To determine whether multistate insurers avoid state taxes through strategic allocations in the statutory reports, insurance prices are regressed on measures of state taxes, controlling for previously documented price determinants:
PRICEij = $0 + $1 tax measuresij + $2 control variablesj + control variablesi + eij for all insurers i and states j.
The dependent variable, PRICE, is the insurer’s premiums less policyholder dividends divided by losses incurred (including estimated unpaid claims at year-end) as reported in Schedule T of the annual regulatory reports. This premium-to-loss ratio is recognized within the insurance industry as the ex post price of a dollar of coverage. A superior dependent variable would be the ex ante price of insurance; however, we are unaware of readily available data that would permit us to compute ex ante prices. Using an ex post measure is only problematic if unanticipated catastrophes occur disproportionately in states with similar tax structure. To mitigate this concern, the study is limited to 1993, a year without major earthquake or hurricane damage. loans
For the insurer-states that do not face income taxation, an ideal dependent variable would include only premiums because losses do not affect tax liabilities in those states. However, if the sample is restricted to those insurer-states and premiums are used as the dependent variable, a scalar is needed to control for heteroskedasticity, and the most appropriate scalar is losses. Thus, the premium-to-loss ratio probably is the best dependent variable for assessing premium tax avoidance. Moreover, for the insurer-states facing income taxes, losses are the largest deduction. The ratio also controls for the possibility that some premiums are inseparable from their associated losses for shifting purposes and thus of limited use in avoiding income taxes. On the other hand, to the extent insurers mask premium tax avoidance by shifting both losses and premiums, PRICE may reduce the power of the tests to detect premium tax management.