Changes and Continuities in the Determinants of Older Adults' Voter Turnout 1952–1996

Abstract

Purpose: The goal of this project was to examine the factors related to the high level of voter turnout among older adults and how these factors have changed across the past 50 years. The study builds on previous research efforts by combining individual level analyses from three nationally representative surveys. Design and Methods: We examined voter turnout among older citizens using the American National Election Studies (ANES) 1952–1996, the Current Population Studies (CPS) 1972–1996, and the General Social Surveys (GSS) 1972–1996. Logit regression identified significant factors that contribute to or detract from voting in presidential elections. Results: Included in the extensive results, we found that being married, attending church, and contact by political operatives (since the 1980 election) contributed positively to older voters' turnout, whereas living in the South was a negative predictor. Implications: Triangulating data sources, we are able to utilize the strengths of each study—providing an overview of the determinants of voter turnout and understanding of the changes related to older Americans' turnout in presidential elections.

Voting behaviors, Presidential elections, Political participation

Decision Editor: Laurence G. Branch, PhD

The increasing number of older adults relative to younger adults participating in the political arena makes the elderly population an ever-growing important voting block, as witnessed in the many appeals that presidential candidates made to older voters during the 2000 election. As the number of elders grows in American society, it becomes increasingly important that we better understand their patterns of voter turnout and the factors contributing to their involvement in electoral politics (Rollenhagen 1984). Indeed, the political rhetoric of today clearly suggests that older voters are essential for either party to win an election. The candidates' focus on Medicare and Social Security indicates politicians' beliefs that they must carry the "senior" vote to win any election (Alwin 1998; Bazargan, Barbre, and Torres-Gil 1992; Binstock 1997b, Binstock 1983, Binstock 2000; Cornman and Kingson 1996; Jirovec and Erich 1995; Williamson 1998). Yet, we have limited understanding about the factors causing the comparatively high levels of voter turnout among older voters. We know even less about how (or even if) these factors have changed over time or why turnout among older adults rises and falls in different presidential election cycles. In the present study, we begin to answer these questions by combining analyses of voter turnout from three nationally representative surveys extending over a time period of more than 50 years of presidential elections. Our goal was to explain the factors related to voter turnout of older adults as well as to examine how (or if) these factors have changed over time.

Importance of the "Senior Vote"

Binstock 2000 suggests three primary reasons why political leaders focus so much attention on issues they believe to be important to older adults. First, those over 65 years of age compose a "substantial proportion of voting-age Americans today" (p. 18). This relative percentage will continue to grow throughout the first half of this century. Secondly, older adults have voted at higher rates than other age groups in the 1970s, the 1980s, and the 1990s. In fact, the voting rate for younger cohorts has declined, whereas that of older adults has increased slightly (Miller and Shanks 1996). The final factor is the assumption that politicians can control older voters by focusing on "old-age policies" and that they cannot win without the "senior vote." Torres-Gil 1992 concurs that these factors have increased political notice of older voters. In addition, he contends that the recognition of older adults as an "identity group" and the growing numbers of age-based organizations that have developed in recent years contribute significantly to the attention paid to potential older voters in recent elections. In coming presidential elections, because of the continued growth of the older population, basic issues related to economic and policy decisions such as long-term care, the solvency of Social Security and Medicare, intergenerational transfers, health insurance coverage, and support for caregivers of older adults will become even more prevalent in political debate and in the elections themselves. Politicians' efforts to win the senior vote without alienating younger voters will become even more intense. The rationale many politicians, journalists, academicians, and others use to emphasize the importance of the senior voting block is explained through what Binstock 1997b, Binstock 2000 calls the Senior Power Model.

The Senior Power Model

The basis for this framework is the continuing numeric and proportional growth of individuals over the age of 65 years (Binstock 2000; Cornman and Kingson 1996). In addition, many political operators assume that all older adults are more politically active than their younger counterparts (Bazargan, Kang, and Bazargan 1991; Hudson 1999; Quadagno 1989). Further, those guiding political campaigns believe that older voters are only motivated to vote by issues directly impacting their lives, such as Medicare, Social Security, and such (Binstock 1997a, Binstock 1997b, Binstock 2000; Hudson 1999; Quadagno 1989). These assumptions lead to the conclusion that older adults are a homogeneous, self-focused group wielding tremendous political power. Thus, with each local and national election, politicians focus a great deal of time and energy on issues they believe will guarantee them the senior voting block.

These assumptions have contributed to fears among younger adults, resulting in recommendations derived from what some have called intergenerational warfare. Fear of a loss of generational equity has brought forth terms such as "greedy geezers" (Fairlie 1988). Smith 1992 suggested that "the tyranny of America's old" is "one of the most crucial issues facing U.S. society" (p. 68). Policy recommendations have included disenfranchising retired persons and adults over the age of 70; providing parents of minor children with an additional vote for each dependent child; and giving working taxpayers two votes compared to older, retired voters (Binstock 2000).

Binstock 1983, Binstock 1992, Binstock 1997a, Binstock 2000 and others (e.g., Jirovec and Erich 1995; Torres-Gil 1992) contend that many of the assumptions forming the very basis of the concept of "senior power" are erroneous. Although the relative proportion of older adults is indeed growing rapidly, the diversity of these elders prevents them from voting as a unified block. Further, empirical efforts have failed to support the assumption that the majority of older voters cast their votes based on their own self-interest (Binstock 1997a, Binstock 1997b). Nevertheless, one aspect of the senior power model does appear valid: the steady or growing rates of voter turnout among elders. As Fig. 1 demonstrates, over the past several Presidential election cycles, voter turnout rates of the older population, defined by the Federal Election Commission as people over 65 years of age, have grown slightly since the 1970s, whereas the turnout rates of other age groups have remained stagnant or declined (see Table Aa , Note 1).

A close examination of Fig. 1 suggests a possible cohort effect for the group of individuals 45 to 64 years of age in 1976 who have continued to vote as they aged past 65 years, a possibility proposed by Miller and Shanks 1996. In addition, the 45–64-year-old group remains the largest voting age group from 1976 through 1984. It remains at a consistent level until 1996, when it declines slightly. This observation might suggest that voting behavior is a developmental issue in that individuals move into a more active political role as they negotiate midlife transitions. Regardless of the cause, the figure clearly shows that over the last 20 years individuals 45 years of age and older have voted at a higher rate than those under the age of 45 years, and in more recent elections those 65 years of age and older have voted at a higher rate than any other age group.

Voting Behaviors

Because registration requirements and suffrage laws became more lenient in the decades since the 1950s (Table Aa , Note 2), and because of the overall increase in educational levels of the American electorate, many political scholars expected voter turnout to increase in the latter half of the 20th century. The overall decline in voter turnout rates from a high in 1960 of approximately 65% to the low of 49% in the 1996 presidential election, has produced much research attempting to explain this "puzzle of participation" (e.g., Brody 1978; Rosenstone and Hansen 1993; Teixeira 1992). Teixeira 1992 argues that socioeconomic and educational upgrading of the general population has actually prevented the turnout decline from falling even further. He argues that "a substantial decline in social connectedness, as manifested in a younger, less married, and less church-going electorate; and a generalized withdrawal from the political world, as manifested by declining psychological involvement in politics and a declining belief in government responsiveness" (p. 49) has contributed to over one third of the turnout decline witnessed since the 1960s.

Teixeira 1992 conclusions about declining voting rates resulting from social disconnectedness are theoretically consistent with Cumming and Henry 1961 social theory of aging called "disengagement theory." Cumming and Henry proposed that aging itself was both an individual and a social process. They contend that older adults' voting rates should be declining (as Teixeira 1992 suggests younger voting rates are) because adaptively aging adults should decrease rather than increase their involvement in all social and political arenas, including voting. The increasing voting rates of older adults, especially when controlling for differences in education and income (Rollenhagen 1984; Williamson 1998), add to empirical evidence (Hochschild 1975) that disengagement may not be appropriate as a theoretical basis upon which to understand the later years of life for a majority of older adults today.

Rosenstone and Hansen 1993 "mobilization theory" of voter participation contends that the main culprit for the decline in voter turnout in the later half of the 20th century was the steady decline of recruitment efforts from political parties and political candidates. Yet, voter turnout among older voters has remained relatively high in spite of the changes in the general electorate (MacManus 2000).

None of the explanations used to explain voter turnout among the general population seems to fully explain the voting rates of older adults. The number and variation of factors found to be catalysts or deterrents of voter turnout are vast (Verba, Schlozman, and Brady 1995). Our goal in this project was to map the individual level determinants of voter turnout among elders and to examine if and how these factors changed over time. In order to accomplish our goal, we identified existing nationally representative data sets that provided us with breadth and continuity across several presidential election cycles.

Theoretical Foundations

Growing interest among gerontologists in the political activity of older Americans has resulted in a spate of research investigating the causes of turnout among this growing segment of the American electorate. In fact, the growing importance of studying political activity among elders has given rise to a surprisingly sophisticated and rich literature uncovering the causes of voter turnout among this group of Americans. The theoretical framework for the vast majority of prior work on voter turnout has been the rational actor model, or the theory of rational choice (Downs 1957). Central to this framework is the recognition that the decision to vote, at least for most citizens, is a marginal (low cost/low benefit) action (Aldrich 1993; Jackman 1993). Consequently, answering the question of "who votes?" has generally focused on factors that increase or decrease the costs of voting (Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980). In their work contributing to the mobilization theory of political participation, Verba and colleagues 1995 argue that the factors increasing or decreasing those costs are divided into three broad categories: resources (ability to vote), psychology (desire to vote), and recruitment (were you asked to vote). Resources and demographic variables influencing turnout include income, education, state and region of residence, and being a minority (Rosenstone and Hansen 1993). Internal and external political efficacies are among the psychological factors reported to be influential in determining voting behaviors. Internal political efficacy refers to the degree to which potential voters feel capable of understanding politics. External efficacy is the degree to which they believe the political powers care about their personal attitudes and needs (Campbell, Converse, Miller, and Stokes 1960). Recruitment and mobilization may come through formal or informal networks at work, religious institutions, or political organizations. Rosenstone and Hansen 1993 take a similar approach, dividing the factors that affect participation into "individual influences" and "political influences." Individual influences are represented by personal costs and resources (the demands that participation makes on one's "money, time, skill, knowledge, and self-confidence"), which interact with personal rewards, interests, and beliefs regarding participation. Although these help predict which people participate, political influences—social networks and strategic mobilization efforts—help predict when they participate.

Age and Voter Turnout: Linear or Nonlinear Relationship?

Previous work has increased the understanding of the importance of these factors as determinants of turnout among older voters. Binstock 1983, Binstock 1997b, Binstock 2000 has written extensively on this subject and argues that while the older population may be voting at higher rates than other age or demographic groups, they are not the monolithic, homogeneous, self-interested voting block often suggested by the popular conventional wisdom expressed in the Senior Power Model. In addition, MacManus 1996, MacManus 2000 describes in detail the voter turnout rates of America's elders and proposes several possible reasons for their comparatively high rates of turnout (see also Alwin 1998; Glenn and Grimes 1968; MacManus and Tenpas 1998; Williamson 1998; Table Aa , Note 3). Several studies have argued that, once controls are made for socioeconomic factors such as the often lower educational levels of older citizens, the probability of voter turnout consistently increases with age up to the late 50s or early 60s, then continues to increase but at slower rates (Campbell 1971; Converse and Niemi 1971; Curtis and Lambert 1976; Glenn and Grimes 1968; Gubrium 1972; Jirovec and Erich 1992; Nie, Verba, and Kim 1974; Rosenstone and Hansen 1993; but see Jennings and Markus 1988; Strate, Parrish, Elder, and Ford 1989; Table Aa , Note 4).

In general, there are three main hypotheses accounting for the relationships between age and political participation. The life-experience hypothesis suggests that, as people age, they acquire resources and have learning experiences that promote participation (Rosenstone and Hansen 1993; Strate et al. 1989). Second, proponents of the life-cycle hypothesis argue that young citizens are less likely to vote because they lack the community involvement necessary to believe that politics is an important endeavor. In addition, the life-cycle hypothesis suggests that there will be a gradual decrease of political and social involvement among the oldest voters as physical limitations begin to increase and intensify (Cumming and Henry 1961; Milbrath and Goel 1977; Rosenstone and Hansen 1993). The disengagement social theory of aging (Cumming and Henry 1961) contends that such withdrawal is not only due to physical limitations but that it is required by society at large in order to support the well-being of society as well as that of older adults themselves. Disengagement theory contends that elders must step aside and remove themselves from social and political activity as they age so that the younger generation can take over the primary determining roles within the family and within society as a whole. Finally, proponents of the generational hypothesis suggest that socializing experiences influence each generation differently, and these generational experiences have helped explain the gradual decline in national voter turnout rates since the 1960s (Lyons and Alexander 2000; Miller 1992). For example, older women may vote less because they were socialized before the 19th amendment permitted women to vote (Firebaugh and Chen 1995). Similarly, young people today may vote less because they did not experience the politically turbulent events of the Great Depression or the World Wars (Miller 1992; Miller and Shanks 1996).

Although each perspective has some face validity, Rosenstone and Hansen have recently presented evidence that supports the life-experience hypothesis. As they argue (1993, p. 140)

[our] evidence strongly favors the life-experience hypotheses… the relationship between age and electoral participation definitely persists even when life-cycle and cohort effects have been taken into account … consistent with the life-experience explanation, participation in electoral politics increases throughout life… Correspondingly, the findings offer little support for either the generational or the life-cycle hypotheses (emphasis added).

Further evidence comes from gerontological studies examining voter turnout exclusively among older adults. For example, according to Bazargan and colleagues 1991(p. 18) "Although middle-aged citizens appear to vote at higher rates than older adults, once other demographic variables have been held constant, aging itself produces not a decline but an increase in turnout. The rate of increase in voting begins to level off at around age fifty-five, but turnout continues to rise. This rise in voter turnout, however, occurs at an increasingly slower pace through the seventies." Further, Wolfinger and Rosenstone 1980 suggest that once education, gender, and marital status have been held constant, the rate of turnout actually increases with age (see also, Campbell et al. 1960; Glenn and Grimes 1968). Conversely, Bazargan and coworkers 1991 find that, given controls for socioeconomic factors, age was not even a significant predictor of turnout—at least among their comparatively small sample of older women.

Decline in Turnout Since the Late 1950s

Turnout has been found to be strongly and positively related to the resources of education and income, as well as to employment, which can represent connections to society, financial rewards, and recruitment networks (Cassel and Hill 1981; Leighley and Nagler 1992a, Leighley and Nagler 1992b; Miller and Shanks 1996; Shaffer 1981; Teixeira 1987; Wolfinger and Rosenstone 1980). Psychological factors are also clearly important in the decision to vote. In particular, the decline in voter turnout witnessed since the 1960s has been attributed, in large part, to declining levels of political interest and efficacy (Teixeira 1987, Teixeira 1992). These factors have also been found to be predictors of turnout among the older population, along with measures of partisan affiliation and trust in government (Bazargan et al. 1992; Firebaugh and Chen 1995; Glenn and Grimes 1968; Kam, Cheung, Chan, and Leung 1999; Peterson and Somit 1994). Employment and marital status also predict voter turnout, which may reflect a combination of resource, recruitment, and psychological factors (Leighley and Nagler 1992a, Leighley and Nagler 1992b; Miller and Shanks 1996; Stoker and Jennings 1995; Teixeira 1987; Wolfinger and Rosenstone 1980). Bazargan and colleagues 1991 and Peterson and Somit 1992, 1994) demonstrate the importance of marriage as a catalyst of turnout among older voters by showing that older adults were less likely to vote once they became single, either through death or divorce.

Despite the benefits, both material and psychological, associated with being employed (Quadagno and Hardy 1996), few studies have examined the impact of employment on the voter turnout of older adults. Nevertheless, several studies have identified increased education and church attendance as important predictors of older voter turnout (Bazargan et al. 1991, Bazargan et al. 1992; Jirovec and Erich 1995; Peterson and Somit 1992, Peterson and Somit 1994). Some studies of older voter turnout suggest that there are significant differences between men and women (Curtis and Lambert 1976; Hudson and Gonyea 1990; Jirovec and Erich 1995; Peterson and Somit 1992, Peterson and Somit 1994; Nie et al. 1974) as well as between Caucasians and African Americans (Bazargan et al. 1991, Bazargan et al. 1992; Peterson and Somit 1992, Peterson and Somit 1994; Powell and Thorson 1990). General studies of voter turnout have long found significantly lower probabilities of voter turnout among residents of the southern states (Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980), but such regional variations have rarely been included (Lyons and Alexander 2000) in prior analyses of older voter turnout. Several studies have investigated the importance of subjective evaluations of personal health status but have reached different conclusions. For example, Peterson and Somit 1994 find mixed evidence of the impact of health on a number of attitudinal and behavioral measures. Similarly, Kam and associates 1999 find mixed results for various measures of health status on an index of political participation. In addition, Jirovec and Erich 1992 find no significant impact of health status on voting, given controls for other variables such as income and education. Bazargan and colleagues 1992 find evidence that the zero-order correlation between health status and voter turnout among older African Americans may become insignificant after controlling for other socioeconomic factors. Nevertheless, Bazargan and coworkers 1991 found that health status was a significant predictor of turnout among Caucasians.

Finally, despite the importance of the mobilization theory of voter turnout, few studies have examined the role of elite mobilization efforts on the turnout of older people. Kam and colleagues 1999 find that candidate and party recruitment are significant factors in the political participation of older adults living in Hong Kong. Further, Hudson and Gonyea 1990 argue that recent changes in several senior political organizations indicate that political groups have begun to realize the importance of mobilizing and addressing the specific needs of older women. Finally, MacManus 1996 reports evidence that candidates and political parties strategically contact "likely" voters, such as the wealthy, educated, and elderly (Lyons and Alexander 2000; Rosenstone and Hansen 1993). Whether or not candidate or party contact has had a significant impact on the voter turnout of American elders over time, or if there is a significant impact controlling for other relevant factors, remains to be seen. In this regard, nearly all prior research examining voter turnout among older adults has been based on small samples or single cross-sectional surveys. The extent to which the findings hold up across election years and data sets, if the impacts persist or change, or if the influence disappears given the addition of new controls introduced in this study (such as party recruitment and living in the South), are open questions.

Methods

The Current Population Survey

The Current Population Survey (CPS; 1972–1996) involves a monthly household telephone survey conducted by the Bureau of the Census for the Bureau of Labor Statistics. A representative sample of the civilian noninstitutionalized American population, it provides a comprehensive body of information on the employment and unemployment experiences of the U.S. population, classified by age, sex, race, and a variety of other characteristics. The CPS is a large-scale survey for which the data are collected by face-to-face and telephone interviews of more than 60,000 households. The CPS contains only measures of respondents' socioeconomic backgrounds such as education and income levels. Following every national election, the CPS contains a "voter supplement" that also asks respondents if they voted in the presidential election. The strength of the CPS is the sample size, which includes thousands of older voters in every presidential election from 1972 through 1996.

The American National Election Studies

The American National Election Studies (ANES; 1952–1996) are national surveys conducted by the Center for Political Studies of the Institute for Social Research at the University of Michigan. Data collection began in 1952 and continues to the present. The surveys are based on multistage representative cross-sectional samples of citizens of voting age, living in private households. Each study contains information from interviews conducted with 1,000 to 2,000 respondents. The samples are representative of the four major regions (Northeast, North Central, South, and West) of the contiguous United States as defined by the Census Bureau. The strength of the ANES is the wide range of politically relevant variables for predicting voter turnout. While the ANES does not include measures of physical health, it does include political attitudes such as whether or not one believes the presidential race will be close; whether one cares about the outcome of the presidential race; an individual's overall interest in politics; retrospective evaluations of personal financial situations; whether or not political parties or candidates attempted to contact them during the campaign; and measures of both internal and external political efficacy.

The General Social Survey

The General Social Survey (GSS; 1972–1996) is an almost annual personal interview survey of U.S. households conducted by the National Opinion Research Center. The first survey took place in 1972. Although the GSS does not include as wide a range of politically relevant variables as the ANES, it does contain substantially more measures of political attitudes and positions than the CPS. In addition, the GSS is the only survey to include measures of subjective well-being. In fact, the GSS has consistently asked respondents questions about their current health status. However, the GSS has the drawback that respondents are sometimes not asked about voting in previous elections until several months after the election. Such time delays increase the potential for misreporting actual voting behavior. Nevertheless, some studies investigating the potential for biases when using reported measures of voter turnout rather than validated measures have found that while misreporting is a problem, it does not bias results dramatically. In fact, Katosh and Traugott 1981 and Sigelman 1982 argue that nearly all inferences based on reported voting behavior remain true when validated vote data are used. While the GSS has a substantially smaller sample size, the addition of some variables unavailable in the other data sets, specifically measures of health status, provide more certainty concerning the impact of other factors. While variables such as education, family income, and regional residence may play important roles in a purely demographic model, the findings may be misleading as the demographic variables may provide proxies for attitudinal differences across geographical areas. For example, a negative impact of living in the South could be the result of fewer people living in this area who maintain strong affiliations with the political parties rather than something more generally associated with the physical location.

Analysis

The analysis procedures for this study were limited to those in each data set over the age of 55 years (see Table Aa , Note 5). We begin the analysis by examining basic demographic changes among the older population over time. We constructed basic frequency tables to allow us to examine what demographic changes experienced by this group may be related to changes in the probabilities of voter turnout. For example, if older Americans are experiencing greater levels of family income, an outcome the Senior Power Model might lead us to believe, then the increasing levels of family income may not only increase the probability of voter turnout for the group overall, but also may have become a more important individual level predictor of turnout over time. Of course, the opposite could be true if older Americans are experiencing decreasing standards of living. In order to identify the individual level determinants of voter turnout, and if the impact of these factors changed over time, we estimated encompassing models of voter turnout including controls for all variables identified in prior research as important predictors of voter turnout (contingent upon availability across data sets). Because the dependent variable in our analysis, voter turnout, is dichotomous, we use logistic regression to conduct our multivariate analysis. Furthermore, we analyze each election year separately because pooling data across years requires strong assumptions "that the effects of demographic characteristics on turnout do not change from year to year" (Leighley and Nagler 1992b, p. 725). If the impact of various factors does change from year to year, which is one of the primary questions raised in this study, then pooling the data set will result in biased and inconsistent estimates (Johnston 1984). For more in-depth discussions of the hazards of pooling data across presidential election years, see Green 2000 and Table Aa , Note 6.

Demographic Changes

In Table 1 and Table 2 we see that the senior population (defined here as those over 55 years of age) has experienced several changes relevant to voter turnout across the last several decades.

According to both the CPS and the ANES, the number of elders living in the South has slightly increased, with just under one third in the CPS and slightly more than one third of all the elders in the ANES living in the southern part of the United States. According to the ANES, the number of older adults living in the South was at a peak during the 1960s and 1970s, when the percentage reached 38.6% during the election of 1976. There appears to be a slight increase in income across time as well as a consistent increase in educational attainment. The number of older African Americans in the electorate has increased slightly since the 1950s, but has remained relatively stable since the 1970s. Marriage rates have also remained relatively stable among the older electorate, and they reached their highest levels during the late 1950s and 1960s. The average age of the older population has grown consistently but only slightly from the mean of approximately 66.3 in the 1972 election to 68.2 in the 1996 presidential election (according to the CPS data). Perhaps the most dramatic changes have occurred in the levels of efficacy, voters' sense of their ability to make a difference or to understand the political process, among older voters and the number of elders who were contacted by a political candidate or party. Looking first at internal political efficacy, voters' belief that politics was not too complicated for them to understand, we see that the levels appear to change dramatically from election to election. For example, internal efficacy reached its lowest level in the 1996 presidential election, when only 34% of older voters displayed a sense of internal efficacy. Comparatively, in the election of 1984, 61% of older voters believed that they were able to understand political life. Similarly, levels of external political efficacy, the general belief among potential older voters that government is responsive to their needs and desires, have varied greatly, ranging from a high of 38% in 1960 to a low of 15% in 1980. It is interesting to note that the percentages of the older population with high feelings of external efficacy are relatively small, reflecting the general belief among the older population that government is not very responsive to their needs and desires.

Results

As shown in Table 3 , the analysis of the CPS data sets reveals some interesting patterns. First, there is a significant, consistent, and negative coefficient associated with living in the South. Despite controls for other socioeconomic factors, southern elders had a significantly lower probability of voting in presidential elections between 1972 and 1996. Nevertheless, the size of this coefficient is steadily decreasing from –.58 in 1972 to a low of –.27 in 1996. Being a participant in the labor force and increased levels of educational attainment were associated with significantly greater probabilities of turnout—although the marginal impact of being in the labor force appears to be declining from a high of .41 in 1972 to a low of .15 in 1996. Increased levels of family income, as well as additional increases in age, were associated with an increased probability of voting. Although the influence of age appears relatively stable, the marginal effect of income indicates an increasing impact from a low of .05 in 1976 to a high of about .10 in 1988. Further, while the predicted probability of voting increases with age, the increase significantly attenuates at the oldest age levels (represented by the significant and negative age-squared coefficients). Once controls are made for the generally lower socioeconomic standing (e.g., lower educational attainment, lower income levels, etc.), older African Americans were significantly more likely to vote than similarly situated older Caucasians. In addition, the impact of being an African American appears to be increasing from a low marginal impact of .07 in 1972 to a high of .52 in 1996. Finally, the impact of marriage is significant and positive whereas there appear to be no significant differences between older men and women once we control for socioeconomic factors.

The analysis of the ANES data set reveals that once controls are made for socioeconomic standing and political attitudes, older African Americans were less likely to vote in the early presidential election years of the 1950s and early 1960s (see Table 4 ). This pattern is likely the result of Jim Crow laws and Black Codes preventing African Americans in the South from voting until the passage of the Civil Rights Act in 1965. The impact of education, while not significant in some of the years, has a positive and significant impact on the probability of turnout in 7 of the 12 elections. Once controls are made for political attitudes, such as "Do you care about the outcome of the presidential election," the impact of family income among elders also appears to be a nonsignificant predictor until the 1996 election. Concerning the impact of age, we find that the coefficients attached to both the age and the age-squared term are consistently in the same direction as the findings from the CPS, but they do not reach statistical significance in most of the presidential elections.

As in the CPS, living in the South is associated with a significantly lower probability of voting in presidential elections among elders. Despite the substantial number of attitudinal, economic, and demographic controls, there is a reoccurring difference between elders living in the South compared to older adults anywhere else in the nation. The increased migration of the older population into the South has apparently not altered this phenomenon. As found in the CPS, there appear to be few years with significant differences between older men and women when controlling for the other factors included in the model. A relatively consistent finding from the ANES data is the significant positive impact of being interested in politics. Alternatively, the strength of the respondent's partisanship is a significant positive predictor in only 2 of the 12 presidential elections included in the analysis. As in prior research including voters of all ages, the impact of church attendance is generally a consistent and positive predictor of voter turnout among elders (Table Aa , Note 7). The impact of candidate and party recruitment has a significant and positive impact on the probability of voting in every election after the 1976 presidential election. Finally, external political efficacy apparently has no consistent impact on the probability of turnout among older voters. The impact of internal efficacy consistently has a positive sign but only reaches statistical significance in 4 of the 12 elections.

Table 5 presents the results from the analysis of the GSS that reveal very similar patterns. Again, we see the consistent and substantial positive impact of education and the negative impact of living in the South (although the impact of living in the South does not reach conventional levels of statistical significance as consistently as in the analysis of the CPS or the ANES data sets). The impact of church attendance was positive and significant in every election. There appear to be few years when there were significant differences between men and women. The patterns associated with the age and age-squared terms show the standard directional impact as found in our analysis of the CPS data but, like the findings from the ANES, once controls are made for political attitudes, the variables do not reach statistical significance in this analysis. The number of people living in the home is not a consistently significant variable. Finally, poor health is a significant deterrent in only four of the seven elections. Apparently, there are factors in some elections that make it worth the struggle for elders in poor health to vote. This finding could be related to the mobilization by political operatives in some elections who encourage high voter turnout among supporters by providing transportation or facilitating absentee voting for elders with health conditions that make it difficult to go to the polls. Promising areas of future research concern disentangling the contextual factors that help elders overcome the deterrent effects of poor health (e.g., candidate issue positions and appeals, mobilization efforts, close elections) and under what circumstances such factors have their greatest impact and among whom they have that impact.

Discussion

In general, the analyses from all three data sets reveal similar findings. Table 6 presents the variables in each data set that were significant in at least a majority of the elections. The findings seem clear regarding the negative impact of living in the South, the positive impact of education, and the positive impact of church attendance. Although the findings from the CPS suggest that there are significant relationships between voter turnout and marriage, income, and being an African American, the findings from the ANES and the GSS suggest that once controls are made for political attitudes such as interest in politics, behaviors such as church attendance, and political factors such as being recruited, some of the relationships found in the analysis of the CPS disappear. Nevertheless, despite numerous controls, these analyses support the notion that there is a fundamental difference in the turnout patterns of older voters living in the South, those who attend church regularly, those who are more highly educated, and those who have been explicitly recruited or mobilized into political life through contact by political parties or candidates (see also MacManus 2000).

Although the impact of education is consistently significant in the analyses of the GSS and the CPS data sets, its significance fluctuates in the findings from the ANES. In fact, the impact of education is not significant at conventional levels in five of the years included in the ANES. Prior research focusing on the overall population has created the conventional wisdom that education is the primary predictor of voter turnout in the United States (Verba, Nie, and Kim 1978; Wolfinger and Rosenstone 1980). The findings of this study support the general conclusion that education is an important factor in determining voter turnout but also suggests that, among older adults, other factors may be stronger and more consistent determinants of voter turnout. In fact, while educational attainment is an important predictor of voter turnout among elders included in these surveys, it is neither the strongest nor the most consistent finding. Similarly, controlling for other factors, the impact of family income among the older population is also less important than found in previous research.

There were only a few years with significant differences between older men and women, controlling for the other factors included in the models. A relatively consistent finding in the analysis of the ANES data is the significant positive impact of being interested in politics and the importance of being contacted by a political party or candidates. These findings imply that as older adults continue to become a larger part of the eligible electorate, candidates and political elites may attempt to relate and address relevant issues in ways that render politics more interesting to older citizens. The result may be an increase in the number of older adults who are interested in political life and are therefore more likely to vote. Nevertheless, this outcome may depend on the willingness of political activists to actually mobilize and recruit older voters.

Prior research suggests that the strength of partisanship should be a consistent predictor of those who are likely to vote, but among the elders in the ANES and GSS surveys it is generally not significant once controls are made for other demographic, economic, and attitudinal factors. In comparison, the impact of candidate and party recruitment has a significant positive impact on the probability of voting in every election after the 1980 presidential election. Prior investigations examining changes in candidate and party recruitment among the general population have shown that the overall number of candidate and party recruitment efforts increased during the presidential election years occurring in the 1950s and 1960s, then fluctuated during the 1970s, and then fell substantially during the 1980s (Rosenstone and Hansen 1993). Nevertheless, the marginal impact of recruitment among older people is one of the strongest predictors of voter turnout after the 1980s. In other words, while candidates and parties may be currently recruiting fewer people compared to the 1950s and 1960s when political parties had much greater organizational strength (Beck and Sorauf 1992), the effectiveness of these recruitment efforts, at least among senior voters, has apparently increased.

External political efficacy apparently has no consistent impact on the probability of turnout among the older population. However, the impact of internal efficacy more regularly has a positive sign but only reaches statistical significance in four of the elections included in the ANES. Although some prior research (examining the entire population) suggests that a loss of political efficacy is one of the main reasons voter turnout has declined since the 1960s (Teixeira 1992), political efficacy, both internal and external, does not seem to play a consistent role in the individual level determinants of voter turnout among the nation's elderly population.

Contrary to the Senior Power Model, which suggests that older Americans are a self-interested voting block with ever increasing political strength evidenced by their increasing levels of turnout, our analysis suggests that elders are far from the image of a monolithic voting juggernaut. In fact, our analysis suggests that over the past 50 years the older population has experienced declining levels of political efficacy, substantially reduced incomes during the 1980s, lower marriage rates, increased percentages of minority population, and greater percentages of elders living in southern states. All of these trends suggest increased heterogeneity of the senior voting block and reduced turnout rates. Our analyses also suggest that these deterrent factors have been offset by the increased levels of education, income (during the 70s and 90s), interest in political affairs, and mobilization/recruitment efforts by political activists. Far from the stereotype created by the Senior Power Model, older adults are experiencing lower levels of both internal and external political efficacy and, according to the analysis of the ANES and CPS data, those living in the South are generally much less likely to vote than their peers living elsewhere.

Overall, the findings of this study seriously question the applicability of the Senior Power Model. In fact, Binstock's contention (1983, 2000) that older adults are a more heterogeneous group than many think is supported by the findings of this study. The fact that neither external nor internal efficacy were consistently significant in predicting voter turnout suggests that elders do not perceive themselves to be a "power block" that can make a difference in the outcome of a presidential election. Again, it is important to note that the percentages of the older population with high feelings of external efficacy are relatively small, reflecting the general belief among the senior community that government is not very responsive to their needs and desires. Further, the behaviors of elders living in the South appear to be anything but "senior power." In fact, it has long been noted that legal, institutional, and cultural factors in the South have been deterrents to voter turnout (Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980), and the overall voting rates among older adults may primarily rest on the changing political environment of the South.

Nevertheless, with the continuing substantial growth of the older population and the number of advocacy and mobilization groups, older voters may continue to gain more political power at the ballot box. The growth of political advocacy and mobilization groups in the past three decades has been phenomenal. Three types of age-based groups have developed (Torres-Gil 1992), and the organizations represented by these groups are numerous. The following list includes a only small representation of each type. The three types of groups include mass membership groups such as the American Association of Retired Persons (AARP) and the Older Women's League (OWL); professional organizations composed of scholars and practitioners focused on issues related to aging such as The Gerontological Society of America (GSA), the American Society on Aging (ASA), and the Gerontological Health Section of the American Public Health Association; and public interest organizations and agencies that reflect concerns for older adults and related service providers such as the Alzheimer's Association, the National Arthritis Foundation, the National Association of Area Agencies on Aging (NAAAA), the National Association of State Units on Aging (NASUA), and the National Coalition of Aging Groups. However, because of the heterogeneity of older Americans in education, health, income, and such, even these organizations together cannot create a monofocused mobilizing group with the power to elect public officials based on candidates' support of specific policies relative to older adults (Binstock 1997a, Binstock 2000). Thus, as with younger voters, these organizations may contribute to determining when older adults vote but not necessarily how.

Whether or not the Senior Power Model has any validity, and if the voting behaviors of older adults are tied to the mobilization of these voters by political operatives, older voters will remain an important segment of the voting population whom researchers cannot ignore. In fact, a full understanding of voter turnout in contemporary American democracy, as well as a full understanding of the civic lives of elders in the United States, will remain incomplete without many more investigations into the political behavior of older Americans. Particularly productive investigations will focus on disentangling the contextual factors that help elders overcome the deterrent effects of poor health (e.g., candidate issue positions and appeals, mobilization efforts, close elections) and under what circumstances such factors have their greatest impact and among whom that impact is realized.

Practice Concepts

The Forum

Book Reviews

Table 1.

Simple Statistics From the Current Population Study

Variable	1972	1976	1980	1984	1988	1992	1996
Living in the South, %	31.28	33.06	31.56	30.70	33.29	31.42	32.04
African American, %	8.44	9.14	8.14	8.06	8.08	8.06	8.21
Married, %	65.08	65.09	65.33	64.52	63.37	63.07	63.59
Education, M (SD)	1.23	1.40	1.60	1.72	1.85	2.02	2.18
	(1.4)	(1.4)	(1.4)	(1.4)	(1.4)	(1.4)	(1.4)
Income, M (SD)	6.06	6.54	7.81	4.97	5.31	6.07	7.90
	(2.8)	(3.3)	(3.2)	(3.8)	(3.6)	(3.7)	(3.8)
Age, M (SD)	66.29	66.43	66.68	67.14	67.80	68.15	68.20
	(8.4)	(8.5)	(8.6)	(8.7)	(8.7)	(8.9)	(9.1)
Sample size (raw number)	23,847	21,322	30,037	26,961	25,402	24,907	19,511

Note: Variable categories and operational definitions for all variables across data sets are discussed in Table Ab b.

Table 2.

Means and Frequencies From the American National Election Studies

Variable	1952	1956	1960	1964	1968	1972	1976	1980	1984	1988	1992	1996
African Americans, %	7.17	5.83	7.87	7.74	9.39	7.94	9.96	10.92	9.81	10.53	11.65	10.14
Married, %	59.43	62.47	66.50	57.20	54.08	55.68	53.21	54.97	54.03	53.00	50.87	49.83
Living in the South, %	27.25	27.27	31.98	33.12	37.14	33.70	38.66	37.23	33.09	31.77	32.80	36.01
Internally efficacious, %	58.40	62.24	62.18	61.94	39.18	52.26	45.22	44.44	61.20	40.52	47.39	34.09
Externally efficacious, %	23.98	33.10	38.07	25.81	19.59	19.78	20.97	15.40	22.40	15.56	17.14	19.76
Contacted, %	—	17.81	19.78	20.57	23.11	26.75	30.45	30.05	28.47	30.36	22.91	35.45
Income, M (SD)	1.32	1.22	1.3	1.24	1.20	1.29	1.46	1.32	1.5	1.48	1.55	1.53
	(1.3)	(1.2)	(1.2)	(1.3)	(1.1)	(1.2)	(1.2)	(1.2)	(1.1)	(1.1)	(1.1)	(1.1)
Age, M (SD)	65.3	65.1	65.6	66.1	66.37	66.5	66.97	66.89	67.77	68.01	68.85	68.47
	(8.2)	(7.4)	(7.8)	(8.2)	(8)	(8.4)	(8.4)	(8.3)	(8.5)	(8.9)	(8.6)	(9.2)
Education, M (SD)	1.88	2.10	2.21	2.44	2.4	2.41	2.67	2.92	3.03	3.13	3.19	3.49
	(1.4)	(1.5)	(1.6)	(1.7)	(1.7)	(1.6)	(1.7)	(1.7)	(1.7)	(1.8)	(1.7)	(1.6)
Sample size	315	268	261	285	241	172	366	250	340	339	439	443

Note: Variable categories and operational definitions for all variables across data sets are discussed in Table Ab b.

Table 3.

Voter Turnout in Presidential Election Years: Current Population Studies 1972–1996

Variable	1972	1976	1980	1984	1988	1992	1996
Intercept	−12.64***	−11.91***	−14.02***	−12.31***	−14.47***	−11.88***	−12.7***
	(.92)	(.95)	(.81)	(.86)	(.94)	(.96)	(1.03)
Living in the South	−.58***	−.43***	−.31***	−.37***	−.35***	−.29***	−.27***
	(.03)	(.03)	(.03)	(.03)	(.03)	(.03)	(.04)
In labor force	.41***	.37***	.28***	.31***	.21***	.23***	.15**
	(.04)	(.04)	(.04)	(.04)	(.04)	(.05)	(.05)
Education	.44***	.45***	.44***	.44***	.41***	.45***	.44***
	(.01)	(.02)	(.01)	(.01)	(.01)	(.02)	(.02)
Family income	.05***	.05***	.07***	.06***	.10***	.09***	.08***
	(.01)	(.01)	(.005)	(.005)	(.006)	(.01)	(.006)
Age	.36***	.33***	.39***	.35***	.41***	.34***	.35***
	(.03)	(.03)	(.02)	(.02)	(.027)	(.03)	(.03)
Age squared	−.002***	.002***	−.003***	−.002***	−.003***	−.002***	−.002***
	(.0002)	(.0002)	(.0002)	(.0002)	(.0002)	(.0002)	(.0002)
African American	.07	.20***	.18***	.32***	.52***	.35***	.52***
	(.05)	(.05)	(.05)	(.05)	(.06)	(.06)	(.07)
Married	.64***	.58***	.55***	.57***	.68***	.67***	.69***
	(.03)	(.04)	(.03)	(.04)	(.04)	(.04)	(.04)
Women	−.13**	−.12**	−.04	.01	−.02	.03	.05
	(.03)	(.04)	(.03)	(.03)	(.03)	(.04)	(.04)
Sample size	23,847	21,322	30,037	26,961	25,402	24,907	19,511
−2 log	26069.01	23230.86	30420.23	26220.58	25057.36	22615.87	19112.44

Note: Cell entries are logistic regression estimates with standard errors in parentheses. All measures of the −2 log likelihood are significant at the .05 level.

**

p < .01; ***p < .001.

Table 4.

Voter Turnout in Presidential Election Years: American National Election Studies 1952–1996

Variable	1952	1956	1960	1964	1968	1972	1976	1980	1984	1988	1992	1996
Intercept	2.7	3.8	−4.36	9.44	−6.71	−12.99	−4.74	−30.58***	−9.8	−16.12	−5.35	−11.32
	(9.49)	(13.7)	(13.48)	(9.43)	(11.14)	(12.3)	(7.96)	(11.7)	(10.1)	(8.37)	(9.56)	(8.8)
Family income	.09	−.16	.005	.02	.16	.17	.27	.29	.33	.15	.23	.73***
	(.14)	(.19)	(.21)	(.15)	(.23)	(.24)	(.19)	(.21)	(.19)	(.17)	(.18)	(.22)
Age	−.05	−.09	.06	−.20	.16	.19	.13	.84*	.27	.45	.11	.19
	(.28)	(.41)	(.39)	(.28)	(.32)	(.35)	(.23)	(.34)	(.29)	(.24)	(.27)	(.25)
Age squared	.000069	.001	−.00039	.001	−.001**	−.001	−.001	−.006*	−.002	−.003	−.001	−.001
	(.002)	(.003)	(.003)	(.001)	(.002)	(.002)	(.002)	(.002)	(.002)	(.002)	(.002)	(.001)
African American	−1.19*	−2.53***	−1.16*	−1.45**	1.54*	.86	.56	.17	.28	−.83	.83	−.41
	(.53)	(.71)	(.59)	(.54)	(.68)	(.82)	(.52)	(.59)	(.66)	(.49)	(.49)	(.54)
Education	.29	.44**	.49**	.14	.27	.37*	.21	.37**	.19	.38***	.28*	.44*
	(.15)	(.17)	(.19)	(.11)	(.14)	(.18)	(.12)	(.14)	(.12)	(.1)	(.12)	(.14)
Married	−.19	.31	.98*	−.17	.46	.94*	.17	.24	−.02	.09	.59	.76
	(.33)	(.39)	(.49)	(.34)	(.39)	(.47)	(.34)	(.42)	(.36)	(.36)	(.36)	(.41)
Living in the South	−1.39***	−1.9***	−.86*	−1.02**	−1.49***	−1.32**	−.19	.35	−.74	−.95	−1.13***	−.43
	(.32)	(.38)	(.41)	(.32)	(.38)	(.45)	(.3)	(.39)	(.34)	(.32)	(.32)	(.36)
Women	−1.17***	−.91	−.59	−.92**	−.31	.03	−1.41*	−.84*	−.29	−.17	.35	.36
	(.33)	(.39)	(.43)	(.34)	(.38)	(.47)	(.34)	(.42)	(.37)	(.32)	(.33)	(.39)
Church attendance	.09	.35**	.24	.19*	.46***	.22	.32*	.36**	.27**	.39***	.18*	.42***
	(.1)	(.12)	(.13)	(.09)	(.11)	(.13)	(.09)	(.12)	(.09)	(.09)	(.09)	(.11)
Years in community	N.A.	N.A.	.16	.09	.02**	.004	.01	.02*	.01*	.01	.008	.002
			(.13)	(.11)	(.01)	(.008)	(.01)	(.007)	(.005)	(.006)	(.005)	(.005)
Internal efficacy	.36	.31	.99*	.77*	.53	.89*	.38	1.04**	.26	.13	.11	.03
	(.3)	(.37)	(.39)	(.33)	(.39)	(.42)	(.29)	(.38)	(.33)	(.32)	(.30)	(.37)
External efficacy	−.22	.54	.58	−1.18**	−.32	.37	.85	−.49	.43	.59	.14	.29
	(.43)	(.48)	(.52)	(.43)	(.47)	(.65)	(.42)	(.57)	(.48)	(.49)	(.45)	(.46)
Not working	−.51	.26	.97	−1.77	−.65	1.1	.29	−.58	.91	.43	−.27	−1.05
	(.84)	(3.01)	(1.5)	(.98)	(.97)	(.97)	(.54)	(.64)	(.63)	(.56)	(.49)	(.61)
Political interest	.76***	.39	.54*	.26	1.13***	.75*	.46	.74**	.62**	.75***	.819***	.628*
	(.22)	(.23)	(.26)	(.21)	(.26)	(.3)	(.19)	(.25)	(.24)	(.21)	(.2065)	(.247)
Strong partisan	.70*	.45	1.07*	.38	.26	.41	.04	.04	.01	.45	.4243	.576
	(.35)	(.39)	(.46)	(.41)	(.44)	(.5)	(.35)	(.43)	(.36)	(.33)	(.3207)	(.355)
Financial situation	N.A.	.32	.24	−.14	−.07	.72	−.21	.25	.05	−.35	−.136	−.329
		(.29)	(.32)	(.27)	(.35)	(.39)	(.27)	(.31)	(.32)	(.28)	(.2628)	(.311)
Contacted	N.A.	.47	−.40	.71	.83	.52	.84*	1.09*	.96*	1.04***	1.1841**	1.029*
		(.48)	(.49)	(.46)	(.52)	(.48)	(.35)	(.51)	(.42)	(.39)	(.4575)	(.407)
Trust in government	N.A.	N.A.	N.A.	−.14	−.03	.45	.25	.05	.05	.13	.0764	.175
				(.24)	(.27)	(.37)	(.26)	(.33)	(.28)	(.23)	(.2274)	(.29)
Own home	.56	N.A.	N.A.	−.4	.33	1.29**	.3	.79	.51	−.22	.2301	.361
	(.32)			(.39)	(.39)	(.46)	(.33)	(.44)	(.37)	(.40)	(.3286)	(.372)
Care about presidential race	−.14	.34	.009	1.01**	−.09	.92*	.59	.29	.33	.32	.7871*	.084
	(.35)	(.36)	(.41)	(.34)	(.39)	(.46)	(.31)	(.39)	(.35)	(.32)	(.3296)	(.39)
Presidential race will be close	−.22	.84*	.25	.14	.15	.95	.33	−.29	−.11	.16	.5584	.261
	(.31)	(.34)	(.38)	(.33)	(.36)	(.49)	(.29)	(.40)	(.33)	(.32)	(.3362)	(.329)
2–Log likelihood	315.25	242.74	192.27	293.26	227.93	167.97	347.52	213.51	281.06	316.72	335.968	262.09
Sample size	315	268	261	285	241	172	366	250	340	339	439	443

(Table continues on next page)

Notes: Cell entries are logistic regression estimates with standard errors in parentheses. All measures of the –2 log likelihood are significant at the .05 level. N.A. = Not available.

*

p < .05; **p < .01; ***p < .001.

Table 5.

Voter Turnout in Presidential Election Years: General Social Survey 1972–1996

Variable	1972	1976	1980	1984	1988	1992	1996
Intercept	11.39	−9.81	−7.06	−6.58	−16.06	−4.69	−8.22
	(10.36)	(7.64)	(6.39)	(8.34)	(9.05)	(9.41)	(5.65)
Age	.28	.27	.21	.17	.43	.09	.27
	(.30)	(.22)	(.18)	(.24)	(.25)	(.26)	(.16)
Age squared	−.002	−.002	−.001	−.001	−.003	−.0005	−.002
	(.002)	(.002)	(.001)	(.002)	(.002)	(.002)	(.001)
Church attendance	.14**	.12*	.14***	.23***	.17**	.21***	.17***
	(.05)	(.05)	(.04)	(.05)	(.05)	(.06)	(.04)
Number of persons in household	.02	−.08	−.13	−.45**	−.34	−.16	−.21
	(.13)	(.14)	(.09)	(.17)	(.19)	(.22)	(.13)
Education	.19***	.15***	.06*	.22***	.16**	.20***	.092**
	(.04)	(.04)	(.03)	(.04)	(.05)	(.05)	(.033)
Health status	−.08	−.13	−.27*	−.46**	−.41*	−.16	−.31**
	(.14)	(.14)	(.12)	(.14)	(.16)	(.15)	(.11)
Income	−.007	−.001	−.001	.001	−.004	−.01	.002
	(.005)	(.005)	(.005)	(.006)	(.006)	(.01)	(.005)
Married	.9**	.49	.62*	.88**	.81*	1.05*	.17
	(.3)	(.29)	(.26)	(.33)	(.38)	(.42)	(.25)
Strong partisan	.15	−.14	−.03	−.29	.52	.45	.05
	(.26)	(.25)	(.21)	(.26)	(.31)	(.31)	(.20)
African American	−.23	−0.14	.67*	.012	1.33*	1.09*	−.33
	(.39)	(.39)	(.29)	(.42)	(.63)	(.55)	(.31)
Living in the South	−.54*	−.17	−.41	−.21	−.14	−.66*	−.28
	(.27)	(.27)	(.23)	(.27)	(.32)	(.33)	(.21)
Women	–.23	–.33	–.51*	–.39	–.34	–.67	–.49*
	(.3)	(.27)	(.24)	(.29)	(.34)	(.36)	(.22)
In labor force	.74*	.62	.34	.14	.47	.03	.27
	(.35)	(.34)	(.28)	(.34)	(.43)	(.45)	(.27)
-2 Log likelihood	387.38**	426.53***	569.11***	404.43***	286.91***	275.73***	638.71***
Sample size	433	456	582	494	325	308	738

Note: Cell entries are logistic regression estimates with standard errors in parentheses.

*

p < .05; **p < .01; ***p < .001.

Table 6.

Variables Significant in Majority of Years From Each Data Set

Variable	CPS	ANES	GSS
Living in the South	X	X	nsa
Health status	—b	—	X
Education	X	X	X
Church attendance	—	X	X
Contactedc	—	X	—
Interest in politics	—	X	—
Income	X	ns	ns
In labor force	X	ns	ns
Marriage	X	ns	X
Age	X	ns	ns
African American	X	ns	ns

Note: CPS = Current Population Survey; ANES = American National Election Studies; GSS = General Social Survey.

a

Variable not significant, but included in the data set and estimated models.

b

Variable not included in the data set, not included regularly, or changes in question wording across years rendered the variable inappropriate for comparisons.

c

Contact is not significant in a majority of the years included in the ANES. In fact, it is significant only after the 1980 presidential election. Consequently, it is significant in only 5 of the 12 presidential elections included in the ANES. It is included here, however, because of the consistency and size of the impact after the 1980 election.

Table Aa.

	Notes
1.	Figure taken from the Federal Election Commission homepage at http://www.fec.gov/pages/agedemog.htm
2.	Most striking in this regard are the Civil Rights Act of 1965, the 26th Amendment passed in 1971 allowing those 18 years and older to vote, and the National Voter Registration Act of 1993.
3.	According to MacManus 2000, older Americans are more likely to vote because, among other things, they are very partisan; they are interested in politics; they are more willing to volunteer for political parties and candidates and are therefore likely to be recruited into politics; they are more likely to believe that voting is part of their civic duty; and they are more likely to feel connected to government. In this study, we explore many of these determinants, as well as others such as church attendance and geographic region.
4.	There is growing evidence that the higher probability of participation among older Americans does not extend to older people living with disabilities (MacManus 2000). The data sets used in this study identified people with disabilities by asking only those individuals who were currently unemployed if a disability prevented them from entering the work force. Those individuals who identified themselves as out of the labor force as a result of their disability are much less likely to register or vote (Schur 1998; Schur and Kruse 2000; Shields, Schriner, and Schriner 1998).
5.	We chose to examine those aged 55 and older because these "younger seniors" have often been included in prior studies; it provides for large sample sizes; and supplementary examinations showed little difference in the substantive findings if the age cutoff was raised to 60 or to 65.
6.	Given that there has been some debate concerning the pattern of voter turnout across age, we include age and age squared in our model of voter turnout. If voter turnout does decrease across age levels, or even if the increase in voter turnout simply slows-down in the latter years, but continues to rise, the age-squared term allows us to model such relationships. Although some studies have not included the squared value of age (e.g., Bazargan et al. 1991), it is appropriate to test for such relationships (Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980).
7.	The American National Election Studies slightly changed the wording and response of their measure of church attendance in 1972; see Table Ab b for more details.

Table Ab.

Variable Descriptions
Voted	Coded 1 for respondents who reported voting and 0 otherwise.
Living in the South	Coded 1 for respondents in southern states (AL, AR, DE, DC, FL, GA, KY, LA, MD, MS, NC, OK, SC, TN, TX, VA, WV); 0 = Otherwise.
In Labor Force	Coded 1 for those currently employed or looking for work; coded 0 for respondents who were not in the labor force because they were currently not employed or looking for employment.
Education	For CPS data, 6 categories ranging from 1 = Respondent completed less than 9th grade, to 6 = Respondent graduated from college.
	For ANES and GSS data, 1 = 8 grades or less, 2 = 9–12 grades no diploma, 3 = 12 grades + diploma, 4 = 12 grades, diploma, + nonacademic training, 5 = Some college, no degree; junior/community college level, 6 = BA level degrees, 7 = Advanced degrees including LLB.
Family Income	For CPS data, 14 categories ranging from 1 = less than $5,000 per year, to 14 = $75,000 or more.
	For the ANES data, the variable ranges from 0 to 4 based on what quartile the individual was in during that year. For example, in 1952, the quartiles were 0 = None–$1999, 1 = $2000–2999, 2 = $3000–3999, 3 = $4000–4999, 4 = $10,000 +; in 1980, the quartiles were 0 = None–$6999, 1 = $7000–11,999, 1 = $12,000–24,999, 3 = $25,000–49,999, 4 = $50,000 +; and in 1998 the quartiles were 0 = none–$10,999, 1 = $11,000–19,999, 2 = $20,000–39,999, 3 = $40,000–89,999, 4 = $90,000 +.
	For GSS data, the variable ranges from 0 to 20 in increments of approximately $3,000 at incomes below $30,000 and $5,000 increments above this level.
Age	Age in years ranging from 18 to 90.
African American	Coded 1 for African American respondents; 0 otherwise.
Married	Coded 1 for married respondents; 0 otherwise.
Women	Coded 1 for male respondents; 0 otherwise.
Church Attendance	Before 1970, 4 = Regularly, 3 = Often, 2 = Seldom, 1 = Never; after 1970, 5 = Every week, 4 = Almost every week, 3 = Once or twice a month, 2 = A few times a year, 1 = Never
Years in Community	1 = 4 years or less, 2 =5–9 years, 3 = 10–19 years, 4 = 20–29 years, 5 = 30 or more years, 6 = "All of life"
Internal Efficacy	Politics too complicated for someone like me to understand. 0 = Agree, 1 = Disagree
External Efficacy	People like me have no say in what government officials do. 0 = Agree, 1 = Disagree
Political Interest	Interest in the following campaigns, 1 = not much, 2 = somewhat, 3 = very much
Strong Partisan	Coded 1 for respondents who were strong Democrats or 7 for those who were strong Republicans; 0 = Otherwise. 1 = Strong Democrat, 2 = Weak Democrat, 3 = Leans Democrat, 4 = Independent, 5 = Leans Republican, 6 = Weak Republican, 7 = Strong Republican.
Financial Situation	Do you think your finances will … 1 = Improve, 2 = Stay the same, 3 = Get worse over the next few years.
Party Contacted	Did someone from a political party contact you about the election? 1 = Yes, 0 = Otherwise
Trust in Government	How much of the time do you think you can trust the government in Washington to do what is right? 1 = None of the time, 2 = Some of the time, 3 = Most of the time, 4 = Just about always
Own Home	1 = Yes, 0 = Otherwise
Care About Presidential Race	Do you care about who wins the election? 1 = Care a good deal, 0 = Don't care
Presidential Race Will Be Close	Will the presidential election be close? 1 = Close race, 0 = Win by quite a bit
Number of Persons in Household	Household size, ranges from 1 to 16
Health Status	Would you say your own health, in general, is 1 = Excellent, 2 = Good, 3 = Fair, 4 = Poor?

Figure 1.

Changing size of voting blocks.

The authors would like to express their appreciation to Whitney Brosi, MS, for her assistance with the preparation of the manuscript.