Abstract
This article addresses the appropriate modeling of losses for the insurance sector. In fact, solvency 2 framework has suggested some formulas to evaluate losses and solvency capital using an internal approach. However, these formulas where derived under the assumption of independent losses. Thus, the amount of capital may be inaccurate when losses are dependent, which is the case in practice. The aim of this paper is to investigate temporal dependence structure among claim amounts (losses). For that, a novel model named autoregressive conditional amount (ACA) model handling the dynamic behavior of claim amounts in insurance companies is proposed. Results show that ACA, models allow to predict accurately the future claims. Moreover, a measure of risk namely value at risk (VaR) ACA that could hedge daily dependent losses is provided. By backtesting techniques, empirical results show that the new VaR ACA can efficiently evaluate the coverage amount of risks. (C) 2016 Elsevier B.V. All rights reserved.