Page 121 - 34-1

NTU Management Review Vol. 34 No. 1 Apr. 2024

All of the variables (except the in-house indicator and the interaction terms with the in-
house indicators) are used in the propensity score estimation. In the second step, for each
insurer employing an in-house actuary, we identify an insurer with the closest five digits
35
of the propensity score that has not used an in-house actuary. This approach enables us to
form matched pairs with the smallest propensity score differences (i.e., most similar along
a set of firm characteristics) but the greatest difference in the in-house actuary choice
among the three subsamples. Third, we combine insurers using in-house actuaries and their

matched observations for all three subsamples. Eventually, we are able to identify 1,411
observations employing in-house actuaries and 1,411 matched observations. With the PSM
sample, we estimate equation (1) to control for any potential remaining differences in
36
insurer characteristics between the treatment and control groups.

5. Results and Discussion

This section discusses the results of the regression models. But first, the results of the

Chow tests are presented, and descriptive statistics are discussed.

5.1 Results of Chow Tests for Weak versus Healthy Insurers
As indicated previously, weak firms may behave differently than healthy ones (e.g.,
Gaver and Paterson, 2014; Petroni, 1992). If this were the case, then making inferences
from a regression that includes both weak and healthy insurers in the sample could be

misleading. If weak firms systematically behave differently from healthy ones, the results

35 If the selected insurer employs an actuary from an external independent firm, we find a match. If not,
we next try to match on four digits of the propensity score. This process continues down to a one-
digit match on propensity score for those that remain unmatched. Some observations could not be
matched.
36 A simple univariate t-test (Wilcoxon rank-sum test) of the differences in means (medians) suggests
that there are no significant differences in most variables between matched pairs except that insurers
employing in-house actuaries have a larger size (larger premiums written) than those that do not at the
5% significance level. Insurers employing in-house actuaries have less reinsurance usage and lower
values for the net income smoothing variable than those that do not at the 5% significance level. Also,
insurers employing in-house actuaries have a higher mean (but lower median) for the line of business
Herfindahl index, and the difference is significant at the 5% level.

113