Inaccurate Statistical Discrimination: An Identification Problem

Abstract

Discrimination has been widely studied in the social sciences. Economists often categorize the source of discrimination as either taste-based or statistical—a valuable distinction for policy design and welfare analysis. In this paper, we highlight that in many situations economic agents may have inaccurate beliefs, and demonstrate that the possibility of inaccurate statistical discrimination generates an identification problem for attempts to isolate the source of differential treatment. We introduce isodiscrimination curves —which represent the set of preferences and beliefs that generate the same level of discrimination—to formally outline the identification problem: when not accounted for, inaccurate statistical discrimination can be mistaken for taste-based discrimination, accurate statistical discrimination, or their combination. A review of the empirical discrimination literature in economics, spanning 1990-2018, reveals the scope of this issue. While most papers discuss and attempt to distinguish between taste and statistical discrimination, a small minority—fewer than 7%—consider inaccurate beliefs in the analysis. An experiment illustrates a methodology for differentiating between the three sources of discrimination, demonstrating the pitfalls of the identification problem while presenting a portable solution.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 147 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.