I briefly review the emergence of “design‐based” research methods in labor economics in the 1980s and early 1990s. These methods were seen as a partial solution to the problems of credible inference identified by Ashenfelter (1974), Leamer (1978), Hendry (1980), and others.  Designed‐based studies typically use a simplified one‐equation model of the outcome of interest – in contrast to model‐based studies that specify a data generating process for all factors determining the outcome.  I discuss some of the strengths and weaknesses of the design‐based approach and the value of such research in the field.