Data Sparseness and Variance in Accounting Profitability
研究了数据稀疏性如何导致方差分解中特定成分被高估，通过元回归和方差分解分析发现，以往研究结果的差异很大程度上受每组观测数影响，为战略管理文献中的相关争论提供了启示。
A central question in strategic management is why some firms perform better than others. One approach to addressing this question empirically is to decompose the variance in firm-level profitability into firm, industry, location, and year components. Although it is well established that data sparseness in variance decomposition studies can lead to overestimating particular variance components, little attention has been paid to sample size requirements in strategic management studies that have examined the nature of differences in firm profitability. We conduct a meta-regression and variance decomposition study and conclude that the variation in the results from previous studies is driven—to a considerable extent—by the number of observations per group within a component. Based on these findings, we draw conclusions regarding the validity and reliability of previous variance decomposition studies and provide implications for current debates in the strategic management literature.