Forecasting Regional Employment with the ENTROP Method

The paper provides an outline of a method useful for forecasting problems. The approach is based on a combination of top-down and bottom-up techniques. It is applied to project employment in all 327 (western) German districts for a time span of two years. The most important step in the preparation of the forecast uses the ENTROP method, which is an entropy optimizing procedure, a generalization of common RAS techniques, newly developed for the estimation of matrices from heterogeneous information. In a defined sense the estimated matrix is the most probable one. The method chosen is very flexible and uses any available information extensively. Therefore, the estimates are reliable as is shown in an ex-post forecast. There is a double purpose for the forecast of employment. First, it helps to gain insights in the causal processes generating regional developments and spatial disparities on labour markets. Second, it is useful for regional labour market policies, e.g. the budgetary planning of the Federal Employment Services. Cet article cherche a esquisser une methode qui repond aux problemes de la prevision. La methode est fondee sur une combinaison des techniques descendantes et ascendantes. On s'en sert afin de prevoir sur une periode de deux annees l'emploi dans chacun des 327 districts a l'ouest de l'Allemagne. Le stade preliminaire le plus important utilise la methode ENTROP, une methode qui optimise l'entropie, une generalisation des techniques RAS courantes, nouvellement developpees pour estimer les matrices de l'information heterogene. Du point de vue de sa definition, la matrice estimee s'averera la plus probable. La methode choisie est tres flexible et emploie beaucoup toute information disponible. Ainsi, les estimations sont sures, dont fait preuve une prevision ex post. La raison d'etre de la prevision de l'emploi est double. En premier, on arrive a mieux connaitre les causes du developpement regional et des ecarts regionaux du marche du travail. En deux(This abstract was borrowed from another version of this item.)