Generalizability in the Social Sciences

• In Event: Poster Session: Formal Political Theory; Political Methodology & Canadian Politics
  In Poster Session: Poster Session: Political Methodology

Abstract

o Social scientists widely agree that research findings ought to be “generalizable.” However, this term lacks a consistent definition. I show that methods textbooks often use the same term to capture two different underlying ideas about generalizability. The first is statistical generalizability, or the ability to generalize from a point estimate of a variable in a sample to the population. The second is cumulative generalizability, or the ability to identify variables which affect outcomes of interest across different populations, locations, and times. Research need not be statistically generalizable to be cumulatively generalizable. I explore the differences between these terms using examples from American politics, comparative politics, and development economics.

Author

• Rachel Strohm, UC Berkeley