The Measurement Characteristics of Electoral Contestation

2026 Western States Political Psychology Conference
October 22–24, 2026 · University of New Mexico · Albuquerque

Christopher R. Weber

University of Arizona

Juliet E. Carlisle

University of Utah

Christopher F. Karpowitz

Brigham Young University

Fabian G. Neuner

University of Colorado Boulder

Mark D. Ramirez

Arizona State University

Contesting Election Results

  • The importance of contestation in American politics.
  • Contestation is important to democracy, but also a potential threat to the peaceful transfer of power.
  • 2020 brought it to the forefront: a siege on the Capitol on January 6, “stop the steal” protests, an automatic recount in Georgia and a requested one in Wisconsin, and more than fifty lawsuits challenging state and federal electoral actions.
  • A longer history. Dissatisfied candidates contested outcomes in 2000, 1960, 1888 and 1876 (Foley 2019). After 2016, opponents of the president-elect demonstrated, called for faithless electors, and pursued recounts in several states.

Three goals

  1. What do citizens think is a legitimate way to contest an election?
  2. How are contestation beliefs structured?
  3. Are these beliefs malleable, shifting based on electoral fortunes, or are do they represent stable orientations towards American elections?

Traditional and Subversive Forms

TRADITIONAL

Routinely deployed, regular, institutionally permissible.

  • Ballot recounts
  • Litigation
  • Public criticism/dissatisfaction
  • Public protest

SUBVERSIVE

Informal, uncommon, sometimes anti-democratic.

  • Refusing to certify ballots
  • Occupying buildings
  • Political violence
  • Public criticism/dissatisfaction with election integrity

Another way of considering this is who acts: some are predominantly elite behaviors (candidates, parties, officials), others are citizen behaviors.

Contesting an outcome is both a key attribute of democracy (Coppedge 2008; Harvey & Mukherjee 2020) and a potential threat to to the peaceful transfer of power (Przeworski 1991; Albertson & Guiler 2020). The interpretation may depend on the observer’s perspective and stakes, the legality of the action, and the broader political context (Beaulieu 2014).

The Core Questions

Many people are unhappy with the outcomes of elections. How much do you support or oppose each of the following behaviors when people are unhappy with the outcome of an election?

  • March Attend a march or demonstration
  • Flag Burn the American flag
  • Courts Contest the outcome in the courts
  • Recounts Support ballot recounts
  • Criticize Publicly criticize the integrity or fairness of the election

All questions are asked on a five-point scale, from strongly oppose to strongly support. These questions are in the seven surveys we analyze.

The Surveys

Year Survey Design Sample N
Cronbach's α
All five Excluding Flag Burning
2020 Western States Survey Cross-section, pre/post AZ, CO, NV, NM, UT; Latino oversample 3,600 0.69 0.64
2022 CES – ASU Module Two-wave pre/post panel National 1,000 0.61 0.69
2022 CES – BYU Module Two-wave pre/post panel National 1,000 0.70 0.72
2023 Arizona Voter Project Panel wave 1 (AZ) Arizona likely voters 1,000 0.65 0.73
2024 Arizona Voter Project Panel wave 2, 403 recontacts Arizona likely voters; 403 recontacts 803 0.63 0.68
2024 CES – BYU Module Two-wave panel National 1,000 0.58 0.64
2024 Western States Survey Cross-section, randomized pre/post Same five states; Hispanic oversample 4,000 0.59 0.62

alpha is computed on the pre wave for panel surveys. Excluding flag burning increases alpha for six of seven surveys: flag burning is virtually uncorrelated with every other item except attending a march or demonstration.

Additional Items

Item Asked in
Engaging in violent acts CES ASU 2022, CES BYU 2022
Engage in violent protest AVP 2023, AVP 2024
Express support for politicians who refuse to concede WSS 2024, CES BYU 2024, AVP22/24
Occupy buildings to prevent certification WSS 2024, CES BYU 2024
Object to the eligibility of individual voters WSS 2024, CES BYU 2024
Public officials choosing to certify the results AVP 2023, AVP 2024
State legislators refusing to certify the results AVP 2023, AVP 2024
Call a new election AVP 2024

“Violent acts” and “violent protest” are different questions on different response layouts and are not pooled.

Support for Electoral Contestation

Mean Support by Item and Survey

Mean Support, All Items

Structure of Contestation Belief

Dimensionality

ONE DIMENSION

A single willingness to contest, with items differing only in how hard they are to endorse.

  • Items are interchangeable indicators
  • Recounts and flag burning sit at opposite ends of one continuum

SEPARABLE CONSTRUCTS

Distinct orientations that happen to be asked together.

  • Items cluster, and the clusters behave differently
  • One score averages away real variation
  • Endorsing a recount tells you little about burning a flag

Factor Structure: Six Common Items

Item
WSS 2024
CES BYU 2024
AVP 2023
AVP 2024
F1 F2 F1 F2 F1 F2 F1 F2
Publicly criticize integrity 0.79 0.00 0.88 0.04 0.90 −0.02 0.84 0.04
Refuse to concede 0.70 −0.12 0.83 −0.10 0.89 −0.10 0.80 −0.18
Contest outcome in court 0.62 0.15 0.36 0.46 0.75 0.20 0.69 0.22
Support ballot recounts 0.47 0.12 0.53 0.16 0.69 0.09 0.67 0.06
Attend march or demonstration 0.00 1.00 0.01 0.87 0.09 0.78 0.10 0.85
Burn the American flag 0.00 0.38 −0.23 0.52 −0.25 0.52 −0.21 0.63

Two-factor EFA, oblique rotation. Factor 1 is oriented to the traditional factor in every survey. The four surveys that carry all six items; CES BYU 2024 uses its pre wave.

The Larger AVP Battery

Item
Two factors
Three factors
F1 F2 F1 F2 F3
Publicly criticize integrity 0.84 0.08 0.83 0.06 0.04
Refuse to concede 0.84 −0.04 0.75 −0.14 0.30
Legislators refuse to certify 0.84 −0.11 0.76 −0.19 0.24
Contest outcome in court 0.67 0.19 0.78 0.24 −0.22
Call a new election 0.81 −0.07 0.77 −0.13 0.17
Support ballot recounts 0.60 −0.05 0.76 0.02 −0.36
Burn the American flag −0.08 0.99 −0.11 0.93 0.17
Attend march or demonstration 0.23 0.53 0.30 0.59 −0.14
Engage in violent protest 0.40 0.45 0.17 0.29 0.76
Officials certify results 0.26 −0.22 0.17 −0.31 0.25

AVP 2024, n = 803. Oblique rotation. Two factors: TLI 0.745, RMSEA 0.170. Three factors: TLI 0.964, RMSEA 0.064. Note: The second dimension does not replicate in AVP 2023.

Stability of Contestation Belief

Test–Retest Stability

Item AVP 2023 → 20241 CES ASU 2022 (Control)2 CES BYU 20243
Observed items
Attend March or Demonstration 0.556 (387) 0.524 (425) 0.530 (799)
Burn the American Flag 0.684 (386) 0.710 (424) 0.786 (799)
Contest Outcome in Court 0.519 (387) 0.516 (424) 0.393 (801)
Support Ballot Recounts 0.647 (387) 0.519 (424) 0.303 (798)
Publicly Criticize Election Integrity 0.661 (387) 0.623 (425) 0.537 (801)
Refuse to Concede 0.679 (387) – 0.542 (802)
Latent stability, corrected for measurement error
Five core items 0.853 (387) 0.767 (425) 0.508 (802)
Excluding flag burning 0.852 (387) 0.769 (425) 0.495 (802)
Two factors: traditional 0.865 (387) 0.787 (425) 0.553 (802)
Two factors: expressive 0.810 (387) 0.892 (425) 0.798 (802)
1 No intervening election; 17 months apart.
2 Spans the 2022 midterm election; weeks apart.
3 Spans the 2024 presidential election; weeks apart.

Note: The two-factor model (with two items) is identical in model fit to a one-factor model with a march–flag residual correlation. Flag burning correlates with nothing but attending a march, but march correlates with the other traditional items, a bridge.

Persistent Winner Loser Effects

  • 1,400 Arizona respondents, four waves. 1000 from AVP 2022 and 803 AVP 2024, joined for 403 pairs; the Western States waves reaches 170 and 173 of them.
  • Stability is highest where no election intervenes. The 17-month AVP-to-AVP gap returns 0.85 [0.79, 0.91], higher than the six-month gap that straddles the 2024 election despite being almost three times as long.
  • The path from 2020 is near zero?
  • Not a winner–loser estimate. The AVP waves carry no pre/post structure (Dec 2022 follows the midterm, May 2024 precedes the general), and WSS 2024 mixes 2000 pre- with 2000 post-election interviews.
Transition1 Gap Election n linked2 Latent stability
WSS 2020 → AVP Dec 2022 25 months 2022 midterm 125 0.11 [-0.10, 0.31]
AVP Dec 2022 → AVP May 2024 17 months none 403 0.85 [0.79, 0.91]
AVP May 2024 → WSS Nov 2024 6 months 2024 presidential 156 0.76 [0.64, 0.87]
1 The AVP is treated as the universe of data: every AVP respondent is kept and the Western States waves are joined on where the YouGov crosswalk links one. AVP 2023 fielded 1,000 and AVP 2024 fielded 803, overlapping on 403.
2 Respondents contributing both waves of the transition. All paths are estimated by FIML on the whole panel, not on these subsets.
Model fit: χ²(98) = 222.0, CFI = 0.952, TLI = 0.941, RMSEA = 0.030, SRMR = 0.114.

Flag Burning Does Not Move

  • Every pair looks the same. All six wave-to-wave correlations run 0.55 to 0.69, with no relation to the length of the gap or to whether an election falls inside it. The four-year 2020-to-2024 pair is 0.64; the six-month 2024 pair is 0.69.
  • Compare the contestation factor on the previous slide: 0.11, 0.85, 0.76. Those paths swing with the electoral calendar; this item does not.
  • Flag burning behaves like a standing disposition, not a contestation attitude – which is why it scales poorly with the procedural items and raises the scale’s reliability when dropped.
Wave WSS 2020 AVP Dec 2022 AVP May 2024 WSS Nov 2024
WSS 2020 — 0.65 0.69 0.64
AVP Dec 2022 125 — 0.69 0.55
AVP May 2024 129 402 — 0.69
WSS Nov 2024 38 94 156 —
Correlations above the diagonal, pairwise n below it in grey.

Group Norms

Conformity, Not Reactivity

  • Items WSS 2024: recounts, occupying buildings, refusing to concede – asked as own view and as what a randomly assigned partisan does.
  • Perceived norm: [Party members] “support ballot recounts”; “support occupying buildings to prevent the certification of election results”; “express support for politicians who refuse to concede their election” (1 = not at all, 5 = a great deal).
  • The Self, the Ingroup, the Outgroup. *Predict one’s own contestation using perceived contestation of in- and out- partisans.
  • The Measurement Model. Same-content residuals covary; model fits well. 0.995, RMSEA 0.030, SRMR 0.015.
  • Seeing your own side contest predicts contesting (β = 0.44). Seeing the other side predicts slightly less (β = -0.09). The two paths are not equal. χ²(1) = 85.7, p < .001.
In-party referent
Out-party referent
b (SE) β 95% CI b (SE) β 95% CI
Perceived Contestation Norms +0.299 (0.042) +0.444 [+0.36, +0.53] -0.050 (0.023) -0.088 [-0.17, -0.01]
Covariates
Ideology (conservative) +0.072 (0.011) +0.248 [+0.18, +0.31] +0.078 (0.012) +0.261 [+0.20, +0.32]
Authoritarianism +0.033 (0.009) +0.115 [+0.05, +0.18] +0.077 (0.011) +0.254 [+0.19, +0.32]
Age -0.064 (0.009) -0.225 [-0.29, -0.16] -0.092 (0.010) -0.304 [-0.36, -0.25]
Female +0.005 (0.007) +0.017 [-0.03, +0.07] +0.001 (0.008) +0.003 [-0.05, +0.06]
College -0.016 (0.008) -0.056 [-0.11, -0.00] -0.015 (0.009) -0.052 [-0.11, +0.00]
Two-group SEM, n = 1535 / 1629. Loadings equal across arms; intercepts and covariate paths free. χ²(64) = 755.5, CFI = 0.843, TLI = 0.779, RMSEA = 0.083, SRMR = 0.070. Equality of the norms path rejected: χ²(1) = 85.7, p < .001. Constraining the covariates equal is also rejected, χ²(5) = 20.6, p = 0.001.

Conclusion

Concluding remarks

  1. Support varies systematically from expressive to traditional forms of contestation.
  2. Flag burning does not clearly scale with three of the five or six core items.
  3. Attend march is a bridging measure. It loads on both expressive and traditional forms of contestation.
  4. Contestation beliefs are stable over time, especially in the absence of an intervening election.
  5. Winning and losing effects persist until the next election.

Limitations: small n; assumptions about missing data; different samples across surveys.

Thank You!

Questions: chrisweber@arizona.edu

Supplementary Material

Measurement Model for Stability

One Factor With a Correlated Error

Missing Data in a Four Wave Panel

Variable In a WSS wave AVP only Difference p
Age 56.33 49.54 6.79 0.006
Female 0.56 0.53 0.03 0.284
White 0.77 0.76 0.00 0.865
Black 0.03 0.03 0.01 0.578
Latino 0.16 0.14 0.01 0.550
College 0.36 0.45 −0.09 0.004
Married 0.50 0.48 0.02 0.571
Ideology (5 pt) 3.19 3.07 0.12 0.146
Party ID (7 pt) 4.20 3.97 0.23 0.120
Trump vote 2020 0.54 0.47 0.07 0.052
Authoritarianism 0.72 0.65 0.07 0.071
Contestation (AVP) 3.31 3.15 0.16 0.011
  • Linkage is sparse: 305 of 1,400 AVP respondents link to at least one Western States wave; only 3 have complete data.
  • Why it matters: listwise deletion is impossible, so stability estimates rest on MAR – the likelihood uses every observed case.
  • The assumption: linkage depends only on the covariates above, not on unobserved contestation. This cannot be tested directly.

Item Wording

Item Wording Surveys
Attend march Attend a march or demonstration All seven
Burn flag Burn the American flag All seven
Court Contest the outcome in the courts All seven
Recount Support ballot recounts All seven
Criticize Publicly criticize the integrity or fairness of the election All seven
Concede Express support for politicians who refuse to concede / Refusing to concede an election defeat WSS 2024, CES BYU 2024, AVP
Occupy buildings Occupy buildings to prevent the certification of election results WSS 2024, CES BYU 2024
Object eligibility Object to the eligibility of individual voters WSS 2024, CES BYU 2024
Officials certify Public officials choosing to certify the election results AVP
Legislators certify State legislators refusing to certify the election results AVP
Violent protest Engage in violent protest AVP
New election Call a new election AVP 2024

Stems. WSS 2020 prefixes every item “When unhappy with election results”; WSS 2024 and the CES modules label the battery “Electoral Contestation”; the AVP labels it “Contestation”. The codebooks record item labels rather than the full preamble, so the stems above are the recorded text, not a verbatim transcript of the question screen.

Response scales differ, and the midpoint differs with them. WSS and the CES modules use a five-point support/oppose scale whose middle category is “Neither support or oppose”. The AVP uses four substantive options plus “Not sure”, and the recode places “Not sure” at the midpoint – so the AVP midpoint is a don’t-know, not a considered neutral. All items are scored so that higher values mean more support for contestation; public_official_certify is reversed to run with the battery.

Question Wording Experiments

Survey What was randomized Arms
WSS 2020 Item wording -- march plain | "even if it might turn chaotic or dangerous"
WSS 2020 Item wording -- criticize plain | "on social media"
CES ASU 2022 Post-wave prompt Control (425) | Fraud (198) | Norms (222)
CES BYU 2022 Pretest wording support/oppose (503) | participation (497)
CES BYU 2022 Post-wave prompt Control (497) | Fraud (252) | Norms (251)

The two WSS 2020 splits were assigned independently. Fraud prompt adds “especially when there are concerns about election fraud”; Norms prompt prepends “The peaceful transfer of power is a foundation of American democracy.” WSS 2024 and CES BYU 2024 carry no wording experiment on this battery – WSS 2024 randomizes interview timing instead.

Do the Surveys Differ?

Item n
Survey means
One-way ANOVA
Lowest Highest Spread F df p
Burn the American Flag 11,903 1.56 1.98 0.41 23.77 (6, 11896) < .001
Support Ballot Recounts 11,902 3.62 3.93 0.31 12.63 (6, 11895) < .001
Publicly Criticize Election Integrity 10,123 2.59 2.88 0.29 5.12 (6, 10116) < .001
Attend March or Demonstration 10,091 3.12 3.35 0.23 3.46 (6, 10084) 0.002
Contest Outcome in Court 11,905 2.94 3.15 0.21 5.92 (6, 11898) < .001

Standard-wording arm only, ordered by spread. Every item differs significantly across surveys, but the differences are small: the widest spread is 0.41 scale points on a 1–5 scale. Degrees of freedom are given as (numerator, denominator).