Publication Bias in the Social Sciences

Abstract Publication bias is essentially a form of selection bias. If published studies are a non-random and systematically biased sample of the population of conducted studies, then meta-analyses will not be able to recover true population parameters of social phenomena. This chapter discusses two types of publication bias—selection for statistical significance, and selection on inference—and provides evidence that both types of publication bias have occurred in the academic literature. The chapter reviews how these types of publication bias can bias inferences from the literature, and the chapter discusses procedures that authors, journal editors, and others can adopt to limit the negative effects of publication bias.