Static and Dynamic Panel Data Analysis of Insurance for Escalating Climate Change (US)

This paper examines the relationship of climate shocks impact on private climate insurance by a number of econometric models. The least squares estimator produces biased estimates of such impacts because it ignores time-invariant unobservable influences; the FD estimator has a similar drawback with a more parsimonious modeling by using lagged dependent variables in first differences since the FD lagged are also endogenous. We employed a two-stage GMM dynamic Arellano and Bond (1991) model with internal lagged instruments defined in levels for endogenous FD lags. We applied these models to annual panel data sets of climate shocks of fire and water disasters and insurance prices for the US states over 2003-2013. Our best estimates of climate effects obtained from the dynamic panel model suggest the frequency of both water- and fire-based disaster events are positively and significantly related to insurance premiums over the period. In most models, the correlation with disaster incidence is strongest at the second lag, indicating that there is a reaction period after the occurrence of disaster events before they are fully absorbed in insurance prices. The dynamic panel test for the null hypothesis of second and third-order serial correlation, and that for weak instruments are both rejected. There is some suggestive evidence from descriptive data that the positive empirical relationship between environmental disasters and the price of climate insurance is likely to be driven by increasing volatility of climate shocks, lending support to the concern that the growing impact of climate change will be accompanied by increasing insurance costs and weaker private insurance coverage