Portal de Periódicos da CAPES

Recurrent unemployment, welfare benefits and heterogeneity

2004; Taylor & Francis; Volume: 18; Issue: 4 Linguagem: Inglês

10.1080/0269217042000266436

1465-3486

José María Arranz Muñoz, Juan Muro,

Employment and Welfare Studies

Abstract The objective of this paper is to investigate the causes of the recurrences of individuals in unemployment during benefit periods. So as to attain this objective, we use administrative data from the Spanish Employment Agency to estimate a duration model with multiple spells that allows for unemployment state dependence through lagged unemployment duration in order to distinguish the heterogeneity and scarring effects. We find that an increase in the duration of previous unemployment benefit periods lengthens the expected duration of future unemployment benefit periods. True state dependence and heterogeneity, intensity of job search and local labour market conditions are among the elements that explain this unemployment state dependence. Keywords: State dependenceunemployment benefitsmixed proportional hazard modelunobserved heterogeneity Acknowledgements We are grateful to the comments received at the 54th International Congress of the Institute for Fiscal Studies in Cordoba (Argentina), Symposium in Barcelona and a seminar at CEMFI (Bank of Spain). We are indebted to M. Arellano, S. Bentolila, R. Blundell, C. Flinn, C. García‐Serrano, J. I. García‐Pérez, J.F. Jimeno, J.P. Hutton, L. Toharia, Y. Omori and an anonymous referee of the International Review of Applied Economics for their comments and suggestions on a previous version of this paper. A substantial portion of this research was carried out while J. Ma Arranz was a visitor in the Department of Economics at University College London (UCL). He would like to thank them for their hospitality and the Foundation Ramón Aréces for its financial support. J. Muro acknowledges research support from the Comunidad Autónoma de Madrid (Spain). Any remaining errors are our sole responsibility. Notes Correspondence Address: José Ma Arranz, Departamento de Fundamentos de Economía e Historia Económica, Facultad de Ciencias Económicas y Empresariales, Universidad de Alcalá, Plaza de la Victoria 3, Alcalá de Henares, 28803 Madrid, Spain. Email: josem.arranz@uah.es; Tel.: + 00‐34‐91‐8854271; Fax: + 00‐34‐91‐8854239 Our concern is about younger workers because they normally present more recurrences in unemployment in the Spanish Labour market. For example, in our database more than 75% of younger workers present at least three experiences in unemployment under benefits, 80% at least two and 76% only one. In contrast, Omori (Citation1997) developed a theoretical model that predicts that the re‐employment hazard of workers decreases (1) with durations of past and current nonemployment, holding the demand and supply shocks (or more generally the circumstances under which nonemployment was experienced) constant, but (2) decreases more if nonemployment was experienced in a labour market more favourable to workers (or more generally under circumstances more favourable to workers). Webster (Citation1997) presents a novel finding that there has been an unchanging simple relationship between long‐term and total claimant unemployment, at every spatial level, in Britain since the war. This relationship is consistent across countries (see Machin & Manning, Citation1999, for a survey). In fact, we have centred our study in the UI and UA part of the UCS. However, we only consider UA as a mere extension of UI (UI + UA) when the unemployed exhausted UI. In future studies we will study the effect of UA on the unemployed who are not entitled to UI but who could claim UA if they have contributed at least 3 months. This particular sample selection can raise certain doubts about the generality of our results because the selection of a sample of workers with at least three unemployment spells may create a non‐random sample. If this is not properly taken into account, the estimated exit rates may be biased. To correct this potential non‐randomness, we have estimated a reduced form probit model (as one cross‐section) on the presence in the sample. We have included the associated Heckman correction term as a regressor in the hazard rate equation. Results indicate that the correction term is not significant and the estimated coefficients in the equation are unchanged. This seems to point out that this type of selection is random. Alternatively, the distribution of the unobserved heterogeneity can be approximated nonparametrically, see Heckman & Singer (Citation1984). The reference individual is a male, unskilled clerical or unskilled production worker registered in a service region, without family burdens, with a replacement rate equal to mode, with an entitlement period from 6 to 15 months, who entered unemployment for other reasons (not end of contract), with a past unemployment period in a region with high unemployment rate and who exhausted his unemployment benefit period in the past. Although we have estimated our model in the AFT form, it is easy to interpret results from a Weibull specification in the proportional hazard form (Cox regression). Thus, calling the coefficients of the AFT form as b, the Cox coefficients analogue will be –b/σ. For the effects of the magnitude of the parameters, when the jth covariate (dummy) Xj increases by one unit, the hazard rate out of unemployment changes by a percentage equal to (exp(–b/σ)–1)× 100. This result is not a surprise because Cebrián et al. (Citation1996b) also detected with Spanish recipients (and with the HSIPRE database) that the replacement rate is not significant in explaining hazard rates out of unemployment. Additional informationNotes on contributorsJOSÉ Ma ARRANZFootnote Correspondence Address: José Ma Arranz, Departamento de Fundamentos de Economía e Historia Económica, Facultad de Ciencias Económicas y Empresariales, Universidad de Alcalá, Plaza de la Victoria 3, Alcalá de Henares, 28803 Madrid, Spain. Email: josem.arranz@uah.es; Tel.: + 00‐34‐91‐8854271; Fax: + 00‐34‐91‐8854239