Applying Shock-Based versus Panel Data Methods in Corporate Finance and Accounting Research: Evidence from a Case Study of Korea
58 Pages Posted: 15 Sep 2012 Last revised: 23 Nov 2018
Date Written: November 23, 2018
Only rarely do researchers use multiple methods to study the same research question. We provide case-study evidence on the value of doing so, and the limited reliability of both panel methods and simpler causal designs, as guides to actual causation. We study the effect of 1999 Korean reforms, which required large firms (assets over 2 trillion won) to change their board structures, relative to smaller firms, on a number of finance and accounting outcomes which could be plausibly affected by this shock to governance: Tobin’s q, disclosure, investment, growth, and abnormal accruals. We first exploit the Korea shock using a combined difference-in-differences (DiD)/regression discontinuity (RD) design, with annual “leads and lags” estimates of the treatment effect. This design likely comes as close to a randomized experiment as one is likely to achieve, short of an actual experiment. With this design, the shock predicts an increase in a "Disclosure Subindex," but no significant change in other outcomes. We compare these “benchmark” results to those from simpler causal designs exploiting this shock (DiD, RD, and instrumental variables) and to classic panel designs. With panel methods, a Korea corporate governance index also predicts higher Tobin's q, slower growth, lower absolute abnormal accruals, and more extensive MD&A disclosure. However, only the Tobin's q results survive with simpler causal methods, and those weaken with our benchmark design. We thus provide case study evidence that panel methods can often provide apparent false positives; and simpler causal methods also tend, less strongly, to generate apparent false positives. We also illustrate how using multiple causal designs can provide insight not available from a single design.
Keywords: corporate governance, board structure, financial reporting quality, investment, sales growth, earnings management, accrual quality, audit committees, Korea, causal inference, panel data, difference-in-difference, instrumental variables, regression discontinuity.
Suggested Citation: Suggested Citation