Using Medicaid Automatic Voter Registration to Address Persistent Voter Registration Problems: Helping the National Voter Registration Act Achieve its Potential

June 22, 2023
Douglas R. Hess, PhD

Executive Summary

Millions of U.S. citizens are affected each election cycle by difficulties registering to vote and problems created by the maintenance of voter registration records. Many of these problems disproportionately affect young, non-white, and low-income citizens, as well as those with disabilities or who have recently moved.

This report reviews the scope and underlying causes of several persistent voter registration problems in the U.S. and why automatic voter registration (AVR) for Medicaid is a promising innovation in election administration policy that states can introduce to address these problems.

Thirty years ago, Congress extensively reformed voter registration policy by passing the National Voter Registration Act (NVRA) of 1993. Among the Act’s many innovations is the requirement in Section 7 of the NVRA that states actively offer voter registration at agencies serving people with low incomes or disabilities, such as Medicaid, SNAP (“food stamps”), and WIC.

These programs are well-positioned to produce millions of new voter registration applications and, equally important, updates to existing registration records each election cycle:

  • Each year since 2010, from 30 million to more than 40 million adult citizens — or 16 to 19 percent of adult citizens — participated in one or more of the four largest programs covered by Section 7.
  • Voter registration rates among the populations served by these programs are up to twenty points lower than that of wealthier citizens.
  • Moreover, there is a substantial need for agency clients to update their voter registrations. Adult citizens in these programs change their address frequently, even more often than other adult citizens.

Unfortunately, far more often than not, the results from these agencies have been disappointing:

  • When the law is implemented well, roughly 10 to 15 percent of all voter registration applications in a state have come from social services agencies.
  • However, this percentage has been closer to three percent for the average state.
  • Moreover, despite dramatic increases in the size of Section 7-covered programs and the clear need for new and updated registrations, the total number of voter registration applications from these agencies has never surpassed the number submitted in the first election cycle under the NVRA (1995 to 1996).

Underlying these disappointing results are elements in the design of Section 7 voter registration programs, which have two important effects:

  • First, compared to the design in Section 5 for registration programs at Departments of Motor Vehicles (“motor voter” programs), the design of Section 7 decreases the likelihood that agency visitors will accept the offer to register to vote.
  • Second, Section 7’s design increases the likelihood that states will fail to comply with the law.
  • Indeed, since 2000, more than half the states covered by the NVRA have signed agreements with civil rights advocates or the Department of Justice to settle or avoid litigation related to state noncompliance with this portion of the Act.

Without significant changes in state practices, millions of citizens each year will not benefit from their rights under the NVRA:

  • As a result, voter registration rolls will continue to lack both coverage and accuracy.
  • Moreover, states will remain exposed to expensive enforcement litigation by civil rights organizations or the Department of Justice.

In light of these serious concerns, state election officials are exploring more robust systems for managing voter registration services in Section 7-covered agencies, including:

  • Building on successful AVR policies for motor voter programs, which more than twenty states have adopted, several states have passed legislation for AVR in Medicaid.
  • Rather than making individuals provide the same information on two separate forms – one for the agency and one for voter registration – AVR leverages Medicaid’s existing procedures to seamlessly register adult citizens.
  • Significantly, optimal Medicaid AVR policies will register all eligible household members included on agency applications, not just those conducting a transaction covered by the NVRA.
  • Applying optimal AVR practices to Medicaid programs offers a promising way to substantially increase Section 7 performance and improve state compliance with the NVRA in a program that reaches tens of millions of adult citizens.

Based on the size of Medicaid, the characteristics of the population Medicaid serves, and results from states that adopted best practices in motor voter AVR, Medicaid AVR, if adopted nationally, could generate millions of new and updated registrations each election cycle.


Voter registration is a citizen’s ticket to an election: No ticket, no entry.

Unfortunately, problems surrounding how these “tickets” are obtained and maintained affect millions of citizens in the United States each election cycle. Many of these problems disproportionately affect young, non-white, and low-income citizens, as well as those with disabilities or who have recently moved.1 The potential consequences of these issues are grave. Persistent and avoidable inequalities in access to voter registration undermine the norms of democracy. Moreover, citizens may lose confidence in elections if they encounter, or hear of others encountering, obstacles to the ballot box related to the accuracy of registration lists.

Congress extensively reformed voter registration policy with passage of the National Voter Registration Act (NVRA) in 1993.2 Among the Act’s many policy enhancements is the requirement in Section 7 that states offer voter registration opportunities when members of the public interact with government programs primarily serving people with low incomes and disabilities, such as Medicaid.3 A key goal of Section 7 of the NVRA is to reach citizens who are less likely to drive and, thus, may not benefit from “motor voter,” the requirement in Section 5 that states incorporate voter registration into driver’s license and identification card procedures.

Because tens of millions of adult citizens with low registration rates and high residential mobility rates participate in programs covered by Section 7, these programs should produce a substantial number of registration applications from groups most in need of voter registration services. However, more often than not, the results from states’ Section 7 registration programs have been disappointing.

Overall, Section 7 programs have yielded impressive results in some states, but rarely consistently and usually only after advocates or the Department of Justice (DOJ) intervened to enforce the NVRA. In fact, the total number of voter registration applications (VRAs) from social services agencies nationwide has yet to equal the number produced in the first election cycle after states implemented the Act (1995 to 1996) despite dramatic increases in the number of adult citizens served by the four largest programs covered by Section 7 and the continued need for registration opportunities.

Underlying the deficient performance of agency registration programs are elements of its design. Compromises Congress made while drafting the NVRA weakened Section 7 procedures compared to the requirements placed on motor voter procedures, enshrined in Section 5 of the Act (Piven and Cloward 2000). These differences in Section 5 and 7 procedures, it turns out, reduce the likelihood that citizens will register to vote at Section 7 agencies and increase the likelihood that agencies or agency sites will introduce practices that do not comply with the law.

Moreover, the NVRA was designed for the office technology and procedures of the last century. During the last twenty years there have been significant changes in how citizens and governments interact, including changes related to the “e-government revolution.” However, these technological advances have rarely been met with compatible innovations in NVRA implementation, further exacerbating compliance problems.

In its final report, the Presidential Commission on Election Administration (2014) referred to the NVRA as “the election statute most often ignored” (page 17). As evidence of the extent of noncompliance with Section 7 in particular, 23 states — or more than half the states covered by the Act — have, since 2000, faced court orders to comply with the law or signed comprehensive compliance agreements to settle or avoid enforcement litigation. Unfortunately, there is extensive evidence that many states still violate Section 7, particularly in their online applications for covered programs.

Without significant changes in how states incorporate Section 7 requirements into agency practices, millions of citizens each year will not benefit from their rights under the NVRA. As a result, voter registration rolls will continue to lack both coverage and accuracy. Moreover, states will remain exposed to expensive enforcement litigation by civil rights organizations or the DOJ.

In light of these serious concerns, states are exploring more effective and robust systems for managing voter registration services in social service agencies. Building on automatic voter registration (AVR) policies for motor voter programs, which more than twenty states have adopted, several states have passed or are considering legislation for AVR in Medicaid.

By addressing deficiencies in the design of Section 7, applying optimal AVR practices to Medicaid offers a promising way to both increase Section 7 performance and improve state noncompliance with the NVRA in a program of serving tens of millions of adult citizens. Indeed, based on the size of Medicaid, the characteristics of the population Medicaid serves, and results from states that adopted best practices in motor voter AVR, Medicaid AVR could register millions of citizens each cycle and keep their registrations up to date.

This report reviews the scope and underlying causes of recurrent voter registration problems in the U.S., why Section 7’s potential to address these problems is not being met, and why Medicaid AVR is a promising innovation.

Persistent Voter Registration Problems in the U.S.

Numerous problems, affecting millions of citizens, trouble the building and maintaining of voter registration lists in the U.S. Below I briefly review four broad categories of issues needing attention.

First, millions of eligible citizens are not registered, creating substantial disparities in voter registration rates between various groups. Registration rates for non-white, Latino, and younger citizens, as well as those with low incomes, a disability, or who have recently moved, are far lower than that of their fellow citizens (see Table 1, all tables and figures are located starting on page 17). Second, millions of citizens have difficulties registering to vote due to language barriers, a disability, not knowing where or how to register, or because they did not meet the deadline in their state. In 2020, these four barriers affected an estimated five million citizens (see Table 2), roughly one in five unregistered citizens that year.

Third, an alarming number of potential voters are turned away from the ballot box each election cycle due to problems with their registration. The U.S. Census estimates that more than six hundred thousand citizens could not vote in the 2020 November election due to a problem with their registration (Fabina and Scherer 2022). Similarly, surveys conducted by the Massachusetts Institute of Technology’s Election Data & Science Lab have found that two to three percent of in-person Election Day voters encounter registration-related problems that prevent them from casting a ballot (Stewart 2021).

Finally, there are perennial concerns about deadwood — inaccurate records — on the registration rolls. Although deadwood, which largely consists of outdated registration records for otherwise eligible voters, does not threaten election integrity in the manner and certainly not to the degree expressed by some alarmists, overzealous initiatives to reduce deadwood may explain why some citizens are surprised to find that they cannot vote because they are no longer on the registration rolls.

Why do Voter Registration Problems Persist?

Voter registration rates have been rising, albeit slowly, this century. These increases are partly due to reforms instituted by the NVRA, additional reforms adopted by many states, and the tens of millions of dollars spent by political parties, civic organizations, and some states to register voters each election cycle.

Nonetheless, as shown in the prior section, substantial, persistent, and deeply concerning problems with voter registration remain, negatively affecting millions of citizens. Why?

First, the problems identified in the previous section are inherent to the peculiar voter registration system used in the U.S. In most wealthy democracies, governments primarily create and revise voter registration lists by relying on information generated as citizens interact with their governments (Massicotte, Blais, and Yoshinaka 2004). However, in the US, citizens are responsible for both submitting initial voter registration applications and updating their registration records. It is well known that this policy creates barriers to registration that have a greater negative impact on some citizens than others (Leighley and Nagler 2013; Piven and Cloward 2000). For example, this system may hinder registration by citizens with disabilities or facing language or informational barriers.

Particularly disadvantaged by the U.S. registration system are populations that move frequently. Thus, a second cause of the policy problems under discussion is high residential mobility in the U.S. Americans change addresses with remarkable frequency (see Table 3). However, some groups of citizens move more often than others, creating differences in who has an up-to-date registration. For instance, compared to wealthier citizens, citizens with low incomes are twice as likely to have been at their address for less than one year and more than 1.5 times as likely to have been at their address for less than three years (34 percent vs. 20 percent). Table 3 also indicates marked differences in time at an address across age groups. The impact of these demographic differences in residential mobility on registration rates is apparent in Table 4.

Residential mobility also explains why some citizens who thought they were registered cannot vote. Compared to citizens at their address for five or more years, those at their address for less than one year are more than six times as likely to report that a registration problem prevented them from voting. Those who have been at their address for one to two years are more than three times as likely to report such issues (see Table 5).

Another cause of disparities in registration is that the opportunities to register or update a registration are not equally accessible to all citizens. For instance, even though the use of online registration is increasing, citizens with low incomes and disabilities are much less likely to use this option. Likewise, although the most common method for registering to vote is through a Department of Motor Vehicles (DMV), this method does not reach citizens with low incomes or a disability at the same rate as other citizens (see Table 6).4

Finally, many voter registration problems persist because, as detailed later in this report, Section 7 of the NVRA has yet to achieve its full potential. By improving registration programs with the potential to reach millions of citizens who are less likely to be registered and more likely to be residentially mobile, such as those created by Section 7 of the NVRA, states can address disparities in registration rates and problems associated with inaccurate registration rolls.

The following section reviews the requirements of Section 7 of the NVRA and the reasoning behind its inclusion in the Act. The section after that discusses why the NVRA has not lived up to its potential. The report then ends with a discussion of Medicaid AVR and the logic suggesting that this innovation could dramatically improve both the performance of the NVRA and state compliance with this important federal voting rights law.

What is Section 7 of the NVRA & Why does it Matter?

The NVRA is a complex law. Its explicit goals include reversing the harms of unfair and archaic state registration laws, increasing the number of citizens registered, and ensuring that registration rolls are up to date.

To meet these goals, the NVRA significantly reformed U.S. voter registration policy by, among other things, banning registration deadlines of more than 30 days before federal elections, requiring acceptance of mail-in voter registration forms, protecting infrequent voters against hasty removal from the rolls, and expanding motor voter programs to all covered states.5

As Congress drafted the NVRA, election officials and advocates raised concerns about differences in who has access to various voter registration methods (Groarke 2000; Piven and Cloward 2000). Specifically, proponents of the NVRA worried that if motor voter was the only federally mandated program to proactively offer registration services, the Act could, perversely, increase inequalities by only easing registration for groups with higher-than-average registration rates.6

To address this possibility, Congress included in Section 7 the requirement that states offer voter registration services at agencies administering programs serving individuals with disabilities and low incomes. The reason for including these agencies was to reach citizens less likely to drive, and thus less likely to visit a DMV.

The agency registration programs mandated by Section 7 are well-targeted for meeting the NVRA’s goals. Citizens interacting with social services agencies are registered to vote at a rate significantly below the average (see Table 1). Moreover, on average, citizens with low incomes move more frequently than the rest of the population (Table 3). Data on driver’s licenses and participation in Section 7 programs are not readily available. However, the 2020 Survey of Income and Program Participation indicates that adult citizens under 65 who participate in Medicaid, SNAP, or TANF are more than twice as likely to reside in a household without a vehicle owner as their peers who do not participate in these programs (26 percent vs. 12 percent).

Based on these characteristics of participants and the sheer size of these programs — with the four largest programs involving from 30 million to more than 40 million adult citizens (16 to 19 percent) annually since 2010 — we would expect Section 7-covered agencies to generate, nationwide, several million new registrations and updates to registration records each election cycle. During the 1996 election cycle, the first covered by the NVRA, states received more than 2.6 million VRAs from social service agencies. This amounted to 6.3 percent of VRAs received from all sources that cycle and about 6.6 per 100 proxy encounters between adult citizens and these agencies.7

However, performance rapidly declined during the following election cycles, and the national total of VRAs from these agencies has never matched that during the 1996 cycle (see Table 7 and Figure 1).8 This decline in output occurred despite steady increases in the U.S. adult citizen population, dramatic increases in the size of the relevant government programs, and the considerable number of program participants needing to register or update their registration.9 In recent election cycles, the percentage of all VRAs received that are from social services agencies has been a third to a half of what it once was. Moreover, Section 7 performance has varied dramatically among the states and over time for most states, even when controlling for the size of the agency programs (see Table 8).

Why the NVRA’s Potential Remains Untapped

Why has Section 7 performed far below its potential? Below I explore two key reasons. First, compromises Congress made when designing Section 7 made both agency compliance with the law less likely and the offer of registration less effective compared to Section 5 (motor voter). Second, Section 7 was written for 20th-Century government practices. Thus, Section 7 performance has not fared well as agencies moved from in-person interactions with the public to “remote transactions” — that is, transactions by mail, over the phone, or via online agency portals.

Section 7 Design and Poor Performance

The first key issue behind Section 7’s performance woes is the design of the voter registration offer. Ostensibly, implementing Section 7 is straightforward.10 When people initially apply for benefits, recertify their program eligibility, and file changes of address, state social services agencies are to provide VRAs, assist clients with completing the form, and transmit in a timely manner completed VRAs to election officials.

The NVRA created new tasks for agency employees already saddled with significant amounts of paperwork, a likely starting point for implementation problems. Agency clients, of course, are also burdened with agency red tape and requirements to document their incomes, assets, and other information. In the language of cognitive psychology and behavioral economics, the success of the voter registration service envisioned by Section 7 runs into “hassle factors” affecting both clients and agency employees.

Hassle factors are “seemingly trivial tasks like waiting in line, filling out a form, or mailing an envelope [that] can have outsized effects” on decisions or behaviors, and “eliminating these ‘hassle factors’ — for example, by waiving the need for a required form — can have a disproportionate effect compared with their cost” (Richburg-Hayes et al. 2014). In short, during a process that is focused on documenting income, assets, employment, or household needs, the additional paperwork and decisions associated with voter registration risk getting ignored.

Motor voter procedures in Section 5 of the NVRA reduce hassle factors for employees and the public by requiring DMVs to apply relevant information from a driver’s license application directly to a voter registration application without asking clients for redundant information. In short, under Section 5, citizens cannot be asked a question they have already answered. Importantly, this requirement not only reduces the effort required to register but also increases the likelihood that VRAs are provided to the public because VRAs need to be integrated with DMV paperwork to comply with Section 5.

Before the development of electronic interfaces, states complied with Section 5 by using non-carbon paper to streamline the process. As clients printed their name, birth date, etc. On DMV paperwork, any information also necessary to register to vote was transferred to a detachable VRA. As electronic systems were developed, the prohibition in Section 5 against collecting redundant information from clients required states to develop software appropriate to this task. In brief, as long as motor vehicle offices used forms or software compliant with the Act, the costs of compliance by agency employees and the paperwork burden placed on the public were both significantly reduced.

Unfortunately, Section 7 does not require procedural streamlining similar to Section 5, likely reducing the chance that citizens will register to vote. Another possible result of not streamlining Section 7 voter registration procedures is that many states have not included VRAs in agency paperwork. Instead, many Section 7 agencies adopted the policy to only provide VRAs if clients request one, a clear violation of the Act. Eventually, some agency sites stopped complying with the Act entirely. Field investigations found many agency offices without any VRAs (Herman 2008), making registration impossible at these locations.11

Section 7 Challenges in the 21st Century

One reason performance declined over time is the change in how social services agencies interact with the public in the 21st Century. Written in the early 1990s, Section 7 was designed in an era of paper applications and in-person office visits. However, the last two decades have witnessed a transformation in how citizens and agencies interact, including developments stemming from the “e-government” revolution. In brief, health and human services agencies have increasingly encouraged or required clients to interact with agencies, even when first applying for benefits, by “remote transaction” — that is, clients conduct their interactions with agencies by mail, online, or over the phone.

Unfortunately for the NVRA, remote transactions mean that it is less likely that citizens will receive a voter registration application or assistance with it. For instance, a review of states’ online SNAP applications found widespread deficiencies in complying with the NVRA (Ashbrook, Brannon, and Hess 2017). Lawsuits enforcing Section 7 — and related court orders or agreements — frequently mention or correct state noncompliance during remote transactions. Monitoring data provided to plaintiffs per settlement agreements indicate that remote transactions are less likely to result in a registration compared to compliant in-office transactions. (Properly designed, AVR policies should eliminate this distinction in performance.)

The Success and Limits of Enforcement Litigation

Information from NVRA enforcement interventions — such as letters from plaintiffs informing states of noncompliance, legal complaints, depositions, documents and data collected during the discovery process, and court findings — have documented extensive and widespread noncompliance. In addition, the effects of Section 7 implementation agreements on agency registration performance indicate both Section 7’s potential and the harm caused by state noncompliance.

Section 7 agreements typically last between two and four years. When these agreements are in place in a particular state — and, in some cases, while these agreements are negotiated — the output of Section 7 agencies increases dramatically. Table 8 presents this dynamic for a handful of states covered by agreements by dividing the number of VRAs in an election cycle by a denominator serving as a proxy for program size that cycle for that state. Ratios for election cycles wholly or substantially covered by a Section 7 compliance agreement are represented in bold in Table 8. Cycles during negotiation or litigation prior to the signing of agreements are underlined. In these states, there is a several-fold increase in this ratio during agreements compared to prior periods.

Here are four examples of the number of VRAs that social services agencies can produce and the percentage of all VRAs that were from these agencies.

  • After Alabama signed an agreement in 2014, social service voter registration applications jumped from around 5,000 for the 2012 election cycle to nearly 70,000 for 2014 and 120,000 for 2016, accounting for 16 and 10 percent of all applications for those cycles.
  • Missouri agencies filed less than 18,000 registrations in both 2004 and 2006. However, after a court ordered the state to comply in 2008, social service agencies produced about 220,000 registrations over the next four years (approximately 10 percent of all registrations).
  • In Ohio, agency registrations went from nearly 40,000 in the 2004 election cycle to an average of more than 250,000 per cycle after the state settled Section 7 litigation in 2009, or roughly 10 percent of all applications.
  • In Tennessee, following the enforcement of the Act by the Justice Department in 2002, total registrations from these agencies ranged from about 120,000 to 174,000 for the subsequent four cycles, or 16 percent of all registration applications received in 2004.

However, declines in Section 7 performance after agreements expire is common (see Table 8).12 Table 9 presents the combined performance ratio for all 22 states that had agreements in place between 2002 and 2020. Although Table 9 does not account for all factors that may explain performance, it does account for variation in program size and the large declines in performance after agreements end are readily apparent.

As important as enforcement initiatives are, compliance agreements are temporary, may not cover all Section 7 agencies, and require enormous effort from advocates or the DOJ, neither of which have the resources to police all states constantly. In short, enforcement work suggests that improved performance is possible and points to the need for reforms that will make improved performance consistent.

Improving NVRA Performance with Automatic Voter Registration

For states to reduce the registration problems identified earlier in this report, governments will need to implement reforms that address one or more of the following general criteria.

  • First, registration reforms should target sizable populations with lower-than-average registration rates.
  • Second, to increase the accuracy of voter registration lists, as well as their coverage of the eligible population, reforms should target sizable populations with high rates of residential mobility.
  • Third, reforms should reduce hassle factors — i.e., red tape and other burdens, including seemingly minor ones — placed on citizens when registering or updating a registration.
  • Finally, reforms should be built into citizen-government interactions in a manner that improves agency compliance and allows for compliance monitoring.

Medicaid AVR — providing election officials with information from Medicaid clients verified to be eligible to vote based on age and citizenship — addresses these criteria extremely well.13

First, Medicaid reaches a large population with a low registration rate. In recent years, more than 25 to 30 million adult citizens annually participated in Medicaid and related health services. While there are no national surveys on the voter registration rate for Medicaid participants, from the CPS we know that the registration rate for citizens in households participating in SNAP is not only below average generally but also below that of other low-income citizens (see Table 1). Due to significant overlap among SNAP and Medicaid populations, we can assume the registration rate for Medicaid participants is also much lower than average. Second, Medicaid participants are highly mobile. According to the ASEC, between 2010 and 2019, adult citizens participating in Medicaid were 40 percent more likely than non-participants to have changed their address in the last year (14 percent versus 10 percent).

Regarding the last two criteria, there is reason to believe that Medicaid AVR can have a dramatic impact both on client use of the opportunity to register and on agency compliance.

To date, discussions about and research on AVR have focused on its impact on citizens’ behavior — that is, when optimally designed, AVR can increase the likelihood that citizens interacting with a DMV will register or update their registration (Grimmer and Rodden 2022; McGhee and Romero 2021). For instance, Table 10 shows the marked increase in VRAs from DMVs following introduction of AVR in select states. (Not all states have shown such dramatic increases after introducing AVR perhaps because the policy is new, pandemic-related “lockdowns” altered registration patterns in 2020, or some states selected a less than optimal form of AVR.)

What does this mean for Medicaid AVR? Recall that Section 7, as written, is already “less automatic” than Section 5, and that Medicaid is well targeted to groups with low registration rates and high mobility rates. Thus, making the completion of voter registration automatic in Medicaid should have a substantial impact. As with DMV AVR and Section 7 enforcement agreements, I would not be surprised if the number registrations from Medicaid increases several-fold under well-designed AVR policies.

Moreover, AVR can address a notable gap in how Section 7 ordinarily works. Without automatic registration, Section 7 agencies only offer one voter registration application and registration services are directed to the person completing an application, renewal paperwork, or reporting an address change; other householders are left out of the opportunity. However, if the appropriate information-sharing processes are built into Medicaid AVR, any eligible member of the household in the agency system can be registered or have their registration updated. Thus, AVR avoids leaving out other eligible household members, such as spouses, partners, elderly parents, or voting-age children (see appendix on What is Medicaid AVR?).

However, discussions about AVR and its design should also consider how AVR policies can improve compliance. For instance, AVR policies effectively “build” VRAs into agency procedures. By eliminating the need for agency staff to ask clients additional questions or help clients complete an additional form, AVR reduces hassle factors for agency employees as well as clients. Mundane tasks associated with managing paperwork from a separate government agency — stocking forms produced outside the department, training employees on the use of another agency’s forms, or transmitting forms to another agency — have been significant barriers to compliance with the NVRA for both DMVs and social services agencies. To the degree AVR eliminates these tasks, it should increase agency compliance.

For instance, even with the advantages built into Section 5, many motor voter programs have suffered from compliance problems and faced enforcement litigation or the threat of it (Naifeh 2015, 2017). Due to its seamless design, AVR eases compliance by eliminating the chance that agency staff skip the voter registration process.

Finally, both election officials and agency directors should consider how Medicaid AVR benefits performance monitoring, which is central to ensuring compliance. (Mellor and Hess 2009).14 In brief, AVR programs are well situated to improve monitoring by automatically collecting data on the registration process and various Medicaid transactions. If this data is used to ensure that Medicaid AVR is functioning properly, AVR would help protect citizens’ rights under the NVRA and reduce the exposure of states to NVRA-enforcement litigation.


The U.S. voter registration system produces persistent inequalities in who is registered and problems with accuracy in the registration lists that can reduce turnout. Section 7 of the NVRA has design flaws that significantly reduce its effectiveness in addressing these problems. These flaws both reduce registration by the public and increase the likelihood of agency noncompliance.

As a result, Section 7 has not remotely achieved its potential. Enforcement efforts prove this as they dramatically increase the number of VRAs and the percentage of the total VRAs from these agencies. However, as important as enforcement initiatives are, they are not a substitute for additional reforms. Indeed, enforcement work points to the need for reform.

Medicaid AVR can be designed to significantly improve Section 7 performance. Based on program logic, the size of Medicaid, characteristics of Medicaid participants, and evidence from best practices in DMV AVR, Medicaid AVR should substantially increase Section 7 performance. Moreover, states should be motivated to adopt AVR reforms that improve compliance, reducing the threat of more litigation.



What is Medicaid AVR? 

When the public interacts with government agencies, AVR policies streamline the collecting, verifying, and sharing of information necessary for creating and updating voter registration records. Rather than making individuals provide the same information on two separate forms – one for the agency and one for voter registration – AVR leverages the agency’s existing procedures to seamlessly register eligible citizens. Thus, Medicaid AVR would eliminate the number of forms that officials and citizens have to handle and reduce delays in transferring and processing data. Because Medicaid verifies citizenship information, Medicaid AVR allows for secure registration. Significantly, Medicaid AVR will apply to all eligible household members included on agency applications, not just those filing agency forms or visiting an office to conduct a transaction covered by the NVRA.

How does Medicaid AVR Work? 

After Medicaid applicants provide household members’ names, addresses, and dates of birth, the state Medicaid agency verifies citizenship with the Federal Data Services Hub, a citizenship database maintained by the Social Security Administration and other federal agencies.15 For verified adult citizens, the Medicaid agency electronically passes their data to election officials. If the registration application is considered valid by the election agency, three outcomes are possible. First, if a person is already registered and the address provided matches their current registration record, no further action is taken. Second, if a person is already registered and a new address is provided, their registration is updated. Third, if a person is not on the registration list, a new registration record is created.

In the second and third scenarios, individuals are notified by mail and provided an opportunity to decline the registration or address change. This last step ensures that no one is registered or has their address updated without notification and consent. Note that, if designed properly, this registration and update procedure would apply to all eligible household members included in the Medicaid system.

Where is Medicaid AVR Being Implemented?

Medicaid AVR has been passed in Colorado, Nevada, New Mexico, Minnesota, and the District of Columbia.16 However, to ensure compliance with Medicaid regulations, these states are waiting for the “green light” from the Center for Medicare and Medicaid Services (CMS) before implementing these systems.

Maryland and Massachusetts have also implemented a less streamlined model of Medicaid AVR.17 In these states, Medicaid enrollees are instructed that their information will be used for voter registration unless they check a box to decline this option. Unlike the fully automated model passed in other states, the system used in Massachusetts and Maryland only applies to the person completing the application, not the entire household.

Table of Abbreviations

ASEC Annual Social and Economic Supplement of the Current Population Survey

AVR Automatic Voter Registration

CMS Centers for Medicare & Medicaid (US Department of Health and Human Services)

CPS Current Population Survey (US Census Bureau & US Bureau of Labor Statistics)

DOJ US Department of Justice

DMV Department of Motor Vehicles (or similar state agencies)

EAC US Election Assistance Commission

FEC Federal Election Commission

NVRA National Voter Registration Act of 1993 (52 USC Chapter 205)

SNAP Supplemental Nutrition Assistance Program (formerly the Food Stamp Program)

TANF Temporary Assistance for Needy Families

WIC Special Supplemental Nutrition Program for Women, Infants, and Children

VRA Voter Registration Application

Tables and Figures

Table 1. Voter Registration Rates and Gaps, 2010-2020

Source: CPS November and December Supplements. Note: Non-responses and respondents stating they are not eligible to register are excluded.


Table 2. Why Eligible Citizens were not Registered, 2020

Source: CPS November Supplement. Notes: Non-responses and respondents stating they are not eligible to register are excluded. Extrapolating to all unregistered eligible citizens increases the final column by approximately five to ten percent.

Table 3. Time at Address by Income and Age Groups, 2010-2020

Source: CPS November Supplement.

Table 4. Registration Rates and Time at Address by Income Quartile and Age Groups, 2010-2020

Source: CPS November Supplement.

Table 5. Registered Citizens Reporting a Voter Registration Problem as the Reason for Not Voting, 2010-2020

Source: CPS November Supplement.
Note: Percentage is of those reporting a reason.

Table 6. Method of Voter Registration, 2010-2020

Source: CPS November Supplement.

Voter Registration Applications and Program Participation Encounters, 1996 to 2020

Source: VRAs from the FEC (1996 to 2002) and the EAC (2004 to 2020). Participation data from the ASEC. Note: Adult citizen participation in Medicaid, SNAP, WIC, and TANF is used as a conservative estimate of the total unique adult citizen encounters over an election cycle, assuming an average of one encounter per year with an agency in a two-year election cycle (see footnote 7). Restricting the figure to states that have no cycles with missing data results in a similar portrait.

Table 7. Social Service Agency Voter Registration Program Performance, 1996 to 2020

Sources: Voter registration application data from the FEC (1996 to 2002) and EAC (2004 to 2020). Participation data from the ASEC. See footnote 7. Restricting the data to states that have no cycles with missing data results in a similar portrait.

Table 8. Improvements in VRA Performance during Section 7 Compliance Agreements, Select States

Sources: Voter registration application data from the FEC (1996 to 2002) and EAC (2004 to 2020). Participation data from the ASEC. See footnote 7. Notes: Underlined ratios represent cycles when states were either in litigation or negotiating Section 7 compliance agreements with plaintiffs. Bold ratios represent cycles when compliance agreements were in effect for important parts or all of the cycle. The extreme outlier performance by Nevada in 2002 likely reflects an error in the data reported to the EAC.

Table 9. Social Service Agency Performance by Compliance Intervention Periods, Averages per State Election Cycle, 2002 to 202019

Sources: See footnote 7. Note: Analysis is for the 22 states that had Section 7 compliance agreements or similar comprehensive plans in place for some years between 2002 to 2020.

Table 10. VRAs from DMVs in Select States

Sources: VRAs from EAC reports; DMV AVR policy dates from National Conference of State Legislatures (2023). Notes: VRA totals in bold are election cycles when DMV AVR was reported to be in effect. For some recent election cycles, California and Nevada were also parties to memoranda of understanding with private plaintiffs regarding improvements to NVRA compliance in DMVs.



Ashbrook, Alexandra, Sarah Brannon, and Douglas R Hess. 2017. A Review of National Voter Registration Act Compliance in SNAP Applications and How Anti-Hunger Groups Can Advocate for Improvements. Washington, DC: Project Vote & Food Research and Action Center.

Berinsky, A. J. 2005. “The Perverse Consequences of Electoral Reform in the United States.” American Politics Research 33(4): 471–91.

Fabina, Jacob, and Zachary Scherer. 2022. Voting and Registration in the Election of November 2020. Washington, DC: US Census Bureau.

Grimmer, Justin, and Jonathan Rodden. 2022. “Changing the Default: The Impact of Motor-Voter Reform in Colorado.” Unpublished Working Paper.

Groarke, Margaret M. 2000. “Expanding Access to the Vote: An Analysis of Voter Registration Reform in the United States, 1970-93.” PhD Thesis. The City University of New York.

Herman, Jody. 2008. Re: NVRA Public Agency Registration (Section 7) Field Research Results. Washington, DC: Project Vote.

Hess, Douglas R., Michael J. Hanmer, and David W. Nickerson. 2016. “Encouraging Local Compliance with Federal Civil Rights Laws: Field Experiments with the National Voter Registration Act.” Public Administration Review 76(1): 165–74.

Hess, Douglas R., and Scott Novakowski. 2008. Unequal Access: Neglection the National Voter Registration Act, 1995-2007. Washington, D.C: Project Vote & Demos.

Leighley, J.E., and J. Nagler. 2013. Who Votes Now?: Demographics, Issues, Inequality, and Turnout in the United States. Princeton, NJ: Princeton University Press.

Massicotte, Louis, Andre Blais, and Antoine Yoshinaka. 2004. “Registering Voters.” In Establishing the Rules of the Game, Election Laws in Democracies, University of Toronto Press, 66–82.

McGhee, Eric, and Mindy Romero. 2021. Effects of Automatic Voter Registration in the United States. Los Angeles, CA: University of Southern California, Sol Price School of Public Policy.

Mellor, Brian, and Douglas R. Hess. 2009. Improving Voter Registration Performance in Public Assistance Agencies through Modernizing Data Transfer Systems: Testimony for the Committee on State Voter Registration Databases of the National Research Council. Cambridge, MA: Committee on State Voter Registration Databases of the National Research Council.

National Conference of State Legislatures. 2023. “Automatic Voter Registration.”

Piven, Frances Fox, and Richard A. Cloward. 2000. Why Americans Still Don’t Vote: And Why Politicians Want It That Way. Rev. and updated ed. Boston: Beacon Press.

Presidential Commission on Election Administration. 2014. The American Voting Experience: Report and Recommendations of the Presidential Commission on Election Administration. Washington, D.C: Presidential Commission on Election Administration.

Richburg-Hayes, Lashawn et al. 2014. Technical Supplement for Behavioral Economics and Social Policy: Designing Innovative Solutions for Programs Supported by the Administration for Children and Families. Washington, DC: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services.

Rigby, Elizabeth, and Melanie J. Springer. 2011. “Does Electoral Reform Increase (or Decrease) Political Equality?” Political Research Quarterly 64(2): 420–34.

Rogers, Estelle H. 2009. The National Voter Registration Act at Fifteen. Washington, DC: Project Vote.

Stewart III, Charles H. 2021. How We Voted in 2020: A Topical Look at the Survey of the Performance of American Elections. Boston: Massachusetts Institute of Technology Election Data & Science Lab.

US Department of Justice, Civil Rights Division/Voting Section. 2015. “The National Voter Registration Act Of 1993 (NVRA).” The National Voter Registration Act of 1993 (NVRA).

US House of Representatives, Subcommittee on Elections, Committee on House Administration. 2008. Hearing on the National Voter Registration Act, Section 7: The Challenges That Public Assistance Agencies Face. Washington, DC: US Government Printing Office.

1. Data and other support for statements in the Introduction can be found in the main body of the report.
2. The NVRA applies to 44 states and the District of Columbia. Idaho, Minnesota, New Hampshire, Wisconsin, and Wyoming are exempt because they implemented Election Day Registration (EDR) before 1996. (States that have adopted EDR since 1996 are still obligated to follow the Act.) North Dakota is the sixth exempt state because, unique among the states, it does not have a traditional voter registration requirement. States exempt from the NVRA can, of course, implement the voter registration innovation discussed in this report, as Minnesota did this year.
3. In addition to Medicaid, the other major programs covered by Section 7 are the Supplemental Nutrition Assistance Program (SNAP, aka “food stamps”), the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), and Temporary Assistance for Needy Families (TANF). Section 7 also requires voter registration services at military recruitment centers, disability programs, and one or more additional locations states are to select. However, as important as these other registration services are, they produce relatively far fewer registrations. Thus, following the tradition in other NVRA reports, I use “Section 7” or “agency registration” to refer to the NVRA-mandated voter registration programs at social services programs.
4. The data in Table 6 could be interpreted as preferences in how to register. However, the groups presented in the table are not significantly different in expressing disinterest in registering or voting (also from the CPS, not shown). Furthermore, for each pair in the table, one group has a lower registration rate and is more likely to report certain barriers to registration than the other group. Moreover, as discussed later in the report, when the accessibility of a registration method improves, such as when government registration programs are brought into compliance with the NVRA or when AVR is instituted, use of these methods increases dramatically. In short, it is reasonable to attribute some of the demographic differences in the use of registration methods to differences in who they can reach.
5. For summaries of the NVRA, see US Department of Justice (2015) and Rogers (2009). For the full text in the US Code, see NB: USC sections end in a number one less than the section numbers in the legislation (e.g., “Section 7” is 52 USC § 20506). I follow tradition and use section numbers from the legislation.
6. On the risk that some election reforms may increase inequalities, see Berinsky (2005) and Rigby and Springer (2011).
7. For Figure 1 and Tables 7 to 9, the number of VRAs for a state in any one election cycle is divided by the number of adult citizens participating in the four largest programs covered by Section 7 (Medicaid, SNAP, WIC, and TANF). This number, based on ASEC estimates, is first summed over the two calendar years overlapping with each two-year election cycle. This serves as proxy for the number of program participant encounters. Although this estimate does not include duplicates across programs in any one calendar year, summing over two years does introduce duplicates. However, the estimated count for one year is highly conservative due to underreporting of program participation. The estimate is also conservative because agencies are to offer voter registration services to applicants even though some will not become participants. Likewise, registration is to be offered when clients change address, and this is not captured in participation data. In the end, the exact denominator is not crucial. The point is to adjust each state’s number of VRAs for each election cycle by a reasonably meaningful and conceptually consistent proxy for the number of adult citizens these programs encounter.
8. This remains true even when comparing over time only those states that never failed to report VRA totals.
9. See Hess and Novakowski (2008) regarding evidence of widespread noncompliance and refutations of alternative hypotheses on the decline in Section 7 performance.
10. Portions of this section draw on text from Hess, Hanmer, and Nickerson (2016).
11. During Section 7 litigation, depositions of agency officials and the discovery process also revealed the lack of VRAs supplies in offices in several states (author’s files).
12. Monthly agency monitoring data from several states, provided to advocates pursuant to compliance agreements, also indicate that performance generally declines over time, sometimes before agreements expire (author’s files).
13. Medicaid applicants or participants would also be given an opportunity to decline registration.
14. Note the testimony by Jonnie McClean — then Deputy Director of the North Carolina State Board of elections — at a congressional hearing on agency NVRA compliance: “Most importantly [for increasing NVRA compliance], I believe, was tracking … of the voter registration applications received” from agency offices (US House of Representatives, Subcommittee on Elections, Committee on House Administration 2008).
15. See Center for Medicare and Medicaid Services. (Sept. 20, 2012) “Answers to Frequently Asked Questions, Federal Data Services Hub.” Available at
16. Colo. Rev. Stat. 1-2-502.5; Nev. Rev. Stat. Ann. § 293.5768; N.M. Stat. Ann. § 1-4-5.8; Minn. Stat. § 201.161; D.C. Code § 1-1001.07b.
17. Md. Code Ann. Elec. Law § 3-203; Mass. Gen. Laws ch. 51, § 42G½.
18. Self-reports of registration status and excluding non-responses, as I have done with CPS data for this report, bias these estimates upwards. Nonetheless, the CPS remains the best accessible source of registration estimates and the relevant point — that sizeable gaps in registration rates exist — is not obviated by these measurement difficulties. Note that because I exclude non-responses, my estimates will not match those published by the Census Bureau. SNAP registration data is available for respondents appearing in both the November and December CPS Supplements, of which the latter includes questions about SNAP participation. In theory, due to the longitudinal design of the CPS, roughly 75% of survey respondents should match across two consecutive months. I was able to match 73% of respondents (i.e., 97% of the ideal). Income quartiles are approximate and adjusted each cycle.
19. Intervention periods are coded into five categories for the 2002 to 2020 election cycles based on the status of Section 7 compliance interventions by civil rights organizations or the DOJ. Because intervention periods do not line up precisely with election cycles, the coding of cycles is approximate. Negotiation periods start when states are formally notified that they may be sued due to noncompliance with the NVRA. Agreement periods are based on execution and expiration dates in the agreements. In the few cases when agreements do not have an end date, two election cycles are assumed as this is the longest of those agreements with end dates. The use of two post-agreement periods (each covering two election cycles) allows us to see if the impact of agreements decay over time. Four cycles after agreements end, state election cycles are returned to the pre-intervention category.