The goal of this article is to apply panel data approach to the analysis of claim frequency in automobile insurance. The model which is constructed estimates the influence of particular characteristics of the insured on their insurance loss number, but it also enables identification of the hunger for bonus effect. Panel data approach allows for identification of drivers' individual effects that influence their driving quality, but cannot be quantified directly, such as for example tendency to drive fast. This is done thanks to repetitive observation of the same individuals. Having information on their number of losses claimed in different bonus-malus system classes, it is possible to separate their individual skills from the hunger for bonus phenomenon, as well as identify the scale of the latter, which differs in particular classes. Chapter one is an introduction. In chapter two main benefits from the use of panel data have been described. Recent publications considering the topic are mentioned as well, with emphasis on the differences between other authors' approaches and this one. Chapter three contains a brief description of the methods applied, which are Poisson regression mixed models. In chapter four the basic model is adjusted to the conditions of hunger for bonus and it is shown, how this phenomenon is identified. In chapter five empirical analysis based on the real market data of approximately 21 thousand observations is done. The model is estimated and the conclusions are discussed with a short simulation study of the insurance company financial state.
The aim of this article is to present certain problems concerning construction and estimation of the family design models used in the behavior genetics research. As there are only two Polish examples of empirical research applying family studies method, we discuss main methodological and statistical problems regarding the models presented in these works, including certain doubts of ours considering construction and assumptions made. In the above mentioned articles their Authors use path model to describe connections between latent variables: genetic (G), common environment (C) and specific environment (E) and the observed phenotype variables. In this article we discuss the problem of particular parameters' constancy, relationship (or its lack) between particular variables, as well as problems of adequate estimation method, inferences based upon the constructed models and the application of given statistical methods used draw a conclusions on the models' fit.
JavaScript is turned off in your web browser. Turn it on to take full advantage of this site, then refresh the page.