Voter turnout has been studied through field experiments more thoroughly than almost any other area of political behaviour, which means the evidence is unusually good.
The experimental tradition
Randomised field experiments assigning voters to receive different contact and comparing turnout against validated voting records.
Which allows causal estimation, and hundreds of such experiments have been conducted.
Turnout can be verified from public records in many jurisdictions, which removes reliance on self-report.
What works
Personal contact is the most effective, with in-person canvassing producing the largest effects.
Which declines with impersonality — phone calls by volunteers work better than by paid callers, which work better than automated calls, which have essentially no effect.
Effect sizes are modest even for the best methods, generally a few percentage points among those contacted.
Social pressure
The most effective intervention identified, and the most ethically contested.
Messages indicating that voting is public record and that turnout will be observed produce substantially larger effects than standard encouragement.
Which works because voting carries social expectation, and it has been criticised as coercive.
Milder versions emphasising that voting is a public act produce smaller effects without the same objection.
Descriptive norms
Messages emphasising that many people vote outperform those emphasising that turnout is low.
Which follows from social norm research generally — describing a behaviour as common encourages it.
Campaigns emphasising apathy therefore work against themselves.
Planning
Asking people to specify when and how they will vote increases turnout.
Which is the implementation intention effect found across behavioural domains, and it costs nothing.
Registration
Frequently the larger barrier than the vote itself.
Automatic registration, where the state enrols eligible citizens, produces substantially higher registration than application-based systems.
Same-day registration is associated with higher turnout in comparative studies.
Which means administrative design affects turnout more than most persuasion does.
Convenience measures
Postal voting, early voting and extended hours have been studied with mixed results.
Which generally find modest effects, and in some studies effects concentrated among people who would have voted anyway, shifting method rather than increasing participation.
Universal postal voting produces larger effects than optional systems.
Compulsory voting
Produces the highest turnout by a wide margin where enforced.
Which is used in several countries with varying enforcement, and it raises questions about compelled participation and about the informedness of compelled votes.
Research comparing before and after adoption finds substantial effects on turnout and mixed effects on outcomes.
Who does not vote
Turnout differs systematically by age, income, education and residential stability in most democracies.
Which means the voting electorate differs from the eligible population, and policy responds to voters.
Addressing this is the argument for administrative reform rather than for persuasion, since the barriers are largely practical.
Election day arrangements
Whether elections are held on working days or at weekends, and whether the day is a holiday, vary between countries.
Which affects turnout in comparative studies, though the effects are modest and confounded with other differences.
Polling place proximity and waiting times have measurable effects, and inequalities in these have been documented within countries.
Mobilisation and demobilisation
Campaigns can also reduce turnout, deliberately or otherwise.
Negative campaigning effects on turnout have been studied with mixed findings, and the evidence does not support strong claims in either direction.
Habit
Voting once increases the probability of voting again, which has been demonstrated in follow-up studies of turnout experiments.
Which means interventions have effects beyond the election they target, and it strengthens the case for measures that get first-time voters to participate.
The habit effect is one of the more interesting findings, since it means the benefit of a modest one-off effect accumulates.
Youth turnout
Consistently lower than for older groups in most democracies.
Which is partly a life-cycle effect that resolves and partly a cohort effect that does not, and distinguishing them requires long-run data.
Residential mobility is a substantial contributor, since moving requires re-registration in many systems.
Automatic registration and registration at educational institutions address this directly.
Elections at different levels
Local and second-order elections have substantially lower turnout than national ones.
Which affects who is represented locally and produces results driven partly by national politics.
Holding elections concurrently raises turnout and reduces the distinctiveness of local contests, which is a genuine trade-off.
Measuring turnout
Reported as a proportion of registered voters or of the voting-eligible population, which produce very different figures.
Which means international comparison requires knowing which denominator is used, and countries with automatic registration look different under each.
Voting-eligible population is generally preferred for comparison, since registration rates vary with the system rather than with participation.
Practical implications
Registration reform, convenience measures and personal contact are the levers with evidence.
Which are administrative and organisational rather than rhetorical, and they are where effort produces measurable results.
Which is more useful than exhortation, and it is where the experimental evidence points consistently.