Political Behavior, Vol. 27, No. 2, June 2005 (Ó 2005) DOI 10.1007/s11109-005-3341-9 FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US Julianna Sandell and Eric Plutzer How large a role does the family play in civic development? This paper examines an important aspect of family influence by tracing the impact of divorce on voter turnout during adolescence. We show that the effect of divorce among white families is large, depressing turnout by nearly 10 percentage points. Using data from the National Education Longitudinal Study of 1988, we demonstrate that the impact of divorce varies by racial group and can rival the impact of parents’ educational attainment, which is generally regarded as the most important non-political characteristic of one’s family of origin. We attempt to explain the divorce effect by examining the mediating impacts of parental voter turnout, active social learning, income loss, child–parent interaction, residential mobility, and educational attainment. Key words: turnout; political participation; civic engagement; youth; family; divorce; education; residential mobility. How large a role does the family play in civic development? A generation ago, political socialization scholars were divided on the question. Some, like Levin (1961) believed the family, or family structure (Clarke, 1973) was central; others, such as Connell (1972) thought that institutions such as the schools and mass media had supplanted the family and relegated it to a marginal role in political socialization. Since that time, political scientists have shown that family characteristics have important consequences for many aspects of political behavior. Adults’ current marital status has consequences for presidential voting (Weisberg, 1987) and political participation (Burns, Schlozman, and Verba 2001; Stoker and Jennings, 1995); and early parenthood depresses turnout (Plutzer, 2002). Parents’ characteristics are important as well for their children. Parents’ education level and political activity are highly predictive of their chilJulianna Sandell and Eric Plutzer, 219 Pond Laboratory, Department of Political Science, Pennsylvania State University, University Park, PA, 16802-6200. 133 0190-9320/05/0600-0133/0 Ó 2005 Springer ScienceþBusiness Media, Inc. 134 SANDELL AND PLUTZER dren’s political participation (Verba, Schlozman, and Burns, 2005) and parent marital status is associated with political knowledge and external efficacy (Gimpel, Lay, and Schuknecht, 2003). In this paper, we examine the impact of parental divorce on offspring voter turnout. Although divorce is just one aspect of family structure and marital transition, it has several characteristics that suggest its importance for political participation. First, it is the most common marital transition experienced when children are teenagers. It is estimated that more than one million children experience divorce every year (US Bureau of the Census, 1998, Tables 160 cited in Amato, 2000) and Bumpass (1990) estimates that about 40% of all children will experience parental divorce before reaching adulthood. In our data, divorce is two and a half times as common as marriage by a resident parent and more than three times as likely as the death of a parent during the period of adolescence when children are especially likely to internalize a variety of civic orientations and norms (Nie, Junn, and Stehlick-Barry, 1996; Schuman and Scott, 1989). Second, the sociological literature shows that divorce has numerous consequences for children. Compared to those with continuously married parents, children with divorced parents score lower on measures of academic success (Astone and McLanahan, 1991; Teachman, Paasch, and Carver, 1996), psychological adjustment (Forehand, Neighbors, Devine, and Armistead, 1994), self-concept (Wenk, Hardesty, Morgan, and Blair 1994), social competence (Beaty, 1995; Brodzinsky, Hitt, and Smith, 1993), and long term health (Tucker et al., 1997). Divorce also produces economic stress in the short term (Holden and Smock, 1991; Peterson, 1996) and increased residential mobility (Speare and Goldsheider, 1987). Divorce causes a loss of a potential civic role model and, in principle, one might imagine cases in which this improves the home environment and improves political socialization. However, the available research suggests that the impacts are almost always negative, with custodial parents investing less time with school work, being less supportive, and engaging in more conflict with their children (Amato, 2000). Third, divorce rates have risen during the last generation, which corresponds with falling voter turnout rates among 18–24 year olds. Figure 1 plots the percentage of children in single parent households because of divorce against youth turnout rates based on the eligible electorate (as computed by McDonald and Popkin, 2001). The figure shows that divorce rates rose steadily until 1992 and that (with large idiosyncratic shocks) turnout shows a general downward trend through 1996 (we expect a lag because the children represented by the lower line in the figure are not yet eligible to vote). Of course divorce alone cannot explain the broader turnout decline but we show that it is part of the explanation. In this paper, we use longitudinal data on over 6,000 8th graders whose parents were married when initially interviewed in 1988. As these 6,000 8th Percentage FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 135 50 Voter Turnout 18-20 Children in Single Parent Households Because of Divorce 25 0 1960 19641968 1972 1976 19801984 1988 19921996 2000 Year FIG. 1. Children in single parent households because of divorce and youth voter pariticipation trends 1960–2000. graders pass through middle school and High School, through the spring of 1992, over 450 experienced parental divorce. We estimate the impact of these divorces on their electoral participation between 1992 and 2000. Is our exercise simply one of common sense? And why should political scientists care? We noted that divorce has been empirically linked with many negative consequences, so it may seem obvious that divorce would have a negative impact on later political participation. However, such a common sense prediction does not provide much specificity: should we expect a negative impact on all children or just a subset? Are the impacts negligible in magnitude—perhaps dissipating so rapidly that they are difficult to discern? Or are the effects large enough to stand alongside the traditional core variables in political participation research? Do all of these negative consequences mediate the impact of divorce or just one or two? Is the major impact proximate to the family or indirect through the effects of educational attainment or other major determinants of turnout? It is our intention to provide this kind of specificity and thereby show when and where the sociological consequences of divorce also have political consequences. As to importance, we will show that the impact of divorce can be surprisingly large. More generally, we see this as a first attempt to redirect the attention of political behavior scholars back to the family. Family structure does matter and helps us better understand how SES and residential mobility impact participation. The family need not be a ‘‘black box’’ that is generally ignored in participation research but can be locus of study that helps us better understand the mechanisms of political socialization and variation in civic engagement we see in contemporary politics. 136 SANDELL AND PLUTZER EXPECTATIONS We expect divorce to contribute to lower turnout via two broad classes of mechanisms. First are a set of general stresses and dislocations that often accompany divorce and can impact later educational and income attainment. Second are the ways that divorce can disrupt political socialization via the processes of social learning. Stress and Dislocation Divorce typically leads to a loss of income and reduced standard of living in the short term (Peterson, 1996) and contributes to longer term economic loss for children of divorced parents (McLanahan and Sandefur, 1994). Since the SES of one’s family of origin is strongly related to political participation (Verba et al., 2005) and current SES is a powerful predictor of participation, divorce could contribute to lower turnout through decreased family income. More generally, divorce is stressful and results in less parental involvement in children’s lives and the interactions they do have are more likely to be conflictual (Amato, 2000), leaving less opportunity for positive parental involvement in school and the adolescent’s life in general. Divorce is disruptive and often leads to residential mobility, uprooting entire families from their previous neighborhood and political contexts (Speare and Goldscheider, 1987; South, Crowder and Trent, 1998). Residential mobility greatly reduces turnout (Squire, Wolfinger, and Glass, 1987) and, in particular, of voting in one’s first eligible election (Plutzer, 2002). Mobility can have indirect effects as well, because children from families who move frequently have lower academic achievement (Ingersool et al., 1989), more behavioral problems (Wood et al., 1993), and less educational attainment (Aston and McLanahan, 1994; Hagan et al., 1996). Finally, it is possible that these stressful consequences of divorce act on turnout indirectly, through lowered educational attainment, typically the strongest predictor of voting. There is evidence to suggest that parental divorce is associated with low socioeconomic attainment among children (Amato, 1999) and the economic hardship associated with divorce leads some children to abandon plans to attend college (Amato, 2000). Likewise, Keith and Finlay (1988) estimate that white children whose parents divorced before age 16 had significantly fewer years of education than similar children who did not experience divorce. Social Learning Divorce has been shown (and we will show again below) to reduce the turnout levels of adults (Stoker and Jennings, 1995). This is likely due to a FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 137 combination of factors including economic stress (Rosenstone, 1982) and residential mobility (Squire et al., 1987). Whatever the cause, when parents do not vote the potential for teens to identify with a positive civic role model should be diminished (Hess and Torney, 1968; Jennings, Stoker, and Bowers, 2001; Verba et al., 2005). This should ultimately reduce youth voter turnout. In addition, parents may actively transmit norms of political involvement to their children (e.g., Jennings and Niemi, 1981; Jennings et al., 2001; Plutzer, 2002). One important avenue for such direct transmission is through explicit discussion of politics and public affairs. Divorce reduces the opportunities for child–parent interaction because one parent is absent. Research shows that the opportunity of political discussions is enhanced by the presence of adult household members (Straits, 1991), suggesting that children living in divorced families simply have fewer opportunities to discuss politics with their parents than children living in two-parent households. In addition, divorce has also been shown to reduce the quality and amount of youth interaction with the custodial parent (Amato, 2000). Therefore, even with the parent in residence, political discussion should decrease in frequency and quality. The impact of race We have reasons to believe that divorce’s impact may differ across racial groups but the extant literature provides less specific guidance. Marriage is less common among African Americans than it is among whites and married blacks are more likely to divorce. At least since Carol Stack’s (1974) path breaking work on the black family, it is widely accepted that black families are more adaptive to the economic and social consequences of single parenthood due to the support given by extended family networks, which also contribute positively to child development (McAdoo, 1988; Wilson, 1989). More specifically, divorce has a less harmful effect on the psychological well-being of black females than on white females (Gove and Shin, 1989), blacks have fewer problems adjusting to divorce than whites (Monaghan and Lieberman, 1986), and divorced black women receive more support from family members than divorced white women (Cherlin, 1981). Subjectively, married African Americans also perceive fewer negative consequences than whites (Rank and Davis, 1996). Research also shows that negative consequences associated with divorce are less for black children than white children such that family demographic variables contribute little to the achievement orientation for African American youth (Ford, 1993) and family configuration, while important for white students’ achievement, is not 138 SANDELL AND PLUTZER a factor for black students’ academic achievement (Brady, Tucker, Harris, and Tribble 1992). The sociological literature is more mixed for Hispanics, a diverse pan-racial group with multiple cultures. Our sample size does not permit separating Hispanics by national subcultures (e.g., separate analyses for Mexicans and Puerto Ricans) or by race. This too limits our ability to form specific hypotheses about divorce which may have a differential impact on (or among) Hispanics. In addition, Cho (1999) has shown that the impact of standard socioeconomic predictors on voter turnout varies substantially across racial and cultural groups. Using a 1984 survey of CA residents, Cho found that the positive effect of education for Latinos on political participation is twice that of the general population while it is completely absent for Asian Americans suggesting the importance of English proficiency for Latino turnout. An important implication of Cho’s work is that models combining all racial groups must allow the impact of education to vary across groups (either through interactions or by splitting the samples). Since we expect that the impact of divorce and SES will vary across groups, we chose to investigate our expectations for divorce separately for non-Hispanic whites, non-Hispanic blacks, and Hispanics (of all races). Previous study The impact of family transitions generally on political participation was examined by Dolan (1995) who found that no form of family change— whether divorce, remarriage or death of a parent—had a significant impact on an index of voting related activities. Dolan’s conclusions were based on her 1989 survey of 1,352 students enrolled at one of 36 participating colleges and consequently her sample over-represents (future) college graduates. Most of these students had little political experience beyond the previous fall’s presidential election (60% were under 21 years old) and the measures of family structure were composites based on retrospective accounts of family transitions when respondents were as young as 5 years old. While Dolan combined all types of family transitions into one dummy variable, we isolate divorce from other transitions, such as a death of a parent or remarriage, which may have rather different social, economic, residential and psychological consequences (Amato, 2000). For these reasons, we believe that the possible impact of divorce impact merits additional investigation. Our data set is generally better to test these concerns. The sample is much larger, nationally representative (and therefore more racially and economically diverse), has multiple reports of political participation, and nearly contemporaneous reports of family transitions. Moreover, the detailed set of independent variables allows us to explain the effects that we do find. FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 139 RESEARCH DESIGN AND METHODOLOGY To examine the impact of divorce, we utilize the National Education Longitudinal Survey, 1988–2000 (NELS) produced and distributed by the National Center for Education Statistics (NCES). The spring 1988 NELS baseline survey is a nationally representative sample of 8th-graders attending 1,052 schools, both public and private, across the fifty states and the District of Columbia. The completion rate for the initial wave was 93% (Curtin, Ingels, Wu, and Heuer, 2002, p. 195). A random subset of the respondents was selected for follow-up interviews in 1990, 1992, 1994, and 2000 and 79% of those students selected for follow-up actually completed the entire panel (Curtin et al., 2002, p. 205). Students were asked about numerous topics including family situation, family relations, and political participation. In addition to surveying the students, NCES also surveyed one of the child’s parents in 1988 (87% response rate) and again in 1992 (with a 92% retention rate). Like the students, parents were asked about family situation, family relations, school characteristics, and socio-economic status. To isolate divorce’s impact, our sample is comprised of all 8th graders whose parents were married—at risk of divorce—at the time of the 1988 interview, who gave valid responses to the turnout questions, and who were in high school at the time of the second follow-up interview in 1992 (only these respondents were asked family communication questions detailed below). Of these 6,894 respondents 458 reported divorce at least once between 1988 and 1992. In terms of gender, race and parental education, the students whose parents were married at the time of the initial interview are slightly more affluent but otherwise similar to the larger sample (see Table 1). DEPENDENT VARIABLE Our dependent variable is based on questions during the third and fourth follow-ups in the spring of 1994 and 2000. Respondents were asked TABLE 1. Descriptive Statistics for Nationally Representative Sample and the Subset at Risk of Divorce Female Non-white Parents > HS but < 4 Years of College Familes with 1987 Income >$35,000 Entire Sample (%) Respondents at Risk for Divorce (%) 52.0 28.0 40.7 40.8 52.0 28.0 40.6 50.0 140 SANDELL AND PLUTZER about their voter participation in the 1992 and 1996 presidential elections, 1993 local/state elections, and any elections during 1998/1999. Respondents were also asked about their voter registration during the fourth follow-up. Each respondent could report participation in up to four elections plus being registered in the spring of 2000, yielding a zero to five index. We multiplied the index by 20 in order to get a dependent variable that ranges from zero to 100 and reflects the percentage of the five acts with an affirmative report. Table 2 reports the variation in scale scores separately for each racial/ethnic group in the analysis. In general, there is considerable variation within each group, with Hispanics having somewhat lower reported turnout than non-Hispanic whites and blacks. Details on question wording and coding decisions for this and all other variables are contained in Appendix A. INDEPENDENT VARIABLES The main independent variable is reported divorce between 1988 and 1992, a dummy variable coded with a 1 indicating at least one report of parental divorce in the time span. To minimize missing data, we combined questions from both student and parent questionnaires. If either the parent or the child reported a divorce, it is coded as a 1 with all other cases coded zero. This preserves information in cases where a student or parent did not complete all questions or was not interviewed in a particular wave of the survey. Table 3 shows that the incidence of divorce varies by race and ethnicity with non-Hispanic blacks having the highest incidence and Asians the lowest. Because there were 20 or fewer reported divorces among Native TABLE 2. Distribution of the Voting Participation Index by Race among Students at Risk of Parental Divorce in 1988 Whites, not of Hispanic Origin Score 0 20 40 60 80 100 Total Mean SD f 652 695 763 1089 1074 1023 5296 56.3 32.9 Blacks, not of Hispanic Origin Hispanics, Regardless of Race % Score f % Score f % 12.3 13.1 14.4 20.6 20.3 19.3 0 20 40 60 80 100 Total Mean SD 60 74 89 125 154 104 606 56.1 32.3 9.9 12.2 14.7 20.6 25.4 17.2 0 20 40 60 80 100 Total Mean SD 125 86 94 138 76 69 588 47.4 33.7 21.3 14.6 16.0 23.5 12.9 11.7 FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 141 TABLE 3. Incidence of Divorce Among Married Families, By Race and Ethnicity among Students at Risk of Parental Divorce in 1988 Percent reporting Divorce between 8th and 12th Grades N 6.4 10.7 5.7 3.5 8.0 6.6 5,220 590 576 258 250 6,894 White, not of Hispanic Origin Black, not of Hispanic Origin Hispanic, regardless of Race Asian or Pacific Islander American Indian or Alaskan Native Total Americans and Asians, we will restrict our analysis to whites, blacks and Hispanics. Variables Prior to Exposure to Divorce We control for sex (female = 1), 1987 family income, and the average educational attainment of the respondent’s parents. The parental education variable indicates the highest degree earned by a parent, and runs from 1 (less than high school) to 6 (PhD, MD, or other advanced degree). Family income is an ordinal measure with values ranging from 1 (no income) to 15 ($200,000 or more). We also created an index of parent–child communication based on a series of questions beginning ‘‘Since the beginning of the school year, how often have you discussed the following with either or both of your parents or guardians?’’ This was followed by questions about programs at school, school activities, and curriculum topics. Each communication type is coded with a 0 indicating ‘‘Not at all,’’ a 1 indicating ‘‘Once or twice,’’ and a 2 indicating ‘‘Three or more times.’’ The scale ranges from 0 (no communication) to a 6 (very frequent communication on all topics). The reliability of the scale, as indicated by Cronbach’s alpha, is .60. We regard this measure as an indicator of the overall quality of child–parent interactions and expect scores to decline as a result of stresses brought about by divorce or other dislocations. Stresses and Dislocations between 8th and 12th Grade To assess changes that might result from divorce we include a measure of 1991 family income, the same family communication scale asked in spring 1992 (reliability in 1992 is .78), and an indicator of residential mobility. We assess mobility by questions asked of both students and parents in the 1992 interview about the number of moves since 1988. This variable is coded 0 if 142 SANDELL AND PLUTZER the family did not move any times, 1 if there was one move, 2 if there were two moves, and 3 if there were three or more moves within the time span. Resources for Social Learning Hess and Torney (1968) suggest that young people may identify with parents such that parents who are politically active will serve as positive role models. Others have shown that parental political participation is a powerful predictor of offspring participation (Verba et al., 2005) and especially in the first eligible election (Plutzer, 2002). Unfortunately, the NELS-88 did not ask parents any questions about their turnout or campaign participation. We therefore created a proxy variable that we interpret as the probability that parents voted in the most recent presidential election. We start by estimating a logistic regression equation of voter turnout in the previous presidential election from the General Social Survey. The independent variables are the major predictors of adult turnout that are available in both data sets: sex, age, race, education, income, employment status, and southern residence (we were unable to measure homeownership or frequency of religious attendance). All variables were coded in a way that permits precise replication using the NELS-88 (see Appendix A). Since we want to account specifically for the effect of divorce on parents, we also include a GSS divorce variable that asks whether the respondent has gotten divorced within the past 5 years. We limited our analysis to respondents who have a teen living within their household in order to closely match the parent sample of the NELS-88. The results for the logistic regression are reported in Appendix Table A.1 and the results for the major SES and demographic variables are quite similar to more comprehensive models reported by Wolfinger and Rosenstone (1980) and Rosenstone and Hansen (1993) on nationally representative samples. The GSS model indicates that turnout is lowered by recent divorce, living in the south, and being black and is increased with higher age, income and education. We then use the prediction equation estimated from the GSS sample, to yield a prediction equation that we apply to the NELS-88 parental sample. Each NELS-88 parent is then assigned a probability of voting in the most recent presidential election (an out-of-sample Ŷ). These probabilities range from 6% to 100%, with a mean of 72% and a standard deviation of 18%. The high mean reflects overreporting but the measure has considerable variance allowing us to rank parents from least likely to have voted to those who were nearly certain to report a vote, had they been asked. Hess and Torney (1968) also discuss more active socialization by parents that leads to an accumulation of resources that facilitate participation. The FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 143 NELS-88 includes one measure from the 12th grade follow-up that allows us to measure this. Teens were asked how frequently (never, sometimes, often) they discussed current events with their parents, which is our second measure of social learning. Unlike the communication scale, this explicitly taps into political content and, after controlling for the overall level of communication, should provide a good measure of active transmission of political ideas and information. Educational Attainment One possible causal path is for divorce to reduce the later educational attainment of the children. Students who completed 4 years of higher education in 4 years plus a summer would have attained their degree a few months before the 1996 presidential election. We can calculate whether or not each respondent earned a college degree before any particular event in our participation scale but we developed alternative measures that we think do a better job of capturing education at a time it might impact participation. We utilized a detailed set of monthly status questions from the 1994 interview to create a measure of educational attendance during the 1993–1994 academic year—what would be the sophomore year of college if a student continued on to higher education directly after high school. This measure is based on status reports during September, October, and November of 1993 and February, March and April of 1994. A student who reported full-time attendance at a 4-year college or university during all 6 months received a score of 100%. A student, who reported half time attendance during all 6 months, or full time attendance for 3 months, would receive a score of 50%. Students attending less than half time are scored as 25% for that particular month. Thus the scale ranges from 0% to 100%. The scale has a high correlation with parents’ education (r=.45) and, as we will show, is a powerful predictor of turnout. We created a similar score for attendance at two year colleges. This is negatively correlated with parents education (e.g., if a parent completed college, children tend to score zero on this measure) but is also a powerful predictor of turnout. We believe these measures are better than traditional indicators of degrees earned because they capture educational experiences before respondents would have had a chance to graduate from college, which for most respondents would have been after two, or even three of the elections tapped by our dependent variable. As we note below (endnote 4), earning a degree explains no additional variation in political participation beyond that captured by these two measures. 144 SANDELL AND PLUTZER RESULTS We first estimate a baseline regression model separately for whites, blacks and Hispanics. In this baseline model, we control for sex, parental education, family income and frequency of child–parent discussions about school during 1988. Because schools served as primary sampling units, we report Huber-White robust standard errors that adjust for clustering within schools. Finally, it is useful to be cognizant of the fact that the NELS oversamples certain minority groups and school types. Because education researchers were especially interested in having a sufficient number of private schools in the data set, these were oversampled. In addition, Asians and Hispanic students within selected schools had a higher probability of selection. To account for deliberate oversampling of both schools and students, and differential non-response, the NCES calculated a series of weights to apply to different types of subsamples and analyses. We utilize their longitudinal weight (F4PNLWT), which is intended to allow the sample to generalize to the population of all students eligible to be interviewed in 1988 (see Curtin et al., 2002, chapter 3, especially pp. 64–93). Thus, our analytic sample should be representative of that proportion of the roughly 2.9 million children enrolled in 8th grade in 1988 who completed high school. Table 4 presents the OLS regression slopes, robust standard errors, and statistical significance. The results show that divorce has a powerful, negative impact on turnout among whites but has no significant impact among either African Americans or Hispanics. We will return to the non-findings for minorities below. In the following sections, we seek to understand why the experience of divorce during adolescence has such a profound impact on white youth. TABLE 4. Baseline Models Examing the Impact of Divorce by Race on Voter Turnout Whites (N=4,840) B Divorce Parental Education Female Parent–Child Communication Family Income Constant R2 Blacks (N=392) Robust S.E. p B Robust S.E. 9.93 3.64 .54 3.42 2.46 .61 1.25 .47 ** ** 9.70 1.97 1.16 1.57 7.31 2.37 5.14 1.68 1.52 13.65 .08 .40 5.43 ** ** 2.36 32.01 .04 1.01 11.64 ** Hispanics (N=686) p * ** B Robust S.E. p .64 1.88 .53 2.33 5.84 1.60 3.18 1.15 * 1.68 16.86 .05 .84 8.05 * * FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 145 Before we attempt to explain the impact of divorce it is useful to appreciate how large the effect is among white youth. Holding parental education, family income, sex, and parent–child communications constant, divorce decreases voting participation by roughly 10%. For comparison, the estimated effect of having a college educated parent compared to one who only finished high school (two steps on the education measure) would be just over 7%. Thus the effect of divorce is comparable to the net impact of parental education for white youth. We should also note that children who talk frequently with their parents about their school work have considerably higher turnout, an effect that is strong and significant among whites and Hispanics. As we will show, falling communication levels do not account for the divorce effect but suggest that the internal dynamics of family life have important impacts that are independent of socioeconomic status and merit further attention.1 Social Learning, Stress & Dislocation, and Educational Attainment Among white youth whose parents were married in 1988, divorce significantly lowered parental voter turnout by 9% and decreased current events discussion by 8%. Divorce also led to reduced scores on the communication scale by about a fifth of a point, diminished income by more than a full category (roughly $10,000), and increased the probability of moving from 16% to 33%. Divorce was especially associated with multiple moves, more than tripling the likelihood of moving more than once between 8th and 12th grade. Clearly, the white teenagers in divorced families experienced a drop in their home politics and resources for social learning, the stress of mobility, income loss, and found themselves communicating less frequently with their parents than would have been expected based on their base year communication scores. Do these changes account for their lower voter turnout? To see we added these measures to our baseline model. The results are reported in Table 5. The first column reports the social learning model containing all previous regressors plus the estimated probability of parental voter turnout in the most recent presidential election and the frequency of discussion about current events. The parental voter turnout variable, scaled from zero to one hundred, has a positive effect and is statistically significant. Having a parent with a 75% probability of voting results in a 19% increase on offspring turnout compared with a parent with a 25% probability of voting. Likewise, the amount of current events discussion has a large positive effect on voter turnout and achieves statistical significance. Of special interest is that the addition of these variables reduces the divorce effect from 9% B p<.05, * p<.01, ** p<.001. *** Divorce 2.32 Parental Education .62 Female 1.69 Parent–Child Communication 1988 1.96 Family Income 1987 .30 Parental Voter Turnout .35 Current Events Discussion 1992 7.53 Family Income 1991 – Parent–Child Communication 1992 – Number of Moves 1988–1992 – 1993–1994 Enrollment Status in 2 Year College – 1993–1994 Enrollment Status in 4 Year College – Constant 10.79 Adjusted R2 .09 Number of Cases 4081 Independent Variables 2.63 .69 1.30 .47 .42 .08 .93 – – – – – 7.15 – – – – – – – – – – *** *** – *** – – – Robust S.E. p 3.43 2.58 .46 2.17 .79 – – .08 2.73 5.22 – – 27.04 .09 4215 B 2.85 .59 1.18 .46 .60 – – .58 .42 1.17 – – 4.12 – – – – *** – – *** *** – – – – *** – *** – Robust S.E. p 8.37 2.24 .09 2.86 .97 – – .14 – – .15 .12 19.91 .11 4840 B 2.30 .62 1.23 .47 .41 – – .56 – – .02 .02 5.24 – – – – *** *** *** – – – – – * *** – *** *** Robust S.E. p OLS Regression Slopes and Robust Standard Errors TABLE 5. Explaining The Impact of Divorce on the Electoral Participation of White Respondents 1.46 4.04 .10 .07 20.66 .12 4038 .75 .19 .64 1.37 .48 .27 5.56 B .46 1.12 .02 .02 4.87 – – (2.30) .68 1.19 .45 .57 .07 .99 – – *** *** *** *** ** *** *** – ** – – – Robust S.E. p 146 SANDELL AND PLUTZER FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 147 to 2%. To see which variable had a larger mediating effect on divorce, we ran the model separately for parental voter turnout and current events discussion. The results indicate that, though the variables combined have a large effect on divorce, current events discussion has a bigger mediating effect on voter turnout.2 This suggests that divorce, by reducing parental turnout and, especially current events discussion, influences offspring turnout through social learning. The second column reports the stress and dislocation model that adds the 1992 parent–child communication scale, 1991 family income, and residential mobility. The results indicate that the communication scale has a positive impact on electoral participation; that is, more communication is associated with higher turnout. Family income in 1991 has no impact on youth voter turnout, while the number of residential moves has a depressing effect. Most importantly, the addition of these three variables causes the net effect of divorce to decrease from 9% to 3% and become insignificant. To precisely identify which of the three variables mediate the effect of divorce, we estimated three additional models with each explanatory variable added alone. These show that the large impact of divorce among white students acts primarily through the disruptions associated with residential mobility. Including the significant communication scale alone did not appreciably reduce the divorce effect.3 We also explore the impact of the young citizens’ own educational attainment. In our data, 41% of teenagers whose parents did not divorce were enrolled full time in a four year college in 1993–1994; in contrast, only 31% of those who experienced divorce in adolescence were similarly enrolled. To see if this difference further erodes the impact of divorce we added two measures of enrollment status to the model in the third column. Although both educational attainment variables were powerful predictors of turnout, the slope estimate for divorce is relatively unchanged. Thus, education matters a great deal but does not help us explain the impact of divorce.4 The final column in Table 5 presents a comprehensive model. Within the comprehensive model both the social learning variables (parental voter turnout and current events discussion) remain highly significant as does residential mobility. Most notably, the divorce effect decreases close to zero (.75) and is no longer significant. While the education variables remain significant, the divorce effect is mediated most by parental voter turnout, current events discussion, and residential mobility. These results provide strong evidence that divorce acts through two pathways: social learning through decreased parental voter turnout and current events discussion; and stress and dislocation through increased mobility. To more precisely examine the relative importance of these two pathways we also estimated a structural equation model. This model is 148 SANDELL AND PLUTZER described in Appendix B and permitted decomposing the total effect of divorce into 15 specific indirect paths. The total effect of divorce is calculated net of antecedent variables for sex, parent education, 1987 family income and 1988 communication (i.e., comparable to the white sub-sample in Table 4). Our model, estimated by AMOS, fit the data well (GFI = .990, CFI = .981) and shows that 39% of divorce’s impact flows through the stressors, 50% through the two social learning indicators, and 11% through educational attainment (two-step paths through stressors or social learning variables and subsequently through education measures). Like estimates from all structural equation models, this partitioning is sensitive to causal assumptions but is consistent with the inferences that we draw from the sequence of OLS regression models in Table 5. DISCUSSION & IMPLICATIONS Divorce is one of those factors that Miller and Shanks (1996) refer to as ‘‘non-political.’’ Yet it has important political consequences. Earlier in this paper we argued that common sense inferences about divorce were not very specific and, in any case, merited empirical testing. The results show that two common sense inferences turned out to be invalid. Parental divorce is related to lower family incomes and lower educational attainment—the core elements to the traditional SES model of participation. Yet neither accounted for the very large impact of divorce among young white citizens. In contrast, we find strong evidence that divorce acts through three other core variables in participation research: parental political participation, current events discussion, and residential mobility. Jennings et al. suggest that parental participation is directly linked to the frequency of political discussion between parents and children and as political discussion increases so does the transmission rates, ‘‘particularly on topics of general political significance and salience’’ (2001, 21; see also Verba et al., 2005). As a result, children become politically knowledgeable if they live in a highly politicized environment, which encourages adult voter turnout. Our results bear this out, with current events discussions explaining a large portion of the total effect of divorce. We could not measure parental turnout directly but our results suggest at least a modest role for this aspect of the home environment as well. The Role of Residential Mobility Though residential mobility is a primary stressor of divorce leading to decreased voter turnout, we are unsure as to what exactly it is about residen- FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 149 tial mobility that disrupts participation. Here we measure the exact number of moves the respondent’s family experienced between 1988 and 1992, but what is it that is crucial to voting? The extant literature suggests four possible explanations. The literature on mobility and turnout would lead us to imagine that residential mobility would lead to a decline in the parents’ voter participation, which would then be ‘‘inherited’’ by the children. Squire et al. (1987) show that adults who moved in the previous 6 months had turnout rates 16% lower than those who had been in the same residence for more than 6 years. Plutzer (2002) shows that when parents do not report voting, the probability of twenty-something citizens voting in their first eligible election declined by 21%. As a rough approximation, we can combine these two empirical findings and they suggest as much as a 3% decline in turnout via parents’ participation (i.e., P(A and B) = P(A)*P(B) = .16*.21 = .034). We lack a direct measure of parental political participation but it seems plausible that this is a major explanation for the strong mediating impact of residential mobility. Second, moving places considerable stress on the student, decreasing the child’s short and long term well-being and eventual participation. Amato and Sobolewski (2001) cite a dozen recent studies that show that ‘‘adults with divorced parents, when compared with adults with continuously married parents, report greater unhappiness, less satisfaction with life, a weaker sense of personal control, more symptoms of anxiety and depression, and greater use of mental health services.’’ Rosenstone (1982) showed that citizens rebound quickly from the short term stresses of unemployment but there is little empirical work on the impact of chronic hardships and stresses on political participation. We lack the data to empirically assess this idea here but, again, we believe that it may account for at least a part of the impact of residential mobility. Third, sociological literature suggests that children of divorce are much more likely to move to a significantly poorer neighborhood than children in stable, two-parent families due to the decline in family income (South et al., 1998). Consequently, the presence of poor neighborhoods has been shown to reduce educational attainment (Brooks-Gunn, Duncan, Klebanov, and Sealand, 1993; Crane 1991; Duncan 1994; Garner & Raudenbush 1991), school performance (Dornbusch et al., 1991), cognitive development (Brooks-Gunn, Duncan, Klebanov, and Sealand, 1993), increase delinquent behavior (Elliott et al., 1996), and increase teenage childbearing (Billy and Moore, 1992), all of which could also be linked to voter turnout. Lastly, frequent mobility may cut community ties. South et al. (1998) suggest that residential mobility ‘‘places the children of divorce at a dual disadvantage, because it disrupts both the social capital of the family (when one parent leaves the household) and the child’s access to community-based social capital, as neighborhood ties are severed upon 150 SANDELL AND PLUTZER moving’’ (669). We do not have direct reports from the main respondents but have some suggestive data from the parental interviews. Parents who moved during the 1988–1992 period were 10% less likely to report feeling ‘‘a part of the neighborhood.’’ Whether or not this is an indicator of a broader sense of community alienation, and whether this would be transferred to a similar feeling held by the children is impossible to say, but might be illuminated by other data sets. The Puzzle of Minorities We found large effects among whites but no impact of divorce among minorities. We believe this is because divorce has differential consequences across racial and ethnic groups. For example, Amato and Keith (1991) find that divorce has consistently negative impacts among whites but less consistent impacts among African Americans and Hispanics. In the NELS data we find evidence of the conditional impact of divorce. While divorce doubles the likelihood of residential mobility for whites and Hispanics, it is actually associated with greater stability among African Americans (only 11% of African American adolescents experiencing divorce moved compared to 22% of those whose married parents remained married). Similarly divorce leads to a drop in parent–child communication among whites (.40) and blacks (.25) but not among Hispanics who show a statistically insignificant increase. Thus, divorce has more consistently negative consequences for white adolescents and this may account for its greater impact among whites. CONCLUSION White citizens comprise roughly three quarters of the electorate and the turnout rate of young voters has been falling steadily since the mid-1970s. We have shown that the experience of divorce is sufficiently consequential to result in a very large drop in voter turnout among young white citizens. Indeed, the impact of divorce rivals the large impact of parent’s education. As such this research provides a small piece of the puzzle of declining youth turnout. Previous literature on parental voter turnout, education attainment, current events discussion, income, parent–child communication, and residential mobility suggested possible causal paths through which divorce could lower turnout. Only three of these, parental voter turnout, current events discussion, and residential mobility, were shown empirically to mediate the effect of divorce. If policy makers are unable to lower the rate of divorce, those interested in fostering civic engagement might effectively target young people who have recently moved or experienced divorce. Certainly, civic engagement programs based in schools may be able to identify both FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 151 recent migrants and children living with a single parent or, alternatively, target schools with high percentages of children from such families. More broadly, this research illustrates how one aspect of family dynamics has important political consequences. Twenty five years ago, Jennings and Niemi acknowledged that the political consequences of the family ‘‘were more variable than had been assumed’’ (1981, p. 76) and this variability has prevented the development of anything approaching a unified view of how the family operates to enhance or retard civic development. Our research reinforces the conditional nature of family effects—in this case we find that the negative consequences are restricted to whites. It also provides an illustration of why the family merits considerably more attention from scholars of political participation. In a recent paper, Verba et al. (2005) use key aspects of the Civic Voluntarism Model to help clarify how the SES of one’s family of origin influences political participation. We believe it shall be equally fruitful to focus on the complexities of family structure and family transitions. We have examined an important, but not the only, type of family transition and divorce is often seen as the first stage of many family changes (Amato and Sobolewski, 2001). We believe that future work should carefully examine family formation—the timing and the number of children along with whether or not that formation is accompanied by marriage. These events have the potential to also set in motion consequences that could substantially increase or decrease turnout for children raised in those families. In this sense, we see opportunities in developing theories, concepts, and data that allow detailed examination of family characteristics that exert influence well before the establishment of social and economic status in young adulthood and alongside the dynamics of socioeconomic status in the family of origin. Acknowledgments. The authors gratefully acknowledge that this research was supported by a grant from The Russell Sage Foundation and we thank three anonymous referees for their helpful comments and suggestions. Of course, all conclusions and interpretations are that of the authors. APPENDIX A: CODING OF VARIABLES USED IN STUDY Dependent Variable Voter Turnout: VOTEPCT, created by combining answers from NATELEC, VOTEPRES, F4IVPRE, F4IVANY, and F4IRVOTE. NATELEC asked during the F3 follow-up study used the following ‘‘During the past 152 SANDELL AND PLUTZER 12 months, have you voted in a local, state, or national election?’’ 0 = No, 1 = Yes. VOTEPRES asked F3 respondents ‘‘Did you vote in the 1992 Presidential election?’’ 0 = No, 1 = Yes. F4IVPRE asked during the F4 follow-up study asked respondents the following: ‘‘Did you vote in the 1996 Presidential election?’’. 0 = No, 1 = Yes. F4IVANY asked F4 respondents ‘‘In the last 2 years, have you voted in any local, state, or national election?’’ 0 = No, 1 = Yes. F4IRVOTE asked F4 respondents ‘‘Are you currently registered to vote?’’ 0 = No, 1 = Yes. All variables were added together to create, ALLVOTE_, a five point scale and 1 was subtracted from F4IRVOTE in order to separate those who abstained from voting but were registered .1 was added to ALLVOTE_ and then it was multiplied by 20 in order to create VOTEPCT, a 0–100% voter participation range. Independent Variables Parental Divorce from 1988 to 1992: F2DIVOR created by combining answers from F2PDIV, F1S99C, and F2S96B. F2PDIV was created from the variable F2P7, which asked parents in the F2 survey ‘‘What is your current marital status?’’. 0 = Not divorced 1 = Divorced. To save cases, answers from student questions were also used. F1S99C asked students during the F1 wave ‘‘Lots of things happen in families that may affect young people. In the last 2 years, have any of the following happened to your family? C. My parents got divorced or separated’’. 0 = No, 1 = Yes. F2S96B asked students during the F2 survey the same question again and answers were coded identical to F1S99C. The F2DIVOR variable first was equaled to the F2PDIV variable and then was coded as the following: 0 = No reports of divorce and 1 = At least 1 report of divorce: 0 if F2PDIV was missing and F1S99C and F2S96B both were 0, 1 if F2PDIV was missing and F1S99C or F2S96B were 1. Parental Voter Turnout: Created from the GSS. Variables included in the regression analysis to predict parental voter turnout include DIVOR5 indicates divorce in the last 5 years and is measured as a dummy variable where 1 = Yes 0 = No; AGE measured as a dummy variable with 20– 28 years old as the omitted dummy; SOUTH measured as 1 = south 0 = other region; EDUCATION measured as a dummy variable where less than HS diploma is the omitted variable; FULLTIME employment where 1 = Yes, 0 = No; PARTIME employment where 1 = Yes 0 = No; BLACK, FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 153 OTH, and HISP as three separate race dummy variables; and FAMINCOME as a continuous variable measured in $1,000s. The regression analysis from the GSS is shown below. Current Events Discussion: F2S99H, which asked students in 1992 ‘‘In the first semester or term of this school year, how often did you discuss the following with either or both of your parents or guardians? Community, national, and world events?’’ 1 = Never 2 = Sometimes 3 = Often. Total Family Income during 1987/1991: BYP80 and F2P74, the question asked parents during the BY (F2) study the following: ‘‘What was your total family income from all sources in 1987 (1991)? (If you are not sure about the amount please estimate)’’ .1 = None, 2 = Less than $1,000, 3 = $1,000–2,999, 4 = $3,000–4,999, 5 = $5,000–7,499, 6 = $7,500–9,999, 7 = $10,000–14,999, 8 = $15,000–19,999, 9 = $20,000–24,999, 10 = $25,000–34,999, 11 = $35,000– 49,999, 12 = $50,000–74,999, 13 = $75,000–99,999, 14 = $100,000–199,999, 15 = 4$200,000 or more. In the event that F2P74 was missing, the mean was used. Parental Education: BYPARED by the creators of NELS-88 and was constructed using the parent variables BYP30 and BYP31 and the student variables BYS34A and BYS34B, which all asked about parental education. With the parent and student information, BYPARED captured the highest degree earned by either parent. BYPARED was then coded as 1 = Did not finish High school, 2 = High School grad or GED, 3 => High School & <4 Year Degree, 4 = College Graduate, 5 = MA/Equivalent, 6 = PhD, MD, or other, 7 = Don’t Know. BY (F2) Communication Scale: BYCOMM and F2COMM created from BYS36A (F2S99A), BYS36B (F2S99B), and BYS36C (F2S99C), which asked students the following questions in both the BY and F2 waves: BYS36A (F2S99A) = ‘‘Since the beginning of the school year, how often have you discussed the following with either or both of your parents or guardians? A. Selecting courses or programs at school’’. BYS36B (F2S99B) = ‘‘Since the beginning of the school year, how often have you discussed the following with either or both of your parents or guardians? B. School activities or events of particular interest to you’’. BYS36C (F2S99C) = ‘‘Since the beginning of the school year, how often have you discussed the following with either or both of your parents or guardians? C. Things you’ve studied in class’’. All variables were coded as 0 = Not at all, 1 = Once or Twice, 2 = Three or more times. To 154 SANDELL AND PLUTZER TABLE A.1. Logistic Regression of Most Recent Presidential Turnout, GSS (N¼7,448) OLS Regression Slopes and Standards Errors B Divorce in last 5 Years 29–33 Years Old 34–38 Years Old 39–43 Years Old 44–48 Years Old 49–58 Years Old 59 Years and Older South HS Education Junior College College Graduate Female Full Time Employment Part Time Employment Black Other Race Hispanic Family Income in $1,000s Constant .32* .61* .82** 1.22** 1.41** 1.71** 1.76** .16 1.22** 1.53** 2.13** 2.52** .11 .07 .17 .07 .99** .16 13.41** 1.82** S.E. .15 .25 .20 .20 .22 .23 .29 .11 .12 .28 .25 .41 .11 .13 .18 .15 .26 .20 2.45 .25 save cases, if one question was missing, the average of the other two valid questions was used as a replacement. For example, if BYS36A was missing, but BYS36B and BYS36C was valid, then BYS36A = (BYS36B+BYS36C)/2. In the event that any of the three values was missing, the midpoint was used instead. The same process was completed for the F2 communication variables. Residential Mobility from 1988 to 1992: NUMMOVES created from combining F1S99A, F2S96A, and F2S102. F1S99A (F2S96A) asked students during the F1 (F2) wave the following ‘‘Lots of things happen in families that may affect young people. In the last 2 years, have any of the following happened to your family? A. My family moved to a new home’’. 0 = No, 1 = Yes. F2S102 asked students during the F2 survey ‘‘How many times have you moved since January 1, 1988’’ 1 = None, 2 = 1 time, 3 = 2 times, 4 = 3 or more times. NUMMOVES was created first by equaling it to F2S102. If either F1S99A or F2S96A = 1, then NUMMOVES = 2 (meaning 1 move). If F1S99A and F2S96A = 1, the NUMMOVES = 3 (2 moves). NUMMOVES was then recoded to the following 0 = No moves, 1 = 1 move, 2 = 2 moves, 3 = 3 or more moves. FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 155 Sophomore status in 4 year College (Sophomore status in a 2 year College): SOPH4/SOPH2 created from combining answers from ENRL0993 ENRL1093 ENRL1193 ENRL0294 ENRL0394 and ENRL0494, which asked respondents in F3 about their full and part time TABLE A.2. Structural Coefficients Endogenous Variable Predictor Variable Standardized Estimate S.E. p Divorce Parental Education Female Family Income 1987 Parent–Child Communication 1988 0.02 0.06 0.01 0.01 0.00 0.19 0.01 *** 0.00 0.70 0.00 0.58 Parental Education Family Income 1987 Divorce Parent–Child Communication 1992 Parental Education Female Family Income 1987 Parent–Child Communication 1988 Divorce Number of Moves 1988–1992 Parental Education Family Income 1987 Divorce Current Events Discussion 1992 Parental Education Female Family Income 1987 Parent–Child Communication 1988 Divorce Enrollment in 2 Year College Parental Education Female Family Income 1987 Divorce Family Income 1991 Parent–Child Communication 1992 Number of Moves 1988–1992 Current Events Discussion 1992 Enrollment in 4 Year College Parental Education Female Family Income 1987 Divorce Family Income 1991 Parent–Child Communication 1992 0.15 0.58 0.17 0.02 0.01 0.10 0.18 0.16 0.05 0.23 0.05 0.02 *** 0.05 *** 0.02 0.01 0.02 *** 0.10 0.00 0.07 0.15 0.14 0.01 0.01 0.05 0.14 0.01 0.07 0.23 0.04 0.01 *** 0.02 0.75 0.01 *** 0.01 *** 0.05 0.00 0.07 0.03 0.02 0.02 0.02 0.00 0.06 0.02 0.54 1.07 0.40 2.40 0.38 0.40 0.72 0.91 0.25 0.03 0.11 0.01 0.07 0.12 0.70 *** 1.40 0.03 0.53 *** 3.13 0.46 0.50 *** 0.52 *** Family Income 1991 *** *** *** *** *** *** *** 0.09 0.29 0.15 0.50 0.98 *** 0.42 156 SANDELL AND PLUTZER Table A.2. Continued. Endogenous Variable Predictor Variable Standardized Estimate S.E. p Number of Moves 1988–1992 Current Events Discussion 1992 0.12 0.06 0.94 1.19 *** *** Parental Education Family income 1987 Divorce 0.46 0.44 0.11 0.15 0.09 0.66 *** *** *** Parental Education Female Family Income 1991 Parent–Child Communication 1992 Number of Moves 1988–1992 Current Events Discussion 1992 Parental Turnout Enrollment in 2 Year College Enrollment in 4 Year College Divorce 0.01 0.02 0.02 0.08 0.10 0.12 0.12 0.11 0.11 0.00 0.60 0.64 1.00 0.31 0.31 0.31 0.37 *** 0.67 *** 0.85 *** 0.05 *** 0.02 *** 0.01 *** Parental Turnout Electoral Participation TABLE A.3. Unanalyzed Correlations & Correlated Errors Exogenous Variables r Parental Education & Parent–Child Communication 1988 Parental Education & Family Income 1987 Family Income 1987 & Female Parental Education & Female Parent–Child Communication 1988 & Family Income 1987 Parent–Child Communication 1988 & Female .21 .52 .01 .02 .18 .14 Residuals e (1988 communication) & e (1992 communication) e (2 yr college) & e (4 yr college) .44 .43 Table A.3 reports the unanalyzed and residual correlations while table A.4 reports the variance explained for each endogenous variable. TABLE A.4. R2 for Each Endogenous Variable Endogenous Variable Divorce Current Events Discussion Number of Moves 1988–1992 Parent–Child Communication 1992 Family income 1991 Enrollment in 2 Year College Enrollment in 4 Year College Parental Turnout Electoral Participation R2 0.00 0.11 0.06 0.15 0.48 0.01 0.23 0.63 0.11 0.00 0.00 0.00 0.15 0.05 0.58 0.02 0.11 0.44 0.00 0.00 0.16 0.00 0.03 0.03 0.00 0.02 0.00 0.00 0.00 0.23 0.01 0.23 0.01 0.07 0.06 0.01 0.46 0.01 0.25 0.07 0.15 0.18 0.07 0.02 0.14 0.06 0.00 0.12 0.01 0.11 0.00 0.00 0.17 0.02 0.00 0.05 0.02 0.00 0.00 0.00 0.14 0.00 0.04 0.00 0.10 0.12 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.08 0.12 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.07 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.11 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.11 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.12 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Table A.5 reports the direct effects of all variables on one another and these serve as the basis for calculating the contribution of all 15 possible paths from divorce to turnout in Table A.6. Divorce Current Events Discussion Number of moves 1988–1992 Parent–Child Communication 1992 Family Income 1991 Enrollment in 2 Year College Enrollment in 4 Year College Parental Turnout Electoral Participation Family Parent–Child Current Parent–Child Family 2 Year 4 Year Income Communication Parental Events Numer of Communication income College College Parent Female 1987 1988 Education Divorce Discussion Moves 1992 1991 Enroll. Enroll. Turnout TABLE A.5. Standardized Direct Effects FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 157 158 SANDELL AND PLUTZER TABLE A.6. Indirect Paths of Divorce on Voter Turnout Indirect Effect Through FAMINC F2SCOMM NUMMOVES F2S99H VOTEPROB SOPH2 SOPH4 FAMINC*SOPH2 F2SCOMM NUMMOVES F2S99H FAMINC F2SCOMM NUMMOVES F2S99H = .174*.018 = .046*.080 = .140*.097 = .044*.124 = .107*.117 = .024*.107 = .011*.109 = .174*.015*.107 *SOPH2 = .046*.001*.107 *SOPH2 = .140*.064*.107 *SOPH2 = .044*.015*.107 *SOPH4 = .174*.071*.109 *SOPH4 = .046*.121*.109 *SOPH4 = .140*.116*.109 *SOPH4 = .044*.062*.109 Total Indirect Effect = = = = = = = = = = = = = = = = .003 .004 .014 .005 .013 .003 .001 .000 .000 .001 .000 .001 .001 .002 .000 .036 classes in either 2 or 4 year colleges for September, October, and November of 1993 and for February, March, and April of 1994 (the academic school year). Full time attendance = 1, part time attendance = .5, and less than part time attendance = .25. Likewise, variables were divided into 2 and 4 year colleges and coded on a percentage scale ranging from 0% to 100%. A student who reported full-time attendance at a 4 year college or university during all 6 months received a score of 100%. A student who reported half time attendance during all 6 months, or full time attendance for 3 months, would receive a score of 50%. Students attending less than half time are scored as 25% for that particular month. APPENDIX B: STRUCTURAL EQUATION ESTIMATION To estimate a structural equation model, we assumed that parental education, family income in 1987, respondent sex, and base-year child–parent communication were exogenous. All other variables were considered endogenous. The model was estimated by Amos. Base year income and communication were assumed to influence turnout through their lagged counterparts or indirectly through other variables, though relaxing these assumptions had no impact on any estimates or substantive conclusions. In addition, we have no basis for hypothesizing an effect of gender on likelihood of residential mobility and 1991 family income so these effects are constrained to zero. In addition, we compared models with all independent residuals with models that permitted correlated errors among identical questions asked at both FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 159 waves (parent–child communication) and the two college enrollment measures, which are partially constrained to be complements of one another (someone enrolled in a 2-year college is unlikely to be enrolled in a 4 year college). Allowing these correlations improved the fit of the model (Goodness of Fit Index increased from .961 to .987). Since we discovered no direct effect of divorce in our OLS models, we then constrained the direct effect of divorce to be zero. The GFI increased modestly (to .989) and the fit was not significantly worse (chi square = .189, p=.664) so by conventional standards our ‘‘best’’ model constrains the direct effect of divorce on turnout to zero. The standardized coefficients are reported in Appendix Table A.2 and the direct effects on turnout (near the bottom of the table) conform quite closely to reported effects in model 5.4. NOTES 1. We ran the model for the white sample with parental education and income represented by categorical dummy variables, to account for any nonlinear effects that might spuriously get absorbed by divorce. The divorce impact showed little change (B=9.83) and still achieved statistical significance. Hence, the divorce impact is not an artifact of nonlinear independent variables being treated as intervals. In addition, we investigated whether the impact of divorce varied by income, as suggested by a reader of an earlier version of the paper; the interaction was essentially zero and did not approach statistical significance. 2. The divorce effect decreases from 9.93 to 6.00 and is still significant at the 95% level with the addition of the parental voter turnout variable. The divorce effect decreases from 9.93 to 4.83 with a p-value of .061 with the addition of the current events variable. 3. The divorce effect decreases from 9.93 to 9.82 and is still significant at the 99% level with the addition of the 1991 family income variable. The divorce effect decreases from 9.93 to 9.26 with a p-value of .000 with the inclusion of the 1992 family communication scale. The divorce effect decreases to 4.02 and is no longer significant at the 95% confidence level with the inclusion of the residential mobility variable. These results are available from the authors upon request. 4. We ran Model 5.4 with a variable that measured the completion of a college degree by the 1996 Presidential election (4.5 years after completing high school). The results were essentially unchanged and this variable was not statistically significant. The divorce coefficient was nearly identical (B=8.28 compared with 8.37). This is consistent with Plutzer’s (2002) conclusion that enrollment, but not degree attainment, was a crucial predictor of young voters’ turnout. REFERENCES Amato, Paul R., and Keith, Bruce (1991). Separation from a parent during childhood and adult socioeconomic attainment. Social Forces 70: 187–206. Amato, Paul R. (1999). Children of divorced parents as young adults. In E.M.Hetherington (Ed.), Coping With Divorce, Single Parenting, and Remarriage: A Risk and Resiliency Perspective, pp. 147–164. Mahwah, NJ: Eribaum. Amato, Paul R. (2000). The consequences of divorce for adults and children. Journal of Marriage and the Family 62: 1269–1288. 160 SANDELL AND PLUTZER Amato, Paul R., and Sobolewski, Juliana M. (2001). The effects of divorce and marital discord on adult children’s psychological well-being. American Sociological Review 66: 900–921. Astone, N., and McLanahan, Sara S. (1991). Family structure, parental practices, and high school completion. American Sociological Review 56: 309–320. Astone, Nan Marie, and McLanahan, Sara S. (1994). Family structure, residential mobility, and school dropout: a research note. Demography 31: 575–584. Brady, B., Tucker, C., Harris, Y., and Tribble, I. (1992). Association of academic achievement with behavior among black students and white students. Journal of Educational Research 86: 43–51. Beaty, L.A. (1995). Effects of parental absence on male adolescents’ peer relations and self-image. Adolescence 30: 873–80. Billy, John O.G., and Moore, David E. (1992). A multilevel analysis of marital and nonmarital fertility in the United States. Social Forces 70: 977–1011. Brodzinsky, D., Hitt, J.C., and Smith, D. (1993). Impact of parental separation and divorce on adopted and nonadopted children. American Journal of Orthopsychiatry 63: 451–461. Brooks-Gunn, Jeanne, Duncan, Greg J., Klebanov, Pamela Kato, and Sealand, Naomi (1991). Do neighborhoods influence child and adolescent development. American Journal of Sociology 99: 353–395. Bumpass, L. L. (1990). What’s happening to the family? Interactions between demographic and institutional change. Demography 27: 483–498. Burns, Nancy, Schlozman, Kay Lehman, and Verba, Sidney (2001). The Private Roots of Public Action. Cambridge, MA: Harvard University Press. Cherlin, Andrew (1981). Marriage, Divorce, Remarriage. Cambridge, MA: Harvard University Press. Clarke, James W. (1973). Family structure and political socialization among urban black children. American Journal of Political Science 17: 302–315. Connell, R. W. (1972). Political socialization in the American family: the evidence reexamined. Public Opinion Quarterly 36: 323–333. Crane, Jonathan (1991). The epidemic theory of ghettos and neighborhood effects on dropping out and teenage childbearing. American Journal of Sociology 96: 1226– 1259. Curtin, T. R., Ingels, S. J., Wu, S., and Heuer, R. (2002). National Education Longitudinal Study of 1988: Base-Year to Fourth Follow-up Data File User’s Manual (NCES 2002-323). Washington, DC: U.S. Department of Education, National Center for Education Statistics. Dolan, Kathleen (1995). Attitudes, behaviors, and the influence of the family: a reexamination of the role of family structure. Political Behavior 17: 251–264. Dornbusch, Sandford M., Ritter, Philip L., and Steinberg, Laurence (1991). Community influences on the relation of family statuses to adolescent school performance: differences between African Americans and non-Hispanic whites. American Journal of Education 99: 543–567. Duncan, Greg J. (1994). Families and neighbors as sources of disadvantage in the schooling decisions of white and black adolescents. American Journal of Education 103: 20–53. Ford, D. (1993). Black students’ achievement orientation as a function of perceived family achievement orientation and demographic variables. Journal of Negro Education 62: 47–66. Forehand, R., Neighbors, B., Devine, D., and Armistead, L. (1997). Interparental conflict and parental divorce. Family Relations 73: 387–393. FAMILIES, DIVORCE AND VOTER TURNOUT IN THE US 161 Elliott, Delbert S., Wilson, William Julius, Huizinga, David, Sampson, Roberta, Elliott, Amanda, and Ranjkin, Bruce. (1996). The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency 33: 389– 426. Garner, Catherine L., and Raudenbush, Stephen W. (1991). Neighborhood effects on educational attainment. Sociology of Education 64: 251–262. Gimpel, James G., Lay, J. Celeste, and Schuknect, Jason E. (2003). Cultivating Democracy: Civic Environments and Political Socialization in America. Washington, DC: Brookings Institution Press. Gove, Walter R., and Shin, Hee Chon (1989). The psychological well-being of divorced and widowed men and women an empirical analysis. Journal of Family Issues 10: 122–144. Hagan, John, MacMillan, Ross, and Wheaton, Blair (1996). New kid in town: social capital and the life course effects of family migration on children. American Sociological Review 61: 368–385. Holden, Karen C., and Pamela, J. Smock (1991). The economic costs of marital dissolution: why do women bear a disproportionate cost? Annual Review of Sociology 17: 51–78. Ingersool, Gary M., Scamman, James P., and Eckerlin, Wayne D. (1989). Geographic mobility and student achievement in an urban setting. Educational Evaluation and Policy Analysis 11: 143–149. Jennings, M. Kent, Stoker, Laura, and Bowers, Jake (2001). Politics across generations: family transmission reexamined, Paper WP2001-15. Institute of Governmental Studies. Jennings, M. Kent, and Niemi, Richard G. (1981). Generations and Politics. Princeton University Press. Keith, Verna M., and Finlay, Barbara (1988). The impact of parental divorce on children’s educational attainment, marital timing, and likelihood of divorce. Journal of Marriage and the Family 50: 797–809. Levin, Martin, L. (1961) Social climates and political socialization. Public Opinion Quarterly 25: 596–606. McAdoo, H.P. (1988). Transgenerational patterns of upward mobility in AfricanAmerican families. In H.P.McAdoo (Ed.), Black Families, pp. 148–168. Newbury Park, NJ: Sage. McDonald, Michael P., and Popkin, Samuel L. (2001). The myth of the vanishing voter. The American Political Science Review 95: 963–974. McLanahan, Sara, and Sandefur, Gary (1994). Growing Up with a Single Parent: What Hurts, What Helps. Cambridge: Harvard University Press. Menaghan, E. G., and Lieberman, M. A. (1986). Changes in depression following divorce: a panel study. Journal of Marriage and the Family 48: 319–328. Miller, Warren E., and J. Merrill Shanks (1996). The New American Voter. Cambridge, MA: Harvard University Press. Nie, Norman H., Junn, Jane, and Stehlick-Barry, Kenneth (1996). Education and Democratic Citizenship in America. Chicago: University of Chicago Press. Peterson, Richard R. (1996) A re-evaluation of the economic consequences of divorce. American Sociological Review 61: 528–536. Plutzer, Eric (2002). Becoming a habitual voter. American Political Science Review 96: 41–56. Rank, Mark R., and Davis, Larry E. (1996). Perceived happiness outside of marriage among black and white spouses. Family Relations 45: 435–441. 162 SANDELL AND PLUTZER Rosenstone, Steven J. (1982) Economic adversity and voter turnout. American Journal of Political Science 26: 25–46. Rosenstone, Steven J., and Hansen, John Mark (1993). Mobilization, Participation, and Democracy in America. New York: Macmillan. Schuman, Howard, and Jacqueline, Scott (1989). Generations and collective memories. American Sociological Review 54: 359–381. South, Scott J., Crowder, Kyle D., and Trent, Katherine (1998). Children’s residential mobility and neighborhood environment following parental divorce and remarriage. Social Forces 77: 667–693. Speare, Alden Jr., and Goldscheider, Frances Kobrin (1987). Effects of marital status change on residential mobility. Journal of Marriage and the Family 49: 455–464. Squire, Peverill, Wolfinger, Raymond E., and Glass, David P. (1987). Residential mobility and voter turnout. The American Political Science Review 81: 45–66. Stack, Carol (1974). All Our Kin: Strategies for Survival in a Black Community. Basic Books. Stoker, Laura, and Jennings, M. Kent (1995). Life-cycle transitions and political participation: the case of marriage. The American Political Science Review 89: 421–433. Straits, Bruce C. (1991). Bringing strong ties back in interpersonal gateways to political information and influence. The Public Opinion Quarterly 55: 432–448. Tam Cho, Wendy K. (1999) Naturalization, socialization, participation: immigrants and (non-) voting. Journal of Politics 61: 1140–1155. Teachman, J. D., Paasch, K., and Carver, K. (1996). Social capital and dropping out of school early. Journal of Marriage and the Family 58: 773–783. Tucker, J. S., Friedman, H. S., Schwartz, J. E., Critiqui, M. H., Tomlinson-Keasey, C., Wingard, D. L., and Martin, L. R. (1997). Parental divorce: effects on individual behavior and longevity. Journal of Personality and Social Psychology 73: 381–391. U.S. Bureau of the Census (1998). Statistical Abstract of the United States (118th ed.) Washington, D.C.: U.S. Government Printing Office. Verba, Sidney, Schlozman, Kay Lehman, and Burns, Nancy (2005). Family ties Understanding the intergenerational transmission of participation. In S. ZuckermanAlan (Ed.), The Social Logic of Politics, Philadelphia: Temple University Press. Weisberg, Herbert F. (1987). The demographics of a new voting gap marital differences in American voting. Public Opinion Quarterly 51: 335–343. Wenk, D., Hardesty, C. L., Morgan, C. S., and Blair, S. L. (1994). The influence of parental involvement on the well-being of sons and daughters. Journal of Marriage and the Family 56: 229–234. Wilson, M. N. (1989). Child development in the context of the black extended family. Amercian Psychologist 44: 98–104. Wolfinger, Raymond, and Rosenstone, Steven J. (1980). Who Votes. New Haven: Yale University Press. Wood, David, Halfon, Neal, Scarlata, Debra, Newacheck, Paul, and Nessim, Sharom (1993). Impact of family relocation on children’s growth, development, school function, and behavior. Journal of American Medical Association 270: 1334–1338.
© Copyright 2026 Paperzz