Surveys and Quantitative Analysis

The survey is the workhorse of quantitative sociology. A well-designed survey measures attitudes, behaviors, and social positions across a large, defined population, yielding estimates whose uncertainty can be quantified. Mid-twentieth-century sociologists built the infrastructure that still anchors the field: Paul Lazarsfeld's Bureau of Applied Social Research at Columbia pioneered panel studies of voting and media effects (The People's Choice, 1944); the University of Michigan's Survey Research Center conducted the American National Election Studies from 1948; and in 1972 the General Social Survey (GSS), directed by James Davis and later Tom Smith at the National Opinion Research Center (NORC), began its nearly biennial measurement of American social attitudes. The European Social Survey (ESS), launched in 2001 under Roger Jowell's direction, extended a comparable instrument across more than thirty European countries with strict cross-national comparability standards.

These long-running surveys matter because they let sociologists track change over time and across countries: the declining religiosity of European publics, the rise and partial decline of tolerance toward same-sex relationships, the widening partisan gap in American views of climate, immigration, and institutions, and countless other trends. Public-use microdata let thousands of researchers reanalyze the same data, making survey evidence one of the most democratized research resources in the social sciences. Robert Putnam's Bowling Alone (2000), Arlie Hochschild's The Second Shift (1989), and Claude Fischer's America Calling (1992, using Census data) all drew on large survey datasets to develop influential arguments about community, family, and technology.

Surveys are used at many scales. Cross-sectional surveys measure a population at one point in time. Repeated cross-sections (like the GSS or ESS) measure comparable populations at multiple times. Panel surveys follow the same individuals over time — the Panel Study of Income Dynamics (PSID) began in 1968 and has followed families across three generations; the German Socio-Economic Panel (SOEP) has done similar work since 1984. Cohort studies track individuals born in the same period. Each design supports different inferences: repeated cross-sections detect change in populations, while panel designs can distinguish change within individuals from compositional change in the population.

The survey's analytical power depends on whether its sample is drawn in a way that supports inference about the target population. That question — how to sample — has been the central technical concern of survey methodology since Jerzy Neyman's 1934 paper that established probability sampling as the gold standard.