2024 turnout differed by about 2 points between strict-photo-ID and no-ID states — not a statistically significant gap

Summary

In the 2024 general election, states with strict photo-ID voting requirements averaged 64.2% voter turnout (voting-eligible-population basis), versus 66.2% in states that require no ID at all at the polls — a raw gap of about 2 points. That gap is not statistically significant (Welch t-test, p=0.50) given only 9 strict-ID states and 17 no-ID jurisdictions, and it shrinks further, to +0.5 points and still non-significant (p=0.83), once state political lean is controlled for. The honest answer is: the ID-requirement difference between these state groups is small, in the noise, and does not survive as a clean causal story once you account for confounds like partisanship — consistent with the mixed empirical literature on this question, most of which finds effects concentrated among specific subgroups (e.g. racial-minority voters) rather than in aggregate state-level turnout.

Why VEP turnout, not a survey or a raw vote count

Two other 'turnout' measures are available and both would mislead here. The Census CPS Voting and Registration Supplement is self-reported and known to systematically OVER-report actual turnout by roughly 10 points (social-desirability bias); it also uses citizen-voting-age population, not voting-eligible population, as its denominator. Raw ballots-counted totals divided by census voting-age population ignore that voting-age population still includes non-citizens and, in many states, disenfranchised felons. The standard the political-science literature actually uses is voting-eligible population (VEP) — voting-age population minus ineligible non-citizens and felons, plus eligible overseas voters — a measure developed and maintained by Dr. Michael McDonald (originally at the US Elections Project, now hosted by the University of Florida Election Lab). This analysis pairs each state's 2024 certified total ballots counted with McDonald's VEP denominator, taken directly from the Election Lab's published 2024 general-election turnout dataset (v0.4).

How states were classified by voter-ID strictness

Following the National Conference of State Legislatures' (NCSL) classification (as compiled by worldpopulationreview.com, current as of 2026), states fall into four groups: strict photo ID (a voter without acceptable photo ID must vote provisionally and take further action after Election Day for it to count) — Georgia, Indiana, Kansas, Mississippi, Nebraska, North Carolina, Tennessee, West Virginia, and Wisconsin (n=9); strict or non-strict non-photo ID, or non-strict photo ID (all other ID-requiring states, n=26); and no ID required at all — California, DC, Hawaii, Illinois, Maine, Maryland, Massachusetts, Minnesota, Nevada, New Jersey, New Mexico, New York, Oregon, Pennsylvania, Vermont, and Wyoming (n=16+DC=17).

Results

The raw gap between the no-ID and strict-ID groups is 2.03 points (Welch's t=0.70, p=0.50) — well within sampling noise given the small number of states in each bucket. Individual states span a wide range within each group: strict-ID Wisconsin had the highest turnout of all 50 states (76.7%), while strict-ID West Virginia (55.6%) and no-ID Hawaii (50.3%) were both near the bottom — the within-group spread (roughly 20+ points) dwarfs the ~2-point between-group gap.

To check whether the raw gap reflects the ID policy itself rather than a confound, an OLS regression of 2024 VEP turnout on a strict-ID indicator plus each state's Composite Political Index (CPI, a -100 to +100 partisan-lean score built from recent presidential, Senate, and House results, from AskAmerica's officials.state_political_index) was run across all 50 states. The strict-ID coefficient falls to +0.49 points (p=0.83) once political lean is included — political lean itself is the stronger, statistically significant predictor (coefficient +0.025 turnout points per CPI point, p=0.01), consistent with more Democratic-leaning states also tending to have looser ID rules and higher turnout for reasons unrelated to the ID law itself. A leave-one-out sensitivity check (refitting the two-group comparison 51 times, once per state omitted) found no sign flip and no state whose removal crosses the p=0.05 threshold; Wisconsin is the single most influential state (removing it would widen the strict-ID penalty to about -2.9 points, still not significant), consistent with it being both strict-ID and the single highest-turnout state in the country.

What the broader research literature says

The empirical literature on voter-ID laws and turnout is genuinely mixed, and this state-level aggregate result is consistent with that literature's more careful findings, not a contradiction of it. Early studies (e.g. work summarized by the Brennan Center and MSU's IPPSR) found strict ID laws widened racial turnout gaps — one widely cited estimate found Latino turnout roughly 10 points lower and multiracial-American turnout roughly 13 points lower in strict-ID states, even though the AGGREGATE state effect was small, because ID possession rates are not evenly distributed across demographic groups. A 2019 NBER working paper (Cantoni & Pons) similarly concluded that strict ID laws have little detectable effect on aggregate turnout because so few eligible voters actually lack acceptable ID. A 2025 study (Lipkovitz, Research & Politics) found strict ID laws differ by election type and adoption timing, and in some specifications found strict ID laws associated with HIGHER midterm turnout (+2.9 points), plausibly reflecting mobilization/backlash effects or that adopting states already had unusually strong turnout trends. In short: at the state-aggregate level the ID-requirement effect is small and easily swamped by noise and confounds (exactly what this analysis found); any real effect the literature has reliably identified operates at the level of specific subgroups least likely to hold ID, not the state as a whole.

Caveats

• n=9 and n=17 (and n=51 for the regression) are small samples; standard errors are wide and this design cannot rule out a true effect of a few points in either direction — it can only say the state-aggregate 2024 data do not show a significant one.
• This is a single-election (2024), state-aggregate, cross-sectional comparison. It does not test within-state before/after effects of adopting a strict ID law, nor does it isolate the subgroup-level effects (e.g. by race) that the literature has more reliably found — that would require individual- or precinct-level data with demographic breakdowns, which is outside the scope of this analysis.
• Voter-ID classification reflects laws in effect for the 2024 cycle as compiled by NCSL/worldpopulationreview (current as of 2026); a few states (e.g. West Virginia) adopted strict ID requirements only in 2025, after the 2024 election covered here, so the classification used here describes the current (2026) legal landscape applied retroactively to the 2024 election it is closest to, not necessarily the exact law in force on 2024 Election Day in every state.
• Political lean (CPI) is one plausible confound; it is not an exhaustive control. Other correlated factors (competitiveness of the 2024 race in that state, vote-by-mail availability, early-voting window length) were not modeled and could also explain part of the residual variation.