The main source for marital and employment data is a data series that starts with a 1 percent sample of the decennial census for 1960 and continues from 1968 through 2010 using the March editions of the CPS. The CPS surveys for 1961 through 1967 are not used because the coding for occupations in those years was inadequate to identify who met the Belmont and Fishtown occupational criteria.
The 1960 census sample for persons ages 18–65 was 986,917, of whom 402,889 were whites ages 30–49 (the primary sample for analysis). Annual sample sizes for the CPS for persons ages 18–65 from 1968 through 2010 ranged from 130,124 to 209,802, with the samples of whites ages 30–49 ranging from 22,345 to 48,134.
Both the census data and the CPS data are available online through IPUMS (Integrated Public Use Microdata Series), managed by the Minnesota Population Center, which is always cited as the source for the census and CPS data in the graphs. The URL for IPUMS is http://cps.ipums.org/cps/.
The GSS has been conducted since 1972 by the National Opinion Research Center at the University of Chicago. Available online, the GSS surveys include a wide variety of demographic, behavioral, and attitudinal questions, including many that have been asked identically for all of the surveys. It is the most widely used American attitudinal database.
Sample sizes for the GSS are much smaller than for the CPS. From 1972 through 1993, the entire sample ranged from 1,372 to 1,860. From 1994 through 2008, it ranged from 2,023 to 2,992, with a special augmented sample of 4,510 in 2006. The number of whites ages 30–49 has ranged from 413 to 1,176. The URL for the GSS is http://www.norc.org/GSS+Website/.
The NCHS collects data on all births in the United States. For the analyses of out-of-wedlock births in chapter 8, I used random samples of 200,000 cases drawn from every other year. The URL for the Vital Statistics system at NCHS is http://www.cdc.gov/nchs/nvss.htm.
For the zip code analyses of the 2000 decennial census, I employed the American FactFinder tool provided on the website of the Bureau of the Census. This resource enables anyone to download census data broken down by zip code (or many other aggregations). As I write, the Census Bureau is in transition from an old to a new version of American FactFinder, but you can easily find it on the Census Bureau’s main web page, http://www.census.gov/. For locating the geographic borders of zip codes, I used hipcodes.com, supplemented by Google Maps.
For the 1960 data on census tracts, I used the PDFs of the 1960 census volumes, available on the Bureau of the Census website, and the ASCII files for the 1960 census available from the Interuniversity Consortium for Political and Social Research (ICPSR). The ICPSR’s URL is http://www.icpsr.umich.edu/icpsrweb/ICPSR/.
The day after submitting the draft of Coming Apart, I was shown the Social Explorer website, which I recommend to all other scholars who want to do this kind of analysis, and which would have saved me weeks of work. Its URL is http://www.socialexplorer.com/pub/home/home.aspx
The NLS comprise a family of surveys sponsored by the Bureau of Labor Statistics and conducted by the Center for Human Resource Research at Ohio State University. For constructing the trendline for children still living with both biological parents when the mother was age 40, I integrated data from the Young Women sample (initial sample size was 5,159) and the Mature Women sample (initial sample was 5,083), both of which were followed from 1968 to 2003, and the Fertility sample of the 1979 cohort of the National Longitudinal Survey of Youth (initial sample size was 6,283), which is still being followed as I write in 2011. All of the surveys are available online. The NLS’s URL is https://www.nlsinfo.org/.
The UCR is an ongoing compilation of national offense and arrest statistics conducted by the Federal Bureau of Investigation and published annually since 1935. Data for the years from 1995 onward are available online. The UCR’s URL is http://www.fbi.gov/about-us/cjis/ucr/ucr.
Inmate surveys, designed by the Bureau of Justice Statistics and conducted by the Bureau of the Census, have been conducted in 1974 (state), 1979 (state), 1986 (state), 1991 (federal), 1997 (state and federal), and 2004 (state and federal). The samples of males used in the analysis were 8,741 in 1974, 9,142 in 1979, 11,556 in 1986, 11,163 in 1991, 14,530 in 1997, and 11,569 in 2004. The surveys are available online at http://www.icpsr.umich.edu/icpsrweb/ICPSR/.
As noted in chapter 8, the vertical axis in a graph is based on the minimum and maximum values of the variables being plotted in a given graph, with a minimum range of 20 percentage points.
The smoothed curve that runs through the actual data points on each graph using the CPS serves the same purpose as a moving average, to give a visual sense of the overall trend. They are created by a procedure known as “locally weighted scatterplot smoothing,” originated by W. S. Cleveland in 1979 and abbreviated as LOESS or LOWESS.1 In an ordinary moving average, the smoothed value is the mean of however many data points are specified. In a LOESS plot, the smoothed value is calculated by giving the most weight to the adjacent points in the data and less weight to distant ones (or no weight, depending on the bandwidth that the analyst has specified).
For the CPS data, LOESS serves the cosmetic purpose of smoothing the annual data. For the GSS, with its much smaller sample sizes, LOESS serves a more important function of maximizing the available information, producing trendlines that are more confidently interpretable than ordinary least squares (OLS) regression trendlines or moving averages. I follow the procedure used by Claude Fischer and Michael Hout in Century of Difference. It has three steps, as summarized by Fischer and Hout in their appendix A, which describes the procedure in detail:
In graphs using GSS data, I do not show the survey-by-survey data points. The sample sizes are too small, especially for Belmont, to produce reliable estimates for a given survey.
ONE FINAL NOTE on presentation: In Coming Apart, I continue to treat singular third-person pronouns according to a rule I have been unsuccessfully advocating for more than a quarter of a century: Absent a reason to do otherwise, use the gender of the author or, in a coauthored book, the gender of the principal author.