Forecasting the Truck Fleet Size by Regions of the Russian Federation

<p>This paper is devoted to forecast of the truck population in the regions of the Russian Federation. Task may be solved using simple classic linear model with technique of ordinary least squares. However in this paper sophisticated procedures of data analysis are used to improve precision of the forecast. Input data looks like a panels of regions with time and place dimensions. Short description of the territorial and economic features of the regions and useful mathematics methods for panel data processing are given in the paper. In the paper some modern econometric models based on panel data processing were developed. For example dynamic simultaneous linear equations system, spatial econometrics, nonparametric methods. Search of the best model consists in the sequential complication of considered models. We try to improve precision of the forecast in different ways, e.g. trucks characteristics clusterization of the regions, keeping of the spatial correlation and including nonlinear nature of the relation between truck population and the main factors. Input data have set of drawbacks: nonuniformity of the regions, short nonstationary time series including periods of economics instability(economic crisis), small spatial sampling. In those conditions use of complicated econometric tools is associated with certain difficulties and limitations. Despite of described features of the task we provide the model which make it possible to obtain more precise forecast of truck population in the region compared with classic linear model. Forecast the truck population allows to predict so significant indexes as oil and petroleum consumption, workload and the wear and tear of roads, requirements to infrastructure and automobile services. Work has practical importance and has been carried out with support of research group &laquo;Petromarket&raquo;.</p>