Overview
Donald T. Campbell (1916–1996) was an American social scientist and philosopher of science whose work on experimental and quasi-experimental research design became foundational across psychology, education research, and economics. His systematic taxonomy of "threats to validity" — distinguishing internal validity (did the treatment cause the outcome within the study?) from external validity (do the results generalize?) — is the standard framework for evaluating causal claims in non-randomized settings. His collaboration with Julian Stanley (Experimental and Quasi-Experimental Designs for Research, 1966) and later with Thomas Cook (Quasi-Experimentation, 1979) codified the quasi-experimental design tradition that natural experiment economists explicitly built on.
Key Contributions / Features
- Threats to internal validity: Catalogued the systematic reasons why a before-after comparison might yield a spurious treatment effect — history (omitted variables), maturation (trends in outcomes), testing, instrumentation (mismeasurement), regression to the mean (selection on extreme values), selection, attrition, and interactions among these. Meyer (1995) adapted this list for economics as 9 internal threats.
- Threats to external validity: Population × treatment interaction, setting × treatment interaction, and history × treatment interaction — the three fundamental limits on generalizability from any particular study.
- Quasi-experimental designs: Together with Stanley (1966) and Cook (1979), described designs that approximate randomization without random assignment — interrupted time series, regression discontinuity, untreated comparison group (DiD). This literature predates the economics natural experiment movement by decades and is its direct intellectual antecedent.
- "Reforms as experiments" (1969): Argued that government policy changes can and should be evaluated as natural experiments; anticipated the quasi-experimental turn in applied economics.
- Validity taxonomy refined: Cook and Campbell (1979) extended the list to include construct validity (what was actually manipulated?). Cook and Shadish (1994) further updated and refined the framework.
Related
Sources