The article analyses the historical course of quarterly time series of employed and unemployed population of the Czech Republic at the age of 15 years and more for the period 1993-2018. The objective of the contribution is to describe the statistical distribution of the number of employed and unemployed persons, characterise time course of relevant time series and subsequently propose at least two optimal forecasting models and compare their statistical properties and their accuracy. The contribution deals with forecasting macroeconomic variables of the number of employed and unemployed persons. For forecasting, Brown’s linear exponential smoothing, and Winters smoothing model are used. The accuracy of final forecasts is characterised by Mean Squared Error (MSE) and Mean Absolute Error (MAE). The results are interpreted only in the Czech context, not on the European level. The analysis is based on decomposition, seasonal adjustment, determination of empirical seasonal index and determination of optimal point and interval forecasts for the period 2019-2021. Since the null hypothesis of the series independence on the quarter was rejected, both time series had to be seen as seasonal.
Authors: Iveta Kmecová, Jaroslav Stuchlý, Lukáš Polanecký, Michal Šuta
Keywords: employment, unemployment, Brown’s linear exponential smoothing, Winters model of smoothing