Evaluation of the Matching Process of Disabled People Through a Macroeconomic Approach: the Italian Case

Abstract

This paper estimates a spatial matching function for disabled people by using a panel of 20 Italian regions over the period 2006–2011. According to the results obtained, in the regional labour market analysed there is a congestion effect among unemployed disabled people, due to an excess of unemployed disabled people compared to available vacancies.

In terms of policy, it is necessary to promote policy actions aimed at supporting private firms in their production processes, especially for those with a high number of temporary layoff hours.

Appendices

Appendix A

Law 68/1999: Some Clarifications

Law 68/1999 is addressed to: “the working-age people suffering from physical, mental or sensory and intellectual disabilities, resulting in a reduced capacity to work by more than 45 %…”, “Disabled from work with a degree of incapacity of more than 33 %…”, “the blind or the deaf-mute…”, “war invalids, civil war invalids and disabled for service with disabilities ascribed from the first to the eighth category…”.

The prerequisite to take advantage of the benefits provided by Law 68/99 is the inclusion in the compulsory employment lists, that are held by the Employment Services of the provincial governments. Employment services usually enroll the applicant in the lists of compulsory employment conditionally to further assessment of disability by health care bodies. Next to the entering, the disabled person is then able to join job opportunities that come to the Employment Service from both public bodies and private companies, by filling out the reservation form.

Appendix B

In this Appendix we present the limitations of our analysis on the estimation of the matching function.

First of all, this matching function specification follows the hypothesis of the random matching approach. This approach, defined stock-stock approach, assumes that agents are matched randomly at any point in time, regardless of the duration of the research. It is opposite to another approach defined stock-flow approach, which assumes that unemployed people have complete information about possible jobs. Either they find a job instantaneously or they wait until new vacancies arrive on the market. Consequently, the stock-flow approach does not take into account only stocks but also inflows into unemployment and vacancies (Lottman 2012). The stock-stock approach has been widely criticized (see Coles and Smith 1998; Gregg and Petrongolo 2005; Petrangolo and Pissaridies 2001; Fahr and Sunde 2009), because it implicitly assumes an underlying undirected and random search process leading to matches between homogeneous unemployed and homogeneous vacancies.

In our case, we refer to the stock-stock approach due to unavailability of data on unemployed disabled people’s inflows and on specific vacancies for disabled people. The choice of the stock-stock approach represents a first limit of our analysis.

Another limitation, also highlighted by empirical literature, is represented by the choice of the dependent variable, as the variable used to measure matches tends to exclude the amount of employed job searchers. In many empirical studies this variable is omitted due to data unavailability (Broersma and Van Ours 1999). Many studies use total hires as a proxy for the number of matches, as hires include not only the unemployed who find a job but also persons out of the labour force flowing into a job and the flow of employed workers who are looking for another job. As a result, matching concerns the vacancies and all job seekers and we no longer have the matching of vacancies and unemployed job seekers. In particular, Broersma and Van Ours (1999) conclude that it is important to distinguish between employed people looking for jobs and unemployed people looking for a job, as if this distinction is not taken into account it produces a biased estimate of the parameters of the matching function. This distinction is relevant when we consider the spatial lag of vacancies among the regressors of the matching function. In particular, Fahr and Sunde (2002, 2003) find that the sign of the coefficient of spatially lagged vacancies depends on the type of dependent variable used: the effect is positive for all hires and hires of people who are already employed, while it is negative for unemployment outflows into employment. In our case we do not have information on the construction of the matching variable for disabled people, as ISFOL data does not explain whether the matching variable includes total hires (hires of those who are already employed and outflows of unemployed into employment) or only unemployment outflows into employment; therefore, the sign of the spatial lag of vacancies allow us to identify which type of dependent variable we are using.

Another issue of our empirical analysis is the temporal aggregation of data (Hynninen 2005). Since the matching function describes a process that is continuous in time, the use of data referring to discrete time introduces a problem of bias in the parameters estimates of the matching function. In particular, a greater aggregation of data will result in a greater bias (Burdett et al. 1994). It is advisable, therefore, to use highly disaggregated data over time (such as monthly or quarterly data). In fact, data with high temporal frequency are rarely available, especially when evaluating a matching function for disabled workers. In our case, as in other empirical works on the matching function, we have no data with less than annual frequency.

A final issue is related to the aggregation of spatial data (Hynninen 2005; Kano and Ohta 2005). The spatial aggregation of data was discussed and rejected by the empirical literature: at the level of local labour markets, it is preferred to work with disaggregated data (Burgess and Profit 2001; Coles and Smith 1996; Ilmakunnas and Pesola 2003; Hynninen 2005). Spatially disaggregated data are justified by the idea that the aggregate economy is a collection of distinct and heterogeneous spatially labour markets which suffer from many frictions. Spatially disaggregated data allow to examine the spatial aspects of the matching function and the weight of the interactions between the different local labour markets within an economy. On the basis of these considerations, many scholars have enriched the basic matching function with the spatial lag of regressors (unemployment, vacancies) with the objective to evaluate spillovers between different local labour markets (Burda and Profit 1996; Burgess and Profit 2001; Ilmakunnas and Pesola 2003). However, in our study disaggregated data are available only from regional level.