Does Electoral Reform Increase (or Decrease) Political Equality?

Transcription

1 Policy Studies Organization From the SelectedWorks of Elizabeth Rigby 2010 Does Electoral Reform Increase (or Decrease) Political Equality? Elizabeth Rigby, University of Houston - Main Melanie J. Springer Available at:

2 Does Electoral Reform Increase (or Decrease) Political Equality? Elizabeth Rigby University of Houston Melanie J. Springer Washington University in Saint Louis November 1, 2009 Forthcoming at Political Research Quarterly Acknowledgement: The authors are listed alphabetically and share equal responsibility for both interpretation and any errors. We thank Robert Erikson, Jonathan Nagler, Caroline Tolbert, Daniel A. Smith, Thad Kousser, and other colleagues who provided helpful feedback on earlier versions of this paper presented at the 2007 Annual Meeting of the American Political Science Association and the 2009 State Politics and Policy Conference. We also thank the anonymous reviewers at PRQ, Sarah Bruch, Tony Stenger, and Joel Sievert for their research assistance, and the University of Houston Faculty Research program for financial support of the first author.

3 Abstract Over recent decades, the American states have implemented electoral reforms making it easier for citizens to register and vote. This paper examines the equality effects of these reforms: the degree to which reform serves to equalize or further skew participation rates between the rich and poor. Using the Voter Supplement to the Current Population Survey, we generate state-level estimates of income bias in registration and voting for elections from Findings support our theory that some electoral reforms promote equality, while others further stratify the electorate particularly when state registration rolls are already unrepresentative in terms of income groups. 2

4 As levels of income inequality continue to climb in the United States, many question the ability of the political system to uphold its normative commitment to political equality among citizens with vastly different resources (Bartels 2008; Jacobs and Skocpol 2005). Although concerns about income-based political inequality are long-standing (e.g., Schattschneider 1960), concerns have been reinvigorated by recent work identifying public officials greater responsiveness to wealthier constituents (Bartels 2008; Gilens 2005; Rigby and Wright forthcoming), the growing role of money in the electoral process (Ansolabehere, de Figueiredo, and Snyder 2003), and political parties tendency to engage in targeted mobilization strategies by disproportionately focusing their efforts on the most advantaged and engaged members of the electorate (Rosenstone and Hansen 2003). In response, many reformers have directed their efforts toward equalizing the political playing field. One primary target for reform has been state electoral institutions, which determine the participatory rules of the game governing registration and voting in the American states (Cain, Donovan, and Tolbert 2008; Lijphart 1997; Piven and Cloward 1988, 2000). Motivated by the clear differences in participation rates among advantaged and disadvantaged citizens (Leighley and Nagler 1992; Rosenstone and Hansen 1993, 2003; Verba, Schlozman, and Brady 1995), electoral reforms aim not only to increase overall voter turnout rates, but also and perhaps more importantly to equalize political participation across different socio-demographic groups (Lijphart 1997; Mitchell and Wlezien 1995; Piven and Cloward 1988, 2000; Wolfinger and Rosenstone 1980). Despite the normative importance of this goal, most evaluations of electoral reforms tend to draw conclusions about their effects on the representativeness of the electorate indirectly, from studies primarily designed to evaluate the effect of reform on overall turnout rates. Instead, in this paper, we focus more directly on the equality consequences of 3

5 several popular electoral reforms. The theory we develop and test is unique in that it specifies not only the types of electoral reform, but also the characteristics of state electorates in which reforms are enacted. These distinctions allow us to identify the reforms most likely to enhance equality, as well as those that may serve to further skew the representativeness of the electorate. This theory differentiates among three common but distinct approaches to electoral reform, each expected to affect the balance of participation across income groups in different ways: reforms such as mail-in registration that make voter registration easier, reforms like Election Day Registration that reduce the barrier registration poses to voting, and reforms such as early voting that make voting more convenient. Of course, this approach is not new; scholars commonly differentiate among different reforms (i.e., motor voter or mail-in registration). Yet, they rarely theorize different effects for distinct approaches to electoral reform as we do here. This is not problematic for studies of turnout since recent reforms are all presumed to be costdecreasing and therefore are all expected to increase participation. However, distinguishing among reforms becomes critical for understanding political inequality as an effect of reform, particularly in light of evidence that some reform approaches risk perverse consequences, exacerbating political inequalities rather than remedying them (Berinsky 2005). Also central to our theory is an expectation that the degree to which electoral reforms equalize participation rates between the rich and poor depends on the level of pre-existing bias in the state s registration rolls. This is based on the notion that registration-based reforms can only have an equalizing effect on voting if there is something to equalize. In addition, reforms that make it easier for registered voters to vote will only benefit those who are already registered. Therefore, if the registration rolls are particularly skewed by income, then registration-focused reforms have a greater potential to equalize the electorate (since there is so much to equalize), 4

6 while, in this same context, voting-focused reforms could actually exacerbate income bias in political participation by making it easier for those already engaged in the system to vote. To test this theory, we aggregate individual-level data from the U.S. Census Bureau s Current Population Survey, November Voter Supplement to estimate state-level participation rates among the rich versus poor. From these average values, we generate a relative measure of vote bias (the ratio of participation among rich versus poor) for each state during midterm and presidential elections from Then, we examine how the implementation of each approach to electoral reform shifts this relative measure of participation bias. And finally, we examine the role of skewed registration rolls in limiting or enhancing the equality effects of each type of reform. The empirical findings support our theory that electoral reforms have the potential to serve as equality-promoting as well as equality-inhibiting institutions and that preexisting registration bias plays an important role in shaping the consequences of electoral reform. Background: Consequences of State Electoral Reform As the American states have adopted different electoral laws, researchers have capitalized on these natural experiments as opportunities to advance our understanding of the way institutional structures shape citizens political behavior (for a recent overview see Cain, Donovan, and Tolbert 2008). Of course, the central question underlying most electoral law proposals is whether the reform has the capability of increasing turnout and, by extension, representation. Although less explicit in reform proposals, the expectation of enhanced political equality stems from the assumption that restrictive, costly electoral institutions disproportionately affect low-income citizens, who bear higher voting costs in general. Thus, reforms that reduce these costs will increase participation rates among these classically disadvantaged groups at a greater rate than for other citizens. Building on this assumption, 5

7 studies finding increased turnout typically assume increased political equality as well. Intuitively, turnout and representativeness are linked objectives; it would seem that greater participation would be, by definition, more representative. Yet, Hill and Leighley (1992) tested this assumption directly and found only a modest relationship between the level of turnout and the level of class bias within the electorate. Of course, some of the turnout-focused studies have considered issues of inequality indirectly through the use of sub-group analyses examining whether voter turnout rates increase more among some groups than among others. Studies, such as these, that document greater increases in participation for one group than another can provide important, if indirect, evidence that electoral reform may affect the distribution of voters. Unfortunately, the magnitude and direction of the inequality effects presented in this research vary. Many studies highlight the potential for differential effects of electoral reform across groups but they fail to reach consensus on the direction of the effect or to explicitly examine whether these effects are big enough to shift the balance of participation between income groups. For example, some have identified greater increases in turnout among more disadvantaged citizens (Avery and Peffley 2005; Highton 2004; Mitchell and Wlezien 1995; Wolfinger and Rosenstone 1980) or infrequent voters (Stein and Vornahme 2008), while others have found little or no significant difference across income groups (Knack and White 2000; Stein 1998). Some studies even warn of the potential for perverse consequences following electoral reform in which turnout among advantaged groups rises the most, leading to a presumed net increase in political inequality (Berinsky 2005; Berinsky, Burns, and Traugott 2001; Karp and Banducci 2000). These mixed findings provide little guidance for policymakers aiming to use electoral reform to equalize political participation in their states; however, it is possible to model 6

8 participation bias more directly as the ratio of participation among the rich versus the poor. This measurement strategy is frequently adopted by those finding substantial policy consequences (in the form of decreased social welfare provision) of income bias within the electorate (Avery and Peffley 2005; Hill and Leighley 1992). Further, previous studies do not account for differences in state electorates prior to reform most critically, the degree of income bias already embedded in the state registration rolls. Although research on participation typically adopts the familiar conceptualization of political participation as a two-stage process in which individuals must first register and then vote recognizing the long-held concern that the need to participate in the first stage may serve as a barrier to participation in the second stage (Brown, Jackson, and Wright 1999; Highton 1997; Jackson, Brown, and Wright 1998; Wolfinger and Rosenstone 1980), the potential for extant registration bias to alter the consequences of electoral reform is rarely explored. Expectations: When Electoral Reform Will Increase (or Decrease) Equality Informed by previous, conflicting research on the subject, we do not expect all electoral reforms to have the same effect on income-based inequality in political participation. Nor do we expect electoral reforms to have the same effect when adopted in states with different levels of pre-existing income bias in their registration rolls. Instead, we theorize an interactive effect. First, we distinguish among three different approaches to electoral reform, expecting the equality effects the balance of participation across income groups to vary for each approach to electoral reform. Second, we maintain that the ability of the reforms to promote or hinder equality in participation is conditioned by the state s level of pre-existing registration bias at the time the reform is adopted. The following section develops our theory more thoroughly. 7

9 Three Approaches to Electoral Reform In order to distinguish among types of electoral reform, we adopt the familiar conceptualization of political participation as a two-stage process in which individuals must first register and then vote (e.g., Brown, Jackson, and Wright 1999; Jackson, Brown, and Wright 1998; Highton 1997). As shown in Figure 1, we distinguish among electoral reforms that affect each stage of the participation process. We identify three distinct approaches to electoral reform: (1) reforms that make the first stage (registration) easier; (2) reforms that eliminate the potential for the first stage (registration) to serve as a barrier to the second stage (voting); and (3) reforms that make the second stage (voting) easier. Our expectations for each reform approach are discussed below. [Figure 1 about here] Approach 1: Registration Reform. The first reform approach focuses on making voter registration easier. Since the 1960s, a majority of states have adopted some form of mail-in registration that allows individuals to simply send in a postcard to register rather than requiring a trip to the registrar s office, notarization, or witnesses (Mitchell and Wlezien 1995; Springer 2006). Additionally, motor-voter programs were adopted by a few states before the 1993 National Voter Registration Act (NVRA) required all non-exempt states to implement motorvoter programs prior to the 1996 presidential election (Mitchell and Wlezien 1995; Springer 2006). Because these types of registration reforms make it easier for individuals to enter the electoral system, it seems reasonable to expect these sorts of registration-focused reforms to disproportionately benefit those who are underrepresented in the electorate prior to their enactment. This expectation is supported by Mitchell and Wlezien (1995), Avery and Peffley (2005), and others who found that liberalized registration laws have a greater impact on 8

10 increasing the registration rates of the disadvantaged; however, registration reforms are routinely shown to have only an indirect effect on voting rates. Although cost-reducing registration reforms may make it easier for people to register, they do not necessarily ease people s ability to actually vote once registered (Brown, Jackson, and Wright 1999; Brown and Wedeking 2006). Further, the lower correspondence between registering and voting for low-income citizens compared to higher income individuals (Brown, Jackson, and Wright 1999) suggests that equalizing the representativeness of registered individuals may not do much to equalize the representativeness of voters. As such, we propose Hypothesis 1: Registration-focused reforms (Approach 1) will decrease inequality yet, the effect is likely to be small in magnitude due to the indirect nature of this reform. Approach 2: Eliminating Registration Barrier. The second approach, which also pertains to registration laws, decreases the importance of registration as a gateway (or barrier) to voting and effectively eliminates the two-stage process. This is a more substantial reform than the sort described above. The clearest version of this approach is North Dakota s complete elimination of the registration process in Although no other state has adopted this unconventional policy, several states have enacted Election Day Registration (EDR) programs, which allow the greatest length of time for individuals to become engaged in an election while still remaining eligible to register to vote. EDR allows non-registered, potential voters the opportunity to register to vote when they show up at the polls on Election Day, thereby transforming the traditional two-stage participation process into one concise step. This electoral reform has been shown by numerous scholars to have disproportionate effects on different sociodemographic groups (Highton 1997; Knack 2001; Knack and White 2000). As a result, 9

11 Hypothesis 2 posits that reforms that effectively eliminate the registration barrier (Approach 2) will decrease inequality in voting. Approach 3: Convenience Voting. The third reform approach alters voting procedures to make it more convenient for those who are already registered to actually vote (Gronke, Galanes-Rosenbaum, and Miller 2008). One such method is allowing unrestricted, no-excuse, absentee voting. Another convenience voting reform is the enactment of in-person, early voting procedures, whereby public locations are open for established periods of time to allow individuals to cast their ballot in-person before Election Day. Since both of these reforms make it easier to vote, they only advantage those who are already registered. Additionally, since lowincome citizens are less likely to be registered, the benefit that these groups gain from this sort of electoral reform is likely to be smaller than the benefit to wealthier citizens (who are more likely to be registered). Not surprisingly, it was these types of reforms that Berinsky (2005) linked to perverse consequences in which political inequality increases post-reform. Therefore, Hypothesis 3 expects that reforms making voting procedures more convenient (Approach 3) will increase inequality in participation by making it easier for already registered voters to vote. Variation in States Pre-Existing Registration Bias In addition to our expectations about these distinct approaches to electoral reform, we maintain that reform effects will vary across state electorates. Namely, the level of bias in the state registration rolls prior to the enactment of any electoral reform is fundamental in conditioning the effects of any of the three reform approaches. We expect that the degree to which voting reforms disproportionately benefit wealthier or poorer citizens is a function of each group s under- or over-representation among registered voters in the state. For example, in a state where the rich and poor are registered at similar levels, electoral reforms focused on 10

12 increasing registration (Approaches 1 and 2) will, at best, have only a small equalizing effect on participation rates. In contrast, in a state with skewed registration rolls, that same reform can do more to equalize voting rates across income groups. Conversely, reforms that make it easier for registered voters to vote (Approach 3) have less potential to skew voting rates or increase inequality when income groups are registered at similar levels. But, if the rich are registered at much higher rates than the poor, a reform that makes it easier for the registered voters to actually vote has the potential to increase voting inequality. More concretely, for the reforms that focus on altering the registration process either by making registration easier (Approach 1) or making registration less of a barrier to voting (Approach 2) any equality-enhancing effects require the existence of inequality in the state registration rolls. If, however, the rich and poor are registered at the same rate, then making registration easier is unlikely to have a substantially larger impact on one group relative to the other. This expectation leads to Hypothesis 4: For reforms making it easier (or less necessary) to be registered (Approach 1 and Approach 2), inequality reduction will occur primarily in states where the pre-existing registration rolls are most stratified or skewed by income. Conversely, electoral reforms that ease voting procedures (Approach 3) may have considerable potential to increase equality in a state with relatively equal registration rolls, since any bias in voting stems from different decisions to turnout, independent of any barriers to registration. However, that same change to voting procedure, if enacted in a state in which the rich are much more likely than the poor to be registered voters, has the potential to further benefit those more advantaged citizens (who also have disproportionately higher registration rates), serving to increase political inequality. Therefore, Hypothesis 5 expects that voting 11

13 reforms (Approach 3) are likely to produce the largest increase in inequality in states where the pre-existing registration rolls are highly skewed by income. Empirical Strategy To test our hypotheses regarding the balance of political participation among the rich and poor, we developed a state-election-year dataset combining midterm and presidential elections from 1978 to After aggregating individual-level data on participation rates for each income group to the state-election level, we merged in data on the state-by-state adoption of five electoral institutions: two capturing the registration approach (universal mail-in registration and motor-voter provisions), one capturing the second approach of eliminating the registration barrier (election day registration), and two capturing the third convenience voting approach (universal absentee voting and in-person early voting). We estimated fixed-effects models with panel-corrected standard errors. 1 State dummy variables were used to account for any time in-variant unit effects (such as population density within each state, which could affect the distribution of precincts). Year effects were included to capture national trends in participation (such as the heightened attention and turnout in the 2008 presidential election) or policy adoption (adoption of motor voter following the federal NVRA). This modeling strategy is well-suited to our dataset, in which multiple cross-sections of data are pooled in order to estimate changes over time (for a more complete discussion, see Wilson and Butler 2007). We also accounted for the time-variant characteristics of each unit (state) with the inclusion of multiple contextual variables that capture shifts in demographic characteristics (such as income inequality) as well as features of the electoral calendar (such as the competitiveness of the presidential election) that may affect the nature of turnout in a particular state-year. By including these fixed-effects and time-varying control variables, the coefficient for each electoral 12

14 law dummy variable captures the average change in income bias following adoption of the electoral reform across states and years, as well as over and above changes in states demographic or election-specific characteristics. Measurement Table 1 presents descriptive statistics and information on the sources of each of our variables described below. [Table 1 about here] Estimating Participation Bias. The dependent variables for all of the models are constructed using individual-level data drawn from the November Supplement of the Census Bureau s Current Population Survey. The Current Population Survey (CPS) is a monthly survey of households conducted by the U.S. Census Bureau. Each month approximately 50,000 households are surveyed with special supplemental modules added in particular months and years. The survey s large samples from each state (more than 1,000 respondents each) are particularly important for this paper s objectives. In November of even-numbered years (i.e., during election years), the CPS includes a short battery of questions on voter participation. Respondents are asked whether they voted in that month s election and whether or not they were registered. Additionally, and critical to our study, information is also collected on the respondents household income level a key factor rarely captured in administrative datasets. Despite the strengths of these data, an analysis of social desirability bias among respondents found that CPS respondents often over-report participation. This over-reporting is primarily concentrated among regular church attendees, those with higher education, those living in more diverse districts, and among whites living in the Deep South (Bernstein, Chada, and Montjoy 2003). The substantial state-level differences in over-reporting were found to be 13

15 relatively stable over time (Bernstein, Chada, and Montjoy2003), however, we are able to absorb much of this variation through our use of fixed-effects models. Further, we see no reason to expect that any difference in over-reporting between wealthy and poor respondents would systematical vary within states or over time: a pattern of systematic reporting bias that could compromise our estimates. The individual-level sample from which we developed our aggregate measures was restricted to citizens at least eighteen years of age for whom we had data on voting and registration. For a typical year, this analytic sample contained an average of 1,801 respondents for each state with a standard deviation of 1,021 (minimum of 801, maximum of 6,738, median 1,512). From these cases, we drew distinctions along one key social characteristic: family income. Income provides resources and opportunities to families; as such, we use self-reported family income as a reflection of the class situation of individuals in the American electorate during the year in question (Ortiz and Stonecash 2008). For each election year, we classified respondents as Rich if their family income was equal to four or more times the poverty line that the federal government identified for a family of that size in that year. For example, a family of four in the year 2008 with a family income greater than $80,000 would be classified as rich in our sample. This cut-point was chosen to divide the sample into (approximate) thirds, which capture basic differences in family resources but retain a large enough sample size in each state to make adequate estimates for each election. Poor is defined as those individuals whose family income was equal to or less than two times the poverty threshold, approximately $40,000 for a family of four in For each state-year, we calculated the percentage of each group (i.e., rich) that were registered to vote and the percentage of each group that reported actually voting. 2 14

16 To construct the dependent variables used in this study, we began by generating a set of state-year-specific aggregated participation measures for the poor and rich sub-samples defined above. These include the Percent of Rich Voting, Percent of Rich Registered, Percent of Poor Voting, and Percent of Poor Registered. These variables were used to establish the general relationship between electoral reform and participation for each sub-group presented in Models 1 and 2. To estimate the measures of participation bias among rich versus poor, the central focus of this paper, we generated a measure of Income Vote Bias: simply the ratio of Percent of Rich Voting to Percent of Poor Voting. We multiplied this ratio by 100 to generate measures with a mean of (sd=29.77) indicating that, on average, rich citizens voted 66 percent more often than poor citizens. Higher values of the Income Vote Bias variable indicate greater inequality in the voting rates between the rich and poor and lower values correspond to less inequality (or more equality) in voting rates between income groups. In order to test our interactive theory, we used a parallel approach to generate measures of Registration Bias in which the ratio of [Percent Rich Registered-to-Percent Poor Registered * 100] was calculated. 3 We lagged this measure of registration by two elections in order to align presidential election years with presidential election years (and midterm election years with midterm election years), which have distinct patterns of voter registration and turnout. Like the dependent variable described above, higher values of the Registration Bias variable indicate greater inequality in the registration rates between the rich and poor and lower values correspond to less inequality (or more equality) in registration rates between income groups. Measuring State Electoral Reform. This paper focuses on five common reforms during the timeframe for which we have data on participation inequality: These institutions provide at least one example of each of the common reform approaches that were hypothesized 15

17 about in the previous section. Though it would be instructive to test our expectations across a longer time-series during which time states adopted other electoral reforms (e.g., literacy tests, extension of poll hours), we deem this timeframe acceptable since it captures the major changes in state electoral institutions that are most common today. Appendix A presents additional information on when each state adopted each reform; and Figure 2 illustrates the scope of institutional change by presenting the number of states with each reform at the beginning, middle, and end of our time period. [Figure 2 around here] The first approach making it easier to register was employed by states that adopted Universal Mail-in Registration or Motor Voter (either on their own accord or as incentivized by federal policy), respectively. Both are measured with a dummy variable indicating whether the state has universal mail-in registration that allows individuals to register to vote by mail without limitation or cause, or whether the state permits individuals to register at motor vehicle offices. 4 The second approach reducing the importance of the first stage (registration) as a gateway to the second stage (voting) is illustrated by one reform: allowing Election Day Registration (EDR). This dummy variable indicates whether the state allowed voter registration at the polls on Election Day. Ideally we would have also included North Dakota s no voter registration policy in this category; however, since voter registration was eliminated in North Dakota in 1951, prior to the time period under study, we are not able to isolate the impact of this innovative electoral reform on levels of political inequality in the state. Further, no other state has adopted North Dakota s no-registration policy. Instead, we coded North Dakota as having EDR (the next most lenient registration rule) for the entire time period under study since they are theoretically equivalent. 5 16

18 The third and final reform approach examined in this paper making it easier to vote once registered is illustrated by the adoption of Universal Absentee Voting or In-Person Early Voting procedures. Again, both were measured as dummy variables indicating whether the state allowed citizens to vote by absentee ballot for any reason without special limitation or cause or if the state provided for in-person, early voting allowing voters to cast their ballots at community centers, fire stations, libraries, and other governmental agencies well before an election. Both reforms offer a means for voters to cast their ballots on, or prior to, Election Day. Controlling for Other Factors Affecting Participation. In all of the models, we also controlled for a number of other variables that could affect the level of income-bias in political participation. The rationale for this highly specified model is to reduce the risk of an endogeneity problem in which we mistakenly attribute inequality effects to electoral reform that may instead be driven by shifts in demographic characteristics or the nature of a particular election. To capture important socio-demographic characteristics predictive of individual-level voter registration and turnout (for review, see Galbraith and Hale 2008), we controlled for changes in states education levels with Percent with a College Degree, changes in states income levels with Per Capita Income in 2008 dollars (in $1,000), and included a measure of state-level Income Inequality. Age is measured as the percentage of a state s population age sixty-five years and over, and we also include Percent Non-Hispanic White in each state-year. We also capture several important characteristics of the electoral calendar, since voting bias may be affected by the level of public discussion, interest, and mass mobilization surrounding particular races, issues, or campaigns (e.g., Smith and Tolbert 2007; Verba, Schlozman, and Brady 1995). Elections that tend to have higher voter turnout rates are controlled for with dummy variables: Presidential Elections and Statewide Elections (if the state has either 17

19 a gubernatorial or senatorial election on the ballot). We also included the number of Ballot Initiatives, as well as a measure of competitiveness: the Margin of Victory (for the highest race on the ballot) which could affect turnout. [Table 2 about here] Results Table 2 presents the results for the models pooled across presidential and midterm elections from 1978 through 2008 (N=800). Before examining our key dependent variable income vote bias we follow the common practice when investigating inequality effects in studies on voter turnout and evaluate sub-group effects: the effect of adopting each reform on turnout among the rich (Model 1) and poor (Model 2) sub-groups. In these models, very little is statistically significant. The one exception is a predicted decrease in voting among the poor after a state adopts in-person early voting. Although this is a curious result, it could be the case that different mobilization efforts in states with early voting are confounding this effect. 6 Of course, an advantage of this work is that it moves beyond basic sub-group analyses. As stated earlier, we believe that more insight can be gained by modeling the relative turnout of these two groups. Accordingly, in Model 3 we focus more directly on our hypotheses about the balance of participation by testing the effect of each type of electoral law on the ratio of participation between the rich and the poor: income vote bias. As expected (Hypothesis 3), inperson early voting has a significant and positive effect on income bias, indicating heightened inequality post-adoption. Based on the negative coefficients for early voting in the sub-group models (Model 1 and 2), we can infer that this heightened inequality results from a dampening effect of early voting on overall turnout, which has a slightly larger dampening effect among the poor. 7 When these two effects are combined the income skew among the electorate grows. 18

20 Although we find no other direct effects of reform on income bias in voting, we note that the sign of the variables are all in the expected directions: negative coefficients (decreased inequality) for the reforms representing the first two registration-focused approaches (Hypotheses 1 and 2) and positive coefficients (increased inequality) for the two convenience voting reforms (Hypothesis 3). This provides weak support for our general expectation that some approaches to electoral reform may promote equality, while others may exacerbate it. Model 4 tests our expectation that the inequality effects of electoral reform depends on the pre-existing level of bias in states registration rolls. We test this by interacting each of the electoral laws with the lagged measure of registration bias by income. We find evidence to support Hypotheses 3 and 4 for three of the five electoral laws one for each of our three approaches to reform. As expected from Hypothesis 4, we find a significant negative interaction between registration bias and two registration-focused reforms: motor voter (Approach 1) and EDR (Approach 2). Since these are both registration-focused reforms, it is quite logical that they have the most potential to equalize registration when registration is most skewed across income groups. Also as expected from Hypothesis 1 and 2, the magnitude of the interaction effect is larger for EDR (a reform that eliminates the registration barrier altogether) than it is for motor voter (a reform that takes a more indirect approach to increasing participation equality by registering more people to vote). 8 The significant positive interaction between registration bias and early voting supports Hypothesis 5, which expects convenience voting reforms (Approach 3) to further increase inequality in states with a greater income skew among the registered electorate. 19

21 For ease in interpretation, these three interactions are illustrated in Figure 3, which presents the marginal effects of adopting Motor Voter (top panel), EDR (middle panel), or Early Voting (bottom panel) when states face different levels of registration bias. 9 [Figure 3 about here] Beginning with the Motor Voter reform (an example of Approach 1), the figure illustrates support for Hypothesis 4 by predicting a decrease in inequality post-reform (evidenced by a downward sloping line) in states with skewed registration rolls. Plotting the marginal effects this way indicates that the effect is statistically significant when state registration rolls exhibit an income skew of 140 or more (meaning that the rich are 40 percent more likely to be registered than poor). This level of income bias is found in 39 percent of the state-election years during our time period. Also consistent with Hypothesis 1, we note a moderate-sized effect (ranging from no effect when registration rolls were not skewed to nearly one standard deviation (30) under the most extremely skewed registration rolls). Supporting Hypothesis 4 as well, we find a similar downward-sloping line indicating a reduction in inequality following state adoption of EDR, which becomes statistically significant when registration bias is 138 or higher (true for 44 percent of state-years we examine). Additionally, as expected (Hypothesis 2), we note a substantially steeper slope for EDR than Motor Voter, a statistically significant difference (Wald 4.15, p<.05), indicating a more substantial decrease in inequality when electoral reforms greatly reduce the registration barrier (rather than simply making it easier to register). The magnitude of EDR s effect in a state with average registration bias is small (less than half a standard deviation); yet, as registration bias increases the marginal effect increases to more than three times the standard deviation for the most highly skewed rolls. 20

22 Finally, the bottom panel provides support for Hypothesis 5. As expected, convenience voting reforms may exacerbate registration bias and, as a result, increase voting bias, depending on the level of pre-existing registration bias in the state. We find this to be true for in-person early voting, which increases inequality in voting if the registration bias is greater than 137 (a level found in 47 percent of the state-election years we examine). At the most extreme, this effect is estimated to be more than one and a half standard deviations in magnitude. Sensitivity Analysis: Higher and Lower Turnout-Elections Throughout this article, we presented expectations for how electoral reform may impact equality across elections. To test this, we pooled data across two distinct types of elections: Presidential election and midterm elections. We did this to test our general theory regarding the effect of electoral reform on political equality. Yet, we acknowledge the differences in patterns of political behavior across these two types of elections (Fitzgerald 2005; Knack 1995; Springer 2006). In particular, Presidential elections are known to have higher turnout rates and attract more media attention and mobilization efforts than midterm election years. To confirm that our decision to pool these two distinct types of elections did not drive our findings, we conducted a sensitivity analysis aimed at distinguishing equality effects across the two types of elections. Table 3 presents the key institutional and interactive coefficients from Table 2, as well as coefficients from parallel models run on sub-samples of the presidential and midterm election years. We find more similarity between the conclusions drawn from the pooled and presidential models in which coefficients for Models 3 and 4 align in direction, magnitude, and statistical significance. For midterm elections, we find a similar pattern of results but fewer significant differences, suggesting that our findings from the pooled models are less robust during midterm elections. The primary exception is early voting, which produces a substantially larger increase 21

23 in vote bias during midterm elections (14.02, p<.05) than during Presidential elections (1.62, not significant). This finding further suggests that early voting reforms have different patterns of effects than the other electoral reforms and that additional investigation into the unique effects of early voting is needed. Conclusion Despite a long-standing normative commitment to equality in political representation (Bartels 2008; Ortiz and Stonecash 2008), the United States has seen growing income inequality in recent decades, which raises the threat of increased political inequality as well. In this paper we examine one tool commonly employed to combat political inequality: state electoral reform. Although electoral reforms are often presumed to increase political equality by reducing the costs of registering and voting, our findings illustrate that electoral reform in and of itself does not have a clear equality-enhancing or equality-impeding effect. Instead, electoral reform is an umbrella term incorporating a number of distinct approaches to reform, each leading to distinct inequality effects. Further, the equality effects associated with electoral reform are conditioned by the level of pre-existing bias in state registration rolls at the time of adoption. Considering a variety of approaches to electoral reform, we find that motor voter, one of the two electoral reforms that make it easier to register, produced a small decrease in voting inequality in states where registration rolls were particularly unequal prior to reform. The second reform approach to eliminate the registration barrier through EDR was found to produce a larger reduction in inequality in those states with high registration bias. In contrast, in-person early voting, one of the two convenience voting reforms, had a net positive increase in inequality on average, as well as the potential to more substantially increase inequality in states with more stratified rolls. These findings highlight the need to consider the conditional effects of different 22

24 approaches to electoral reform in states with different patterns of class-based participation an interactive effect that has been unappreciated until now. In light of current trends in electoral reform (see Figure 2), these findings suggest both cause for concern and an opportunity to enhance equality. First, beyond the expansion of motor voter implementation over a decade ago, today s most popular electoral reforms focus on voting procedures (Approach 3). Yet as our analysis shows, it is this type of reform that has the potential to further skew political participation toward those with more resources. As more states consider adopting some form of early voting, this possibility needs to be seriously considered, particularly in states with highly stratified registration rolls. In fact, we saw this dynamic at play during the 2008 presidential election: among the ten states experiencing the most income bias in voting, eight had adopted early voting and seven of these had registration bias levels more than a standard deviation above the mean. 10 Our data does suggest, however, that Election Day Registration has the potential to reduce participation inequality. EDR is the electoral reform that demonstrated the most potential to promote equality in political participation, yet only a few states currently employ EDR, including only one of the ten high-bias states discussed above. As such, EDR may provide an untapped tool for these states to counterbalance the perverse effect of early voting reforms. Taken together these findings have important implications for understanding political representation among disadvantaged groups and for drawing a more nuanced connection between scholarship on voting behavior and political institutions. 23

25 References Ansolabehere, Stephen, John M. de Figueiredo, and James M. Snyder, Jr Why is There so Little Money in U.S. Politics? Journal of Economic Perspectives 17: Avery, James, and Mark Peffley Class Bias in State Electorates and the Adoption of Restrictive Welfare Policies after TANF. State Politics and Policy Quarterly. Bartels, Larry M Unequal Democracy: The Political Economy of the New Gilded Age. New York: Russell Sage Foundation and Princeton University Press. Beck, Nathaniel and Jonathan N. Katz What to Do (And Not to Do) with Time-Series Cross-Section Data. American Political Science Review 89(3): Berinsky, Adam The Perverse Consequences of Electoral Reform in the United States. American Politics Research 33: Berinsky, Adam, Nancy Burns, and Michael Traugott Who Votes by Mail? A Dynamic Model of the Individual-level Consequences of Vote-by-Mail Systems. Public Opinion Quarterly 65(2): Bernstein, Robert A., Anita Chada, and Robert Montjoy Cross-State Bias in Voting and Registration Over-reporting in the Current Population Surveys. State Politics and Policy Quarterly 3(4): Brambor, Thomas, William Roberts Clark, and Matt Golder Understanding Interaction Models: Improving Empirical Analysis. Political Analysis 14(1): Brown, Robert, Robert Jackson, and Gerald Wright Registration, Turnout, and State Party Systems. Political Research Quarterly 52(3): Brown, Robert D., and Justin Wedeking People Who Have Their Tickets but Do Not Use Them: Motor Voter, Registration and Turnout Revisited. American Politics Research 24

26 34: Cain, Bruce E., Todd Donovan, and Caroline J. Tolbert (Eds.) Democracy in the States: Experiments in Electoral Reform. Washington, DC: The Brookings Institution. Fitzgerald, Mary Greater Convenience but not Greater Turnout: The Impact of Alternative Voting Methods on Electoral Participation in the United States. American Politics Research 33: Galbraith, James K. and J. Travis Hale State Income Inequality and Presidential Election Turnout and Outcomes. Social Science Quarterly 89(4): Gilens, Martin Inequality and Democratic Responsiveness. Public Opinion Quarterly 69(5): Gronke, Paul, Eva Galanes-Rosenbaum, and Peter A. Miller Early Voting and Voter Turnout. In Democracy in the States: Experiments in Election Reform eds. Bruce Cain, Todd Donovan, and Caroline Tolbert. Washington, DC: Brookings Institution Press. Highton, Benjamin Easy Registration and Voter Turnout. Journal of Politics 59: Highton, Benjamin Registration and Voting in the United States. Perspectives on Politics 2: Hill, Kim Quaile, and Jan Leighley The Policy Consequences of Class Bias in State Electorates. American Journal of Political Science 36(2): Jackson, Robert A., Robert D. Brown, and Gerald C. Wright Registration, Turnout, and the Electoral Representativeness of U.S. State Electorates. American Politics Quarterly 26: Jacobs, Lawrence, and Theda Skocpol (Eds.) Inequality and American Democracy: What 25

27 We Know and What We Need to Learn. New York: Russell Sage. Karp, Jeffrey A., and Susan A. Banducci Going Postal: How All-Mail Elections Influence Turnout. Political Behavior 22(3): Knack, Stephen Does Motor Voter Work? Evidence from State-Level Data. The Journal of Politics 57: Knack, Stephen Election-Day Registration The Second Wave. American Politics Research 29: Knack, Stephen, and James White Election-Day Registration and Turnout Inequality. Political Behavior 22: Leighley, Jan E., and Jonathan Nagler Individual and Systemic Influences on Turnout Who Votes, Journal of Politics 54: Lijphart, Arend Unequal Participation: Democracy s Unresolved Dilemma. American Political Science Review 91: Mitchell, Glenn E, and Christopher Wlezien The Impact of Legal Constraints on Voter Registration, Turnout, and the Composition of the American Electorate. Political Behavior 17: Ortiz, Hector, and Jeffrey M. Stonecash The Class Gap. In Laura R. Olson and John C. Green (Eds.) Beyond Red State, Blue State: Electoral Gaps in the Twenty-First Century American Electorate. New Jersey: Pearson Education Press. Piven, Frances Fox, and Richard A. Cloward Why Americans Don t Vote. New York: Pantheon Books. Piven, Frances Fox, and Richard A. Cloward Why Americans Still Don't Vote and Why Politicians Want It That Way. Boston: Beacon Press. 26

28 Rigby, Elizabeth and Gerald C. Wright. Forthcoming. Whose Statehouse Democracy? In Peter Enns and Christopher Wlezien (Eds.). Who is Represented? New York: Russell Sage. Rosenstone, Steven J., and John Mark Hansen. 1993; Mobilization, Participation, and Democracy in America. New York: Macmillan. Schattschneider, E. E The Semi-sovereign People: A Realist s View of Democracy in America. Hinsdale, Ill.: Dryden Press. Smith, Daniel, and Caroline Tolbert The Instrumental and Educative Effects of Ballot Measures: Research on Direct Democracy in the American States. State Politics and Policy Quarterly 7(4): Springer, Melanie J Electoral Institutions, Voter Turnout, and American Democracy: The States, Doctoral dissertation, Columbia University, and book manuscript. Stein, Robert M Early Voting. Public Opinion Quarterly 62(1): Stein, Robert M., and Greg Vornahme Engaging the Unengaged Voter: Vote Centers and Voter Turnout. Journal of Politics 70(2): Tolbert, Caroline, and Daniel Smith Representation and Direct Democracy in the United States. Representation: Journal of Representative Democracy 42: Verba, Sidney, Kay Schlozman, and Henry Brady Voice and Equality: Civic Voluntarism in American Politics. Cambridge, Mass.: Harvard University Press. Wilson, Sven E., and Daniel M. Butler A Lot More to Do: The Sensitivity of Time- Series Cross-Section Analyses to Simple Alternative Specifications. Political Analysis 15: Wolfinger, Raymond E., and Steven J. Rosenstone Who Votes? New Haven, Conn.: Yale University Press. 27