The purpose of this study is to investigate whether property-casualty insurance companies exercise accounting discretion when reporting the claim loss reserve, in response to regulatory, tax, signaling, and financial reporting incentives.; The loss reserve that would have been reported in the absence of managerial discretion is first modeled by using a latent variable technique, the Kalman filter. The model relies on the institutional characteristics of the industry and on the properties of the accounting system, but it is flexible enough to be applied to other accruals in any industry. The results show that the Kalman filter achieves an 80% success rate in detecting reserve manipulation and its direction, and that it also has good predictive properties.; In a second stage, the estimated discretionary component of the loss reserve is regressed on exogenous variables that proxy for the various incentives to exercise accounting discretion. The results of these tests indicate that financially weak insurers understate the loss reserve in order to avoid regulatory scrutiny. The incentive to appear solvent is quite strong and overpowers all the other incentives when firms are financially weak. Second, financially strong insurers overstate the loss reserve in order to pay lower taxes. Third, unlike weak insurers, strong insurers use the loss reserve to signal future profitability. Fourth, insurers understate the loss reserve, or overstate it less, in order to obtain more competitive rates from regulators and attract more business.; This paper differs from previous research on loss reserves along several dimensions. First, it is the first study to consider jointly the incentives just mentioned. It is also the first time that signaling is analyzed as an incentive in the context of loss reserving behavior. Second, more powerful tests of discretionary behavior have been designed by recognizing that the financial condition of an insurer can greatly influence its incentives. Third, this study tests for discretionary behavior on current data. Previous research designs can only conduct tests of discretionary behavior occurring five or more years in the past. Finally, the paper contributes to the accruals management literature by providing a new method to model discretionary accruals.
展开▼