the assessment requirement has NO effect on attitudes or on participation. Campbell, writing later, groups this with Green et al. 2011 as the pair of studies showing civic education moving knowledge and nothing else, and asks the right question: 'Is it enough to boost what students know without changing what they think?'

Medium to high for the causal reading. States are not randomly assigned and family background is not held constant within state; the defence is a plausibility argument about policy salience plus controls for state maths and reading performance. The compensation pattern (effects concentrated where home political socialisation is weakest) is at least the pattern a genuine schooling effect should have, since it is where school has the most environmental slack to fill.

The knowledge finding has not been directly replicated. Its BEHAVIOURAL implication has been tested and rejected: Jung and Gopalan 2024 apply a difference-in-differences design to the same class of policy (state-mandated civics tests) and find no effect on youth turnout. Campbell and Niemi themselves report no effect on attitudes or participation, so the two papers are consistent - together they are the cleanest single demonstration that the same policy lever moves knowledge and not behaviour. Note also that Carlson 2012 and Kawashima-Ginsberg and Levine 2014 had found state-level civics policies to have NO effects on civic outcomes.