Data on race, inequality, and social capital in the U.S. counties

This article presents data on social capital at the United States’ county-level. Following Rupasingha et al. (2006), the social capital index captures the common factor among density measures of 10 different types of associations, voter turnout rates, U.S. decennial census participation rates, and the number of non-profit organizations. Based on Knack (2003), we create associational densities measures as a proxy for both bridging and bonding social capital. Including data on income inequality, racial diversity, minority group size, average household income, educational attainment, the ratio of a family household, the size of migration population, and female labor market participation rates, the data covers 3,104 U.S. counties for both 2009 and 2014. This paper includes descriptive statistics and figures. This data article is associated with the article “Race, Inequality, and Social Capital in the U.S. Counties.”


Value of the Data
• Social scientists who are interested in the dynamics created by income inequality, racial diversity, and social capital in the U.S. Counties can easily utilize the dataset. • This dataset also provides other county-level covariates that can be utilized by social science and humanities research. • This dataset provides the most comprehensive measure of social capital of U.S. Counties for two time periods.

Data Description
This Data in Brief article is associated with the article "Race, Inequality, and Social Capital in the U.S. Counties." [2] The data provided in this article were constructed to understand the variations of social capital across U.S. counties by examining the interaction between income inequality and ethnic diversity. Although the concept of social capital has been much debated, it can be largely defined as intangible social assets that individuals can utilize or enjoy by engaging with others. In that regard, Putnam [6] defines the concept as "networks, norms, and trust that enable participants to act together more effectively to pursue shared objectives." Following Rupasingha et al. [7] , the social capital index measures the extent to which individuals engage with others at the county-level.
The social capital index measures the common factor among four different types of variables: (1) the associational density of 10 different types of organizations (civic organizations, bowling centers, golf clubs, fitness centers, sports organizations, religious organizations, political organizations, labor unions, business organizations, and professional organizations), (2) the turnout rates for the previous presidential elections, (3) the response rate to the Census Bureau's decennial census, and (4) the number of non-profit organizations. The data are provided by the Northeast Regional Center for Rural Development at Penn State University. The index data has been updated four times since 1990. As the index has adopted a new associational typology for the 20 0 0s data points, we only included data with a consistent typology. Thus, we have a social One of the independent variables is racial diversity. From data provided by the U.S. Census Bureau's American Community Survey (ACS), the racial diversity index was calculated as one minus the Herfindahl index of 7 ethnic groups (Non-Hispanic white, Hispanic, Black, Indian, Asian, Hawaiian, and two-more). It measures the probability that two people randomly chosen from a county belong to different ethnic groups (see, e.g., Alesina et al. 1999). Fig. 2 displays each county's average scores of diversity index on the map. Another key independent variable for the associated article is income inequality. Based on data also provided by the ACS, the variable measures the Gini index, which takes 0 for a perfectly equal distribution of income and 1 for perfectly unequal income distribution. Fig. 3 shows the geographical distribution of income inequality across the U.S. Counties. Table 2 shows the list of counties at both the top and the bottom ten ranks for these three key variables in 2014.
We further measured two different types of social capital by utilizing ten associational density variables. Scholars in the literature suggest that social capital has two different types:  bridging and bonding social capital [3][4][5] . According to Putnam [5] , bridging social capital can be defined as an open network that crosscuts, thus bridges, the existing social cleavages while bonding social capital is an inward-looking network that fortifies existing social interests. We labeled the former as 'Putnam-type' and the latter 'Olson-type' following Knack [3] . Based on Knack [3] and Rupasingha et al. [7] , we measured bridging social capital (Putnam-type) with the associational density of the first six organizations (religious organizations, civic organizations, bowling centers, fitness centers, golf clubs, and sports organization) and bonding social capital (Olson-type) with the same density of the remaining associations (business organization, labor union, political organizations, and professional organizations). Both Fig. 4 and Fig. 5 display each variable on the map respectively.
We included other correlates of social capital in the dataset. Following the typology provided by the U.S. Department of Agriculture, the Rural-Urban Continuum Codes (RUCC), the urban and rural variables were dummy coded by taking suburban counties as a reference category. The RUCC scheme provides nine categories that distinguish metropolitan counties by population, and nonmetropolitan counties by population and adjacency to the metro area. We utilized three categories of metropolitan counties to construct a dummy variable for urban counties while using two categories of nonmetropolitan counties that are not adjacent to the metro area to construct the rural indicative variable. It is often believed that rural areas provide a favorable environment for social capital. In the statistical estimation of the associated article, the remaining category was considered as suburban areas and omitted in the regression analysis.
All other county-level variables are compiled by utilizing the ACS database. For the income variable, we used the mean income in the past 12 months with the inflation-adjusted dollars. Then, we transformed the average household income with the natural logarithm. The dataset also has the educational attainment variable that measures the percentage of residents who have at least some college education per county. It is well known that socioeconomic status is positively associated with social capital. Because social capital would be difficult to form in a fluid county, we include the share of the non-migratory population in our dataset. From the ACS's county-to-county migration flow data, we calculated the percentage of non-movers out of the county population. In a similar vein, it is expected that the family-oriented community would

Experimental Design, Materials and Methods
Data construction for the associated article was constrained by the availability of data on social capital. Given social capital data for both 2009 and 2014, all relevant variables were compiled utilizing various data sources. Table 4 provides detailed information about all variables included in the dataset, including primary sources. These raw data are publicly available. However, putting them together to create correlates of social capital at the county-level requires careful handling of the data to align both temporal and geographical units. The Federal Information Processing Standards (FIPS), a four-digit county code, were used to match data points across different data sources. Furthermore, all data provided by the U.S. Census Bureau's ACS utilize the 5-year average estimates so that the dataset contains the least amount of missing values. With the constructed dataset, the associated article examined the variations of social capital at the county-level by utilizing two-stage multilevel regression analysis with year fixed effect [1 , 8 , 9] . Table 4 Variable description and data sources.