Citation:
Gill MZ, Hall AB. Causal Inference with Text: Judicial Identity and Appellate Opinions. 2012.
Notes:
We present a simple non-parametric method for causal inference with textual data---i.e., causal inference when texts serve as the response variable---by unifying longstanding methods from statistics and information retrieval. Using the method, we investigate how the random assignment of female judges to US appellate panels influences the language of majority opinions. We find that the random assignment of a female judge shapes the vocabulary of majority opinions in ways that are statistically differentiable from those produced by all-male panels. In line with cross-sectional studies that find a conditional effect of identity on various legal issue areas, treatment effects are greatest in areas of law with gendered significance.