Statistické metody modelování a odhadu budoucího vývoje plodnosti

As computer software develops, it provides more space for the statistical modelling of demographic processes. Yet the statistical ex-post evaluation of the success rate of these models has been almost completely neglected in publications. The aim of this dissertation thesis is to evaluate, based on statistical tools, the usefulness of the most widespread models used in projecting age-specific fertility rates in a country showing volatile fertility rates. Then, based on the findings of this analysis, use these models to estimate the most likely future developments. Data from the Human Fertility Database and models often used in publications and practice were used, namely the Lee-Carter model, functional demographic model and quadratic spline model to project fertility rates. In the thesis, there is evaluated the success rate of ex-post projections by comparing age-specific fertility rates, total fertility rate, other synthetic indicators of fertility and loss functions. There are also compared results with period projections compiled by The Czech Statistical Office. Most scenarios are able to correctly capture the basic fertility trends, but the realness of the intensity of these changes varies, depending on the scenario. The work provides a quantitative proof, that the selected variants of the models alone in a country with a volatile fertility level cannot be used to project fertility without further expert intervention. In the final chapter, the models are modified through the use of other approaches to modeling time series of parameters in order to create realistic population forecasts in the Czech Republic. It still primarily depends on the projection author’s expertise in deciding how to model time series and how to assess the realness of the model results. Based on the correctness of his estimate, the model can then serve as a very useful tool. The practical contribution of the dissertation to the monitored issue is crucial, given that it evaluates and expands the possibilities of statistical modeling of fertility in the Czech Republic. There is no doubt that it can serve as a valuable source of information for authors of future population estimates and those interested in evaluating the development of fertility using models.