Page 114 - 34-1
P. 114
In-House Provision of Corporate Services: The Case of Property-Casualty Insurers and In-House Actuarial
Loss Reserve Certification
Stock Indicator and for External ActuaryPost SOX×Publicly-traded Stock Indicator for
healthy insurers, under Hypothesis 3b.
3.2.3 Specification of Other Independent Variables
We include independent variables to test the three principal hypotheses that exist
21
in the literature concerning reserve errors. These hypotheses relate to (1) taxes (Grace
and Leverty, 2012; Nelson, 2000; Petroni, 1992), (2) income smoothing (Beaver et al.,
2003; Grace and Leverty, 2012), and (3) rate regulation (Grace and Leverty, 2010; Nelson,
2000). We also include other independent variables reflecting firm characteristics and
economic variables as described below. In order to streamline the following discussions,
we make no further distinction in the expectations for weak versus healthy insurers.
Overestimates of incurred losses understate total net income of the insurer and reduce
its current income tax liability. So everything else held equal, the tax hypothesis is that
firms will shelter income by overstating losses to reduce current taxes. Nelson (2000) and
Petroni (1992), among others, use an indicator variable to capture this effect. The indicator
variable is defined as one if the firm has a high tax rate (i.e., it pays positive income tax
in the current year) and is set to zero otherwise. We expect this variable to have a positive
coefficient in the regressions if insurers overstate losses to reduce current taxes.
Loss reserving errors also have been shown to be related to income smoothing. In
years with high unbiased income, insurers are expected to over-reserve more (under-reserve
22
less). Conversely, in years with low unbiased income, insurers would be expected to
under-reserve more (over-reserve less). We include a variable to capture effects of net
income smoothing in the model, which is defined as the difference between unbiased
net income in year t and reported net income in year t-1 divided by the absolute value of
reported net income in year t-1 (Gaver and Paterson, 2014). That is, unbiased net income
in the current year reflects net income without the loss reserving error; it is an estimate of
actual net income. We expect this variable to be positively related to the loss reserve error.
21 The literature also contains hypotheses about weak insurers. We take this into account by estimating
separate regressions for weak versus healthy insurers.
22 Unbiased net income is net income gross of the reserve error. This variable is used in prior research
(e.g., Gaver and Paterson, 2014).
106