Regression里面确实是用Benchmark interest rate回归的。
下面句子截自原版书:
Empirical duration is a measure of interest rate sensitivity that is determined from market data. A common way to calculate a bond's empirical duration is to run a regression of its price returns on changes in a benchmark interest rate.
因为衡量的是对Benchmark interest rate变动的敏感度,所以回归出来的Empirical duration是一个Curve duration数据,他只能和Effective duration相比,因为Effective duration也衡量的是对Benchmark interest rate变动的敏感度,也是Curve duration数据。
严格说是不能和Modified duration比,因为Modified duration是对YTM的敏感度,是一个Yield duration。
所以我们才有以下和Effective duration比大小的结论:
empirical duration is smaller than the theoretically based effective duration.
因为是Empirical duration和Effective duration相比,而实际的债券价格变动幅度又是小于计算Effective duration时的变动幅度的,所以Empirical duration自然比Effective duration要小。