In this methodological note, I provide additional information on the empirical analysis in chapter 4. It includes information on both of the surveys that I used to track the civic habits of homeowners, as well as the specifications of the regression models used to identify the impact of homeownership and residential stability.
The main analyses in chapter 4 draw on the Social Capital Community Survey, which investigates the ways American citizens build and sustain social capital resources. The survey was first fielded in 2000, and a follow-up survey, which I use in this analysis, was fielded in 2006. The Saguro Seminar at Harvard University’s Kennedy School of Government commissioned the survey to measure the civic and social habits of Americans. The survey includes a nationally representative sample of approximately 3,000 American adults, with a dozen community surveys to supplement the nationally representative survey. The analysis in chapter 4 includes only the nationally representative sample of survey respondents.
One of the main benefits of the Social Capital Community Survey is the breadth of indicators available to measure social capital, community engagement, and political participation. While I report about two dozen measures in chapter 4, the survey includes dozens of additional indicators to evaluate how citizens engage in their communities, build social capital, and become involved in politics. In fact, the Social Capital Community Survey offers the widest range of survey measures that is available to understand the ways citizens engage in their communities, making it the ideal source for studying homeowners’ civic habits. It also includes a unique measure of residential stability that asks respondents how long they have lived in their communities.
At several points throughout chapter 4, I supplement the analysis of the Social Capital Community Survey with analyses of the Current Population Survey, a monthly survey conducted by the U.S. Bureau of Labor Statistics. Where possible, I separately validate the findings from the Social Capital Community Survey with a parallel analysis of the Current Population Survey, as many of the measures from the latter are drawn directly from the former.
(p.154) The primary purpose of the Current Population Survey is to measure employment trends across the country. However, several times each year, the survey asks respondents about their participation in community activities, political affairs, and civic life. These variables are measured in three supplements of the Current Population Survey: the Volunteer Workers, the Civic Engagement, and the Voting and Registration supplements. These supplements are measured either biennially or annually. For the supplemental analyses, I draw on the most recent supplement available at the time of the analysis (i.e., the November 2012 Voting and Registration supplement, the November 2013 Civic Engagement supplement, and the September 2014 Volunteer Workers supplement).
While the Current Population Survey includes a substantially larger number of respondents than the Social Capital Community Survey, there are two drawbacks that limit the utility of the Current Population Survey as the main source of data for chapter 4. First, the number of indicators available in this survey is limited, meaning that most of the measures analyzed in this chapter are not available in it. Second, it does not include a unique indicator of residential stability, and it is therefore impossible to separate the impact of residential stability from the effect of homeownership.
For the regression analysis behind the even-numbered figures in chapter 4 (e.g., fig. 4.2, fig. 4.4, etc.), I estimate a logistic regression from the Social Capital Community Survey to identify the change in the odds associated with the measures of homeownership and residential stability. The regressions are weighted using the respondent weights in the survey data. For homeownership, I utilize a dichotomous measure indicating whether or not the respondent reported owning his or her home. For residential stability, I utilize a dichotomous measure identifying whether or not the respondent reported living in his or her community for at least five years. The regression models also include several other standard demographic variables, including race, ethnicity, education level, marital status, gender, age, employment status, income, the presence of children in the household, and metropolitan status (e.g., rural, suburban, urban).
For each pair of graphs (e.g., fig. 4.1 and fig. 4.2), I first report the raw comparison between homeowners and renters in the odd-numbered graphs (e.g., fig. 4.1). These graphs simply compare the percentage of homeowners who report participation in each activity to the percentage of renters who report participation without controlling for other respondent characteristics. Then, the second graph adjusts those comparisons by accounting for various social and demographic characteristics, as reported above. This analysis acknowledges that homeowners and renters differ from each other in meaningful ways, and these differences may account for the increased likelihood of homeowners participating in their communities.
From the logistic regression analyses, the even-numbered graphs (e.g., fig. 4.2) present the exponentiated coefficients to understand the way that homeownership and residential stability shape civic involvement. These odds ratios identify how homeownership and residential stability influence the likelihood of civic participation, controlling for other characteristics in the models.