Open Access

In this study we are concerned with the impact of vocational training on the individual’s unemployment duration in West Germany. The data basis used is the German Socio-Economic Panel (GSOEP) for the period from 1984 to 1994. To resolve the intriguing sample selection problem, i.e. to find an adequate control group for the group of trainees, we employ matching methods which were developed in the statistical literature. These matching methods uses as the main matching variable the individual propensity score to participate in training, which is obtained by estimating a random effects probit model. On the basis of the matched sample a discrete time hazard rate model is utilized to assess the impact of vocational training on unemployment duration. Our results indicate, that training significantly raises the transition rate of unemployed into employment in the short but not in the long run. JEL classification: C40, J20, J64

This paper provides a review of empirical evidence relating to the impact of training on employment performance. Since a central issue in estimating training effects is the sample selection problem a short theoretical discussion of different evaluation strategies is given. The empirical overview primarily focuses on non-experimental evidence for Germany. In addition selected studies for other countries and experimental investigations are discussed.

In recent econometric work, most analyses of female labour supply consider married women, whereas the results for unmarried women are provided rather as a by-product (Burtless/Greenberg, 1982, Johnson/Pencavel, 1984, Leu/Kugler, 1986, Merz, 1990,). When the particular interest is focused on unmarried women, data of the seventies or rather simple econometric models are used (Keeley et al., 1978, Hausman, 1980, Coverman/Kemp, 1987) . Often very specific populations are examined, like for example lone mothers in Blundell/Duncan/Meghir (1992), Jenkins (1992), Staat/Wagenhals (1993) or Laisney et al. (1993). Analysing the economic behaviour of unmarried women, one is confronted with the problem that the term ‘unmarried’ is not clearly defined. It includes single, divorced, separated and widowed women. They live in different types of households, like one-person households or family households, where they occupy different economic positions as for example head of the household or relative of the head. The present work considers unmarried female heads of household. We assume that the dominant economic position as head of household, voluntarily or involuntarily occupied, forces these women to a similar behaviour independent from their family status. Thus they are taken together in the analysis from the different family statuses: single, divorced, separated and widowed. Being unmarried often is regarded as a temporary state, voluntarily or involuntarily, for example in the case of young women before marriage or in the case of divorced women after their separation. Nevertheless the demographic development shows the increased importance of unmarried women in the population during the last decades. In the USA the portion of female headed households raised from 21,1% in 1970 to 26,2% in 1980 and 29,0% in 1992 (Statistical Abstracts of the United States, 1993. Own calculations). In the FRG, female headed households constitute 26,4% of total households in 1970, 27,4% in 1980 and 30,1% in 1992 (Stat.Bundesamt, FS 1, Reihe 3, 1970, 1980, 1992). Therefore it seems an interesting topic to analyse the labour supply behaviour of unmarried female heads. Especially the question whether the labour supply of unmarried women resembles rather that of married women or of prime-age males is of particular interest. Another purpose of this analysis is to apply modern econometric panel data models with special emphasis on the problem of unbalanced panel data. Most panel data analyses are carried out using balanced panel data, which is no problem if the selection process could be ignored and if enough cases are available to guarantee efficient estimation. Especially the last point was crucial for the present analysis of unmarried females. In the available panel data sets the unmarried female heads constitute only a rather small population. Therefore the estimation techniques were modified to take missing observations of the individuals into account. The paper is organized as follows: In section 2 the underlying theoretical model of intertemporal labour supply under uncertainty is shortly presented. Section 3 deals with the econometric specification and estimation techniques where the use of unbalanced panel data is considered. Section 4 contains the data description with a particular look on the unbalancedness of the samples. In the last section 5 the empirical results are presented. We compare the estimated parameters for the unmarried women between the USA and the FRG and also analyse the differences between unmarried and married women. Moreover a comparison between different samples of unmarried women is provided.