PROCEDURAL POLARIZATION AND CONGRUENCE

Especially following the 2000 election, partisan elites have increasingly promoted divergent narratives about procedural changes needed to improve the integrity and/or accessibility of election (Hasen Reference Hasen2012). Democrats typically champion reforms that expand access, for example all-mail voting, while Republicans tend to emphasize concerns about fraud and advocate for stricter rules like voter ID laws (Bentele and O’Brien Reference Bentele and O’Brien2013). A large literature tracks procedural changes over time (e.g., Bentele and O’Brien Reference Bentele and O’Brien2013; Schraufnagel, Pomante, and Li Reference Schraufnagel, Pomante and Li2022) with the aim of assessing how procedural changes affect the cost of voting (Pomante Reference Pomante2024) or level of “democratization” (Grumbach Reference Grumbach2023). Others show how voters’ perceptions of electoral fairness depend on opinions about election procedures (Bowler et al. Reference Bowler, Brunell, Donovan and Gronke2015). I build on these works and ask: regardless of why they are enacted (Burden et al. Reference Burden, Canon, Mayer and Moynihan2011; Caron Reference Caron2022; Grumbach and Hill Reference Grumbach and Hill2023; Smith, Hill, and Ancheva Reference Smith, Hill and Ancheva2023), how does the portfolio of election procedures being used in a state affect voters’ confidence in the accuracy of the state’s vote count?

To assess public support for election procedures, I combine data from the Survey on the Performance of American Elections (SPAE) (Stewart Reference Stewart2023)—a postelection survey of nearly 200 voters in each state in every election cycle since 2008—with data from a postelection survey I conducted in 2024 that asked comparable questions about actual and hypothetical election procedures (see Appendix A of the Supplementary Material for question wording and sample sizes). Figure 1 presents the average support for each policy pooled over time by self-reported partisanship (with 110,000 respondents, the 95% confidence intervals are smaller than the plotted points).Footnote ¹

Figure 1.

Pooled Average Partisan Support for Election Procedures, 2008–24

Note: To minimize the effects of changing demographics, each year is weighted to match the demographics of the 2020 electorate using the Voter Supplement of the Current Population Survey. With 110,000 respondents, the 95% confidence intervals are smaller than the plotted points.

Figure 1. Long description

A dot plot with the x-axis labeled Percent Overall Support from 10 percent to 90 percent. A vertical dashed line marks the 50 percent threshold. The y-axis lists 18 election procedures. Data points are represented by black circles for Democrat, dark gray diamonds for Republican, and light gray squares for Other.

* Require paper ballot backup: All groups cluster near 85 to 90 percent.

* Allow early in-person voting: High support across groups, with Democrats highest at 85 percent and Republicans at 78 percent.

* Require govt issued photo I D: Republicans lead at over 90 percent, Others at 78 percent, and Democrats at 65 percent.

* Registration portability: Democrats at 88 percent, Republicans at 62 percent.

* Allow mail voting by choice: Democrats at 90 percent, Republicans at 60 percent.

* Non-partisan election officials: Democrats at 71 percent, Republicans at 68 percent.

* Hand count votes: Republicans at 75 percent, Democrats at 58 percent.

* Allow vote drop boxes: Democrats at 82 percent, Republicans at 50 percent.

* Re-register released felons: Democrats at 76 percent, Republicans at 54 percent.

* Make Election Day a national holiday: Democrats at 78 percent, Republicans at 51 percent.

* Automatically reg all citizens over 18: Democrats at 77 percent, Republicans at 39 percent.

* Allow Election Day reg at the polls: Democrats at 75 percent, Republicans at 33 percent.

* Move Election Day to a weekend: Democrats at 63 percent, Republicans at 40 percent.

* Use ranked choice voting: Democrats at 59 percent, Republicans at 31 percent.

* Prohibit computers from counting votes: Republicans at 56 percent, Democrats at 36 percent.

* States remove inactive voters: Republicans at 50 percent, Democrats at 35 percent.

* Allow internet voting: Democrats at 48 percent, Republicans at 20 percent.

* Run all elections by mail: Democrats at 41 percent, Republicans at 15 percent.

* Allow cellphone voting: Democrats at 30 percent, Republicans at 11 percent.

In general, Republicans express greater support for election security-related policies, whereas Democrats express greater support for policies aimed at expanding access (Gronke et al. Reference Gronke, Hicks, McKee, Stewart and Dunham2019; Pew Research Center 2024). Some policies receive broad bipartisan support: both parties favor requiring a paper ballot backup, allowing in-person early voting, and requiring photo identification to vote. Others face bipartisan opposition, including cell-phone or internet voting, conducting all-mail elections, and removing inactive voters from registration rolls. Support for other policies divides along partisan lines: most Democrats support automatic and same-day voter registration, while a majority of Republicans oppose both. (See Appendix B of the Supplementary Material for trends in public opinion over time; with the exception of a temporary increase in Democratic support for all-mail voting during the 2020 election, opinion shifts are small and show no clear relationship to party positioning or policy adoption.)

To track changes in state election procedures, I focus on policies that both exhibit temporal variation and have corresponding measures of public support. I construct an additive index where states receive one point each for using default all-mail voting, online registration, in-person early voting, no-excuse absentee voting, same-day registration, and post-sentence felon re-enfranchisement and lose one point for imposing any voter ID requirement and an additional point if the requirement is classified as strict.Footnote ²

To characterize growing procedural polarization, I plot state-election procedure scores against 2024 presidential vote shares, illustrating how election procedures have changed over time relative to a fixed partisan baseline.Footnote ³

Figure 2 highlights three relevant trends. First, procedural changes accelerate in the late 2000s. Reflected in the growing dispersion of the index, the standard deviation of scores increases from 0.98 in 1996 to 2.03 by 2024. Second, the fact that more movement over time occurs in states with margins greater than zero means that many more changes in these policies occurred in states that supported Vice President Harris over President Trump in the 2024 election. Third, election procedures have become increasingly aligned with state partisanship, shifting from almost no association in 1996 to a correlation of 0.69 by 2024.

Figure 2.

Election Procedures and State Partisanship, 1996–2024

Note: Political environment is measured by the 2024 Democratic–Republican presidential vote share. See Appendix C of the Supplementary Material for additional characterizations, for example, Figure A10 in the Supplementary Material uses contemporaneous presidential margin to measure partisan leanings and Figure A11 in the Supplementary Material using the annual cost-of-voting index scores to measure election procedures (Pomante Reference Pomante2024).

Figure 2. Long description

The grid consists of two rows and four columns. The Y-axis for all panels is Procedure Index, ranging from negative 2.5 to 5.0. The X-axis is Democratic Vote Share Margin in 2024, ranging from negative 40 percent to 20 percent, with a vertical dashed line at 0 percent.

* Top Row (1996 to 2008): The correlation coefficient R increases slightly from 0.033 in 1996 to 0.13 in 2008. The regression lines show a very shallow positive slope, and the data points are widely dispersed with little visible clustering.

* Bottom Row (2012 to 2024): The correlation coefficient R shows a significant upward trend. In 2012, R equals 0.3. In 2016, R equals 0.43. In 2020, R equals 0.58. By 2024, R reaches 0.69.

* Trend Analysis: The regression lines become progressively steeper over time. In the 2024 panel, the data points show a much tighter linear alignment, indicating that states with higher Democratic vote shares in 2024 are increasingly associated with higher scores on the election procedure index compared to earlier years.

PROCEDURAL CONGRUENCE AND VOTER CONFIDENCE

Given partisan divisions in both public opinion and the election procedures states employ, a foundational question emerges: does voter confidence in election outcomes depend on whether voters agree with the procedures being used (Bowler and Donovan Reference Bowler and Donovan2016)? On the one hand, theories of legitimacy suggest that individuals are more likely to view outcomes as legitimate when they result from rules they support and trust (Easton Reference Easton1975; Hibbing and Theiss-Morse Reference Hibbing and Theiss-Morse2002; Tyler Reference Tyler2006). On the other hand, election procedures are often obscure, raising questions about whether voters hold sufficiently developed views for procedural agreement to meaningfully shape confidence, and about the salience of those views relative to considerations, such as whether one’s preferred candidate won, the state’s broader partisan context, or individuals’ ideological commitments.

What it means for voters to “agree” with the election procedures being used is not self-evident. Debates over election administration routinely focus not only on which rules are adopted, but also on which rules are absent—for example, disputes over the presence or absence of voter identification requirements (Adeosun Reference Adeosun2025) or same-day registration (Williamson and Rhodes Reference Williamson and Rhodes2021). To address this ambiguity, I consider two complementary conceptions of procedural congruence.

First, I conceptualize congruence as the extent to which a state’s overall portfolio of adopted and unadopted election policies aligns with a respondent’s preferences. This approach allows the absence of a policy (such as the lack of voter identification requirements) to matter for voter confidence in the same way as its presence.

Second, I examine congruence in terms of respondents’ support for—or opposition to—the specific election procedures currently in use in their state. Defining congruence in this way makes it possible to compare whether opposing a policy being used has a larger negative association with confidence than the positive association of supporting that policy.

To assess whether and how procedural congruence relates to voter confidence, I employ two empirical strategies. First, I leverage within-state changes in election procedures over time to examine how voters’ confidence in the accuracy of their state’s vote count varies with the procedures in use and voters’ support for those procedures. While causal interpretations are limited by the nature of the data, analyses across multiple measures and specifications reveal a consistent association between procedural agreement and voter confidence which also replicate the well-established relationship between voter confidence and electoral outcomes at the state and national levels.

Second, to address key limitations of observational data—including concerns about voter awareness, endogenous opinion change, and unobserved confounding—I field a survey experiment prior to the 2024 election that varies procedural conditions and political context. Randomizing the election procedures and partisan context used in hypothetical electoral scenarios corroborates the association between procedural agreement and voter confidence observed in the observational analyses.

OBSERVATIONAL ANALYSES

The need to measure public support for election procedures whose use varies over time limits the observational analysis to four policies: automatic registration, same-day registration, photo ID, and default all-mail voting. Although these policies represent a subset of the procedures used by states, they capture substantively meaningful variation in election administration. Not only do partisans differ in their support for these policies, but the policies themselves vary in their partisan associations: some are more closely aligned with Democratic elites (e.g., all-mail voting), others with Republican elites (e.g., photo identification), while elite positioning on still others (e.g., same-day registration) is less clearly defined (Burden and Gaines Reference Burden and Gaines2015).

I measure individual-level support for each procedure using postelection survey items summarized in Figure 1, coding respondents as supportive (1) if they report “strong” or “somewhat” agreement and as not supportive (0) otherwise.

Building on the conceptual discussion above, I then operationalize procedural congruence in two ways by combining these respondent-level policy preferences with indicators for whether each procedure was used in the respondent’s state in the most recent election, as coded in the Cost of Voting Index (COVI) dataset (Pomante Reference Pomante2024).Footnote ⁴

Under the first measure, a respondent is considered congruent with a given policy if they either support a procedure that is used in their state or oppose a procedure that is not used. Thus, a respondent who favors photo identification in a state that requires it is treated as equally congruent as a respondent who opposes photo identification in a state that does not. Incongruence arises when a respondent supports a policy not in use or opposes a policy their state implements.

I also construct an alternative measure of congruence based solely on respondents’ agreement with the procedures currently used in their state. This measure counts only support for policies in use and does not treat opposition to absent policies as congruent.

The outcome of interest is whether respondents are “very” or “somewhat” confident when asked: “Now, think about vote counting throughout [STATE]. How confident are you that votes in [STATE] were counted as voters intended in the 20XX general election?”Footnote ⁵

Figure 3 plots average voter confidence against average procedural agreement over time by partisanship. Procedural congruence is positively, but modestly, associated with confidence in the accuracy of a state’s vote count, especially among Republicans. Consistent with established effects of winning on voter confidence, large intercept shifts in partisans’ average confidence in 2016, 2020, and 2024 track national presidential outcomes.

Figure 3.

Trust in State Election Results by Procedural Congruence

Note: Average confidence in the accuracy of state vote counts is plotted against average agreement with the election procedures in use, by partisanship and election cycle. To reduce the effects of compositional changes in the sample over time, each year is weighted to match the 2020 electorate. Figure A16 in the Supplementary Material reports unweighted results.

Figure 3. Long description

A multi-panel line graph with six panels arranged in two rows of three, representing election years 2012, 2014, 2016, 2020, 2022, and 2024. The Y-axis represents percentage confident in accuracy of state results, ranging from 60 percent to 100 percent. The X-axis represents percentage procedural agreement, ranging from 0 percent to 100 percent. Each panel contains three lines with shaded confidence intervals.

* Democrat (black line with circles): Consistently maintains the highest confidence levels, generally between 88 percent and 96 percent across all years, showing a slight positive correlation with procedural agreement.

* Republican (dark gray line with diamonds): Shows significant volatility. In 2012 and 2014, confidence is stable around 80 percent. In 2020, confidence drops sharply to a low of 57 percent at 75 percent procedural agreement before rebounding. By 2024, it shows a strong positive linear trend from 77 percent to 92 percent.

* Other (light gray line with squares): Generally tracks between the Democrat and Republican lines. In 2020, it shows a notable dip to 72 percent at the 75 percent agreement mark, similar to the Republican trend but less extreme.

Across all panels, higher procedural agreement generally correlates with higher confidence, though the gap between partisans widened significantly starting in 2020.

To better assess the relationship between procedural congruence and voter confidence, I leverage within-state changes in election procedures to estimate how policy-induced changes in procedural congruence are associated with the probability that a voter expresses confidence in the accuracy of their state’s vote count. Because partisanship is known to shape how individuals interpret political environments, I estimate the relationship separately for Democrats, Republicans, and unaffiliated voters. This approach allows the estimated associations to vary flexibly across partisan groups, reflecting possible differences in baseline trust, responsiveness to electoral outcomes, and the salience of election procedures.Footnote ⁶

More specifically, I estimate:Footnote ⁷

⁽¹⁾ $$ {\displaystyle \begin{array}{l}Pr({Y}_i=1)={\theta}_0{\mathrm{PctCongruent}}_i+{\alpha}_1{\mathrm{StateWon}}_i\\ {}+{\beta}_0{\mathrm{Personal}\ \mathrm{Ideology}}_i+{\beta}_1{\mathrm{AllMail}}_i\\ {}\hskip5.13em +{\beta}_2{\mathrm{AutoReg}}_i+{\beta}_3{\mathrm{PhotoID}}_i\\ {}\hskip5.2em +{\beta}_4{\mathrm{SameDayReg}}_i\\ {}\hskip5.25em +{\beta}_5{\mathrm{Dem}\hbox{-}\mathrm{Rep}\ \mathrm{Margin}}_j+{\mathbf{X}}_i\Gamma \\ {}\hskip5.2em +{\delta}_{time}+{\delta}_{state,}\end{array}} $$

where Y _i is a binary variable indicating whether respondent i believes that elections are accurately counted in their state j. The primary independent variable, PctCongruent, measures the proportion of the four state-level voting policies that align with a respondent’s expressed support, scaled from 0 to 1. The associated coefficient, $ {\theta}_0\hskip-0.15em $ , reflects the average difference in voter confidence between a respondent who is perfectly congruent (1) and one who is fully incongruent (0) all else equal. Given the importance of winning for voter confidence, the coefficient on StateWon captures the effect of living in a state won by presidential candidate of the respondent’s party (coded only for Democrats and Republicans), and $ {\delta}_{time} $ are election fixed effects that account for the impact of winning at the national level along with other time-varying effects.Footnote ⁸ As confidence itself may be related to the overall ideological orientation of a voter, $ {\beta}_0 $ measures whether confidence covaries with self-reported liberalism (1) and conservatism (5). Because voter confidence may also be related to voters’ opinions about specific policies—that is, those who support all-mail voting by default may be more trusting of election outcomes than those who support photo identification—the coefficients β ₁ through β ₄ allow voter confidence to vary by a voter’s support for each procedure. Individual demographics ( $ {\mathbf{X}}_i $ ) allow for additional variation in voter confidence related to variation in race/ethnicity, gender, education level, and age. State fixed effects $ {\delta}_{state} $ account for between-state differences—meaning that the identified effects of procedural congruence are a result of within-state differences across co-partisans. Standard errors are clustered by state to account for shared contextual factors.

Figure 4 summarizes the estimated coefficients from models estimated separately for Democrats, Republicans, and unaffiliated voters.

Figure 4.

Correlates of Confidence in Own-State Election Results

Note: Estimated coefficients for Equation 1. 95% confidence intervals are based on standard errors clustered by state. Complete results are reported in Table A2 in the Supplementary Material. See Appendix E of the Supplementary Material for alternative specifications.

Figure 4. Long description

A coefficient plot with a vertical dashed line at zero. The X-axis is labeled Coefficient Estimate and ranges from negative 0.1 to 0.3. The Y-axis lists variables from top to bottom.

* Legend at the bottom: Black circles represent Democrat, dark gray diamonds represent Republican, and light gray squares represent Other.

Data points with 95 percent confidence interval bars include:

* % Procedural Agreement: All groups are positive, with Other being the highest at approximately 0.21.

* Party Pres. Cand. Won State: Republicans show a strong positive correlation at 0.31, while Democrats are near 0.05.

* Personal Ideology (1–5): All groups are slightly negative, clustered near negative 0.02.

* Supports Photo I D: Republicans are slightly negative, while Democrats are slightly positive.

* Supports All-Mail Voting: Republicans and Others are positive (0.11), while Democrats are near 0.02.

* Supports Auto Reg.: All groups are positive, with Other and Republican around 0.07.

* Supports Same Day Reg.: All groups are positive, with Other and Republican around 0.05.

* Year 2014: All groups are clustered at zero.

* Year 2016: Republicans are positive (0.07), while Democrats are negative (negative 0.05).

* Year 2020: Republicans are strongly negative (negative 0.12), while Democrats are positive (0.03).

* Year 2022: Republicans are slightly negative, while Democrats are slightly positive.

* Year 2024: All groups are slightly positive, clustered between 0.02 and 0.07.

Consistent with prior work, winning has a large association with voter confidence at both the state and national levels, with particularly pronounced effects among Republicans. Republicans living in states carried by the Republican presidential candidate are 0.32 more likely to express confidence in the accuracy of the state vote count than Republicans living in Democratic-carried states, whereas the corresponding difference among Democrats is considerably smaller (0.06). Confidence also tracks national presidential outcomes: Republican respondents report higher confidence following the Republican victories in 2016 and 2024 and lower confidence following the Democratic victory in 2020, while Democratic respondents exhibit lower confidence in 2016 and higher confidence in 2020.

The coefficients on personal ideology and policy support reveal respondent-level differences in voter confidence that are independent of the procedures used in a respondent’s state. Voters who self-report as more conservative express lower confidence in the accuracy of their state’s vote count on average, as do respondents who support photo identification laws. In contrast, respondents who favor ballot access expanding procedures—such as same-day registration, automatic voter registration, and all-mail voting—report higher average levels of confidence.

Of primary interest, however, is the relationship between procedural congruence and voter confidence. Figure 4 shows a positive association between procedural congruence and voter confidence that varies across partisan groups. The effect of procedural congruence is largest among unaffiliated voters, for whom moving from complete disagreement to full agreement increases the average predicted probability of expressing confidence in their state’s results by 0.21.

Among partisans, moving from complete disagreement to full agreement over the use of these four procedures is predicted to change the probability of expressing confidence by 0.15 among Republicans and 0.05 among Democrats. Compared to the effect of winning, the total effect of procedural congruence is roughly half as large among Republicans and comparable in magnitude among Democrats.

The asymmetric partisan associations are intriguing, but the underlying mechanisms are difficult to identify given the nature of the data. Three of the four policies—all-mail voting, automatic voter registration, and same-day registration—are more commonly supported by Democrats. As Figure 3 shows, Democrats report higher baseline confidence than Republicans, which may attenuate the marginal effect of procedural congruence due to ceiling effects. Finally, Republicans and unaffilated respondents may be more responsive to procedural (mis)alignment because of heightened elite rhetoric surrounding election administration; Republican leaders have been significantly more vocal than Democratic leaders in criticizing election procedures and raising doubts about their legitimacy, for example, Republican messaging in 2024 regarding “Stop the Steal” and “Too Big to Rig” (Cortellessa Reference Cortellessa2024). Regardless of the mechanism(s), the results suggest that voters’ confidence is not merely a function of who wins, but it also matters whether voters agree with the procedures being used.Footnote ⁹

Voter confidence may depend less on broad alignment with a state’s overall policy portfolio than on reactions to the specific election procedures actually in use. To evaluate this possibility and to distinguish the effects of supporting a procedure from opposing one, I construct an alternative measure of procedural congruence based solely on respondents’ agreement with the procedures used in their state. I then estimate the probability that respondent i expresses confidence in the accuracy of their state’s vote count by leveraging within-state procedural variation and estimating:

(2) $$ {\displaystyle \begin{array}{l}Pr\left({Y}_i=1\right)={\alpha}_0+{\beta}_1{\mathrm{SupportPhotoID}}_i+\hskip2px {\beta}_2{\mathrm{SupportAllMail}}_i\\ {}+{\beta}_3{\mathrm{SupportMailChoice}}_i\\ {}+\hskip2px {\beta}_4{\mathrm{SupportInPersonEarly}}_i+{\gamma}_1{\mathrm{PhotoID}}_i\\ {}+\hskip2px {\gamma}_2{\mathrm{AllMail}}_i+{\gamma}_3{\mathrm{EarlyMail}}_i\\ {}+\hskip2px {\gamma}_4{\mathrm{InPersonEarly}}_i\\ {}+{\delta}_1\left({\mathrm{PhotoID}}_i\times {\mathrm{SupportPhotoID}}_i\right)\\ {}+\hskip2px {\delta}_2\left({\mathrm{AllMail}}_i\times {\mathrm{SupportAllMail}}_i\right)\\ {}+\hskip2px {\delta}_3\left({\mathrm{EarlyMail}}_i\times {\mathrm{SupportMailChoice}}_i\right)\\ {}+\hskip2px {\delta}_4\left({\mathrm{InPersonEarly}}_i\times {\mathrm{SupportInPersonEarly}}_i\right)\\ {}+\hskip2px {\delta}_{time}+\delta state.\end{array}} $$

Because $ \boldsymbol{\gamma} $ captures the conditional association between a procedure and voter confidence among respondents who oppose that procedure, and $ \boldsymbol{\gamma} +\boldsymbol{\delta} $ captures the corresponding association among those who support it, I expect $ \boldsymbol{\gamma} <0 $ and $ \boldsymbol{\gamma} +\boldsymbol{\delta} >0 $ . However, there is little theoretical guidance regarding the relative magnitudes of these effects; it is unclear whether opposition or support for a procedure in use has a larger association with voter confidence and whether this varies by policy.

Figure 5 summarizes these relationships and shows that, in general, opposing policies in use is associated with larger reductions in voter confidence than the increases associated with supporting those policies. Regardless of partisanship, voters who oppose same-day registration, automatic registration, or photo identification report lower confidence in the accuracy of state election results when those policies are used compared to similar voters in states where the policies are not in effect. In contrast, supporters of each policy express similar levels of confidence regardless of whether the policy is in use. The largest positive association occurs among Republican supporters of all-mail voting, but only 14% of Republicans express such a preference and those preferences are contrary to most Republican elite messaging (Clinton et al. Reference Clinton, Lapinski, Lentz and Pettigrew2022).Footnote ¹⁰

Figure 5.

Effects of Supporting and Opposing Election Procedures on Voter Confidence

Note: The figure plots the conditional marginal association between policy use and voter confidence for respondents who oppose a procedure ( $ \boldsymbol{\gamma} $ ) and for those who support it ( $ \boldsymbol{\gamma} +\boldsymbol{\delta} $ ), as estimated in Equation 2. Standard errors are clustered by state.

Figure 5. Long description

The two panels are stacked vertically. The top panel is titled Oppose and the bottom panel is titled Support. Both share a horizontal x-axis at the bottom labeled Effect of Policy Presence ranging from negative 0.2 to 0.2 with a dashed vertical baseline at 0. The y-axis for both panels lists four election procedures from top to bottom. Same-Day Reg. dot, Photo I D, Automatic Reg. dot, and All-Mail Voting. Data points are categorized by three markers. black circles for Democrat, dark gray diamonds for Republican, and light gray squares for Independent. Each point includes horizontal error bars representing 95% confidence intervals.

In the Oppose panel, most estimates for Same-Day Reg. dot, Photo I D, and Automatic Reg. dot are negative, falling to the left of the zero line. For All-Mail Voting, the estimates cluster near zero.

In the Support panel, the estimates for Same-Day Reg. dot and Photo I D are near zero or slightly negative. For Automatic Reg. dot and All-Mail Voting, the Republican and Independent estimates shift significantly to the right of the zero line, indicating a positive effect, while the Democrat estimates remain slightly negative or near zero.

The asymmetric relationship between policy support and opposition is notable, but the mechanisms generating this asymmetry are difficult to isolate given the observational nature of the data. As noted earlier, a limitation in generalizing the estimated effects of policy support and opposition is that three of the four policies I am able to examine expand voting access. Combined with Democrats’ higher baseline confidence (92% compared to 81% among Republicans), this creates potential ceiling effects for policy supporters while leaving considerable room for confidence declines among opponents. The asymmetry may also reflect differential measurement error: respondents who oppose a policy may be more aware of whether it is being used than those who support it, which could attenuate estimated effects among supporters. Despite these complications, the central finding is clear: congruence between individual views and election procedures is associated with voter confidence, particularly among unaffiliated voters and Republicans who oppose the procedures in use.

EXPERIMENTAL ANALYSES

Two limitations of the observational analyses motivate the experimental design. First, some respondents in postelection surveys may lack accurate information about the procedures used in their state. Second, policy support may be endogenous to institutional change if voters adjust their attitudes in response to procedural adoption or related elite messaging. Although partisan procedural attitudes appear relatively stable over time and across states (Appendix B of the Supplementary Material), any opinion shifts induced by procedural change would bias estimates of the relationship between procedural agreement and voter confidence.

To address these limitations and expand the analysis to other procedures (e.g., no-excuse mail voting and early in-person voting), I conducted a survey experiment using 15,000 respondents in the month prior to the 2024 election. Respondents were first asked their opinion toward various election procedures and then asked to evaluate five hypothetical election scenarios, one at a time, by indicating their confidence in the accuracy of the vote count of each (see Appendix I of the Supplementary Material for details). Because procedural opinions were elicited prior to the evaluation of the election scenarios, the experiment avoids concerns about endogenous procedural preferences. To maximize awareness of the randomized procedures, scenario descriptions were intentionally sparse and centered on the procedures themselves.

The procedures in each scenario were randomized subject to two constraints. First, scenarios excluded logically inconsistent combinations (e.g., all-mail voting paired with strict photo identification). Second, each respondent evaluated five of six randomly selected procedural portfolios under a fixed, randomly assigned partisan environment: a nonpartisan condition (33%), a Democrat-led state (33%), or a Republican-led state (33%). Holding the partisan environment constant within respondent ensures that within-person variation in confidence reflects differences in election procedures rather than changes in partisan context. This design limits satisficing based on partisan cues, though the effect of partisan control is therefore identified through between-subject comparisons.Footnote ¹¹ Because procedural portfolios were assigned independently of partisan context, an average of 1,394 respondents evaluated each of the 6 × 3 = 18 experimental conditions.

The estimated effects capture how procedural congruence shapes respondents’ beliefs about electoral legitimacy in a setting that abstracts from additional contextual information such as elite messaging. Given the experiment’s emphasis on procedural salience and isolation from competing cues, these estimates are perhaps best interpreted as an upper bound on real-world effects (Brutger et al. Reference Brutger, Kertzer, Renshon, Tingley and Weiss2023).

Figure 6 reports the average confidence expressed by Democrats, Republicans, and unaffiliated voters across the experimental conditions. Consistent with the partisan differences documented in Figure 1, Democrats express higher confidence than Republicans in every scenario. Confidence also varies by procedure, with respondents expressing higher confidence in conditions that include photo identification and lower confidence in those featuring all-mail voting.

Figure 6.

Average Confidence by Condition by Partisanship

Note: Averages and 95% confidence intervals are reported for the randomized scenarios. See Appendix I of the Supplementary Material for question wording and details.

Figure 6. Long description

The x-axis represents Percent Trusting Vote Count ranging from 10 percent to 70 percent. The y-axis lists 17 voting scenarios. Data points include 95 percent confidence intervals.

Partisan groups are represented by symbols: black circles for Democrat, dark gray diamonds for Republican, and light gray squares for Other.

Key trends include:

* Scenarios requiring Photo I D show the highest trust levels across all groups, generally between 45 percent and 75 percent.

* Democrats consistently show higher trust than Republicans in almost all scenarios, particularly in No I D conditions.

* The No I D, All Mail scenario shows the lowest trust for Republicans at approximately 15 percent, while Democrats are near 38 percent.

* In Rep State, No I D scenarios, the gap between Democrats and Republicans narrows significantly compared to Dem State scenarios.

* The baseline condition, No I D, In-Person, shows Democrats at roughly 44 percent, Republicans at 20 percent, and Others at 25 percent.

* Scenarios involving Dem State and No I D result in the widest partisan gaps, with Democrats near 50 percent and Republicans near 15 percent.

To assess the contribution of each randomized scenario feature to voter confidence, I construct indicators for the presence of each feature and estimate their effects on the probability that a respondent expresses confidence in the accuracy of the vote count. Effects are estimated relative to a baseline scenario in which the state does not require photo identification and voting only occurs in-person on Election Day. Specifically, I estimate the following specification, clustering standard errors by respondent:Footnote ¹²

(3) $$ {Y}_i={\displaystyle \begin{array}{l}\alpha +{\beta}_1\mathrm{Dem}\ \mathrm{led}+{\beta}_2\mathrm{Rep}\ \mathrm{led}+{\gamma}_1{\mathrm{PhotoID}}_i\\ {}+\hskip2px {\gamma}_2{\mathrm{InPersonEarly}}_i+{\gamma}_3{\mathrm{EarlyMail}}_i+{\gamma}_4{\mathrm{AllMail}}_i.\end{array}} $$

Figure 7 presents the estimated coefficients for each partisan specification. Republicans express higher confidence when scenarios are described as occurring in a Republican-led state and lower confidence when they are described as occurring in a Democrat-led state, relative to scenarios in which state partisanship is not specified. Democrats respond in the opposite direction. The similarity in the magnitude of in-party and out-party effects across Democrats and Republicans indicates broadly symmetric partisan responses to the political context.

Figure 7.

Procedural and Partisan Determinants of Voter Confidence

Note: Standard errors are clustered by respondent. Coefficients are reported in Table A11 in the Supplementary Material. See Appendix I of the Supplementary Material for similar results using respondent fixed effects. Effects are relative to the baseline of in-person Election Day voting in a state without photo ID.

Figure 7. Long description

The graph uses a horizontal axis ranging from negative 0.1 to 0.4, with a vertical dashed line at 0.0 representing the baseline. Six categories are listed on the y-axis from top to bottom.

* Republican state: Democrat point is furthest left at negative 0.15. Other is at negative 0.05. Republican is furthest right at positive 0.05.

* Democrat state: Republican point is furthest left at negative 0.1. Other is at negative 0.02. Democrat is furthest right at positive 0.06.

* Photo I D: Democrat point is at positive 0.23. Other is at positive 0.34. Republican is furthest right at positive 0.44.

* All-mail voting: Republican and Other points are clustered near negative 0.05. Democrat point is slightly to the right at negative 0.03.

* Early voting by mail: Republican point is at negative 0.04. Other is at negative 0.02. Democrat is at positive 0.03.

* In-person early voting: Republican point is at negative 0.02. Other is at 0.0. Democrat is at positive 0.01.

Legend at the bottom indicates black circles for Democrat, dark grey diamonds for Republican, and light grey squares for Other. All points include horizontal error bars representing standard errors.

Across policies, photo identification has the largest effect on voter confidence for every partisan group. Its magnitude exceeds both the combined positive effects of all other procedures and the effect associated with residing in a state governed by co-partisans by roughly a factor of 10 among Republicans and a factor of 3 among Democrats. By contrast, default all-mail voting uniformly reduces confidence, while other procedures exert smaller and more polarized effects: in-person early voting and optional vote-by-mail modestly increase confidence among Democrats but reduce confidence among Republicans.

To decompose the effect of procedural congruence by policy agreement, I allow the effect of each procedure to vary by respondents’ policy support, using an interactive specification analogous to that used in the observational analyses (Equation 2). For each policy, I allow the respondent’s evaluation to vary based on their support for the policy, whether the policy is present in the experimental vignette, and an interaction between support and presence (e.g., $ {\beta}_1{\mathrm{SupportPhotoID}}_i+{\gamma}_1{\mathrm{PhotoID}}_i+{\delta}_1({\mathrm{PhotoID}}_i\times {\mathrm{SupportPhotoID}}_i)\hskip-0.55em $ . If congruence matters, we should expect to find negative effects for $ {\gamma}_1 $ —the effect of the policy on voter confidence among those who are opposed to the policy—and positive effects for $ {\gamma}_1+{\delta}_1 $ —the effect of a policy among those who approve of that policy.Footnote ¹³

Figure 8 graphs the estimated change in a voter’s probability of expressing confidence in the accuracy of the vote count relative to the baseline condition among supporters and opponents for each policy by partisanship.

Figure 8.

Average Marginal Effects by Respondent Support

Note: Standard errors are clustered by respondents.

Figure 8. Long description

The image consists of two vertically stacked panels. The x-axis represents the Average Marginal Effect of Policy Use, ranging from negative 0.1 to 0.5. A vertical dashed line marks the zero point. The y-axis lists four policies: Photo I D, In-Person Early Voting, Early-Mail Voting, and All-Mail Voting. Data points are color-coded: black circles for Democrat, dark gray diamonds for Republican, and light gray squares for Other, each with horizontal error bars.

Top Panel (Oppose):

* Photo I D: Democrat is slightly negative; Other and Republican are positive, with Republican reaching approximately 0.11.

* In-Person Early Voting: Republican is most negative at negative 0.07; Other and Democrat are near negative 0.02.

* Early-Mail Voting: Other and Republican are near negative 0.08; Democrat is near zero.

* All-Mail Voting: All three groups are clustered between negative 0.08 and negative 0.1.

Bottom Panel (Support):

* Photo I D: All groups show strong positive effects. Democrat is at 0.3, Other at 0.39, and Republican at 0.47.

* In-Person Early Voting: Republican and Other are near zero; Democrat is slightly positive at 0.02.

* Early-Mail Voting: Republican is slightly negative; Other is near zero; Democrat is positive at 0.04.

* All-Mail Voting: All groups are positive, clustered between 0.04 and 0.06.

Mirroring earlier observational results, respondents express less confidence in scenarios that use procedures they disagree with. Partisan asymmetries exist, however. Among Democrats, only the use of all-mail voting reduces voter confidence among those opposed to the policies, but Republican confidence declines for every policy apart from photo ID.Footnote ¹⁴

In contrast to the observational analyses, the experimental results indicate that procedural agreement can increase voter confidence when the supported procedure is present. This pattern is driven primarily by photo identification: among supporters in every partisan group, the confidence gain associated with photo identification exceeds all other policy effects and is substantially larger than the effect of partisan control of the state (by nearly a factor of three). That this confidence increase appears among both Republicans, whose elites typically champion the policy, and Democrats, whose elites largely oppose it, is consistent with an interpretation emphasizing shared beliefs about electoral integrity rather than partisan elite messaging. Support for all-mail voting is also associated with higher confidence when the procedure is present across partisan groups, though the magnitude of this effect is considerably smaller than that of photo identification. More generally, Democrats exhibit positive confidence effects across all procedures when they align with their policy preferences.

Overall, the experimental findings reinforce the observational evidence that procedural agreement affects voter confidence. As in the observational analyses, partisan context matters: respondents express different levels of confidence depending on whether elections are conducted in partisan-dominated states. At the same time, the experimental results reveal substantively meaningful effects of procedural agreement that operate above and beyond partisan context. The consistency in the direction and relative magnitude of these effects across observational and experimental analyses is reassuring, even if the estimates are not directly comparable in scale. Although photo identification appears more consequential in the experimental setting than in the observational data, both approaches show that procedural agreement can rival—and in some cases exceed—the effects associated with partisan control and electoral outcomes.Footnote ¹⁵

CONCLUSION AND IMPLICATIONS

We care about voters’ confidence in elections not only because acceptance of electoral outcomes by losing candidates and their supporters is critical to the peaceful transfer of power (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005; Norris Reference Norris2014), but also because confidence in election outcomes is closely tied to support for broader democratic processes and institutions (Carey et al. Reference Carey, Helmke, Nyhan, Sanders and Stokes2019; Layman, Lee, and Wolbrecht Reference Layman, Lee and Wolbrecht2024). Confidence grounded primarily in electoral victory is not a sustainable foundation for electoral legitimacy, as it offers little basis for acceptance among those who lose. Assessing whether voter confidence is shaped by agreement with electoral procedures is therefore central to understanding the potential for creating and sustaining a shared foundation of electoral legitimacy (Tyler Reference Tyler2006).

The observational and experimental analyses show that election procedures are not merely background institutions, but central objects of political evaluation. Winning is an important determinant of voter confidence, but so too is agreeing with the rules used to conduct those elections. This association appears consistently across research designs and empirical specifications, suggesting that procedural agreement constitutes a substantively important source of electoral confidence.

The association between procedural agreement and voter confidence, however, is not symmetric. Disagreement with election procedures is associated with substantially larger losses in confidence than the gains associated with agreement, indicating that procedural conflict is particularly damaging to electoral confidence. These losses are most pronounced among Republicans and unaffiliated voters. Among the procedures examined, photo identification stands out as especially consequential for voter confidence, with an association that in some cases rivals that of winning and which also exceeds the combined associations of the other policies examined.

Together, the findings clarify the importance of distinguishing between outcome-based and procedure-based legitimacy. Confidence grounded primarily in electoral victory is inherently zero-sum: it confers legitimacy on winners while offering little basis for acceptance among those who lose. Confidence rooted in procedural agreement, by contrast, offers a more generalizable foundation because it is not tied to a particular electoral outcome. Viewed in this light, the findings offer a limited basis for optimism: if voter confidence depends in part on agreement with the electoral procedures being used, then using commonly accepted election rules can help sustain confidence among those who lose.

The conditional nature of procedure-based legitimacy also highlights a paradox at the heart of democratic representation and electoral legitimacy. Election rules that are responsive to partisan preferences may enhance representational alignment, yet that same responsiveness can undermine shared confidence in electoral outcomes when procedural agreement hardens along partisan lines or becomes politicized. Under such circumstances, there may no longer be a shared procedural basis for confidence in electoral outcomes among those who lose.

In a federal system of election administration, this paradox takes on particular significance. Decentralized election administration allows states to adopt procedures that are responsive to partisan preferences, but that same responsiveness can make it harder to sustain shared confidence among those who oppose the resulting rules. Within states, partisan majorities that enact procedures favored primarily by their base risk eroding confidence among political minorities. Across states, the problem can intensify when closely contested national outcomes hinge on results produced under divergent and increasingly polarized procedures, inviting skepticism about the results of elections conducted under different rules. In this sense, procedural responsiveness, while normatively attractive from a representational standpoint, can undermine electoral legitimacy if it produces divergent portfolios of election rules and procedural agreement hardens along partisan lines.

Taken together, the findings show that agreement with electoral procedures is a meaningful source of electoral confidence. At the same time, they highlight how difficult it may be to sustain such confidence when opinions about election rules are, or can be, polarized (Biggers Reference Biggers2019; Biggers and Bowler Reference Biggers and Bowler2022; Graham and Svolik Reference Graham and Svolik2020; Kingzette et al. Reference Kingzette, Druckman, Klar, Krupnikov, Levendusky and Ryan2021; Layman et al. Reference Layman, Carsey, Green, Herrera and Cooperman2010; Lenz Reference Lenz2012). Precisely because procedural agreement is both consequential and fragile, understanding how election rules shape voter confidence remains central to debates over democratic legitimacy.