摘要
We examine the regression model used in the literature to test the signaling model in Leland and Pyle [7] and find that the model may produce positive and significant coefficients for the independent variables even when the inputs are randomly generated. The problem is illustrated by two examples: a theoretical proof based on a uniform distribution and a bootstrap simulation based on actual data. Our studies show that the misleading results are caused by some natural constraints in the regression model.
- 出版日期2014-6