Working Paper: NBER ID: w28718
Authors: Catherine Balfe; Patrick Button; Mary Penn; David Schwegman
Abstract: Audit correspondence studies are field experiments that test for discriminatory behavior in active markets. Researchers measure discrimination by comparing how responsive individuals ("audited units") are to correspondences from different types of people. This paper elaborates on the tradeoffs researchers face between sending audited units only one correspondence and sending them multiple correspondences, especially when including less common identity signals in the correspondences. We argue that when researchers use audit correspondence studies to measure discrimination against individuals that infrequently interact with audited units, they raise the risk that these audited units become aware they are being studied or otherwise act differently. We present the result of an audit correspondence study that demonstrates how this detection can occur when researchers send more than one correspondence from an uncommon minority group. We show how this detection leads to attenuated (downwardly biased) estimates of discrimination.
Keywords: No keywords provided
JEL Codes: C93; J15; J16; J71; K42; Z13
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Sending multiple pieces of correspondence (L87) | Increases likelihood of detection (Y50) |
| Increased likelihood of detection (Y50) | Biases estimates of discrimination downward (J79) |
| Sending more than one correspondence from infrequent identity groups (Z13) | Heightened risk of detection (Y50) |
| Heightened risk of detection (Y50) | Alters behavior of audited units (C92) |
| Exposure to multiple infrequent identity signals (C32) | Significantly raises likelihood of detection (Y50) |
| Receiving more than two pieces of correspondence from same-gender couples (J12) | Changes behavior of audited units (M42) |
| Changes in behavior of audited units (C92) | Biases discrimination estimates towards zero (J79) |
| Restricting analysis to instances with at most two correspondence pieces (C38) | Discrimination estimate was greater (J79) |
| Detection risk (D80) | Undermines internal validity of audit correspondence studies (C83) |
| Number of correspondences sent (L87) | Mitigates detection risks (G18) |