Underwriting cycles in non-life insurance: A study of the dynamics of the combined ratio and the determinants of premiums.

Summary Despite the considerable amount of theoretical and empirical work aimed at analyzing non-life insurance underwriting cycles. No single clear conclusion seems to emerge. The lack of consensus seemed to us to be due to the inadequacy of the adapted linear modeling and the neglect of the cointegration properties applied to both time series and panel data. This thesis has therefore aimed at re-examining the underwriting cycles while focusing on the dynamics of the combined ratio and on the determinants of premiums for the case of France and for other countries. In the first chapter, we presented a review of the literature on the subject. In the second chapter, we analyzed the underwriting cycles and the determinants of premiums for the French aggregate sector using non-linear time series econometrics. In the third chapter, we used recent developments in panel data econometrics, taking into account non-stationarity and non-linearity, to conduct a disaggregated analysis by French line of underwriting cycles and premium determinants, and to conduct a comparative analysis of premium determinants in an international framework. The results obtained for the French aggregate sector led us to conclude that the cyclical phenomenon disappeared in France as early as 1989 and that the dynamics of the combined ratio is instead modeled by a smooth transition regression (STR) model. The pricing of premiums seems to change from 1985 onwards and the growth rate of premiums seems to be represented by a smooth transition error correction model (STECM). Second, the estimation of static and dynamic panels allowed us to detect similarities in the dynamics of the combined ratios of the non-life insurance lines. Similarly, the estimation of cointegration relationships on panel data allowed us to identify differences in the determinants of premiums of non-life insurance lines. Finally, the comparative analysis by country showed that the growth rates of the countries' premiums are reproduced by a smooth transition error correction model on panel data (PSTECM).