Finding DORI: Using Item Response Theory to Measure Difficulty of Registration in the U.S. and Its Impact on Voters

Abstract

How do states differ in how difficult they make voter registration, and what effect does this have on voters? We propose and validate a new Difficulty of Registration Index (DORI) calculated via an item response theory (IRT) model of five key dimensions of registration (automaticity, portability, deadline, mode, and preregistration) for each state from 2004 to 2020. Since 2004, most states eased registration processes, with Democratic statehouses in racially diverse and young states leading the way. Using CCES data, we find that DORI is associated with increased probability that voters experience problems registering and failing to turnout (in both self-reported and validated turnout data). These effects are pronounced for young voters. This study holds lessons for how restrictive registration procedures can change the shape of the electorate and make it harder to achieve political equality.

Keywords

Voter registration youth turnout state politics item response theory election laws

How do states differ in the difficultly of voter registration procedures, and what effect does this have on voters? Extant research typically examines the effect of specific registration dimensions on turnout, such as preregistration (e.g., Holbein & Hillygus, 2016), deadlines (e.g., Grumbach & Hill, 2020), or portability (e.g., McDonald, 2008), or examines the overall costs of voting on turnout, which does not distinguish between registration and voting procedures (e.g., Juelich & Coll, 2020). Indeed, when scholars have attempted to measure registration difficulty, they have typically done so with indices that include both registration and voting laws (e.g., Li, Pomante, and Schraufnagel 2018; Hill & Leighley, 1999; King, 1994). While these approaches have substantially advanced our understanding of voter registration in the US, insights from these studies may be limited by either their treatment of registration laws in isolation or their inability to separate the effect of registration difficulty from voting difficulty.

Measuring the overall level of registration difficulty in each state and over time is critically important for understanding voter turnout, particularly among young voters. We know from the foundational works of Downs (1957) and Riker and Ordeshook (1968) that higher costs of voting, or how much time and resources one must commit to successfully voting, should depress voter turnout. Registration is a costly step in the process of voting that falls disproportionally on the young, as they are most likely to be new voters or voters who have recently moved (e.g., Wolfinger & Rosenstone, 1980). Confusing or difficult procedures can create registration problems for young voters; indeed, Millennials are more likely to report registration problems than other generations (CCES, 2018). Since voting is habitual, creating confusion for young voters could turn them off from participating for their lifetime and can distort the electorate (Plutzer, 2002; Coppock & Green, 2016; Fowler, 2017). In competitive elections, such as the 2016 presidential election where the outcome was determined by a fraction of voters in a handful of states, or the 2020 Iowa second congressional race where just 6 votes separated the winning and losing candidates, any number of voters being turned away have the potential to change election outcomes.

To examine the effect of registration procedures on voters, we create a Difficulty of Registration Index (DORI) using a hybrid item response theory (IRT) model for five major dimensions of registration—automaticity, deadline, mode, portability, and preregistration. These dimensions govern if an eligible voter must take action to register, when and how an eligible voter can register, whether the registration is valid across the state rather than a specific domicile, and whether registration is available to voters who are currently underage. These standards are well-established in the literature as key dimensions of registration procedures. The IRT approach allows us to take into account the unique contribution of each dimension to the overall difficulty of registration procedures in each state and over time.

This approach gives us leverage on variation across the states in registration difficulty and how states with higher levels of difficulty can cause problems for voters that prevent them from casting a ballot. We first identify the major dimensions of registration laws across the states and then describe how we use these dimensions to create DORI scores for each state using IRT. We then show how DORI scores have changed over time, with many states making registration easier since 2004. We validate DORI by assessing its state-level correlates and by comparing its predictive performance to an additive index of registration laws. We use DORI to explore the consequences of registration difficulty for voters. We find that voters in states with higher DORI scores are more likely to report experiencing problems registering to vote, and are less likely to vote as a result. Troublingly, we find that these effects are most pronounced for young voters. In all, the creation of DORI and our demonstration of its impact on voters provide a new tool for studying voter registration in the U.S. federal system and its electoral impact.

Five Dimensions of Voter Registration Procedures in the States

In the U.S. federal system, voter registration laws are mostly determined by state governments. National laws such as the Voting Rights Act of 1964 (VRA), the National Voter Registration Act of 1993 (NVRA), and the Help America Vote Act of 2002 (HAVA) create basic voting and registration regulations across the U.S., but state prerogatives have created significant spatial and temporal variation in five parameters, or dimensions, of voter registration laws. These dimensions are automaticity (i.e., whether action is required for voters to become registered), mode (i.e., how a state allows voters to register), deadline (i.e., whether and when voters must be registered before election day), portability (i.e., whether voters’ registration is honored state-wide), and preregistration (i.e., whether registration procedures are available to otherwise eligible voters who are underage).

We identified these five dimensions after a review of the extant research. Our goal was to identify the major dimensions of registration procedures and then to use this information to create a single measure of registration difficulty in each state over time. We used two criteria to determine what constitutes an important dimension of voter registration law. First, the dimension and the underlying laws that shape it must be primarily focused on registration procedures and not on determining voting procedures, including who is eligible to vote. Second, it must shape the steps that individual voters must take to fully register, and not govern how organizations or political parties that may seek to register voters.¹ Although these dimensions do not capture all possible voter registration laws, scholars focus on these dimensions above all others in determining how difficult it is for an eligible individual to register to vote (Leighley & Nagler, 2014; Holbein & Hillygus, 2016; Fowler, 2017; Yu, 2019). Below, we define each dimension and summarize its importance for shaping registration costs for individual voters.

Automaticity

Whether action is required for voters to become registered is a key dimension that varies across the states. As of 2020, eighteen states² had instituted mechanisms for automatically registering eligible citizens to vote (NCSL, 2019a). These AVR provisions allow state agencies to share with election administrators the basic information they gather from eligible citizens. Election administrators use the information to create or update voter registration records. In this way, voters are registered without having to spend time and resources completing the specific actions required to become registered. AVR is thus viewed as a way to reduce the costs of voting (Li et al., 2018). AVR is a relatively new innovation, so few comprehensive studies have been conducted on their effect, though McGhee and Romero (2020) find that AVR laws greatly increases the voter registration rates among Latinos, and modestly increase voter registration rates among Asian Americans and young people as well.

Mode

The mode dimension captures how a state allows a voter to register. All states—whether due to the NVRA or by their own regulations—have paper voter registration forms that can be submitted in-person or by mail. Arizona was the first state to allow voters to register online in 2002. By 2020, 40 states had enacted online voter registration (OVR), while 10 states kept paper-only procedures (NCSL, 2019b). OVR is a popular mode of registration because it makes registration accessible on laptops or mobile devices, and it decreases administrative costs to the states (Hicks et al., 2016). OVR is associated with increased turnout especially among young voters (Yu, 2019).

Deadline

Whether and when voters must be registered before election day is another key dimension of variation in registration laws across the American states. Some states require registration to be completed before election day, typically 10–30 days prior. Most states’ deadlines have not changed over time, though many states have begun adopting election day registration (EDR) which allows voters to register and vote on the day of the election. EDR allows people to register and vote when interest in voting peaks near the election (Vonnahme, 2012). Having a deadline closer to the election is strongly associated with higher overall voter turnout and higher likelihood that an individual votes (Highton, 1997; McDonald, 2008; Springer, 2012; Vonnahme, 2012; Burden et al., 2014), and having no deadline at all has a greater effect on younger voters than older voters and on movers than non-movers (Knack & White, 2000; Larocca & Klemanski, 2011; Leighley & Nagler, 2014; Grumbach & Hill, 2020).

Portability

Portable voter registration means that a person’s registration is valid across their state of residency rather than tied to their specific domicile at a given time. Only eight states³ explicitly specify that registration is permanent and portable, meaning voters who register and move within the state may show up at their new polling place and update their address during early voting or on election day. Three of these states allow voters to cast a full ballot while the other five allow voters to cast a provisional ballot. Portability is an important dimension as it governs whether people are able to re-register and cast a full ballot regardless of their transience within the state. In his study of the 2004 election, McDonald (2008) found that portability laws boosted turnout, especially among recent movers.

Preregistration

Preregistration provides registration procedures for otherwise eligible voters who are currently underage. Twenty-two states allow young people to preregister to vote at least 60 days prior to turning 18 years old (Li et al., 2018; NCSL, 2019c), with some allowing preregistration for people as young as 16. Offering preregistration has been shown to increase turnout among young voters (Holbein & Hillygus, 2016; Fowler, 2017). One reason why is because preregistration allows people to register when they get their first drivers’ license. Many young people do not interact with government agencies that offer registration services after they obtain their first drivers’ license (Grumbach & Hill, 2020). Another reason why preregistration boosts participation is because it provides a pathway to registration when interest in participation is relatively high as students learn basic civics in high school (Holbein & Hillygus, 2016). Students who are able to preregister also have teachers and peers to help them, providing clarity to an often cloudy process and helping to boost turnout among young people when they become age eligible (Holbein & Hillygus, 2016; Fowler, 2017).

Estimating and Validating a Difficulty of Registration Index

Using the major dimensions of voter registration procedures, we construct a single Difficulty of Registration Index (DORI) using a hybrid item response theory (IRT) model. DORI is structured to capture state variation in registration difficulty and to allow researchers to estimate the effect of registration on individual turnout. Existing measures are limited in their ability to achieve these goals by either lack of separation of registration dimensions from voting dimensions, or the exclusion of some major dimensions of voter registration. In their Cost of Voting Index (COVI), Li et al. (2018) include three dimensions—deadline, restrictions (which includes AVR, OVR, and other provisions), and preregistration—that measure the hurdles an individual could face when attempting to register to vote.⁴ These dimensions are included in the measure alongside regulations on group registration activity and voting rules and hours. We separate these registration dimensions from voting ones, break the restrictions dimension into the more specific automaticity and mode dimensions, and include the portability dimension, which is not included in COVI despite empirical evidence suggesting it significantly raises the level of registration difficulty for certain voters. Other extant indices such as those from King (1994) and Hill and Leighley (1999) tap the mode and deadline dimensions, though they are limited to these dimensions only.

In being estimated via IRT, DORI also has the advantage of not assuming that each dimension makes registration equally difficult. Instead, IRT allows each dimension to contribute differently to our estimate of difficulty. This allows for the parsimonious study of voter registration laws. Laws often duplicate the effects of one another, such as EDR being structurally equivalent to having a registration deadline of 0 days. Treating each dimension as wholly separate and equal, either by modeling them as so in a multivariate model or weighting them equally in an index, can pose multicollinearity problems or would fail to meaningfully reduce registration procedures to a single difficulty measure.⁵ IRT, though, allows researchers to account for several dimensions while allowing those dimensions to contribute differently to the model. Sometimes, that means items (in this case, registration laws) that do not significantly distinguish between the units of analysis (in this case, states) get dropped from the estimation. This is simply because there is not enough meaningful variation to help classify units on a single dimension. We see this as a feature and not a bug, as it allows us to extract a single dimension from the data rather than forcing noise (i.e., registration laws that do not really distinguish between states) into the measure. In the end we are left with a single, meaningful measure that classifies states by their level of difficulty in space and time. The flexibility of the hybrid IRT model has led it to be used in a number of similar applications, such as the extraction of a single measure of state-level democracy (Grumbach, 2021).

We begin constructing DORI by collecting data for each of the above major dimensions. We do this for each state for each election cycle from 2004 to 2020. We measure the deadline dimension using a dichotomous indicator that distinguishes which states have EDR and which do not (NCSL, 2019d; Book of the States, 2020). We also attempted to capture this dimension using various measures of the non-EDR deadline for registration. We tried continuous (in number of days prior to election day) and dichotomous (1 for greater than 21 days prior to election day and 0 otherwise) measures, but these dropped out during the estimation of the IRT model; there simply was not enough variation across states or over time for the model to use non-EDR deadline as a discriminating item. Similarly, we attempted to include a measure of EDR location, with states coded as 1 if voters can register at a central location on election day, 2 if voters can register at their polling place on election day, and 0 otherwise. Again, this measure dropped out during the estimation as there is not enough variation to meaningfully distinguish between states.

We capture variation in mode by measuring which states allowed OVR and in which years (NCSL, 2019b). We also attempted to include a measure of which states allow voters to check their registration online, but there was near perfect overlap between these two measures and therefore including both did not allow the IRT model to converge. We measure portability ordinally, giving states that provide for portability on full ballots a 2, on provisional or limited ballots a 1, and states that do not allow portability a 0 (McDonald, 2008; Harmon et al., 2015; Brennan Center for Justice, 2017).

We measure preregistration ordinally, giving states with the most generous preregistration provisions the highest score on a five-point scale and states with no preregistration the lowest (Li et al., 2018; NCSL, 2019c). Finally, we capture automaticity by measuring whether and when each state adopted AVR laws using information from NCSL (2019a). Although states differ in which state agencies are allowed to transfer information to election administrators (for example, Alaska uses its Permanent Fund Dividend while others use the state’s Department of Motor Vehicles), we chose not to include this information because there is no clear coding scheme for which agencies are easier and which are more difficult for voters.

Table 1 provides a summary of each dimension definition and its coding process. After coding, each dimension was modified so that high scores indicate greater difficulty (e.g., the absence of AVR, rather than presence of AVR) in order to create an overall difficulty score. The IRT model we fit to the data is derived from a two-parameter logistic regression (for the model’s binary inputs) combined with a graded response model application (for the model’s ordinal inputs) with standard errors clustered at the year level. Although nine different measures of registration laws across the five dimensions were tried, only one measure from each dimension contributed to the extraction of state-year difficulty scores. Low scores on the resulting index indicate relatively easy registration procedures in a given state-year while high scores indicate relatively difficult registration procedures. We provide the full model output and additional technical information in the Supplementary Material, and Figures 1 and 2 below describe the model results in brief.

Table 1.

Coding Key Dimensions of Voter Registration Regimes in the United States

Dimension name	Definition	Source	Operationalization
Deadline	Whether the state allows voters to register on election day or else requires registration to be completed some number of days before an election	NCSL (2019d)	Binary; 1 if deadline is effectively 0 days (i.e., EDR allowed) and 0 if greater than 0 days
Mode	Whether or not the state allows voters to register online (all states allow by mail or in person, so online is the key variant in mode state-to-state and over time)	NCSL (2019b)	Binary; 1 if online registration allowed and 0 if no online registration
Preregistration	Whether or not the state allows otherwise eligible voters who are currently underage to register to vote	Li et al. (2018) and NCSL, 2019c	Ordinal; 5 = may preregister at 16 years old, 4 = 17 years old, 3 = 17.5 years old, 2 = within 90 days of turning 18, 1 = within 60 days of turning 18, and 0 = no preregistration
Portability	Whether or not registration is permanent and portable to any precinct within state rather than tied to a voter’s domicile at the time of registration. In effect, these laws allow voters to appear at their new precinct on election day, change their address, and cast either a full or provisional ballot	McDonald (2008); Harmon et al. (2015); Brennan Center for Justice (2017)	Ordinal; 2 if registration is portable and voter can cast regular ballot, 1 registration is portable and voter can cast a provisional or limited ballot, and 0 if no portability
Automaticity	Whether or not an individual is able to be automatically registered to vote when they interact with any other state agency	NCSL (2019a)	Binary; 1 if state has AVR and 0 if not

Operational definition provided in previous column reflects data as collected by the researchers. These variables were flipped so that high scores indicate greater difficulty before application of IRT model.

Figure 1.

Results of the IRT Procedure Used to Produce State-Year DORI Scores. Note: (a) plots characteristic curves denoting the probability (y-axis) that states implementing each dimension, at a particular point in time, earn a particular score on the latent DORI measure (theta; x-axis). (b) Plots information curves for each dimension; that is, how much (y-axis) each dimension differentiates between where states fall on DORI (theta; x-axis). (c) Bar chart plotting mean DORI scores (rescaled from 0 to 1) by state pooled over time.

Figure 2.

Changes in DORI over time for five states.

Figure 1(a) depicts the item (two-parameter) and category (graded response) characteristic curves for each dimension used in the hybrid IRT model. On the y-axis is the probability that states are scored “difficult” on each dimension at a particular point in time, while the x-axis is the value of DORI (i.e., theta). The s-shaped curves demonstrate that the probability that states earned a high (low) DORI score is nearly guaranteed if they are coded as difficult (easy) for the corresponding registration dimension.

Figure 1(b) plots information curves for each dimension, which is an estimate of how much each dimension differentiates where states fall on DORI. High peaks, therefore, indicate that the particular dimension strongly differentiates whether states fall above the corresponding DORI score on the x-axis. The results show that lacking AVR, and to a lesser degree lacking OVR, helps discern whether states earn low DORI scores, while the remaining three dimensions are comparatively better at discerning which states earn a high score.

Figure 1(c) is a bar chart plotting the DORI scores for each state, averaged across time and re-scaled to range from 0 (least difficult) to 1 (most difficult). North Dakota is on average the lowest scoring state on DORI. This makes sense as North Dakota has no formal registration requirements. It was coded as in the “easiest” category on every dimension. A diverse set of states like South Dakota, Mississippi, Arkansas, Michigan, Tennessee, and Pennsylvania tend to score high on DORI. Georgia, despite newly enacted restrictions on absentee and early voting, has relatively low DORI score owing to having both AVR and OVR. New York, by contrast, has a relatively high average DORI score since it has neither EDR, AVR, nor portability. These examples highlight the importance of separating voting and registration restrictions.

Figure 2 plots changes in DORI scores over time for five states: North Carolina, Georgia, Mississippi, North Dakota, and Oregon. North Carolina and Georgia are two of the few states that increased DORI over the time period studied. From 2010 to 2012, North Carolina allowed 16 year olds to preregister to vote. From 2008 to 2012, the state also allowed for EDR during early voting. These laws were repealed in 2013—thereby increasing the state’s score on DORI—and remained repealed until 2016 when federal courts reinstated them for the 2018 elections. Georgia reduced access to preregistration for 16 and newly-turned 17 year olds starting with the 2010 election. This increase in DORI was off-set with the adoption of OVR for the 2014 election and AVR for the 2016 election. While some states have remained consistently difficult, like Mississippi, or easy, like North Dakota, most states saw a reduction in DORI scores since 2004. Oregon is the best example of falling DORI scores. In 2004, the state was considered one of the most difficult with no EDR, AVR, portability, preregistration, or OVR—a DORI score of 1. By 2020, Oregon achieved a DORI score of 0, tied with North Dakota. Many of the states followed Oregon’s trend and, as a result, there was a precipitous drop in DORI for most states over time.

Validity Check: The State-Level Correlates Of DORI

As a validation exercise, we assess whether DORI is associated with factors the literature suggests it should be (Carmines & Zeller, 1979). We know from previous scholarship that states tend to adopt more restrictive or permissive voting and registration laws based on the partisan control, electoral competition, and the racial and age makeup of the state (Bentele & O’Brien, 2013; Rocha & Matsubayashi, 2014; Hicks et al., 2015; Biggers & Hanmer, 2015; 2017; Hicks et al., 2016). Given the heterogeneity of the parties’ base voters, strategic parties will seek to use control of state government to shape voting policies in a way that helps their base voters and/or hinders their opponents’ voters from turning out (Bentele & O’Brien, 2013; Rocha & Matsubayashi, 2014; Hicks et al., 2015; Biggers & Hanmer, 2017). Indeed, partisan efforts to shape turnout has a long history in the U.S., as it was the impetus behind Democratic Party efforts to curb Black, pro-Republican turnout in the South after Reconstruction (Bentele & O’Brien, 2013; Levitsky & Ziblatt, 2018). Sometimes, though, voting and registration policies get adopted before becoming clearly associated with one party over the other. For example, OVR diffused across the states with backing from Democratic and Republican statehouse majorities (Hicks et al., 2016) and states with significant elderly populations tend to allow no-excuse absentee and early voting as a convenience, regardless of which party controls state government (Biggers & Hanmer, 2015).

DORI scores should be associated with the direction parties want to move registration stringency and the factors that determine how willing they are to do so. Specifically, unified Republican control of state government should be associated with higher DORI scores, especially in racially diverse, young, and competitive states, while Democratic control should be associated with lower DORI scores, especially under those same circumstances. If the registration dimensions that constitute DORI are not polarized, racial and age demographics should still be associated with DORI.

Thus, we model DORI as a function of party control of state government,⁶ minority population,⁷ youth population,⁸ and electoral competition.⁹ The model is a two-way fixed effects (FE) model on data from all 49 of the 50 states (excluding Nebraska for its nonpartisan legislature) for election cycles from 2004 to 2020. Time (election cycle) and unit (state) FE are included. We interact party control with the latter three variables because the effect of party control will depend in part on demographic and competitive conditions.

The full model output is presented in Table A1 of the Appendix. We summarize the results here and provide a depiction of the key marginal effects in Figure 3. The results of our validation exercise show that states that elect unified GOP governments (compared to divided governments) on average are associated with higher DORI scores (B = 0.08, p < 0.05; Table A1, Model 1), while unified Democratic governments are marginally associated with lower DORI scores (B = −0.04, p < 0.10; Table A1, Model 1). States that are racially diverse are significantly associated with lower DORI scores (B = −2.37, p < 0.05; Table A1, Model 1). Although most states have trended toward lower levels of difficulty, unified Democratic governments in high minority population states are most clearly associated with lower DORI scores (B = −0.33, p < 0.05; Table A1, Model 2). A similar interactive effect is detected for Democratic government and youth population (B = −5.87, p < 0.05; Table A1, Model 3). Unified Republican control of state government in diverse (B = 0.74, p < 0.05, Table A1, Model 5) and young states (B = 4.14, p < 0.05; Table A1, Model 6) in particular are significantly associated with higher DORI scores. Each model explains a substantial amount of variation with adjusted model fit parameters above 0.777 in all models.¹⁰ Overall, the results comport with what would be expected from the established literature, providing a key validity check for DORI.

Figure 3.

The effect of unified party control on DORI at different levels of minority and youth population.

Using DORI to Assess Impact of Registration On Voters

If we have properly measured registration stringency with DORI, we should also detect an effect of DORI on the propensity for individuals to report problems registering to vote and, as a consequence, failure to turnout to vote. A test of this kind, at the individual level, is different from typical uses of voting and registration law indices. We are not merely examining whether DORI is associated with overall turnout, but if individuals in states with higher DORI scores are more likely to experience problems registering, and the consequences of those negative experiences on their propensity to vote.

We know that registration imposes time and information costs on would-be voters and can depress turnout (Wolfinger & Rosenstone, 1980; Highton, 1997; Springer, 2012; Leighley & Nagler, 2014; Holbein & Hillygus, 2016; Stockemer & Rocher, 2017; Grumbach & Hill, 2020). Indeed, it has long been established that higher costs of voting in general is associated with lower turnout (e.g., Riker & Ordeshook, 1968) with recent empirical showing that states that impose higher overall costs of voting see lower turnout, particularly among young voters (Juelich & Coll, 2020). When states ease registration costs, through preregistration (e.g., Holbein & Hillygus, 2016), EDR (e.g., Grumbach & Hill, 2020), or portability (e.g., McDonald, 2008), turnout tends to increase among young voters.

Young voters are disproportionately affected by the process of voter registration because they are (1) more likely to be mobile, thus requiring frequent re-registration (Timpone, 1998; Highton, 2000; Ansolabehere et al., 2012) and (2) more likely to have to register for the first time (Holbein & Hillygus, 2016; Fowler, 2017). At various points in young adulthood, people may move into or out of college, move out of their parents’ house, relocate for career opportunities, change rental units, or buy a first home.¹¹ For movers, re-registering may be necessary if they leave their voting precinct, which is likely since precincts are quite small. It is also mostly young voters who need to register for the first time. The NVRA enables people to register when they get their drivers’ license, but this is not likely to help many young voters unless they are allowed to preregister at 16 years old. Political interest can help people navigate the registration process (Stockemer & Rocher, 2017), but young people tend to be the least interested in politics due to lack of outreach from political parties (Leighley & Nagler, 2014) and lack of encouragement to participate while growing up (Lawless & Fox, 2015).

In the first step toward casting a ballot, particularly difficult procedures can lead would-be voters to misstep, get confused, and experience problems registering to vote. Thus, since registration is a cost that affects all voters (e.g., Juelich & Coll, 2020), those living in high DORI states should be more likely to experience problems when registering to vote. But it is young voters who are most substantially affected by the ease or difficulty of a states’ registration regime. The absence of preregistration or online modes of registration can make it more difficult for teachers, parents, mentors, or older friends and siblings to guide young people through the process of registering. It may be unclear by when voters need to register, or if they have to re-register if they recently moved. Inconvenient registration regimes that do not allow preregistration, online registration, portable registration, or generous deadlines can create problems for otherwise eligible voters. Consequently, compared to older people, young people in states where it is more difficult to register to vote should be more likely to report problems registering to vote.

If the registration process itself creates problems for people seeking to get registered, these burdened voters may in turn be less likely to vote. Social psychological research suggests that if people hold negative attitudes toward a behavior, they will be less likely to intend to carry it out, and less likely to perform the behavior itself (Ajzen, 2005; Ajzen & Fisbein, 2005). People who experience registration problems may therefore come to hold negative views about voting because they have been signaled that the process itself is overly complex. This could decrease their intention to turn out by signaling that voting is burdensome and not worth navigating. Indeed, negative attitudes toward voting have been shown to decrease intentions to vote on election day (e.g., Netemeyer & Burton, 1990; Hansen & Jensen, 2007). Greater incidences of problems registering may cast voting in a negative light, making it less likely that people actually vote. Thus, we expect that people in states where it is more difficult to register to vote will be less likely to report actually voting if they report a problem registering to vote, and that these effects are likely most pronounced for young people compared to older people.

Measuring Individuals’ Problems Registering and Failures to Vote

We gathered data on our two-fold outcome of interest: first, whether or not an individual reported experiencing registration problems and second, whether or not the individual reported actually voted, given that they experienced registration problems. To measure these outcomes, we use pooled CCES survey data from 2008 to 2018.¹² On the CCES, respondents were asked (1) whether or not they voted¹³ and (2) if they experienced a problem with their registration when attempting to vote.¹⁴ Using the information provided by these questions, we construct two dependent variables. The first is a dichotomous measure of whether or not respondents reported issues with registration when trying to vote in each cycle. The second is an ordinal measure of whether or not respondents either successfully voted without registration issues (j = 0), experienced a registration issue but voted anyway (j = 1), or experienced a registration issue and were unable to vote (j = 2).¹⁵ Note that we provide two versions of this ordinal outcome variable in our analyses; one based on self-reported inability to vote, and one which uses the CCES′ respondent-level vote validation meta-data to assess whether or not those who self-reported experiencing issues with registering (the binary outcome variable discussed first) were able to successfully cast a ballot in each election cycle. This is an ideal measure for testing our hypotheses, since it gives us the opportunity to observe whether DORI scores or increased DORI scores will be more likely to report registration difficulties and be less likely to vote as a result.

Pooling the results from 2008 to 2018, we find that 97.74% of CCES respondents who attempted to vote did so without registration issues, 1.94% reported a registration issue but voted anyway, and 0.32% experienced a registration issue and were unable to vote. Although a small percentage of respondents experienced difficulties, which may or may not have prevented them from voting, the large number of respondents (226,799) over 10 years means we have sufficient power to detect even small effects.¹⁶

We use the CCES data in a pooled mixed-effects multi-level logistic regression with RE for respondents’ state of residence and year of survey administration. The multi-level approach allows us to account for the possibility of correlated errors within states and across election cycles. We estimate two models in which the dependent variable is the binary measure of self-reporting a registration problem (columns 1 and 2 in Table 2) and two models in which the dependent variable is the ordinal measure of not voting given a registration problem (columns 3 and 4 in Table 2). We also re-estimate the ordered logit models using validated CCES voting data (columns 5 and 6 in Table 2). The first model in each establishes a baseline effect for DORI on registration problems and turnout, respectively. The second model in each set interacts whether a respondent is young (18–35 years old; 80,102 respondents or 24.76% of the CCES sample) with DORI.¹⁷ We also include a number of demographic controls from the CCES such as gender and race to capture individuals’ identification with underrepresented groups that have a history of being disenfranchised.

Table 2.

Effect of DORI on Registration Problems and Turnout.

	Y_{J = 1} = (Exp. problem)		Y_{J = 2} = (Not turnout\|problem)
	Self-reported		Self-reported		Validated
Black	0.53*	0.75*	0.52*	0.76*	0.53*	0.76*
Black	(0.07)	(0.15)	(0.07)	(0.15)	(0.07)	(0.15)
Hispanic	0.17*	0.38*	0.19*	0.39*	0.17*	0.38*
Hispanic	(0.09)	(0.19)	(0.08)	(0.19)	(0.09)	(0.19)
Female	0.05	0.19⁺	0.06⁺	0.20⁺	0.05	0.20⁺
Female	(0.03)	(0.12)	(0.03)	(0.12)	(0.03)	(0.12)
Young	0.95*	0.75*	0.98*	0.78*	0.96*	0.77*
Young	(0.03)	(0.08)	(0.03)	(0.08)	(0.03)	(0.08)
DORI	0.32*	0.37*	0.33*	0.38*	0.30*	0.36*
DORI	(0.15)	(0.18)	(0.15)	(0.18)	(0.15)	(0.18)
DORI X Young	–	0.27*	–	0.27*	–	0.26*
DORI X Young		(0.12)		(0.12)		(0.12)
DORI X Black	–	–0.32	–	–0.34	–	–0.31
DORI X Black		(0.24)		(0.24)		(0.24)
DORI X Hispanic	–	–0.31	–	–0.30	–	–0.32
DORI X Hispanic		(0.22)		(0.22)		(0.22)
DORI X Female	–	–0.19	–	–0.18	–	–0.20
DORI X Female		(0.15)		(0.15)		(0.15)
β ₀	–4.42*	–4.46*	–	–	–	–
β ₀		(0.13)
τ ₁	–	–	4.43*	4.47*	4.42*	4.46*
τ ₁			(0.11)	(0.13)	(0.11)	(0.13)
τ ₂	–	–	6.37*	6.41*	5.16*	5.21*
τ ₂			(0.14)	(0.15)	(0.11)	(0.13)
State	0.04*	0.04*	0.04*	0.04*	0.04*	0.04*
State	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
State : Year	0.08*	0.08*	0.08*	0.08*	0.08*	0.08*
State : Year	(0.02)	(0.02)	(0.02)	(0.01)	(0.02)	(0.02)
N	226,799	226,799	223,656	223,656	226,777	226,777

Mixed-effects multi-level logistic (columns 1–2) and ordered logistic (columns 3–6) parameter estimates reported in each cell with standard errors in parentheses. ⁺p < .1, *p < .05.

Results for Impact of DORI on Voting

We find a direct relationship between self-reports of registration problems and state-level DORI scores (B = 0.32, p < 0.05; Table 2, Model 1). This means that would-be voters in states with higher DORI scores are more likely to experience problems registering to vote, all else equal. This is also true of the relationship between DORI and reports of people failing to vote given they experienced a problem (B = 0.33, p < 0.05; Table 2, Model 3). The significant impact of DORI on problems registering holds whether using self-reported (Models 1 and 3) or validated voting data from the CCES (B = 0.30, p < 0.05; Table 2, Model 5).

Consistent with expectations, we also find that young voters are disproportionately impacted by the difficulty of registration. When our young voter indicator (ages 18–35) is interacted with DORI, we obtain consistently positive and statistically significant results across Models 2, 4, and 6. This means that young voters in states with relatively high DORI scores are more likely than other voters to self-report registration problems (B = 0.27, p < 0.05; Table 2, Model 2), and failures to vote given those problems (B = 0.27, p < 0.05; Table 2, Model 1). Like the above effects, this holds when using validated voter data (B = 0.26, p < 0.05; Table 2, Model 6). We do not find similar interactive effects for Black, Hispanic, or women voters. We also note that these effects are robust to the inclusion of voter ID as a control variable. These supplemental analyses can be found in the Online Supplementary Material.

We can get a clearer picture of the meaningfulness of these effects by examining the predicted probability of experiencing a problem registering or in failing to vote for different levels of DORI. These probabilities are listed in Table 3. The likelihood that respondents experience a registration problem increases from about 2% in low DORI states, to 2.6% in high DORI states (Table 3, Row 1). High DORI scores are also associated with significantly more voters being unable to vote after experiencing registration problems. While about 1% of the sample failed to turn out due to registration problems in low DORI states, 1.24% did so in high DORI states (Table 3, Row 3). These effects are most pronounced for young voters. Whereas just under 1.7% of young people failed to vote as a result of registration issues under low DORI conditions, 2.4% did so under high DORI conditions (Table 3, Row 6).

Table 3.

Predicted Effects of DORI on Registration Problems and Turnout.

Y	X ₁	X ₂	Pr(Y)_–2SD	95% CI	Pr(Y)_+2SD	95% CI	ΔPr (Y) (%)
Reg. Prob.	DORI	–	2.01	[1.65, 2.36]	2.60	[2.29, 2.87]	+0.59
Vote^r\|Prob.	DORI	–	0.30	[0.23, 0.37]	0.39	[0.32, 0.45]	+0.09
Vote^v\|Prob.	DORI	–	0.98	[0.80, 1.15]	1.24	[1.10, 1.39]	+0.26
Reg. Prob.	DORI	Age: < 35	3.41	[2.77, 4.05]	4.92	[4.32, 5.53]	+1.51
Vote^r\|Prob.	DORI	Age: < 35	0.52	[0.39, 0.65]	0.77	[0.64, 0.89]	+0.25
Vote^v\|Prob.	DORI	Age: < 35	1.68	[1.36, 2.00]	2.40	[2.09, 2.71]	+0.72
Reg. Prob.	DORI	Age: ≥ 35	1.57	[1.27, 1.86]	1.84	[1.60, 2.07]	+0.27
Vote^r\|Prob.	DORI	Age: ≥ 35	0.23	[0.17, 0.28]	0.27	[0.22, 0.32]	+0.04
Vote^v\|Prob.	DORI	Age: ≥ 35	0.75	[0.61, 0.90]	0.88	[0.76, 0.99]	+0.13

Predicted probabilities derived from FE portion of the MLM models in Table 2 using margins commands in Stata 15. Y is the outcome variable from each model in Table 1. X₁ is the sole or primary independent variable over which the predicted probabilities are calculated. X₂ is the moderator variable; in this case either younger (age < 35) or older (age > 35) respondents. Pr(Y), or the predicted probabilities, are calculated at 2 SDs below (−2SD) and above (+2SD) DORI’s observed sample mean and are presented alongside their 95% confidence intervals. ΔPr(Y) is the change in predicted probabilities observed across values of X₁ or X₁ conditional on X_2. The superscripts r and v denote whether or not voting issues are self-reported (r) or validated (v).

Results Compared to Additive DORI Index

The results above demonstrate that an IRT-estimated DORI yields statistical estimates that match what we would expect to happen to voters in under more difficult registration regimes: they have a harder time getting registered and casting ballots, especially younger voters. Our approach also demonstrates the potential issues associated with the use of additive indices—that is, scales that are summations of different dimensions—to study the effects of registration difficulty on voter behavior. We assess the relative benefits of our IRT approach by creating an alternative, additive indicator of DORI that sums across each policy used to construct the original index.¹⁸

In Appendix B, we present the results of an analysis that underscores the comparative benefits of an IRT approach to estimating DORI. Although scores on the IRT and additive indices are highly correlated with one another (r = 0.89), the two scales present substantive points of disagreement. In Figure B1, bars extending beyond the zero line indicated differences between measures, averaged across state-years. The additive index tends to over-estimate difficulty in states that are least difficult on the IRT-based index. This over-estimation is often in excess of 10 percentage points in states like Hawaii and Delaware. The additive index also tends to over-estimate difficulty in middle-ranked states, like South Carolina and Idaho, also in excess of 10 percentage points. These raw differences translate into inconsistencies in states’ rank-ordering on the additive index compared to the IRT index, as shown in Figure B2. Although the two measures are correlated, additive DORI scores do not rise uniformly as a function of the IRT DORI scores. This is indicative of substantive asymmetries in state-year scoring across procedures. For example, the additive index classifies states like Wisconsin (among the most difficult by the IRT-based DORI) as being less difficult than state like Rhode Island (among the least difficult by the IRT-based DORI), even though Wisconsin’s only innovation over the time period studied was the adoption of OVR, while Rhode Island adopted OVR and AVR—though later than other Democratic-controlled states—and maintained a generous preregistration policy. When these are weighed equally to Wisconsin’s long-standing EDR program, it makes the states look on average more similar to one another by the additive method than by the IRT method, which weights OVR and preregistration as particularly good at distinguishing easy states from more difficult states (see Figure 1).

More problematically, the additive index fails to recover the expected effects on voters and that we detect with the IRT-based DORI. Table B1 presents the same models of voter behavior presented in Table 2 above, but swaps in additive index, rather than the IRT-based index, for DORI. Although the additive version of DORI was positively associated with increased problems registering to vote and failing to turn out to vote among general population and young voters, these effects are generally smaller in magnitude than those detected using the IRT-based index. Correspondingly, the estimates fail to attain conventional levels of two-tailed significance. In other words, the additive approach might have led us to commit a Type II error; rejecting an effect of registration difficulty—well-established by studies of individual registration laws—on voter behavior. Taken together, the results suggest that estimating DORI using the additive approaches more common in past research leads to both descriptive and predictive inconsistencies in studying the effects of registration difficulty on voter behavior.

Discussion and Conclusion

Introducing a novel measure of registration difficulty (DORI), we find that when states make it harder to register to vote people report more registration problems and more failures to cast ballots. This effect is especially pronounced among young voters, who are the most likely to have to navigate the registration process. The size of the effect we find is meaningful, as the half-percentage point reduction in youth voting in high DORI states that we document is significantly greater than that of other voting populations. An effect of the magnitude we detect compounds the underlying tendency for young voters to stay home and can be quite impactful in increasingly competitive states. The 2008 presidential race in North Carolina was decided by just 14,000 votes, while Georgia—which also experimented with increasing and decreasing DORI—was decided by just 11,000 in the 2020 election. In this way, registration difficulty shapes the propensity for voters, especially young voters, to successfully exercise the franchise.

Further, the magnitude of the effects of DORI on turnout could be larger if other conditions among voters were minimized, such as low levels of political interest and the lack of mobilization from political parties and candidates. If these conditions were reduced there would there be more people seeking to register and thus subject to the roadblocks of registration regimes. Boosting turnout among voters requires a multipronged solution in which registration laws are eased and paired with outreach from parties, candidates, and causes, improved civic education, and encouragement from friends, families, teachers, and coaches to get involved in the political process.

To quantify the effect of registration on voters, we used a hybrid IRT model to scale different dimensions of registration regimes and obtain unique DORI scores for each state over time. DORI covers key dimensions of voter registration and DORI captures the relative contribution of these dimensions. DORI is also associated with factors we might expect from the extant research, such as racial and age demographics and party control of state government. The measure helps to isolate and uniquely quantify the overall difficulty of registration in each state in each election cycle since 2004. We are confident that our measure isolates the primary hurdles an individual could face when attempting to register to vote.

But, in creating DORI, we—as any research team must—had to make decisions on which items to scale together. Some of these decisions were driven by data availability, such as the exclusion of detail on states’ OVR systems (i.e., which are user-friendly and which are clunky or difficult to find online). An ordinal measure of OVR quality would be ideal, but these data are not readily available by state and by election cycle. Further, nuance could be added to the automaticity dimension such that states that require voters to opt-out of electronic transfer of information among agencies would obtain the “easiest” score, while states that make voters opt-in to electronic transfer would earn a “medium” score, and states that do not allow electronic transfer would receive a “difficult” score. Other decisions were necessitated by the mathematics of the IRT model; as noted above, we attempted to include EDR location and non-EDR deadline as measures of deadline but there simply is not enough unique variation in these indicators for the IRT method to make distinctions between states. The addition of more years of data with each election cycle will help to increase the statistical power of the model. Other decisions were made to adhere to our goal of isolating the barriers to individuals successfully registering to vote; regulations on organizations conducting voter registration drives and states’ membership in the Electronic Registration Information Center (ERIC)—which could lead to voters receiving communications about updating their registration—were excluded for this reason. Above all else, our paper provides a new method for measuring registration difficulty, and we welcome attempts from others to broaden, customize, and perfect measures DORI. Data and code for this “new tool” is available online at https://osf.io/7fkzp/.

States, after all, have not stopped experimenting with registration laws. For the 2022 cycle, several states have begun eliminating deadlines and allowing online registration. DORI should prove useful for future analyses into why states differ in their registration stringency. For example, future analyses may explore the role of legislative ideology in shaping DORI. As states adopt new registration procedures, these experiments may be picked up by ideologically similar states and begin to diffuse across the country. The most liberal Democratic state governments—such as Oregon Democrats—appear to be the vanguard of this process while Democratic governments in Rhode Island and New York lag behind. We also note that young people were not the only ones who disproportionately faced problems registering to vote. Across all models, Black and Hispanic voters were significantly more likely to report problems and subsequently to fail to vote. We hope DORI can be used to help kickstart new research of the effect of voter registration on historically marginalized voters.

Our findings hold lessons for states that may be considering increasing registration difficulty. Most states sought new and easier paths to registration over the past 16 years and continue to do so, with some notable exceptions. But since the 2020 election, Republican-controlled state legislatures appear poised to adopt new voting restrictions, including some on registration procedures. Making it harder to register to vote, whether in the name of election security or in order to prevent people from turning out, directly impacts voters. Ultimately, what is most troubling about difficult registration laws is that they can disenfranchise Americans who want to vote. Difficult registration regimes risk disenfranchising people who are interested in having their voice heard at the ballot box. Young people are most vulnerable to increased difficulty in voter registration processes. Creating confusion for young people as they seek to vote can harm political equality, a central tenet of American democracy. When new burdens of registration are put in place, they tend to fall disproportionately on some voters, which ultimately distorts who participates. In this way, the electorate shrinks and the quality of democracy in the states suffers.

Supplemental Material

sj-pdf-1-apr-10.1177_1532673X211055050 – Supplemental Material for Finding DORI: Using Item Response Theory to Measure Difficulty of Registration in the U.S. and Its Impact on Voters

Supplemental Material, sj-pdf-1-apr-10.1177_1532673X211055050 for Finding DORI: Using Item Response Theory to Measure Difficulty of Registration in the U.S. and Its Impact on Voters by Joshua M. Jansa, Matthew Motta and Rebekah Herrick in American Politics Research

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Joshua M. Jansa

Matthew Motta

Rebekah Herrick

Supplemental Material

Supplemental material for this article is available online. Replication data is available at .

Notes

Appendix A: Full Results for Correlates of DORI Validity Check.

Table A1.

Two-Way FE Model Predicting DORI Scores, 2004–2020.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Unified GOP govt	0.08*	0.08*	0.08*	0.09*	−0.10*	−0.78*	0.09
Unified GOP govt	(0.02)	(0.02)	(0.02)	(0.02)	(0.05)	(0.35)	(0.13)
Unified Dem govt	−0.04+	0.07+	1.17*	0.14	−0.04+	−0.04+	−0.04+
Unified Dem govt	(0.02)	(0.04)	(0.29)	(0.16)	(0.02)	(0.02)	(0.02)
Minority population	−2.37*	−2.48*	−2.30*	−2.29*	−2.62*	−2.38*	−2.37*
Minority population	(0.55)	(0.55)	(0.54)	(0.56)	(0.55)	(0.55)	(0.56)
Youth population	−0.35	−0.26	0.71	−0.42	0.38	−2.43	−0.35
Youth population	(1.68)	(1.66)	(1.67)	(1.68)	(1.65)	(1.87)	(1.69)
Electoral competition	0.26	0.14	0.21	0.36	0.05	0.21	0.26
Electoral competition	(0.21)	(0.21)	(0.21)	(0.23)	(0.21)	(0.21)	(0.25)
Dem govt * min. pop.		−0.33*
Dem govt * min. pop.		(0.10)
Dem govt * youth pop.			−5.86*
Dem govt * youth pop.			(1.42)
Dem govt * competition				−0.41
Dem govt * competition				(0.37)
GOP govt * min. pop.					0.74*
GOP govt * min. pop.					(0.17)
GOP govt * youth pop.						4.14*
GOP govt * youth pop.						(1.68)
GOP govt * competition							−0.00
GOP govt * competition							(0.31)
Constant	1.47*	1.53*	1.22*	1.42*	1.45*	1.93*	1.47*
Constant	(0.40)	(0.40)	(0.40)	(0.41)	(0.40)	(0.44)	(0.41)
State FE	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES
N	442	442	442	442	442	442	442
Adj. R²	0.777	0.783	0.786	0.777	0.787	0.780	0.777

Standard errors in parentheses; ⁺ p < .1, *p < .05

Appendix B: Comparing “IRT DORI” to “Additive DORI.”

Table B1

Replication of Table 2 Using Additive DORI.

	Y_{J = 1} = (Exp. problem)		Y_{J = 2} = (Not turnout\|Problem)
	Self-reported		Self-reported		Validated
Black	0.52*	0.67*	0.52*	0.68*	0.53*	0.67*
Black	(0.07)	(0.17)	(0.07)	(0.18)	(0.07)	(0.17)
Hispanic	0.17+	0.43*	0.19*	0.44*	0.17*	0.43*
Hispanic	(0.09)	(0.12)	(0.08)	(0.12)	(0.09)	(0.12)
Female	0.05	0.10	0.06+	0.09	0.05	0.11
Female	(0.03)	(0.12)	(0.03)	(0.12)	(0.03)	(0.12)
Young	0.95*	0.81*	0.98*	0.83*	0.96*	0.83*
Young	(0.03)	(0.10)	(0.03)	(0.10)	(0.03)	(0.10)
DORI—Additive Version	0.21	0.22	0.23	0.22	0.19	0.21*
DORI—Additive Version	(0.15)	(0.17)	(0.15)	(0.18)	(0.15)	(0.17)
DORI X Young	–	0.18	–	0.20*	–	0.17
DORI X Young		(0.15)		(0.15)		(0.15)
DORI X Black	–	–0.18	–	–0.20	–	–0.18
DORI X Black		(0.24)		(0.25)		(0.24)
DORI X Hispanic	–	–0.37*	–	–0.35*	–	–0.37*
DORI X Hispanic		(0.15)		(0.14)		(0.15)
DORI X Female	–	–0.06	–	–0.04	–	–0.07
DORI X Female		(0.13)		(0.13)		(0.13)
β ₀	–4.35*	–4.36*	–	–	–	–
β ₀	(0.12)	(0.13)
State	0.04*	0.04*	0.04*	0.04*	0.04*	0.04*
State	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
State : Year	0.08*	0.08*	0.08*	0.08*	0.08*	0.08*
State : Year	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
τ1	–	–	4.36*	4.35*	4.34*	4.35*
τ1			(0.12)	(0.14)	(0.12)	(0.13)
τ ₂	–	–	6.31*	6.30*	5.08*	5.10*
τ ₂			(0.14)	(0.14)	(0.11)	(0.13)
N	226,799	226,799	223,656	223,656	226,777	226,777

References

Ajzen

(1985). From intentions to actions: a theory of planned behavior. In Action control (pp. 11–39). Berlin: Springer.

Ajzen

Fishbein

(2005). The influence of attitudes on behavior. The handbook of attitudes. Editors. Albarracín

Johnson

B. T.

ZannaMahwah

M. P.

, NJ: Lawrence Erlbaum Associates, pp.173–221.

Ansolabehere

Hersh

Shepsle

. (2012). Movers, stayers, and registration: Why age is correlated with registration in the U.S. Quarterly Journal of Political Science. 7:1–31.

Bentele

K.G.

O’Brien

E.E.

(2013). Jim Crow 2.0? Why states consider and adopt restrictive voter access policies. Perspectives on Politics 11(4), 1088–1116.

Berry

W. D.

Fording

R. C.

Ringquist

E. J.

Hanson

R. L.

Klarner

C. E.

(2010). Measuring citizen and government ideology in the US states: a re-appraisal. State Politics & Policy Quarterly, 10(2), 117–135.

Biggers

D.R.

Hanmer

M.J.

(2015). Who makes voting convenient? Explaining the adoption of early and no-excuse absentee voting in the American states. State Politics and Policy Quarterly 15(2): 192–210.

Biggers

D.R.

Hanmer

M.J.

(2017). Understanding the adoption of voter identification laws in the american states. American Politics Research, 45(4), 560–588.

Brennan Center for Justice . (2017). VRM in the states: Portability. https://www.brennancenter.org/our-work/research-reports/vrm-states-portability

Burden

B.C.

Canon

D.T.

Mayer

K.R.

Moynihan

D.P.

(2014). Election laws, mobilization, and turnout: the unanticipated consequences of election reform. American Journal of Political Science 58(1):95–109.

10.

Carmines

E.G.

Zeller

R.A.

(1979). Reliability and validity assessment. Beverly Hills, CA: Sage Publishers

11.

Coppock

Green

D.P.

(2016). Is voting habit forming? New evidence from experiments and regression discontinuities. American Journal of Political Science 60, 4: 1044–1062.

12.

Downs

(1957). An economic theory of democracy. New York: Harper.

13.

Fowler

(2017). Does voter preregistration increase youth participation? Election Law Journal: Rules, Politics, and Policy 16(4):121. DOI: 10.1089/elj.2017.0436

14.

Fraga

Holbein

(2020). Measuring youth and college student voter turnout. Electoral Studies, 65, 102086.

15.

Grumbach

J.M.

(2021). Laboratories of democratic backsliding. Working paper.

16.

Grumbach

J.M.

Hill

(2020). Rock the registration: same-day registration increases turnout of young voters. Journal of Politics, 13, 164. DOI: 10.1086/714776

17.

Hansen

Jensen

J.M.

(2007). Understanding voters’ decisions: a theory of planned behavior approach. Innovative Marketing, 3(4), 87–94.

18.

Harmon

Lauren

Posner

Charles

Jawondo

Michele

Dhaiti

Matt

(2015). The Health of State Democracies. Center for American Progress Action Fund.

19.

Herrick

Pryor

(2021). Forthcoming. Gender and race gaps in voting and over-reporting: an intersectional comparison of CCES with ANES data. Social Science Journal, 12, 147. DOI: 10.1080/03623319.2020.1809901

20.

Hicks

W.D.

McKee

S.C.

Sellers

M.D.

Smith

D.A.

(2015). A principle or a strategy? Voter identification laws and partisan competition in the American states. Political Research Quarterly, 68(1), 18–33.

21.

Hicks

W.D.

McKee

S.C.

Smith

D.A.

(2016). A bipartisan election reform? Explaining support for online voter registration in the American states. American Politics Research 44(6), 1008–1036.

22.

Highton

(1997). Easy registration and voter turnout. The Journal of Politics 59(2): 565–575.

23.

Highton

(2000). Residential mobility, community mobility, and electoral participation. Political Behavior 22(2), 109–120.

24.

Hill

K.Q.

Leighley

(1999). Racial diversity, voter turnout, and mobilizing institutions in the united states. American Politics Quarterly 27(3), 275–295.

25.

Holbein

J. B.

Hillygus

D.S.

(2016). Making young voters: The impact of preregistration on youth turnout. American Journal of Political Science 60(2), 364–382.

26.

Juelich

C.L.

Coll

J.A.

(2020). Rock the vote or block the vote? How the cost of voting affects the voting behavior of American youth. American Politics Research 48 (6), 719–724.

27.

Kaiser Family Foundation . (2020a). Population Distribution by Race/Ethnicity. https://www.kff.org/other/state-indicator/distribution-by-raceethnicity/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D

28.

Kaiser Family Foundation . (2020b). Population Distribution by Age. https://www.kff.org/other/state-indicator/distribution-by-age/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D

29.

King

J.D.

(1994). Political culture, registration laws, and voter turnout among the American states. Publius: The Journal of Federalism 24(4), 115–127.

30.

Knack

White

(2000). Election-day registration and turnout inequality. Political Behavior, 22(1), 29–44.

31.

Larocca

Klemanski

J.S.

(2011). US state election reform and turnout in presidential elections. State Politics & Policy Quarterly 11(1), 76–101.

32.

Lawless

J.L.

Fox

(2015). Running from Office: Why young Americans are turned off to politics. New York: Oxford University Press.

33.

Leighley

Nagler

(2014). Who votes Now? Demographics, issues, Inequality, and turnout in the United States. Princeton: Princeton University Press.

34.

Levitsky

Ziblatt

(2018). How Democracies Die. New York: Broadway Books.

35.

Pomante II

Schrafnagel

(2018). Cost of Voting in the American states. Election Law Journal 17(3), 234–247.

36.

Mann

C.B.

Gronke

Adona

(2020). Framing automatic voter registration: partisanship and public understanding of automatic voter registration. American Politics Research 48(6), 693–699.

37.

McDonald

M.P.

(2008). Portable voter registration. Political Behavior 30 (4), 491–501.

38.

McGhee

Romero

(2020). Registration effects of automatic voter registration in the United States. Unpublished manuscript.

39.

NCSL . (2019a). Automatic voter registration. National Conference of State Legislatures. https://www.ncsl.org/research/elections-and-campaigns/automatic-voter-registration.aspx

40.

NCSL . (2019b). Online voter registration. National Conference of State Legislatures. https://www.ncsl.org/research/elections-and-campaigns/electronic-or-online-voter-registration.aspx

41.

NCSL . (2019c). Preregistration for young voters. National Conference of State Legislatures. https://www.ncsl.org/research/elections-and-campaigns/preregistration-for-young-voters.aspx

42.

NCSL . (2019d). Voter registration deadlines. National Conference of State Legislatures. https://www.ncsl.org/research/elections-and-campaigns/voter-registration-deadlines.aspx

43.

NCSL . (2020). State partisan composition. National Conference of State Legislatures. https://www.ncsl.org/research/about-state-legislatures/partisan-composition.aspx

44.

Netemeyer

R. G.

Burton

(1990). Examining the relationships between voting behavior, intention, perceived behavioral control, and expectation. Journal of Applied Social Psychology, 20(8), 661–680.

45.

Plutzer

(2002). Becoming a habitual voter: Interia, resources, and growth in young adulthood. American Political Science Review 96(1), 41–56.

46.

Riker

W.H.

Ordeshook

P.C.

(1968). A theory of the Calculus of voting. American Political Science Review 62(1), 25–42. DOI: 10.2307/1953324

47.

Rocha

R.R.

Matsubayashi

(2014). The politics of race and voter ID laws in the states: the return of Jim Crow? Political Research Quarterly 67(3), 666–679.

48.

Springer

M.J.

(2012). State electoral institutions and voter turnout in presidential elections, 1920–2000. State Politics & Policy Quarterly 12(3), 252–283.

49.

Stockemer

Rocher

(2017) Age, political knowledge and electoral turnout, a case study of Canada. Commonwealth & Comparative Politics 55(1), 41–62.

50.

Timpone

R.J.

(1998). Structure, behavior and voter turnout in the united states. American Political Science Review 92(11), 145–158.

51.

U.S. Census Bureau . (2015). https://www.census.gov/content/dam/Census/library/publications/2015/acs/acs-31.pdf

52.

Vonnahme

(2012). Registration deadlines and turnout in context. Political Behavior 34(4), 765–779.

53.

Wolfinger

Rosenstone

(1980). Who votes? New Haven: Yale University Press.

54.

(2019). Does state online voter registration increase voter turnout? Social Science Quarterly 100(3), 620–634.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.37 MB