Working Paper: CEPR ID: DP15175
Authors: Gilles Chemla; Christopher Hennessy
Abstract: Causal evidence from random assignment has been labeled "the most credible." We argue itis generally incomplete in finance/economics, omitting central parts of the true empirical causalchain. Random assignment, in eliminating self-selection, simultaneously precludes signaling viatreatment choice. However, outside experiments, agents enjoy discretion to signal, thereby caus-ing changes in beliefs and outcomes. Therefore, if the goal is informing discretionary decisions,rather than predicting outcomes after forced/mistaken actions, randomization is problematic.As shown, signaling can amplify, attenuate, or reverse signs of causal e¤ects. Thus, traditionalmethods of empirical finance, e.g. event studies, are often more credible/useful.
Keywords: signal; random assignment; causal effect; selection; investment; corporate finance; CEO; household finance; government policy
JEL Codes: D82; G14; G18; G28; G3; E6; J24
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| random assignment (C90) | self-selection (C52) |
| absence of signaling (Y70) | incorrect conclusions about causal effects (C20) |
| signaling effects (D85) | alter beliefs and outcomes (D91) |
| random assignment (C90) | misguided investment decisions (G11) |
| traditional methods (C90) | credible estimates of causal effects (C51) |
| signaling (L96) | faulty conclusions about efficacy of various policies (E65) |
| partial causal effects and total causal effects (C32) | inform optimal decision-making (D87) |