Methodological Variation in Empirical Corporate Finance
研究发现顶级金融期刊的公司金融回归中方法论差异很大，这种差异可能导致大量统计显著结果的报告，例如通过10项常规方法选择，研究者可使超过70%的随机变量成为杠杆率的显著决定因素。
Abstract I document large variation in empirical methodology in corporate finance regressions in top finance journals. Although methodological variation allows for customization of empirical tests to fit specific theories, it can also enable excessive reporting of statistically significant results. For example, given discretion over 10 routine methodological decisions, a researcher could report that over 70% of randomly generated variables are statistically significant determinants of leverage at the 5% level. The methodological decisions that affect statistical significance the most are dependent variable selection, variable transformation, and outlier treatment. I discuss remedies that can mitigate the negative effects of methodological variation.