یک رویکرد تجربی در تقسیم بندی کار. برنامه های کاربردی با داده های مدت زمان فردی

This article proposes empirical tools to account for the role of heterogeneities in the labour matching process, and shows an application to the Andalusian labour market which relies on individual data. The central idea of the paper is that the labour market is segmented, and this segmentation can be treated empirically by grouping workers, jobs and matches into labour groups according to their characteristics. In a segmented labour market the probability that a match occurs in a particular job group affects the probability that a match occurs in a particular worker group or vice versa. We propose two empirical measures related to this idea: propensity to match, and segmentation in worker and job groups. The usefulness of this empirical framework is shown by its application to different labour market analyses. Firstly, we use a clustering methodology, based on a similarity measure, to obtain a better overview of the structure of the labour market. Secondly, we propose a measure of mobility based on our similarity measure, and estimate a regression model that relates mobility to worker and job characteristics and to the economic cycle. Finally, these tools are included in an unemployment duration model. The proposed methodology may be useful in labour intermediation by helping seekers to follow a ‘roadmap’ of successful paths.

In the labour market, workers seeking jobs and vacant jobs offered by employers are heterogeneous in aspects as skills, geographical location, gender, age, payment, working time, attitude, taste, and many others. These heterogeneities lead to the concept of mismatch: “Mismatch is an empirical concept that measures the degree of heterogeneity in the labour market across a number of dimensions, usually restricted to skills, industrial sector, and location. Large differences in the skills possessed by workers and those required by firms would lengthen the time that it takes to match a given group of workers to a given group of firms, as agents search for a good match among the heterogeneous group. Industrial sector matters in matching because of industry-specific skills that may not be picked up by generally available measures of skills. Finally, location influences matching because of imperfect labour mobility.” (Petrongolo and Pissarides, 2001, 399–400). The main objective of this paper is to propose, from an empirical perspective, new variables to control for heterogeneities and segmentation in the labour matching process. We begin by dividing the workers, the jobs and the (worker–job) matches into highly detailed groups according to their characteristics (location and skills in our application). Ideally, the detailed division should allow us to consider the groups obtained as homogeneous or almost homogeneous, and the large size of the database should enable data in each group to be sufficiently numerous as to be statistically representative. Based on this segmentation scheme, we propose new empirical variables such as the ‘propensity to match’ between a worker group and a job group, the degree of ‘segmentation’ of a particular (worker or job) group, and the ‘similarity’ in the matching between any two (worker or job) groups. To show the usefulness of these variables, we make use of them in an application to the Andalusian labour market,1 which relies on a database of individual data of considerable size. These data allow us to perform different types of analysis: clustering, mobility and duration. The nature of our data, with information on vacancies, unemployed workers and job placements, links up our work directly with the theoretical concept of matching function. This function is intended to represent heterogeneities, frictions, and information imperfections and to capture the implications of the costly trading process without the need to make the heterogeneities and other features that give rise to it explicit. Instead of representing frictions more specifically according to their origin and their type, we lump them all together into an aggregate function. Therefore, the matching function does not assume that workers and jobs are homogeneous2; it simply omits to make the heterogeneities explicit. Without heterogeneities (zero mismatch), the matching function would not exist and jobs and workers would match instantaneously (Petrongolo and Pissarides, 2001, Pissarides, 2000, Pissarides, 2008 and Shimer, 2007).3 Considerable work has been carried out in an effort to open the ‘black box’ of the matching process and to render the heterogeneities inside the matching function explicit. Island, urn-ball, taxicab, queuing, stock-flow (or marketplace) and mismatch models, have all explored different types of frictions, extending the search theory of the labour market to allow for worker and firm heterogeneity and for micro-foundations of the matching process.4 As a rule, the labour market, or workers and jobs, are divided into parts (local labour markets, locations, islands, queues, worker–job pairs acceptable or unacceptable to match productively, stock (old)-flow (new) workers and jobs), which are then treated as if each part were homogeneous. Our work is not meant to extend or evaluate theoretical models of labour matching, but instead it tries to handle empirically important elements involved in these models — heterogeneities and segmentation. The heterogeneities of workers and jobs cause the segmentation of the labour market5; that is, features such as skills, location, age, gender, etc., make certain jobs only suitable for certain workers.6 We begin our analysis by considering that in a segmented labour market the probability that a match occurs in a particular job group affects the probability that a match occurs in a particular worker group or vice versa (for instance, the municipality, group of occupation or sector of economic activity of a job affects the probability that it matches with a worker corresponding to a particular municipality, occupation or sector of activity). We propose a measure of the degree of segmentation of each group and another measure of the propensity to match between workers and jobs depending on the groups to which they belong. As might be expected, our data show a very high degree of segmentation for the vast majority of groups. Since highly detailed division results in a very large number of groups, which may be difficult to interpret, we use a clustering7 methodology, based on a similarity measure, to obtain a better overview of the structure of the labour market and to obtain a smaller number of clusters (‘groupings of groups’). Cluster analysis enables, as far as possible, subjective or ‘a priori’ grouping criteria to be avoided: in our case, this would be the case, for example, if, for locations, municipalities were grouped in provinces and regions, or if, for skills, classifications with fewer digits for occupations or sectors were used. Instead, we look for a measure of similarity adapted, in the most objective possible way, to the purpose of our research. In the context of the search and matching theories applied to labour economics, we consider that worker (job) groups are more similar the more they resemble in the way they match with job (worker) groups. Using this concept of similarity, we will show in which way the worker–job clusters with high propensity to match that are formed may be considered as labour market clusters. We present results obtained by applying this methodology to our data.8 Mobility and unemployment duration are essential concepts in the search models that make the heterogeneities explicit by dividing the labour market into parts and specifying how workers (and jobs) move from one to another part.9 We propose an empirical measure of mobility directly related to our similarity measure, and then we estimate a multiple regression model that relates mobility in each worker–job match primarily to worker characteristics, and additionally to job characteristics and to the economic cycle. We use the results of the regression to estimate the ‘a priori’ workers' willingness to move. Our analysis ends up showing that the new empirical framework developed in this work can enhance the estimation of unemployment duration models in this field.10 The rest of the paper is organised as follows. Section 2 analyses the concept of labour market segmentation and proposes some related empirical measures: propensity to match and segmentation in worker and job groups. Section 3 develops the clustering methodology and shows the structure of the labour market obtained by applying this methodology. Section 4 proposes a measure of mobility and estimates a regression model that relates this measure to worker and job characteristics and to the economic cycle. The results are used to estimate the willingness of workers to move. Section 5 estimates an unemployment duration model making use of the tools obtained in the previous sections. Finally, Section 6 draws conclusions and suggests a number of possible applications of our methodology to active labour market policies.

In this paper we propose empirical tools to account for the role of heterogeneities in search and matching theories applied to labour economics, and we show an application to the Andalusian labour market, by using a large database of individual data. We have analysed the concept of labour market segmentation and proposed empirical measures related to this concept: propensity to match, and segmentation in worker and job groups. The results of our application show a high degree of segmentation. We use a clustering methodology, based on a similarity measure, to attain a better overview of the structure of the labour market and to reduce the large number of worker and job groups to a manageable number of clusters. We show in which way the worker–job clusters with high propensity to match that are formed may be considered labour market clusters. The clustering again highlights a high degree of segmentation, which is reflected in labour market clusters with high propensity to match, but these clusters are not ‘pure islands’, as shown by the existence of worker–job groups with high propensity to match outside these clusters. A measure of mobility in each worker–job match is proposed, directly related to our similarity measure, and a regression model is then estimated that relates mobility to worker and job characteristics and to the economic cycle. Mobility is higher, the higher the worker mobility in previous matches, the lower the segmentation of the worker or job group, the lower the tightness in the worker group, or the higher the tightness in the job group. With few exceptions, no significantly different effects are obtained for other variables usually included in studies in this field. The results of the regression model are used to estimate the ‘a priori’ workers' willingness to move. We show the usefulness of the tools that we have developed by including them in an unemployment duration model along with other conventional variables. The unemployment duration is higher for workers with lower willingness to move, and for those with higher segmentation or with lower tightness within their worker group. These overall results may change when we take into account the different types of exits. For example, lower willingness to move and higher segmentation in the worker group, which implies less competition from external workers, reduces unemployment duration for matches in the worker's own group (matches without mobility). Unemployment duration is also lower for workers with lower tightness in their worker group when matches take place outside the worker's own group (matches with mobility), which indicates that these workers experience a clear incentive to move. The hazard rate of the worker tends to fall with duration except in the first days of search. Again, no significantly different effects are obtained for other variables commonly included in previous literature in this field. Worker mobility, geographical or occupational, and the availability of relevant information are important requirements for effective labour matching, and constitute a prominent element that should be taken into account to guide the design of active labour market policies. The empirical tools proposed in this paper may be useful in this regard, by helping jobseekers and firms looking for workers to follow successful paths previously used by others. In this sense, the clustering methodology allows past information on matches to be processed in order to generate a ‘roadmap’ of possible routes to different labour market clusters, which can also include the probability of success in each route — a practical example of how this methodology might work can be found in Álvarez de Toledo et al. (2013). The versatility of the methodology proposed makes it possible to enrich the information provided from this perspective and to take into consideration other variables of interest, such as the best search channels for each cluster. Further research is required to test the practical usefulness of this methodology for real labour intermediation.