Registering and Voting with Motor-Voter

Registering and Voting with Motor Voter*

Raymond E. Wolfinger, University of California, Berkeley
Jonathan Hoffman, University of California, Berkeley

The National Voter Registration Act (NVRA) of 1993 (P.L. 103-31) was designed to reduce the cost of voting by incorporating registration into a transaction with a public agency that citizens initiate for another purpose. The act took effect on January 1, 1995; between then and the 1996 election, 18,333,479 people registered in offices they had visited on other business (Federal Election Commission 1998, Table 2).¹ Four years later, answers to even the most elementary questions about this first NVRA election have not been published. We narrow the data gap by describing people who used NVRA to register, and then comparing their participation to that of their fellow citizens who registered to vote using more traditional methods.

Obtaining and renewing driver's licenses and reporting changes of address to departments of motor vehicles (DMV) are the most common and compatible transactions covered by NVRA, hence the "motor voter" nickname.² Seventy-five percent of all NVRA registrations in 1995-96 were completed at DMV offices (FEC 1998, Table 2). Another NVRA title directed states to offer registration to clients in "all offices in the state that provide public assistance." This provision, far less productive than motor voter, accounted for 14% of all NVRA registrations in 1995-96 (6% of all registrations).³

Data Sources

Our individual-level data are from the 1996 Voter Supplement of the U.S. Census Bureau's Current Population Survey (CPS), which provided a total of 78,309 unweighted cases, including 69,960 cases in the 43 states that implemented the NVRA.⁴ Conducted monthly, primarily to provide data on unemployment, the CPS has numerous demographic items and monthly "supplements" on specific topics. Every other November, the CPS asks about citizenship status, registration, and turnout.⁵ In 1996, the Voter Supplement also asked if anyone had registered after January 1, 1995, and if so, whether they had done it when obtaining or renewing a driver's license, or "in some other way."⁶ Respondents in this last category were read five alternatives, one of which was: "At a public assistance agency (for example, a Medicaid, AFDC, or Food Stamps office, an office serving disabled persons, or an unemployment office)" (U.S. Bureau of the Census 1997). The other four codes do not permit us to identify those who used a "voter registration agency" (see note 3). The CPS has a 95% completion rate, in contrast to around 70% for recent National Election Studies (Brehm 1993, 16). Interviews for the 1996 Voter Supplement were completed by November 19.

Taking Advantage of the NVRA

Everyone knows that the NVRA's implementation was not followed in 1996 by the double-digit increases in turnout predicted by some advocates. In fact, "official turnout" in 1996 was described as the lowest in modern times.⁷ One explanation for this disappointment is the common observation that people for whom registration is costless are unlikely to exert themselves to vote; registration might increase, but turnout would not. Therefore, it is perhaps more surprising that responses to successive Voter Supplements, the best available time-series measure, indicate that after a four-percentage-point gain from 1988 to 1992, aggregate registration declined from 78% in 1992 to 77% in 1996.⁸ Reality clearly disappointed those enthusiasts who thought that the NVRA would "raise national voter registration from 60 percent to 95 percent" (Cloward and Piven 1992).

Although aggregate registration did not appreciably change in the first two years of NVRA implementation, many millions of Americans did register at DMVs and public agencies. Thanks to the Voter Supplement, we can describe those who used the new law, comparing them to people who registered by other methods in the same time period and to everyone who registered prior to 1995. In the 43 states that implemented the NVRA, 15% of those who registered in time for the 1996 presidential election registered after January 1, 1995. One would not expect that all kinds of people were equally likely to be "new registrants," as we will call them. As Table 1 shows, they were a notably disproportionate share of all registrants in two categories: the young and the mobile. Among people aged 18 to 21, experiencing their first presidential election, 54% of all those who registered did so after January 1, 1995; this fell to 29% of those just old enough to vote in 1992, 18% of thirtysomethings, and so on. Only 7% of all registrants 50 or older were new registrants. Thirty-nine percent of all registrants who had moved in the year before the election were new, compared to a mere 7% of those who had not moved for at least five years.

Asian Americans and Latinos were particularly likely to be mobilized for the 1996 contest. Twenty-three percent of Asian-American and 22% of Latino registrants registered in the two years before the election, compared to 15% for all other registrants. Multivariate logit analysis does not erase this difference. We found that, with all other demographic variables controlled, Asian Americans were 7% and Latinos 4% more likely to be new registrants. These could be people mobilized by a belief that immigrants were being unfairly treated by laws enacted in Washington, D.C., and in some states and/or new citizens taking advantage of their first opportunity to vote in a presidential election. (Sixty-eight percent of Latino citizens and 36% of Asian-American citizens in the CPS sample were born in the United States, compared to 95% of whites and 96% of blacks.)⁹

We cannot directly test our speculation about recently naturalized people because the CPS did not ascertain when anyone became a citizen, but it did ask when naturalized citizens entered the United States. We used this item to classify Asian-American and Latino registrants by year of entry: those who arrived before 1975, between 1975 and 1979, between 1980 and 1985, and after 1985. There is a pronounced monotonic relationship between recency of arrival and new registration. Among Asian Americans, 40% of the most recent immigrants were new registrants, double the proportion of those who arrived before 1975. The relationship is even more pronounced for Latinos; half of those who immigrated after 1985 were new registrants, compared to just 18% of the longest-term naturalized citizens.

These findings introduce our discussion of what kinds of citizens took advantage of the NVRA to register in motor vehicle and public assistance offices.¹⁰ As Table 2 shows, motor voter was particularly popular among people who had changed residences within two years of the election. More than a third of recently-moved new registrants achieved or maintained eligibility to vote at DMV offices, compared to about a quarter of more settled new registrants. Use of motor vehicle offices for voter registration did not vary much by age, except for the very youngest potential voters and people past their sixtieth birthday. Although these findings might interest political scientists, they have no particular political significance, given the lack of any pronounced partisan or ideological inclinations among the young or the restless.

This could not be said about class and racial patterns in use of the NVRA. From 1989, when the legislation was first introduced, a common argument for motor voter was that driver's licenses were far more prevalent than voter registration cards. While not denying this, advocates for minorities and the poor often said that licenses were considerably less common among their constituents. Testifying before the House Subcommittee on Elections, the late Frank R. Parker, a leading civil rights lawyer, presented figures comparing white and black access to automobiles to conclude that "the provision for driver's license voter registration . . . has a certain class and racial bias" (1989, 238).¹¹ Another civil rights attorney argued that "motor voter registration would not adequately reach low-income voters of any race" (Cunningham 1991, 390).

The data presented in Table 2 confirm these expectations. Only 22% of new registrants who had not graduated from high school qualified to vote through motor voter, compared to 33% who had their diplomas and about the same proportion of college graduates. The disparity is almost as large for people making less than $10,000 a year. By the same token, 32% of white new registrants were motor voter patrons, compared to 23% of blacks and 21% of Latinos.

The figures for public assistance registration go in the opposite direction from motor voter and are more concentrated at the lower end of the income and education distributions. Nine percent of new registrants who were high school dropouts registered in public assistance offices, compared to just 5% of those whose education stopped with high school and barely 1% of anyone who had set foot in a college classroom. The same was true on the income scale; 13% of the poorest, less than half as many in the next highest category, and half again of those in the next highest group registered in public assistance offices. Only 3% of whites and no appreciable number of Asian Americans registered in public assistance agencies, compared to 7% of blacks and 6% of Latinos. The feminization of poverty is illustrated by the higher proportion of female new registrants who registered in public assistance offices: 5% as opposed to just 2% of men. Fully 77% of all public assistance registrants were women.

As intended, NVRA's public assistance agency provision compensated in large measure for class disparities in use of motor voter. This point is illustrated by comparing columns one and three in Table 2 . Although people without a high school diploma were 13 percentage points less likely than college graduates to register at a DMV, the difference between these groups is only five percentage points when the dependent variable is the combined proportion of DMV and public assistance agency registrants. People with incomes below $10,000 were nine percentage points less likely to register at a DMV than those in the $50,000-$75,000 income range, but were three points more likely to register at either a DMV or a public assistance agency. The effect was similar, but more modest, for minorities. Public assistance agency registration only partially compensated for racial disparities; appreciable gaps remained between minorities and whites in use of the NVRA, with a greater difference for Latinos than blacks: eight as opposed to five percentage points.

The bivariate findings in Table 2 are confirmed when we estimate the effects of class and race while controlling for other demographic characteristics. As the first column in Table 3 indicates, people with less than 12 years of schooling were almost six percentage points less likely to register at a DMV than those with at least a high school education, all else being equal. Poor people were about four points less likely to register at a DMV than those enjoying incomes above $10,000, and minorities were over 10 points less inclined than whites to use motor voter. The numbers in the third column in Table 3 show how public assistance agency registration compensated for the social class disparities. Controlling for other demographic variables, people with family incomes below $10,000 were more likely than better-off citizens to register at either a DMV or a public assistance agency, and those who had not graduated from high school were just as likely as the educated to take advantage of NVRA. The availability of public agency registration did not, however, affect the racial disparities shown in Table 2. After controlling for other demographic characteristics, whites were seven points more likely than blacks, nine points more likely than Latinos, and eight points more likely than Asian Americans to register at either a DMV or public assistance agency. In short, while agency registration fulfilled the intentions of the act's authors with respect to poor and uneducated people, it fell short of eliminating racial disparities.

Turnout by NVRA Registrants

The National Voter Registration Act was premised on a belief that easier registration would lead to higher turnout. The Act's principal author, Rep. Al Swift (D-WA), had been a convert to this proposition since learning "that while U.S. voter turnout is far behind European countries, the percentage of registered voters who vote in this country compares favorably to other Western democracies" (Swift 1984, 13).¹² With the implementation of the NVRA, the unanswered question was whether the 87% turnout of the registered in the 1980 election that so impressed Rep. Swift would persist among citizens for whom registration was virtually costless. Putting the question this way suggests that it would not. One journal reported that unnamed "Political scientists expect turnout among motor-voter registrants to be about 10 percent. . . ." (Greenblatt 1996, 233). Walter Dean Burnham expressed the more modest view that people who register through motor voter are "presumed to be relatively high abstainers in the fall" (Greenblatt 1996, 233). These pessimistic expectations should not apply to the incalculable number of motor voter registrants who would have registered in time by some other method, and whose likelihood of voting in 1996 would not be tainted because they took advantage of the convenience offered by this new method.

We used the Voter Supplements to compare the turnout of the registered in recent elections and then analyzed turnout specifically by DMV and public assistance agency registrants. Turnout of the registered in 1988 was 86% -- within a point or two of other elections in the 1980s -- and then rose to 91% in 1992, when official turnout was also appreciably higher than in the earlier contests. These numbers establish the context for assessing our finding that 83% of the registered voted in 1996. This is quite a decline from 1992, and a lesser but still noteworthy difference from elections in the 1980s. However, it should not be thought that the change from 91% in 1992 to 83% in 1996 reflects a denominator swollen by hordes of easily-registered beneficiaries of NVRA. As Table 4 shows, the rate was only 84% in 1996 for those -- over five-sixths of the total -- who registered before the act came into effect. The turnout of the rest of the registered population was appreciably lower; 77% of new registrants went to the polls in November 1996.¹³ Much of the disparity between old and new registrants is due to people who took advantage of NVRA; fully 82% of all other new registrants voted. Seventy percent of people who registered in DMV offices voted in 1996. On the one hand, this is 14 percentage points lower than the turnout of pre-1995 registrants and 12 points below other new registrants. On the other hand, it greatly exceeds the expectations of scholars who thought that motor-voter registrants would be largely abstainers.

The much smaller numbers of citizens -- about 3% of all new registrants -- who registered in public assistance offices came closer to meeting the pessimists' expectations; barely half of them voted. Mostly poor and uneducated, but not notably young or mobile, these people were the intended beneficiaries of the NVRA's public assistance agency provision. Table 4 shows that they were less likely to vote than their demographic counterparts who registered elsewhere. (The modest number of cases leads us to eschew comment here about minority registrants at public assistance offices.)

Because clients of public assistance offices may be multiply disadvantaged, the bivariate analyses in Table 4 are insufficient to exclude the possibility that their lower turnout reflects demographic factors rather than the effects of costless registration. Therefore we returned to logistic regression to analyze turnout differences between three categories of registrants: people who qualified to vote at DMVs, those who registered at public assistance offices, and other new registrants.

In the aggregate, with all demographic variables controlled, people who registered at a public assistance office were 14 percentage points less likely than other new registrants to vote in 1996. Table 5 also shows that, with demographic variables controlled, the disparity between motor-voter registrants and other new registrants was nine points. We venture this speculation: These differences may reflect the weight of the evidence for the proposition that when registration is costless, some of the people thus qualified will find voting too costly. Beneficiaries of a relatively easy process to establish their eligibility to vote, some NVRA registrants might not be inclined to take the marginally demanding final step. Non-NVRA registrants, on the other hand, having exerted themselves to establish their eligibility to vote, might be less likely to be deterred from voting by the challenge of going to the polls or obtaining and using an absentee ballot.

All other new registrants were, by nearly four percentage points, more likely to vote than pre-1995 registrants. This finding is consistent with the proposition that, all else being equal, people drawn to register after January 1, 1995 were a bit more likely to be interested in some aspect of the 1996 election contest.

Conclusions

Our results indicate that critics of the NVRA were correct in predicting class and racial disparities in use of motor voter. Public assistance agency registration fulfilled some of the intentions of the Act's authors by largely compensating for class disparities in motor voter, but fell short of eliminating racial disparities. Black and Latino new registrants were less likely than their white counterparts to take advantage of the NVRA, even after controlling for other demographic differences.

Turnout of motor voter registrants was lower than that of other new registrants and of pre-1995 registrants, but greatly exceeded the expectations of scholars who argued that people who registered by this method would not go to the polls. It seems reasonable to assume that, absent NVRA, many DMV registrants would have registered by some other method before the 1996 election. Although we cannot exclude the possibility that motor voter registration affected turnout in 1996, our findings are consistent with the proposition that costless registration is associated with an increase in nonvoting registrants. Public assistance agency registration came closer to meeting the pessimists' expectations; only half the people who qualified to vote by this method actually took advantage of their opportunity. Some of the turnout disparity between such registrants and others reflected their demographic characteristics, but the 14 percentage point difference that remains after controlling for such variables supports the idea that those who register when the process is costless are less likely to vote. In other words, while registration may be the greatest cost of voting, it is not the only hurdle that must be surmounted.

Notes

* Brian Hancock of the Federal Election Commission generously provided data and guidance that immeasurably aided our research. We benefited also from advice from Margaret Rosenfield and comments on an earlier draft from Benjamin Highton and Scott McClurg. The data set we analyzed was obtained from the Inter-University Consortium for Political and Social Research through UC Data Archive and Technical Assistance (UC DATA) of the University of California, Berkeley. None of these organizations is responsible for our analyses or interpretations. We are grateful for financial support from the Carnegie Corporation and the Ford Foundation.

1. Most of these were new registrations; the remainder were largely changes of address. (We are grateful to Brian Hancock of the Federal Election Commission for his help in interpreting this table.) They are 44% of the total of 41.1 million "registration applications or transactions" of any sort recorded in the 43 states subject to the NVRA in 1995-96. Six states are exempt from the NVRA; five -- Idaho, Minnesota, New Hampshire, Wisconsin, and Wyoming -- because they permit election-day registration at the polling place and North Dakota, which does not require voter registration. Vermont's constitutional provisions kept it from implementing the NVRA by 1996.

2. The NVRA also required states to permit registration by mail. Because most states were doing so already and mail registration has never been shown to have any effect on turnout, we will not be concerned with it.

3. In addition to registration in public assistance agencies, which was intended to bring more poor people to the polls, the NVRA also directed states to designate other public agencies, presumably with more diverse clients, as "voter registration agencies." Four covered states failed to name any additional voter registration agencies and half the states reported designating only one such agency (FEC 1998, 2).

These voter registration agencies accounted for another 9% of NVRA registrations (FEC 1998, Table 2). Neither they nor public assistance agencies provided such a convenient link to registration as a driver's license application. Moreover, the NVRA required them to hand each client a cumbersome and forbidding "declination" form (Highton and Wolfinger 1998, 89).

4. Data on registration and voting in the 2000 election will provide another research opportunity because a complete cycle of driver's license renewals will have occurred. Well over 90% of the population in NVRA states will have had at least one encounter with a DMV that offered a convenient occasion to register.

5. The CPS is based on a household sample in which one person reports on all members of the household. This proxy reporting does not affect estimates of turnout (Jennings 1990).

Voter Supplements are the source of the Census Bureau's biennial reports on registration and voting. Data management in these reports differs from ours in two respects: (1) We deleted cases for which information on registration or voting was not obtained; the Census Bureau codes these individuals as nonvoters. (2) We also deleted noncitizens, in contrast to the Census Bureau, which does this in its detailed reports only after featuring turnout estimates based on millions of people who are ineligible to vote. These adjustments left us with 78,309 unweighted cases.

Computing turnout on a base that includes noncitizens introduces two nonrandom errors. First, because the noncitizen population is growing, national turnout is progressively underestimated. In the 1996 Voter Supplement, 7.1% of the voting-age population were not citizens. Using this number to recompute the official turnout results in an increase from the familiar 48.9% to 52.6% of the voting-age citizen population. Second, turnout in states with disproportionately large noncitizen populations is more substantially underestimated. According to the 1990 Census, 16.8% of voting-age Californians, compared to 5.4% of all adult United States residents, were not citizens (U.S. Bureau of the Census, nd).

6. We deleted from analyses using this item a handful of cases where information about the registration venue was not obtained.

7. The most common measure of electoral participation, "official turnout" is the number of votes cast for presidential candidates divided by the Census Bureau's estimate of the voting-age population in November. See note 5 for a discussion of nonrandom errors associated with this measure.

8. In contrast to this one-point decline, the Federal Election Commission reported a slight increase from 1992. The FEC estimated aggregate national registration by dividing the voting-age population (VAP) into the sum of "active" registrations reported by the states. The result, 72.8% of the VAP in 1996, represents a gain of just over two percentage points from 1992 and "is the highest percentage of voter registration since reliable records were first available in 1960" (FEC 1997, 1). Although this computation excluded inactive registrants in 1996, almost all of whom were believed to be people who had moved, the amount of "deadwood" on many state registration rolls might have grown due to the NVRA's restrictions on purging. Estimates of registration from the Voter Supplement are not affected by such bookkeeping changes.

9. In all our analyses, black Latinos are coded as Latinos.

10. Although one could register in some motor vehicle offices at least 10 years before the NVRA came into force, few states had inaugurated programs similar to what the federal law required (Highton and Wolfinger 1998).

11. Readers doubtless will recognize that one need not own a car to have a driver's license. The disparity between these two indicators is perhaps widest among young people. In 1993, fully 88% of Americans aged 20 to 24 held licenses (U.S. Department of Transportation 1994, 8). This is 35 percentage points higher than the reported turnout of this age group.

12. The table that convinced Swift and was displayed at the hearing where he said this can be found in Glass, Squire, and Wolfinger (1984, 52).

13. Although, in the aggregate, old registrants outvoted new registrants, the opposite was true for the young and mobile new registrants.

References

Brehm, John. 1993. The Phantom Respondents. Ann Arbor: University of Michigan Press.

Cloward, Richard A., and Frances Fox Piven. 1992. "Democracy on Wheels." The New York Times, June 17, A25.

Cunningham, Dayna L. 1991. "Who Are to be the Electors? A Reflection on the History of Voter Registration in the United States." Yale Law and Policy Review 9:370-404.

Federal Election Commission. 1997. "The Impact of The National Voter Registration Act of 1993 on the Administration of Elections for Federal Office 1995-1996." A Report to the 105th Congress. Washington, DC: Federal Election Commission.

-----. 1998. "Implementing the National Voter Registration Act: A Report to State and Local Election Officials on Problems and Solutions Discovered 1995-1996." Washington, DC: Office of Election Administration.

Glass, David P., Peverill Squire, and Raymond E. Wolfinger. 1984. "Voter Turnout: An International Comparison." Public Opinion, December/January: 49N55.

Greenblatt, Alan. 1996. "Court Rejects `Motor Voter' Case, but the Battle Isn't Over." Congressional Quarterly Weekly Report, January 27, 232-34.

Highton, Benjamin, and Raymond E. Wolfinger. 1998. "Estimating the Effects of the National Voter Registration Act of 1993." Political Behavior 20(June): 79-104.

Jennings, Jerry T. 1990. "Estimating Voter Turnout in the Current Population Survey." Studies in the Measurement of Voter Turnout. Current Population Reports, Series P-23, No. 168:21N9. Washington, DC: U.S. Bureau of the Census.

League of Women Voters. 1996. "Absentee Voting." Washington, DC: League of Woman Voters Education Fund.

Parker, Frank R. 1989. Testimony before the Subcommittee on Elections, Committee on House Administration. U.S. House of Representatives. 101st Cong., 1st sess., March 21.

Public Law 103-31. 1993. 107 Stat. 77N89.

Swift, Al. 1984. Task Force on Elections, Committee on House Administration. U.S. House of Representatives. 98th Cong., 2nd sess., June 27.

U.S. Bureau of the Census. nd. 1990 Age, Nativity, and Citizenship for the United States, States, and Counties. CPH-L-114. Washington, DC: Ethnic and Hispanic Branch, Population Division. -----. 1997. Current Population Survey: Voter Supplement File, 1996 [Computer File]. ICPSR version. Washington, DC: U.S. Bureau of the Census [producer]. Ann Arbor, MI: Inter-University Consortium for Political and Social Research [distributor].

U.S. Department of Transportation. 1994. Driver Licenses-1993. HPM. Washington, DC: Federal Highway Administration.