The National Voter Registration Act of 1993 requires states to allow eligible persons to register to vote at various government locations, including public assistance offices. Since the initial reporting period (1995-1996), the number of persons regis­tering to vote at public assistance offices has declined. This trend has led some to speculate that the states are failing to provide welfare recipients with the opportunity to register to vote at public assistance offices. However, other possible explana­tions include declining welfare caseloads caused by welfare reform in 1996.

The analysis presented in this Center for Data Analysis (CDA) report directly tests the hypothesis that the Personal Responsibility and Work Oppor­tunity Reconciliation Act (PRWORA) of 1996 con­tributed to the decline in public assistance voter registrations.[1]  PRWORA replaced Aid to Families with Dependent Children (AFDC) with Temporary Assistance for Needy Families (TANF) and helped to reduce welfare caseloads.

After controlling for factors that influence the number of voter registrations at public assistance offices, CDA analysts found a statistically signif­icant association between AFDC/TANF partici­pation and public assistance voter registrations. For example, a 1 percent decrease in AFDC/ TANF participation is associated with a 0.49 per­cent decline in voter registrations at public assis­tance offices.

Declining AFDC/TANF caseloads from 1996 to 2006 contributed substantially to the decline in public assistance voter registrations. Unlike previ­ous research, this report uses panel regression anal­ysis to estimate the relationship between AFDC/ TANF participation and other factors that influence public assistance registrations.

Members of Congress, policymakers, and the media should not dismiss the major role that wel­fare reform and decreased welfare participation have played in reducing the number of public assis­tance voter registrations.

Background

The National Voter Registration Act of 1993 requires states to allow eligible persons to register to vote at various government locations, including public assistance offices. Starting in 1995, states reported the number of voter registrations by regis­tration location in two-year intervals.[2] 

Since the initial reporting period (1995-1996), the number of persons registering to vote at public assistance offices has declined. This trend has led some to speculate that the states are failing to pro­vide welfare recipients the opportunity to register to vote at public assistance offices.[3] 

A 2005 report by Demos, the Association of Com­munity Organizations for Reform Now (ACORN), and Project Vote-three organizations devoted to voting rights advocacy-used descriptive statistics to describe changes in voter registrations at public assistance offices.[4]  The authors of the paper report that "registrations at public assistance agencies dropped 59% between 1995-1996 and 2003- 2004."[5]  However, the authors do not control for welfare reform and dismiss the potential effect of declines in AFDC/TANF caseloads on public assis­tance voter registrations. The authors state that, "[w]hile caseloads in some public assistance pro­grams have declined overall since the NVRA went into effect, these declines are not sufficient to explain the declines in voter registration applica­tions through public assistance agencies."[6] 

A 2008 report by Project Vote and Demos per­formed another descriptive analysis of trends in public assistance registrations. This report also rejects any possibility that changes in welfare case­loads may help to explain the decline in public assistance voter registrations.[7]  As evidence, it notes:

While welfare reform and the booming economy in the late 1990s contributed to a decrease in participation in some public assistance programs, this trend reversed in the first years of the new century. For instance, the Food Stamp Program-by far one of the largest public assistance programs required to offer voter registration-had several hundred thousand more adult citizen participants nationwide in fiscal year 2006 compared to a decade earlier.[8] 

This study suggests that the number of voter reg­istrations from public assistance offices declined by 79 percent from the reporting period of 1995-1996 to that of 2005-2006.[9]  However, this estimate does not explain why registrations decreased. Moreover, it does not control for factors that influence voter registration rates, such as the passage and imple­mentation of welfare reform in 1996.

Other possible explanations for the decline include voter registration drives by community mobilization organizations, which reduced the need for welfare recipients to register to vote at public assistance offices, and welfare reform, which reduced the number of welfare recipients.

The analysis presented in this report directly tests the hypothesis that the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996 contributed to the decline in public assistance voter registrations. PRWORA replaced Aid to Fami­lies with Dependent Children (AFDC) with Tempo­rary Assistance for Needy Families (TANF). Research by Professors June E. O'Neill and M. Anne Hill of Baruch College strongly suggests that welfare reform accounts for more than half of the decline in AFDC/ TANF participation by single mothers during the 1990s.[10]  Welfare reform led to a substantial decrease in welfare caseloads, which in turn may have led to fewer voters registering at public assistance offices.

Chart 1 plots the trends in average AFDC/TANF participation and the average number of voter reg­istrations at public assistance offices in the states from 1995 to 2006. The decline in voter registra­tions closely follows the decline in AFDC/TANF participation. While the association between wel­fare caseloads and voter registrations seems obvious, other factors that might explain the relationship were also tested.

The Data and Modeling

To check for other possible expla­nations for the decline in voter regis­trations, a state-level panel data set of public assistance registrations, wel­fare participation rates, socioeco­nomic factors, and political election cycles was constructed. Using panel analysis allows this study to test the relative influence of varying AFDC/ TANF participation rates on the number of voter registrations while controlling for other factors that might influence registrations.

The data set contains 12 years of data for 45 states and the District of Columbia. During the time frame of this analysis, several states either failed to report voter registration or were not required to do so. Six states did not report any data during the time frame of the analysis, while 11 states reported public assistance registrations inter­mittently.[11]  The data set is an unbalanced panel because of incomplete voter registration reporting by some states in certain years.

Methodology. Most research on NVRA public assistance voter registrations is filled with assertions about what is or is not responsible for reduced voter registrations at public assistance offices. Many of these assertions are based only on anecdotal evi­dence or descriptive studies that often lack empiri­cal research techniques.

Observing that public assistance voter registra­tions decreased between two time periods does not explain why registrations decreased. Studies relying only on descriptive statistics do not allow research­ers to test competing theories or hypotheses for why registrations decreased because the descriptive methods cannot control for other factors that could influence the observations.

By comparison, the statistical approach used in this report includes control variables and allows for the inclusion of many cases in order to test compet­ing hypotheses.

Regression Analysis. This reportcontains the results of a panel regression analysis of state-level data. The statistical tests used in the regression anal­ysis were conducted to isolate the independent effects of a number of policy, demographic, and socioeconomic factors on public assistance voter registrations in order to explain changes in public assistance voter registrations. Because the regression models in this study include such factors as minor­ity population percentage, unemployment rates, and welfare caseloads, the effect of each variable on public assistance voter registrations can be isolated.

A finding of "statistically insignificant" indicates that the effect of a particular variable is statistically no different from zero. For example, if the relationship between food stamp participants and public assis­tance voter registrations is statistically insignificant, the number of food stamp participants in the states, when combined with other variables, can­not be used to explain changes in public assistance voter registrations.

This analysis uses the 95 percent confidence level as the minimum stan­dard for statistical significance. When a variable is statistically significant at the 95 percent confidence level, there is a 5 percent chance that the variable has no statistically measurable impact on the dependent variable.

Panel Data Analysis. This report also uses panel data analysis. Panel data studies observe multiple units over several periods. The addition of multiple data collection points gives the results of regression analyses using panel data substantially more credibility than studies that use only descrip­tive statistics. By increasing the number of data points compared to simple descriptive studies that only calculate the change in the variable of interest between two time periods, panel data analysis has three important advantages.

First, the longitudinal nature of the panel data allows evaluators to analyze important policy ques­tions that studies using descriptive statistics or cross-sectional and time-series data sets cannot address. The previous research by Project Vote and Demos failed entirely to account for important pol­icy and socioeconomic factors that vary across states and over time and that might affect registra­tion rates.

Second, by increasing the number of data points compared to cross-sectional and time-series analy­ses, panel analysis increases the degrees of freedom and reduces possible collinearity problems among the independent variables, thus improving the effi­ciency of the econometric estimates.

Third, the panel data technique used in the anal­ysis reduces omitted variable bias by introducing state (cross-sectional) fixed effects into the model specification.[12]  By controlling for state fixed effects (individual differences related to each state), the analysis accounts for time-invariant unobserved factors that influence public assistance registration rates in a particular state. The fixed-effects model helps to control for differences in registration rates that are not explained by the independent variables.

Variables. For this analysis, the dependent vari­able is the number of public assistance registrations per 100,000 residents age 18 or over.[13]  The inde­pendent variables are AFDC/TANF recipients per 100,000 residents; food stamp participants per 100,000 residents; Women, Infants and Children (WIC) participants per 100,000 residents; income per capita; unemployment rate; minority popula­tion percent; 18 and older population percent; presidential elections; U.S. Senate elections; guber­natorial elections; off-year congressional elections; and state fixed effects.[14]  Table 1 presents the means and standard deviations for these variables presented.

The independent variables were chosen based on their anticipated influence on public assistance regis­trations. For example, AFDC/TANF, food stamp, and WIC participation rates measure the level of welfare recipients being served by public assistance offices. Increased welfare participation is anticipated to be pos­itively associated with public assis­tance registrations.

State unemployment rates and income per capita help to control for the influence of the economy. Unem­ployment is an especially important variable because it is highly likely that the sharp decline in unemployment during the 1990s reduced welfare par­ticipation. Professors O'Neill and Hill assert that "[t]he true effect of welfare reform cannot be determined without accounting for changes in unemploy­ment and other possible factors affect­ing single mothers' choices."[15]  If decreased unemployment is partially responsible for the decline in AFDC/ TANF participation, then it follows that decreased unemployment would lead to fewer public assistance regis­trations. In addition, the election vari­ables help to control for periods of increased political activity that are also anticipated to be positively associated with public assistance registrations.

Findings
Table 2 presents the findings of Ordinary Least Squares (OLS) panel regressions. Four models were esti­mated, and the regressions are weighted by state population. The standard errors are robust to het­eroskedasticity and autocorrelation.[16] 

After controlling for other factors in Model 1, AFDC/TANF participation has a statistically signif­icant association with public assistance registra­tions. A one-unit increase in AFDC/TANF participants per 100,000 residents is associated with an increase of 0.062 additional registrations per 100,000 adult residents. Another way to inter­pret this finding is to calculate the elasticity. The elasticity represents the percentage change in pub­lic assistance registration rates given a 1 percent change in a particular independent variable. A 1 percent increase in AFDC/TANF participation is associated with a 0.49 percent increase in voter reg­istrations. Conversely, a 1 percent decrease in AFDC/TANF participation is associated with a 0.49 percent decline in voter registrations.

Food stamp and WIC participation do not appear to have any statistically measurable association with public assistance registrations. The results for in­come per capita, unemployment, and the adult pop­ulation percentage are also statistically insignificant.

A state's minority population percentage has a statistically significant and negative relationship with public assistance registrations. A 1 percent increase in the minority population is associated with a reduction of 12.6 registrations per 100,000 adults. Further, a 1 percent increase in the minority population is associated with a 1.1 percent decrease in registrations.

For the election cycle variables, presidential and gubernatorial election years have statistically signif­icant and positive associations with public assis­tance registrations. Registrations increased by 97.4 per 100,000 adults during presidential election years and by 48.8 per 100,000 adults during gubernatorial election years. The elasticity calcula­tions for the election year variables represent the percentage change in registrations during a partic­ular type of election year. The registration rate increased by 0.08 percent during presidential election years and by 0.04 percent during guber­natorial election years. Senate and off-year congres­sional elections appear to have no statistically measurable influence on registrations.

Additional regressions were estimated for Models 2, 3, and 4. Model 2 presents an analysis of data from 1997 to 2006, because the 1995-1996 public assistance registration data may drastically overstate the number of registrations that can reasonably be expected from public assistance offices. During 1995-1996, the debate over welfare reform was at its peak. The political debate likely led opponents of reform to encourage welfare recipients to register to vote in an attempt to influence the policy pro­cess. Average state public assistance registrations dropped 54 percent, from 115,177 in 1995-1996 to 53,552 in 1997-1998. In terms of raw magni­tude, this average decline of 61,625 registrations is the largest drop since the registration data have been collected. However, research by Demos, ACORN, and Project Vote ignores the fact that the largest drop in public assistance voter registrations occurred in 1997-1998 and instead focuses on comparing the initial 1995-1996 reporting period to 2001-2002 and subsequent reporting periods.[17] 

When the data are limited to 1997 to 2006 in Model 2, the coefficient for AFDC/TANF partici­pants remains positive and statistically significant. A one-unit increase in AFDC/TANF participants per 100,000 residents is associated with an increase of 0.062 additional registrations per 100,000 adult residents. A 1 percent increase in AFDC/TANF par­ticipation is associated with a 0.51 percent increase in voter registrations. Differing from Model 1, the unemployment rate is positively associated with increased public assistance voter registrations. A 1 percent increase in the unemployment rate is asso­ciated with 18.3 additional registrations per 100,000 adults. For the elasticity, a 1 percent increase in the unemployment rate is associated with a 0.36 increase in the registration rate.

A state's minority population percentage has a statistically significant yet smaller negative relation­ship with public assistance registrations. A 1 per­cent increase in the minority population is associated with a reduction of 7.2 registrations per 100,000 adults. Further, a 1 percent increase in the minority population is associated with a 0.82 per­cent decrease in the registration rate. The coeffi­cients for the election cycle variables and the other independent variables in Model 2 were not statisti­cally distinguishable from zero.

For Model 3, the regression analyzed data from all years, while individual time-period dummy vari­ables were introduced for the 1997-1998 to 2005- 2006 periods. These time-period variables control for differences in reported public assistance registra­tions between the first reporting period (1995- 1996) and later reporting periods. In this model, the coefficient for AFDC/TANF participation is sta­tistically insignificant, while the time-period dummy variables are statistically significant. Only the coefficients for income per capita and off-year election variables were statistically significant.

A $1 increase in income per capita is associated with a 0.034 increase in public assistance voter registrations per 100,000 adults. For the elasticity, a 1 percent increase in income per capita is associ­ated with a 2.6 percent increase in the registration rate. During off-year congressional election years, public assistance voter registrations declined by 91.6 registrations per 100,000 adults or decreased by 0.14 percent.

The regression for Model 4 used the same vari­ables that were used in Model 3, but the data were limited to the years of 1997 to 2006. The coefficient for AFDC/TANF participation is statistically signifi­cant, while the time-period dummy variable coeffi­cients were not statistically distinguishable from zero. This result for the time-period dummy vari­ables strongly indicates that the reporting of public assistance registrations was unusually high in the 1995-1996 period compared to later reporting periods. Despite the inclusion of time-period fixed effects, the results of Model 4 are remarkably similar to the results of Model 2.

A one-unit increase in AFDC/TANF participants per 100,000 residents is associated with an increase of 0.061 additional registrations per 100,000 adult residents. For the elasticity, a 1 percent increase in the AFDC/TANF participation rate is associated with a 0.5 percent increase in the registration rate.

A state's minority population percentage has a sta­tistically significant yet smaller negative relationship with public assistance registrations. A 1 percent increase in the minority population is associated with a reduction of 7.2 registrations per 100,000 adults. Further, a 1 percent increase in the minority population is associated with a 0.82 percent decrease in the registration rate. The coefficients for the remaining independent variables in Model 4 were not statistically distinguishable from zero.

Across the four models in Table 2, the coefficient for AFDC/TANF participation was statistically sig­nificant in three of the four models. Food stamp and WIC participation had statistically insignificant relationships with public assistance voter registra­tions in all four models.

Changes in AFDC/TANF caseloads appear to have contributed substantially to the decline in public assistance voter registrations.

Conclusion

Declining AFDC/TANF caseloads from 1996 to 2006 contributed substantially to the decline in public assistance voter registrations. Unlike previ­ous research, this report uses panel regression anal­ysis to estimate the relationship between AFDC/ TANF participation and other factors that influence public assistance registrations. Controlling for other factors, a 1 percent decrease in AFDC/TANF partic­ipation is associated with about a 0.5 percent decrease in public assistance registrations.

While voter registrations at welfare offices have declined, this decline does not mean that former welfare recipients are not registering to vote. Low-income Americans have numerous and easy oppor­tunities to register, just like other Americans. For years, Americans have had the opportunity to reg­ister to vote through the mail, at motor vehicle offices, and at other locations. In addition, many "voting rights" and "community mobilization" groups, along with political parties, are actively engaged in making sure that their constituents are registered to vote.

While research on this topic is new and further analysis is needed, Members of Congress, policy­makers, and the media should not dismiss the major role that welfare reform and decreased wel­fare participation have played in reducing public assistance voter registrations.

David B. Muhlhausen, Ph.D. , is a Senior Policy Analyst and Patrick Tyrrell is a Research Assistant in the Center for Data Analysis at The Heritage Foundation.