INTRODUCTION
Campaign finance vouchers have emerged as a popular policy proposal to address inequality of influence in elections. 2020 Democratic presidential candidates Yang, Sanders, and Gillibrand incorporated these innovative campaign financing schemes into their platforms. Gillibrand wrote that vouchers would “amplif[y] the voices of Americans who haven’t been heard for too long.” Yang claimed that by “mak[ing] it possible for all Americans to contribute to candidates they feel strongly about” the program would “allow the will of the people to shine through.” Congress and state and city governments have considered vouchers through legislation and direct democracy, though most efforts have failed. In Seattle, voters approved the Democracy Voucher program in 2015, and voucher distribution began in 2017.
To evaluate whether these programs “amplify the voices” of a diverse citizenry, we must rigorously examine who it is that takes advantage of these innovations. In a new dataset, I identify all individuals registered to vote in Seattle during nine consecutive municipal electoral cycles (2005–21). I link their demographic data to their municipal campaign contributions and voting records, and examine how voucher availability impacts the contribution behavior of Seattle residents. This panel offers the best leverage to analyze the impact of vouchers on donor diversity as voucher-era donation behavior is compared to extensive pre-voucher behavior among the same individuals. Those benefiting from the introduction of vouchers are the same types of people who dominated pre-reform: they are wealthier, whiter, more Democratic, older, and more civically engaged. Although the influence of any given high-dollar donor waned, voucher-era donor pools generally remained similar to their pre-voucher counterparts.
CAMPAIGN FINANCE, INEQUALITY, AND POLITICAL PARTICIPATION
At the national level, participation in the campaign finance process is skewed in favor of certain groups. Racial minorities (Grumbach and Sahn Reference Grumbach and Sahn2020), the non-wealthy (Overton Reference Overton2004), moderates (Broockman and Malhotra Reference Broockman and Malhotra2020), and ideologically unsophisticated Americans (Barber, Canes-Wrone, and Thrower Reference Barber, Canes-Wrone and Thrower2017) contribute at lower rates. This paper considers campaign finance within local elections, where inequalities are more pronounced. Municipal electoral participation tends to skew toward older residents, homeowners, and the educated (Oliver and Ha Reference Oliver and Ha2007).
Inequalities among donors have consequences for policy. People often give to candidates who would be descriptive representatives (Barber, Butler, and Preece Reference Barber, Butler and Preece2016; Grumbach, Sahn, and Staszak Reference Grumbach, Sahn and Staszak2022; Thomsen and Swers Reference Thomsen and Swers2017). If certain groups rarely donate, candidates from these groups may struggle to run a successful campaign. Electing candidates with underrepresented backgrounds can benefit those with similar backgrounds (Chattopadhyay and Duflo Reference Chattopadhyay and Duflo2004) and the lawmaker’s district (Anzia and Berry Reference Anzia and Berry2011). More directly, policymakers are more responsive to campaign donors (Kalla and Broockman Reference Kalla and Broockman2016).
This paper focuses on inequality in campaign finance participation. Though little work in political science studies the reduction of inequality in campaign finance, the larger literature on voting reforms and the cost of participation can help set expectations. How does directly lowering the cost of voting, in monetary terms, impact turnout? Panagopoulos (Reference Panagopoulos2013) finds that cash incentives for voting had null-to-modest effects on overall voter turnout, especially compared to previously established mobilizing methods. This indicates that electoral participation is not primarily determined by direct monetary costs. Other analyses have considered voting reforms that change the indirect costs of voting. Political scientists have theorized that voter identification laws increase the cost of voting, with the biggest cost increases concentrated among racial minorities. However, recent studies have generated controversial and varying estimates (Grimmer et al. Reference Grimmer, Hersh, Meredith, Mummolo and Nall2018; Hajnal, Lajevardi, and Nielson Reference Hajnal, Lajevardi and Nielson2017), sometimes indicating that voter ID laws negatively impact minority turnout, and sometimes indicating that minority turnout remains the same or increases. Some studies on universal mail voting, which eliminates the need to travel to and wait at a polling station, have found that postal voting increases turnout the most among individuals with lower past political participation (Gerber, Huber, and Hill Reference Gerber, Huber and Hill2013), but other work finds that vote-by-mail reforms have negative effects for resource-poor groups such as less-educated, politically uninterested, and younger individuals (Berinsky, Burns, and Traugott Reference Berinsky, Burns and Traugott2001; Kousser and Mullin Reference Kousser and Mullin2007). Berinsky (Reference Berinsky2005) explains that this is because reforms fail to mobilize nonvoters to join the electorate, while making consistent and frequent participation easier for previously “transient” voters (who usually vote but sometimes fail due to the idiosyncrasies of life).
Little work has studied the Seattle case directly. McCabe and Heerwig (Reference McCabe and Heerwig2019) use 2017 data to conclude that vouchers “moved the donor pool in a more egalitarian direction,”Footnote 1 but do find that members of overrepresented groups utilized vouchers more. Additional mobilization efforts within Seattle boosted participation primarily among the overrepresented (Henderson and Han Reference Henderson and Han2022).
DEMOCRACY VOUCHER MECHANICS
In 2015, Seattle voters approved I-122, restricting officials from becoming lobbyists, barring contributions from lobbyists and city contractors, and establishing the Democracy Voucher program. Each election year, registered voters receive four $25 vouchers through the mail, donatable to municipal candidates who agree to lower contribution limits and a spending cap (Table 1), as well as meet a threshold of initial qualifying donations. The vast majority of candidates register to participate in the Democracy Voucher program, though many nonviable candidates fail to earn enough qualifying donations. Physical barriers to utilizing vouchers are low, as participants can assign their vouchers through mail, online, and in-person to a campaign worker.
Table 1. Democracy Voucher Program Limits and Usage, as of 2021
Note: Spending limits are per election (a mayoral candidate who advances to the general election may spend $400k twice) as of 2021. Contribution limits for non-mayoral elections are non-inclusive of Democracy Vouchers, and inclusive for mayoral elections. Without vouchers, the contribution limit is $550. Candidates have successfully petitioned to be released from the spending cap and contribution limits, while maintaining their previously-received voucher hauls. Effectively, this means that vouchers are accepted by most viable candidates, but for competitive races, vouchers cannot be used after candidates reach their spending cap. CC denotes city council.
Campaign finance vouchers aim to diversify the set of candidates who run for office and to transform the contributor pool. Marketing and evaluation material from the city focuses on both of these objectives. What makes vouchers distinct from other public financing systems, which match donations or limit private contributions, is an explicit attempt to create new donors by eliminating the monetary cost of donating. This unique feature enables this paper to both evaluate an objective of the Democracy Voucher program and answer questions on how financial barriers impact political participation.
DATA AND METHODS
Data originate from the Seattle Elections and Ethics Commission, the Washington Secretary of State, L2, the Census, and the Federal Election Commission. Table 2 summarizes the datasets used. Data can be accessed in Yorgason (Reference Yorgason2024).
Table 2. Data Sources
Note: L2 predicted measures (race, income, and party) are generally accurate when compared to administrative data, and results remain significant after correcting for merge-induced bias. Data validity, specific merging procedures, matching rates, sources of possible merge-induced bias, and bias-corrected best estimates are further described in Section 1 of the Supplementary Material.
My panel contains all individuals who were continuously registered to vote in Seattle in the voter files nearest to each of the 2005–21 municipal elections. This amounts to 146,626 individuals (1,319,634 individual-cycles). Over time, the city has grown both in racial diversity and in income, and the panel structure ensures that results are uninfluenced by demographic changes.Footnote 2
This paper focuses on who donates instead of how much is donated. Binarizing campaign finance participation is preferable for two reasons. First, the “democracy”-centered nature of the voucher program emphasizes mobilization of the citizenry. If lawmakers primarily seek to disrupt big donors, they may limit independent expenditures or create low spending caps without exemptions. Second, voucher-accepting candidates agreed to abide by contribution limits, which would likely impact donation totals but not the binary act of donating. Focusing on contribution totals as the dependent variable would conflate the impact of vouchers and contribution limits. Nonetheless, Section 3 of the Supplementary Material replicates the main findings using contribution totals, and recovers similar results.
Models
I estimate two continuous difference-in-differences models. In Model 1, each observation is a census block group-year. Participation, represented by y, is a continuous variable that indicates campaign finance participation in year t in block group b. I utilize four different demographic variables: percentage white, median household income, mean voting frequency, and partisanship within block group b. For clarity, the following models and explanations in this section are described in terms of income.
$ \beta $
is the outcome of interest and measures whether block groups with higher income levels are associated with greater increases in campaign finance participation in voucher years.
$ {\gamma}_b $
and
$ {\delta}_t $
are block group and cycle fixed effects, respectively.
$$ \begin{array}{rl}{\mathrm{y}}_{b t}=\beta ({\mathrm{voucher}\hskip2.22198pt {\mathrm{y}\mathrm{ear}}_t*\mathrm{income}}_b)+{\gamma}_b+{\delta}_t+{\epsilon}_{bt}.& \end{array} $$
Model 2 is similar to Model 1, but utilizes individual i data. Again, Model 2 describes the effects of income (decile),Footnote 3 but models using race, partisanship, and past political participation are presented in the results. I also consider age, but because the relationship between age and participation is quadratic (Figure 1), I utilize two models to convey the behavior of individuals under 71 and over 70.
$$ \begin{array}{rl}{\mathrm{y}}_{it}=\beta (\mathrm{voucher}\hskip2.22198pt {\mathrm{y}\mathrm{ear}}_t*{\mathrm{income}}_i)+{\gamma}_i+{\delta}_t+{\epsilon}_{it.}& \end{array} $$
As all individuals in Seattle are simultaneously treated, the effect of vouchers is identified by comparing participation before and after vouchers among different subgroups. Using Model 1 as an example,
$ \beta $
represents the additional voucher-year disparity in the conditional average treatment effect in between block groups as median income varies. A $1,000 difference in block group median income is associated with a
$ \beta $
%(percentage point) difference in a block group’s voucher-year campaign finance participation. If disparities between groups remain the same in voucher years,
$ \beta $
would be statistically insignificant.
Figure 1. Voucher Uptake is Highest among Demographics Known to Participate Most in Local Politics, Even among Frequent Voters
Note: The population utilized for these graphs is constantly registered individuals in Seattle, 2011–21 (203,685 individuals). Points denote the proportion using Democracy Vouchers when the voucher program is in place.
The assumptions for these models are the same as in a standard difference-in-differences setup (Lueders, Hainmueller, and Lawrence Reference Lueders, Hainmueller and Lawrence2017). In addition to standard OLS assumptions,
$ \beta $
’s identification relies on the assumption that in the absence of campaign finance vouchers, disparities in participation will remain the same. Although this is not possible to observe, parallel pre-voucher trends in between block groups are shown in Section 2 of the Supplementary Material. While block group participation varies each cycle, aggregating block groups into quartiles results in relatively parallel trends over time. In the same section, regressions using quartiles relax the OLS linearity assumption for
$ \beta $
and yield similar conclusions.
RESULTS
Figure 1 presents descriptive data on voucher users, showing that voucher users are skewed toward traditionally overrepresented groups when considering income, ethnorace, age, past political participation, and party. Even among frequent voters, the rich disproportionately participate, a finding that is intuitive when donations are costly but not when donating is free.
Table 3 shows the results of the difference-in-differences regressions outlined in Model 1. From column 1, a 1-unit increase in a block group mean voting level is associated with a 0.93% increase in campaign finance participation during voucher years for members of that block group, beyond already-existing time-invariant disparities. In column 2, a 1% increase in white residents is associated with a 0.05% increase in voucher year participation. Column 3 shows that a $1,000 jump in block group median income is associated with a 0.02% increase in participation in voucher years. Column 4 shows that a 1% increase in the Republican share of the population is associated with a 0.14% decrease in participation in voucher years.
Table 3. Campaign Finance Participation Increases the Most among Already-Overrepresented Block Groups When the Voucher Program Is in Place
Note: Cluster robust standard errors, clustered on block groups. Mean block group participation in voucher years is 7.91%. Observations differ in column 3 because 2010 census lines are used for income, whereas 2020 census lines are used for columns 1, 2, and 4. VPP denotes voucher program in place.
Coefficients in Table 3 are small because of low rates of participation, even in voucher years. However, these differences are substantively significant. Consider the 2021 difference in VPP-donors for block groups that are 55% and 75% white (A and B) with zero block fixed effects. In 2021, for every one hundred new donors in A, B will have about 111 new donors.Footnote 4
A limitation of Table 3 is that it is unclear who is mobilized in voucher years within a census block group. For example, the results from Table 3 could reflect strong voucher-year mobilization among Republicans in block groups where Republicans are particularly outnumbered. Likewise, voucher-year mobilization among nonwhites could be strongest in the whitest block groups. The calculation of voucher-year effects at the individual level allows for the interrogation of these possibilities. Using race, income, partisanship, age and an index of previous voting, the individual-level results of Model 2, shown in Table 4, echo those of Table 3.
Table 4. Individuals in Overrepresented Groups Benefit the Most from the Voucher Program
Note: Robust standard errors, see Section 4.2 of the Supplementary Material for block-bootstrap SE. Mean individual participation in voucher years is 8.77%. A 2011–21 panel is used within the age models to maximize age variation while maintaining the multi-year panel structure. Observation counts differ by regression because income and ethnorace estimates are not available for all individuals. VPP denotes voucher program in place.
In column 1 from Table 4, a white individual has an additional 2.5% higher likelihood of donating in a voucher year relative to a nonwhite individual, on top of already existing disparities in white and nonwhite participation. In column 2, a 1-decile increase in an individual’s income is associated with a 0.21% increase in their voucher-year participation likelihood. Column 3 shows that voting in a single additional election is associated with a 0.95% increase in voucher year participation. Republican identification is associated with a 3.25% decline in voucher year participation. These findings are each combined in column 5, where covariates remain significant. Columns 6 (and 7) show that among individuals under 71 (over 70), each additional year is associated with a 0.05% (0.34%) increase (decrease) in voucher year participation.
Consider two individuals with zero fixed effects in the second (A) and ninth (B) income decile in 2021. For every one hundred new donors like A, there are 117 new donors like B.Footnote 5
Thus far, I have shown that decreasing the costs of political donations results in an influx of new donors who reflect the same types of donor pool inequalities as in previous years. However, because the donor pool already had inequalities in participation, there remains the possibility that these uneven increases in participation still make the donor pool more representative of the population. Figure 2 shows that the demographic breakdown of voucher users is typically similar to the demographic breakdown of both municipal and federal cash donors. In all cases except income, voucher users are similarly representative or less representative than same-year municipal cash donors.
Figure 2. Voucher Donors Are Frequently Similar to Municipal and Federal Cash Donors
Note: The population utilized for these graphs is constantly registered individuals in Seattle, 2011–21 (203,685 individuals). While comparisons treating the same-year cash donors as a counterfactual is causally problematic (see Section 5.1 of the Supplementary Material), these graphs make it clear that campaign finance vouchers are not systematically changing or improving the overall composition of the donor pool across all metrics.
Why are voucher users less wealthy than cash donors, while richer individuals participate more when the voucher program is in place? A potential explanation is that less-wealthy individuals substitute a cash contribution with a voucher contribution, whereas wealthier individuals continue to contribute with cash. In Section 4.1 of the Supplementary Material, I consider the voucher-era contribution patterns of pre-voucher donors. In 2017, less-wealthy 2015 donors were more likely to re-contribute using only vouchers, whereas wealthier 2015 donors were more likely to re-contribute using only cash. Rates of re-donating among these pre-voucher donors are similar across income deciles (this trend also holds before the introduction of vouchers), but the way they chose to re-contribute when vouchers were available differs.
WHY MIGHT INEQUALITY PERSIST?
Voucher availability does not necessarily guarantee knowledge about vouchers, interest in municipal elections, or the expertise to deploy the vouchers correctly and on time. While it is clear that participation comparatively increases among overrepresented groups when contributions become cost-free, it is unclear why this is the case. I utilize descriptive data to consider possible explanations.
Candidates
Past research has found that participation among underrepresented individuals is partially driven by candidate attributes, and this finding suggests two related questions. First, does candidate composition appear to change in the wake of Democracy Vouchers? Figure 3 presents descriptive data on candidates.Footnote 6 While candidate diversity is slowly increasing over time, candidates in the voucher era (most eligible candidates utilize vouchers) are quite similar to candidates who ran in the pre-voucher period along the dimensions of age, gender, ethnorace, and income.Footnote 7 Second, are gaps in voucher-related participation smaller in races with traditionally underrepresented candidates? In Section 4.4 of the Supplementary Material, I find smaller yet positive and significant post-voucher gaps in nonwhite and white participation when nonwhite candidates are on the ballot.
Figure 3. Voucher-Era Candidate Attributes Are Similar to Pre-Voucher-Era Candidate Attributes
Note: To fit the same scale, income is divided by $5,000. Tables in Section 1.4 of the Supplementary Material further show that the largest gender- and race-based shifts in candidate demographics came prior to the introduction of Democracy Vouchers.
Learning at Different Rates
Asymmetric participation could be driven by asymmetric knowledge. Seattle-commissioned reports indicated that in 2019, only about half of those surveyed were familiar with the program. Knowledge was higher among whites and the wealthy. However, the knowledge gap is unlikely to be a primary cause of the disparity: there is likely an association between program knowledge and prior civic engagement. In Figure 1, large discrepancies by subgroup exist among frequent voters. Alternatively, one might ask whether participation among the underrepresented has caught up over time. Section 4.3 of the Supplementary Material compares 2017 voucher uptake to that of 2019 and 2021 by age and race among the whole Seattle population. Relative to 2017, new uptake continues to be dominated by overrepresented groups in 2019–21, despite program expansion to mayoral races and continued marketing efforts from the city.
Funding Rules
Candidates who accept vouchers are subject to fundraising limits. When they hit these limits, they can appeal to the SEEC to continue raising cash, but cannot accept additional vouchers. Does this near-election cash-only period bias the donor pool? Similarly, would the donor pool be more representative if all sent vouchers were accepted? Section 4.2 of the Supplementary Material follows the procedure from Table 4, but excludes post-limit cash donations and includes received-but-not-redeemed vouchers, and results remain virtually identical.
CONCLUSION
The effect of campaign finance vouchers appears to mirror results from the proliferation of vote-by-mail in Oregon in Berinsky, Burns, and Traugott (Reference Berinsky, Burns and Traugott2001): participation increases primarily among those with the highest propensities to donate before reform. Using administrative data, I find that Seattle’s voucher program, which offers all citizens the ability to make a free donation, did not systematically improve the diversity of the donor pool. The participation of individuals in nondominant groups in Seattle politics, such as the young, nonwhite, and Republicans, increased at significantly lower rates than that of groups dominant in Seattle politics. While the donor pool expanded, gaps between the composition of the donor pool and citizenry persisted. These results inform and raise new questions in three broad areas in the study of policy.
First, this study offers evidence on a program with low participation costs and universal and automatic treatment. Most high-profile voucher programs, such as food stamps and school vouchers, are neither universal nor automatic. Should programs that seek to reduce inequalities be targeted or universal? Should programs rely on citizen proactivity instead of universal distribution? While this study hardly provides a definitive answer in the case of campaign finance, let alone vouchers as a whole, the Seattle program provides a large-scale example of heterogeneous effects from a universal program.
Furthermore, this study is among the first to show that the link between personal finances and campaign finance participation is complex. Campaign finance participation is historically high during voucher years, suggesting that money is generally a barrier to making a political contribution. Contingent on contributing, lower-income residents use vouchers more, and the program inherently allows them to substitute something costly with something free. However, is money the primary barrier for groups underrepresented in the donor pool? Even among the most civically engaged, participation gaps persist and widen when contributing is free. This finding suggests that additional obstacles, beyond monetary barriers, dominate donation behavior. Future research should continue to investigate both behavioral and institutional sources of campaign finance inequality.
Finally, this study contributes to the discourse on campaign finance policy. In this study, I find that new voucher-era donors exhibit similar demographic biases as pre-voucher cash donors. Should reformers advocate for vouchers over other policies? Diversity within the donor pool does not substantially improve, but this is just one aspect of the campaign finance ecosystem. Democracy Vouchers reduce the financial burden of political contributions for those with lower incomes, and relative to the pre-voucher era, vouchers increase participation among all groups. The role of vouchers on corporate influence, outside donors, ideological extremism, and candidates has yet to be rigorously analyzed. If well-identified studies show that vouchers are best alternative reforms along other metrics, then voucher programs warrant further consideration, even if donor pool diversity is not improved.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055424000170.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/XWBDPQ. Limitations on data availability are discussed in the appendix.
Acknowledgments
I thank Adam Bonica, Justin de Benedictis-Kessner, Justin Grimmer, Jake Grumbach, Andy Hall, Katie Clayton, Cole Tanigawa-Lau, and members of the Hall Research Group for helpful comments.
CONFLICT OF INTEREST
The author declares no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The author declares the human subjects research in this article was reviewed and approved by the institutional review board at Stanford University (Protocol #61828). The author affirms that this article adheres to the principles concerning research with human participants laid out in APSA’s Principles and Guidance on Human Subject Research (2020).
Open access
