Since the efficiency score lies between 0 and 1, using ordinary least square (OLS) regression in the second stage might lead to estimated coefficients that are biased. As Simar and Wilson [32] point out, direct regression analysis is invalid due to the unknown serial correlation among the efficiency scores. Simar and Wilson [33] confirm that applying OLS in the second-stage estimation is consistent only under very peculiar and unusual assumptions about the data-generating process that limit its applicability. Put differently, truncated regression provides consistent estimation in the second-stage estimation. In the second stage, we thus adopt truncated regression with a bootstrapping approach introduced by Simar and Wilson[32] and [33] to examine whether exogenous factors, i.e., intellectual capital variables in this study, affect the corporate performance of Chinese insurance companies.
This study assumes and tests the following specification: