Holger Spamann, On Inference When Using State Corporate Laws for Identification (Harv. L. Sch. John M. Olin Ctr. Discussion Paper No. 1024, Eur. Corp. Governance Inst. Fin. Working Paper No. 644/2019, Dec. 17, 2019).
Abstract: A popular research design identifies the effects of corporate governance by (changes in) state laws, clustering standard errors by state of incorporation. Using Monte-Carlo simulations, this paper shows that conventional statistical tests based on these standard errors dramatically overreject: in a typical design, randomly generated “placebo laws” are “significant” at the 1/5/10% level 9/21/30% of the time. This poor coverage is due to the extremely unequal cluster sizes, especially Delaware's concentration of half of all incorporations. Fixes recommended in the literature fail, including degrees-of-freedom corrections and the cluster wild bootstrap. The paper proposes a permutation test for valid inference.