Common wisdom dictates that it is easier for governments to provide new social programs than to pare back existing ones. This wisdom rests on decades of research showing that policies can create engaged voter-citizens from among their recipients, who will mobilize to protect new programs (Campbell 2003). There are two mechanisms by which social policies might accomplish this: through the resource mechanism, social programs can increase low-income citizens’ financial resources and thereby their likelihood of voting; and through the interpretive mechanism, social programs serve as cites of “adult political learning” that teach recipients lessons about their government and their own political efficacy (Soss 2000). According to these theoretical accounts, a social program’s design—both its generosity and what it tells recipients about their relationship to government—determines whether it will mobilize recipients.
But in practice, it is often difficult to distinguish between the effect of program design and differences in beneficiary populations that also affect their political participation—for example, the least generous programs often target the poorest groups. Even studies that have been able to isolate a policy’s effect on political engagement have not been able to provide much insight into which mechanism led a policy to generate the observed effects.
In an article forthcoming from the Journal of Politics, I address these questions through the case of Old-Age Assistance (OAA) expansion as part of the Social Security Act of 1935. OAA was a means-tested public assistance program that provided cash payments to destitute elderly applicants. Based on the existing literature, expectations about its effects on political participation are unclear: it provided substantial benefits to recipients, which may have increased participation through the resource mechanism, but also subjected those recipients to extensive and stigmatizing scrutiny by caseworkers, which may have decreased participation through the interpretive mechanism.
Leveraging state-by-state variation in how states chose to implement OAA programs, I find that a certain kind of OAA expansion—increasing the generosity of payments—increased voter turnout, while another kind—increasing the percentage of people covered by the program—did not. I argue that these findings suggest the resource mechanism played a crucial role in mobilizing elderly voters in the wake of OAA expansion, sufficient to overcome what were likely substantial and negative interpretive effects.
Old-Age Assistance
OAA was funded through a federal-state matching program that began its first payments on February 11, 1936. To receive the federal match, states had to gain approval for their plans from the Social Security Administration. Although many states had flagging programs to subsidize the elderly prior to 1936, this influx of federal funds spurred on a rapid expansion in the generosity and coverage of existing programs after 1936, as well as expansion to new states that had previously been unable or unwilling to provide any benefits. Total payments for old-age assistance programs skyrocketed, as shown in Figure 1 below.
OAA was the most predominant form of old-age protection in the U.S. from the time of its initiation in 1936 until Old-Age Insurance (OAI) payments eclipsed it in the 1950s (Amenta, Caren, and Olasky 2005). During the New Deal years it was generous enough to make a material improvement in recipients’ lives: it provided enough funds to lift people out of poverty, resulting in less elderly people working late in life (Fetter and Lockwood 2018) and lower mortality rates among the elderly (Cohen 2007).
But despite its generosity, OAA was a program that was designed to be uncomfortable, and even stigmatizing, for recipients. OAA was meant to be a stopgap measure that provided for the elderly until the earliest benefits from OAI were available. Fearing that it might become too popular in the interim, policymakers chose to model it off of unpopular and antiquated poor laws that subjected applicants to the maximum degree of scrutiny (Cates 1983). As a result, OAA continued to carry the “stigma of charity,” as one contemporary described it (Mettler 1998).
Research Design and Findings
Every state was operating a plan approved by the Social Security Board by 1938, but substantial variation in these plans persisted because states were often faced with trade-offs between extending coverage to more people or increasing their maximum payments to existing recipients (Parker 1936). Although the trade-offs between expanding coverage and increasing generosity were not one-to-one, the fact that states could pull either or both levers in designing their policies is what allows me to gain some insight into the two different mechanisms by which this means-tested program might have affected political participation.
In this vein, I define the treatment in two ways, both using data on states’ annual payments from Social Security Administration reports and publications from the Bureau of Labor Statistics’ Monthly Labor Review: 1) a measure of generosity—each state’s average annual OAA payment per recipient in each year, driven largely by their plan’s maximum monthly payment cap; and 2) a measure of coverage—the portion of the population 65-and-over that was actually receiving OAA benefits in a given year, driven largely by the state’s residency requirements and income and asset limitations. Relative to the generosity measure, the coverage measure should capture a larger role for interpretative rather than resource effects, and vice versa.
Figures 2 and 3 map both treatments by state in 1936: Figure 2 shows each state’s coverage rate and Figure 3 shows the generosity of each state’s plan.
Using these two treatment measures, I compare voter turnout in counties in states with less generous benefits/lower coverage rates to those in states with more generous benefits/higher coverage rates. I also allow the treatment effect to vary according to how many elderly residents were living in the county, as they were the beneficiary population. Counties were coded 1 (“elderly”) if their average elderly population was above the median for Census years 1930 and 1940, and 0 (“young”) otherwise.
Tables 1 and 2 presents the results of these analyses. In Table 1, the treatment measure is the state’s average annual OAA payment per recipient; in Table 2 the treatment is the coverage rate.
In both tables, Model 1 includes only the treatment and the interaction term between the treatment and the elderly variable, as well as the county and year fixed effects. The 95% confidence intervals are shown in parentheses below the coefficient estimates. For each treatment measure, the estimate for the interaction term is in the expected direction—positive—but not statistically distinguishable from zero. Model 2 includes a set of covariates for each county from the U.S. Census, which vary over time: the county’s population density as well as the percent of the population that is Black; unemployed; living on a farm; female; or foreign-born. Including these covariates increases the precision of the estimates in both tables, and the interaction term in Model 2 of Table 1 is now both positive and statistically significant. This conforms to a priori expectations.
Finally, Model 3 in Tables 1 and 2 includes a south-by-year fixed effect, which controls for any shocks specific to region and year that may have affected voter turnout. Given the turmoil in the former Confederacy’s relationship to President Franklin D. Roosevelt during the New Deal, this is my preferred model. The results remain the same: OAA expansion increased political participation in elderly counties; however, it only did so when states concentrated resources on increasing the generosity of their payments rather than making the program more accessible to elderly residents.
And while the effect sizes in Table 1 may seem small, they are substantively significant. The average increase in payments per recipient between presidential elections was $84.42 per person per annum. Based on the estimates from Model 3 in Table 1, the effect of this increase was to increase turnout in elderly counties by 0.93 percentage points.
These county-level results cannot conclusively determine that it was elderly OAA recipients whose turnout increased after program expansion; it is also possible that some increased turnout occurred because recipients’ adult children and caretakers also benefited from their parents’ greater independence. To provide further evidence of the individual-level mechanisms at work, Figure 4 summarizes responses to three Gallup poll surveys conducted in the months following the 1940 presidential election. In each survey, respondents were asked if they voted in the 1936 and 1940 presidential elections. The graphs compare the recalled voter turnout in both years among those between 25–65 years old, those 65 and older who are not receiving OAA, and those elderly who are receiving OAA.[1]
Across all three surveys in 1936, OAA recipients reported lower turnout than did elderly non-recipients and younger voters, suggesting that elderly people who were eligible for OAA had fewer resources and were less likely to vote than elderly non-beneficiaries. But by 1940, every survey indicates that OAA recipients closed the turnout gap with elderly non-recipients, indicating that OAA expansion between 1936 and 1940 may have reduced resource deficits and increased turnout among indigent elderly citizens.
Conclusion
OAA was a deliberately stigmatizing, means-tested social program. According to much of the literature emphasizing the deleterious effects of such programs on political participation, it is somewhat surprising that it was still able to generate increased voter turnout in elderly counties. Given that these effects were generated only when states favored more generous rather than more extensive programs, OAA was likely able to overcome its negative interpretive effects by offering sufficiently generous benefits to recipients. This is not to say that the interpretive mechanism is not an important factor of policy design. Rather, it suggests that even in the case of programs that teach recipients negative lessons about their political efficacy, generous benefits can still motivate political engagement.
[1] Although voter turnout data in surveys are substantially overreported, this would only affect the between-group comparisons if voters in different groups overreport to different degrees. It would only affect the within-group comparisons if voters within each group overreport their two most recent votes to different degrees.
