Demographic Methods for the Sociology of Religion 83
advantage in fertility and migration should have resulted in between 32 and 35 percent
of adults being Catholic in the late 1990s. The steady 25 percent that is observed over
and over in national surveys implies that something is interfering with the growth of
the Catholic population. In fact, 33 percent of American adults interviewed in the late
1990s were raised Catholic (according to the General Social Survey). Ten percent had
left the Church – half to Protestant denominations, nearly half to no religion at all,
and the small remainder to non-Christian religions. The demographic analysis does
not explain the trend in this case. It points to the phenomenon to be explained. But
without reference to demography we are not aware that there is anything to explain.
Once we see the demographic advantages that the Catholic Church had for most of
the twentieth century, its constant proportion in the population becomes a puzzle to
be solved.
DATA NEEDS AND RESOURCES
Demographic research on religion has long been hampered by the lack of religion data
in the census. Demographers thrive on fine-grained comparisons over long periods of
time. The catalogue of religious data is very thin on both counts. Other countries’ cen-
suses routinely record the prevalence of religion in the population. In nations where
religious divisions overlap with political conflict – I already mentioned Israel and North-
ern Ireland and it is true in Canada, Australia, and the Netherlands as well – census re-
turns are anxiously monitored for signs of advantage or disadvantage. The U.S. census
does not ask about religion, initially because census officials and congressional leaders
in the late 1930s thought that it was a bad idea to have lists of Jews stored in one
place and more recently because census items must now be tied to the evaluation of
specific social and economic policies. The U.S. Bureau of the Census did conduct sur-
veys of religious bodies in 1906, 1916, 1926, and 1936. But inconsistent definitions of
membership across denominations and over time limit their usefulness.
The typical survey is sufficient to track the relative sizes of the Protestant and
Catholic populations, the population with no religion, and some of the larger Protestant
denominations (e.g., Baptists, Methodists, and Lutherans). But groups that are less than
5 percent of the adult population – interesting groups like Jews, Muslims, Mormons,
and members of the traditionally African-American churches – are impossible to assess
reliably in a single survey of eight hundred to two thousand adults, and few researchers
have the resources to interview more than two thousand adults.
The General Social Survey, an ongoing project that used to interview about fifteen
hundred adults every year and now interviews three thousand adults in even-numbered
years, has become an invaluable resource for religious researchers interested in these
churches that comprise less than 5 percent of adults (e.g., Smith 1990 and the GSS
website: http://www.icpsr.umich.edu/gss)..) The GSS does not get any more Jews, Muslims, or
Jehovah’s Witnesses than any other survey of that size, of course, but because it has
such high standards of keeping the design and questions the same year after year,
data from several years can be combined to gain insight about these smaller religions
and denominations. Since its inception, but especially since 1983, the GSS also has
taken pains to distinguish precisely among denominations as similar-sounding (but
doctrinally very different) as the United Church of Christ and the Church of Christ, the
Church of God and the Church of God in Christ, and the Southern Baptist Convention,