Miguel Ángel Malo and Dario Sciulli (eds.), AIEL Series in Labour Economics, Disadvantaged Workers, 2014, Empirical Evidence and Labour Policies, DOI: 10.1007/978-3-319-04376-0, © Springer International Publishing Switzerland 2014 AIEL Series in Labour Economics For further volumes: http://​www.​springer.​com/​series/​7370 Editors Miguel Ángel Malo and Dario Sciulli Disadvantaged Workers Empirical Evidence and Labour Policies Editors Miguel Ángel Malo Department of Economics and Economic History, University of Salamanca, Salamanca, Spain Dario Sciulli Department of Economic Studies, University “G d’Annunzio” of Chieti-Pescara, Pescara, Italy ISSN 1863-916X ISBN 978-3-319-04375-3 e-ISBN 978-3-319-04376-0 Springer Cham Heidelberg New York Dordrecht London Library of Congress Control Number: 2014936009 © Springer International Publishing Switzerland 2014 This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer Permissions for use may be obtained through RightsLink at the Copyright Clearance Center Violations are liable to prosecution under the respective Copyright Law The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made The publisher makes no warranty, express or implied, with respect to the material contained herein Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com) List of Referees Tindara Addabbo , Università di Modena—Reggio Emilia, Italy Massimiliano Agovino , Università “G d’Annunzio” di Chieti-Pescara, Italy Roberto Bande , Universidade de Santiago de Compostela, Spain Maurizio Baussola , Università Cattolica del Sacro Cuore—Sede di Piacenza, Italy Filippo Belloc , Università “G d’Annunzio” di Chieti-Pescara, Italy Inmaculada Cebrián , Universidad de Alcalà, Spain Begoña Cueto , Universidad de Oviedo, Spain Sergio Destefanis , Università di Salerno, Italy Verónica Escudero , International Labour Office, Switzerland Carlos García-Serrano , Universidad de Alcalà, Spain Inmaculada García-Mainar , Universidad de Zaragoza, Spain Domenico Lisi , Università de Catania, Italy Elva López-Mourelo , International Labour Office, Switzerland Angel L Martín-Román , Universidad de Valladolid, Spain Marco Mazzoli , Università di Genova, Italy Fernanda Mazzotta , Università di Salerno, Italy Emanuele Millemaci , Università di Messina, Italy Toni Mora , Universitat Internacional de Catalunya, Spain Alfonso Moral , Universidad de Valladolid, Spain Chiara Mussida , Università Cattolica del Sacro Cuore—Sede di Piacenza, Italy Ricardo Pagán , Universidad de Malaga, Spain Giuliana Parodi , Università “G d’Annunzio” di Chieti-Pescara, Italy Vanesa Rodríguez , Universidad de Oviedo, Spain Marcello Signorelli , Università di Perugia, Italy Umut Oguzoglu , University of Manitoba, Canada Mario Veneziani , Università Cattolica del Sacro Cuore—Sede di Piacenza, Italy Acknowledgments This book contains a collection of contributions from authors associated to the Italian Association of Labour Economists (AIEL), the Spanish Association of Labour Economics (AEET), the International Labour Organization (ILO), and the African Development Bank (AfDB) Some of the chapters included in the book have been presented at the thematic session on “Disability” of the XXVII National Conference of Labour Economics of the AIEL, held at the Seconda Università di Napoli in September 2012, while some other chapters have been presented at the IV “Youth at Work” Workshop focused on “Disadvantaged Workers: Short and Long-Term Perspectives,” held at the Università “G d’Annunzio” of Chieti-Pescara in November 2012 Finally, the book includes some invited contributions from AIEL and AEET members All the chapters, before being published, have been submitted to a double blind peer-review process This has been possible thanks to the contributions of the referees (see the list above) that have given a great support to improve the quality of the book with their valuable suggestions We are also grateful for the financial support from the Italian Association of Labour Economists and from the Italian Ministry of Education, University and Scientific Research (PRIN 2009 “Measuring human development and capabilities in Italy: methodological and empirical issues”, prot 2009NM89S5_004, University of Chieti-Pescara Unit, Department of Economic Studies) Part of the editing process was developed while the coeditor Miguel Ángel Malo was on leave at the International Institute for Labour Studies (at the ILO) working as senior economist Contents Introduction Miguel Ángel Malo and Dario Sciulli Part I People with disabilities in the Labour Market Disability and Work:​ Empirical Evidence from Italy Tindara Addabbo, Jaya Krishnakumar and Elena Sarti The Dynamics of Disability and Labour Force Participation in Italy Massimiliano Agovino, Giuliana Parodi and Dario Sciulli Hiring Workers with Disabilities When a Quota Requirement Exists:​ The Relevance of Firm’s Size Miguel Ángel Malo and Ricardo Pagán Sheltered Employment Centres and Labour Market Integration of People with Disabilities:​ A Quasi-Experimental Evaluation Using Spanish Data Begoña Cueto and Vanesa Rodríguez Part II Young Workers in the Labour Market Temporary Contracts and Young Workers’ Job Satisfaction in Italy G S F Bruno, F E Caroleo and O Dessy Youth Unemployment:​ Key Determinants and the Impact of Crises G S F Bruno, M T Choudhry, E Marelli and M Signorelli Characteristics of Parents and the Unemployment Duration of their Offspring.​ Evidence from Italy Salvatore Farace, Fernanda Mazzotta and Lavinia Parisi Youth Employment in Africa:​ New Evidence and Policies from Swaziland Zuzana Brixiová and Thierry Kangoye 10 Understanding the Drivers of the Youth Labour Market in Kenya Verónica Escudero and Elva López Mourelo Part III Women, Migrants and Long-Term Unemployed 11 Disadvantaged Workers in the Italian Labour Market:​ Gender and Regional Gaps Maurizio Baussola and Chiara Mussida 12 Can the Crisis be an Opportunity for Women?​ Emanuela Ghignoni and Alina Verashchagina 13 Differences Between Spanish and Foreign Workers in the Duration of Workplace Accident Leave:​ A Stochastic Frontier Analysis Ángel L Martín-Román and Alfonso Moral 14 Duration of Joblessness and Long-term Unemployment:​ Is Duration as Long as Official Statistics Say?​ José María Arranz and Carlos García-Serrano List of Contributors Tindara Addabbo Dipartimento di Economia, Università di Modena e Reggio Emilia, Modena, Italy tindara@unimore.it Massimiliano Agovino Dipartimento di Economia, Università “G d’Annunzio” di Chieti-Pescara, Pescara, Italy agovino.massimo@gmail.com José María Arranz Departamento de Economía, Estructura y O.E.I, Universidad de Alcalá, Madrid, Spain josem.arranz@uah.es Maurizio Baussola Dipartimento di Scienze Economiche e Sociali, Università Cattolica del Sacro Cuore—Sede di Piacenza, Piacenza, Italy maurizio.baussola@unicatt.it Zuzana Brixiova Development Research Department, African Development Bank, Tunis-Belvedère, Tunisia z.brixiova@afdb.org Giovanni S F Bruno Dipartimento di Economia, Università Bocconi di Milano, Milan, Italy giovanni.bruno@unibocconi.it Floro Ernesto Caroleo Dipartimento di Studi Aziendali ed Economici, Università Parthenope di Napoli, Milan, Italy caroleo@uniparthenope.it Misbah Tanveer Choudry Lahore University of Management Sciences, Lahore, Pakistan misbah.tanveer@lums.edu.pk Begoña Cueto Departamento de Economia Aplicada, Universidad de Oviedo, Oviedo, Spain bcueto@uniovi.es Alfonso Moral Departamento Fundamentos de Teoria Economica, Universidad de Valladolid, Segovia, Spain amoral@eco.uva.es Orietta Dessy Università Cà Foscari di Venezia, Milan, Italy orietta.dessy@unive.it Verónica Escudero Research Department, International Labour Organization, Geneva, Switzerland escudero@ilo.org Salvatore Farace Dipartimento di Scienze Giuridiche, Università di Salerno, Fisciano Salerno, Italy sfarace@unisa.it Carlos García-Serrano Departamento de Fundamentos de Economía e Historia Económica, Universidad de Alcalà, Madrid, Spain carlos.garcia@uah.es Emanuela Ghignoni Dipartimento di Economia e Diritto, Università “La Sapienza” di Roma, Rome, Italy emanuela.ghignoni@uniroma1.it Thierry Kangoye Development Research Department, African Development Bank, Tunis-Belvedère, Tunisia t.kangoye@afdb.org Jaya Krishnakumar Department of Economic Sciences, University of Geneva, Geneva, Switzerland Jaya.Krishnakumar@unige.ch Elva López-Mourelo Research Department, International Labour Organization, Geneva, Switzerland lopezmourelo@ilo.org Miguel Ángel Malo Departmento de Economia y Historia Economica, Universidad de Salamanca, Salamanca, Spain malo@usal.es Enrico Marelli Dipartimento di Scienze Economiche, Università di Brescia, Brescia, Italy emarelli@eco.unibs.it Angel L Martín-Román Departamento Fundamentos de Teoria Economica, Universidad de Valladolid, Segovia, Spain angellm@eco.uva.es Fernanda Mazzotta can be seen, the median (calculated with the data on the time since the last job) is much lower than the mean: 50 % of the unemployed had remained months or less out of work in 2001–2006, months or less in 2008 and 10–11 months or less in 2010–2011 In order to assess clearly the impact of the observations with implausible unemployment tenures on mean and median durations, we have eliminated the extreme cases and calculated both indicators for the rest of the individuals We have considered two scenarios using arbitrary but, at the same time, loose criteria First, we have eliminated the unemployed who declare that they have been out of work for 96 months or more; they account for 5–8 % of the unemployed with previous experience in years of the period 2001–2007 and less than % in 2009–2011 (the mean age of these individuals is about 45, 10 years more than the total) Second, we have eliminated the unemployed who declare that they have been out of work for 48 months or more; they account for 11–14 % of the unemployed with previous experience in 2001–2007 and 7–9 % in 2008–2011 (they are about 43 years old, on average).8 Table 14.2 provides the results of this exercise Once the cases with 96 months or more are excluded, the mean duration is substantially shorter than the one obtained with the raw data and similar to (only slightly shorter than) the one obtained with the assignment method to the duration categories (column [a] in the previous table) Such duration is even shorter when the cases with 48 months or more are excluded, so the average unemployment tenure would have been around months in 2001–2009, with a minimum of months in 2008 and a maximum of 13.5 months in 2011 Obviously, these exclusions also affect the median durations, but to a lesser extent Table 14.2 Mean and median unemployment tenure (Di indicator, in months; time since the last job) for the unemployed with previous labour market experience after excluding extreme cases (Spain, 2000–2011) Unemployed with previous experience Time since the last job Mean Median (a) (b) (c) (a) (b) (c) 2000 30.3 16.2 10.5 2001 25.2 13.3 9.4 7 2002 22.7 13.2 9.2 7 2003 21.6 12.6 9.6 7 2004 21.1 12.3 9.3 6 2005 26.4 13.1 9.5 6 2006 24.0 11.7 9.0 2007 23.9 11.9 8.8 5 2008 19.4 10.5 8.0 5 2009 17.8 11.1 9.0 7 2010 19.9 13.8 11.8 10 9 2011 22.2 16.3 13.5 11 10 10 Source: Own calculations with microdata from the Spanish LFS (second quarters) Note: See Table 14.1 (a) Entire sample (b) Excluding cases of unemployment tenure with 96 months or more (c) Excluding cases of unemployment tenure with 48 months or more Finally, Table 14.3 displays the mean (and median) unemployment tenure (Di) for the unemployed with previous experience classified according to their relationship with the public employment office (PES) and their status of perception of unemployment benefits using the information from the microdata (in months) regarding the time since they left their last job The indicators have been calculated in two ways: without eliminating the cases of individuals answering extremely long tenures—48 months or more—(top panel) and eliminating them (bottom panel) The raw data give mean unemployment tenure longer than years in nearly all the years of the period for those registered in the PES not receiving benefits and for those not registered in the PES, and around year for the benefit recipients These means are clearly affected by the extreme values with implausible long durations When we delete them, the mean durations of the three groups diminish substantially For instance, in 2008 the average durations would have been less than months for the registered unemployed receiving benefits and more than months for the registered unemployed not receiving benefits The effect of the implausible long spells is also apparent, although to a lesser extent, in the median Without them, the median durations of the three groups would have been quite similar Table 14.3 Mean and median unemployment tenure (Di indicator, in months) for the unemployed with previous labour experience classified according to their relation with the PES and the receipt of unemployment benefits (Spain, 2000–2010) Mean Median All Recipients Registered non-recipients Non-registered All Recipients Registered non-recipients Non-registered Panel (a): All unemployed with previous labour market experience 2000 30.3 14.8 34.6 37.0 10 10 2001 25.2 11.7 27.9 32.6 8 2002 22.7 12.7 25.5 27.0 8 2003 21.6 11.6 25.2 24.1 2004 21.1 11.0 24.9 23.9 2005 26.4 12.9 32.2 31.4 2006 24.0 13.3 30.1 23.6 2007 23.9 13.6 30.7 24.7 2008 19.4 10.5 25.8 20.0 2009 17.8 9.2 25.1 21.8 2010 19.9 12.0 26.9 25.5 10 13 Panel (b): Excluding cases with unemployment tenure of 48 months or more 2000 10.5 9.0 11.1 10.8 7 2001 9.4 8.0 10.1 9.3 2002 9.2 7.8 9.8 9.4 6 2003 9.6 8.3 10.6 9.0 2004 9.3 7.7 10.2 9.2 2005 9.5 8.1 10.5 9.3 6 2006 9.0 7.6 9.6 9.2 5 6 2007 8.8 7.8 9.7 8.4 5 2008 8.0 6.6 9.3 7.9 2009 9.0 7.2 10.6 9.9 2010 11.8 10.6 13.6 11.4 10 Source: Own calculation with microdata from the Spanish LFS (second quarters) Note: See Table 14.1 14.4 Measuring Unemployment Duration with Longitudinal Data 14.4.1 Introduction In this section we use a relatively new dataset in order to provide evidence on the complete duration of the spells of unemployment which are observed to start over a given period of time (T indicator) This administrative dataset (the MCVL) contains registered information on all employment and insured unemployment spells of a % random sample of Spanish individuals who ever had any sort of relationship with the social security in any of the years of the period 2004–2010 Thus the population of reference from which the sample is extracted comprises all people who have either contributed to the system (whilst employed or as recipient of unemployment benefits) or received a pension in any moment of the year of reference.9 This data source has a longitudinal design From 2004 onwards, an individual who is present in an edition of the sample and subsequently remains registered with the Social Security stays as a sample member Furthermore, the sample is refreshed with new sample members, remaining representative of the population in each edition.10 The MCVL constitutes a rich but complex dataset It is made up of several files containing diverse information The files on personal details (coming from Social Security records and the Continuous Municipal Register) provide information on personal characteristics (gender, age, province of residence, citizenship and place of birth, etc.) The files on Social Security contributors contain details for each spell of employment on workplace and job attributes (employer size, location, ownership status, industry affiliation, job category, types of contract and tenure -dates of start and end of employment spells-, etc.) Labour market experience of individuals can also be measured since we know the date of their first labour contract One of the main advantages of the MCVL dataset is that the information contained in the personal and contribution files may be matched thanks to the existence of a unique identification number for each person and employer Nevertheless, this procedure is not easy Once all the information contained in those files is linked, it is possible to know the number of days spent in each state: employment, unemployment receiving benefits and non-employment Moreover, for recipients of unemployment benefits it provides information not only on the number of days of receipt but also on the type of benefits (unemployment insurance [UI] or unemployment assistance [UA]).11 When using this longitudinal data, we adopt two points of view: one is annual and the other refers to a period of years From the first perspective, the period of observation is the natural year for all individuals We select all the spells of unemployment benefits in progress in each year, so the information refers to all the spells which (a) start, (b) start and finish or (c) finish in each year of the period, and to the individuals who develop these spells Thus we are able to select all persons who have received unemployment benefits in any given moment of a year This means that this perspective is “crosssectional” in the sense that the observation window is fixed (1 year) but has a longitudinal content in the sense that we know the labour-market status of the individuals during the whole year (not only in particular moments—in a given day or week) From the second perspective, the criterion of selection is the ending of a job (contract) in a given period of time (a natural year), following the individual up during a long period of observation (3 years) In this manner we are able to know the complete duration of nearly all the spells of unemployment benefits starting during a given period of time and therefore to compute the T indicator Furthermore, it is obvious that, in a relatively long observation window there will be more spells of unemployment benefits than unemployed recipients due to the fact that a given person may start several spells (consecutive or not) of unemployment benefits during the period of observation (as shown in Fig 14.1, persons and 6) This implies that we are able to distinguish two types of duration: the duration of spells of covered unemployment and the effective duration of covered (consecutive) unemployment of individuals On the one hand, the duration of the spells of unemployment benefits starting each year does not distinguish whether they correspond to the same person or not (and, in the first case, whether they are consecutive or not) Therefore this duration is called the duration of the spells of covered unemployment This mean duration is calculated dividing the total duration of the spells of unemployment benefits starting in a given period by the total number of spells On the other hand, the duration of consecutive spells of unemployment benefits of the same person will be called the effective duration of covered unemployment of individuals This duration is calculated taking into account that the spells pertaining to the same person are aggregated when they are consecutive; that is, without an intervening spell of employment If a person has a spell of UI starting in 2005 followed by a spell of UA, we consider that they make up one incidence in covered unemployment of the same person (in this case, with two episodes of consecutive covered unemployment, so the duration of both are added up) The same procedure is applied when a person links two consecutive spells of UA For instance, if an individual has five spells with the following sequence UI1 + UA2 + EMP3 + UI4 + UA5, he/she would have two incidences in covered unemployment: the first one is the sum of episodes UI1 and UA2 and the second one the sum of UI4 and UA5; EMP3 would be an intermediate spell of employment The effective average duration of covered unemployment of each person would be the sum of the first incidence (sum of one or more consecutive spells of unemployment benefit receipt) divided by the total number of persons having this first incidence Moreover, individuals may exit benefit receipt (because either they find a new job or they return to their previous employer) before or after benefits exhaust.12 For the ones in this latter case, the duration of joblessness after the exhaustion of benefits can be computed, although we not know whether this additional period is of unemployment or inactivity because the database does not allow one to know whether the individual carries out active job search activities and whether he/she is available to accept a possible job offer Therefore, we would be able to compute the complete duration of joblessness of individuals, which sums the duration of covered unemployment plus the time since the individual exhausts a benefit until he/she finds a job In what follows, we use both perspectives and the distinction between the duration of the spells of covered unemployment, the effective duration of the covered unemployment of individuals and the complete duration of joblessness of individuals to provide evidence on the duration of unemployment and/or joblessness 14.4.2 The “Annual” Perspective When we study the individuals who receive benefits from a yearly perspective, their status may change through time, so it is possible that the same person has had several spells of unemployment benefits This implies that individuals may be classified by the number of spells during the year This type of analysis cannot be undertaken with the LFS because it does not allow one to take into account multiple spells of benefits for the same individual Table 14.4 provides the average number of spells of unemployment benefits for all individuals and for the ones with two or more benefits during each year, using the information from the MCVL The mean, which was stable around 1.5 in 2004–2007, increased to two in 2009–2010 Although the average number of spells of every type of benefit grew, the observed increase in the latter period was a consequence of the large rise in the UI benefits spells due to short-time work following a change in the incentives for workers and firms to be engaged in this sort of scheme to adjust to a fall in product demand Table 14.4 Average number of spells of unemployment benefits per year: all recipients and those with two or more spells of benefits in the same year (Spain, 2004–2010) 2004 2005 2006 2007 2008 2009 2010 All recipients UI (job loss) 1.3 1.3 1.3 1.4 1.4 1.4 1.5 UI (short-time work) 2.5 2.9 2.4 2.3 2.3 7.7 9.1 Subsidies of individuals aged 52 years or more and permanent per-task contracts 1.2 1.2 1.2 1.2 1.2 1.3 1.4 UA (including agriculture workers subsidy) 1.7 1.7 1.7 1.6 1.7 1.7 1.8 Total 1.5 1.5 1.5 1.5 1.6 2.0 2.1 UI (job loss) 2.6 2.6 2.6 2.7 2.7 2.8 2.9 UI (short-time work) 4.6 5.1 4.8 4.2 3.7 9.3 11.9 Subsidies of individuals aged 52 years or more and permanent per-task contracts 2.5 2.5 2.4 2.5 2.6 2.5 2.5 UA (including agriculture workers subsidy) 2.4 2.4 2.5 2.4 2.5 2.5 2.6 Total 2.5 2.6 2.6 2.7 2.7 3.3 3.4 Recipients with two or more spells Source: Own calculations with the MCVL data Note: UI and UA stands for ‘Unemployment Insurance’ and ‘Unemployment Assistance’, respectively Obviously, we also observe that the group of recipients with more than one spell of benefits exhibits larger values than the total on average, since the latter also incorporates the individuals with only one spell Thus the mean number of spells of those with more than one benefit spell was more than 2.5 in the years 2004–2008, increasing to 3.3 spells on average in 2009–2010 There are some categories of benefits that imply a larger number of spells than the rest, such as the UI benefits due to short-time work, whose recipients exhibit a mean of 4–5 spells in 2004–2007 and nearly 12 spells in 2010 How long does a period of covered unemployment last using the “annual” perspective with the MCVL data? Table 14.5 provides the average number of days of receipt of unemployment benefits and of work corresponding to the wage and salary workers in a year window frame for each year of the period 2004–2010 We highlight three issues Table 14.5 Recipients of unemployment benefits (UB) and wage and salary workers by year: average number of days of benefit and employment (Spain, 2004–2010) Wage and salary Recipients of UB NO YES Total Average number of days in receipt of UB Average number of days of employment workers 2004 YES 16,869,825 2,777,025 19,646,850 134 NO 539,325 539,325 Total 16,869,825 3,316,350 20,186,175 159 2005 YES 18,048,900 2,827,450 20,876,350 133 539,975 539,975 282 NO Total 18,048,900 3,367,425 21,416,325 158 2006 YES 18,604,825 2,885,200 21,490,025 132 539,600 539,600 Total 18,604,825 3,424,800 22,029,625 158 2007 YES 19,023,350 3,105,450 22,128,800 129 259 284 NO Total 19,023,350 3,707,725 22,731,075 153 2008 YES 18,245,900 3,987,400 22,233,300 131 765,075 765,075 257 282 NO 602,275 602,275 257 258 277 NO Total 18,245,900 4,752,475 22,998,375 153 2009 YES 16,748,850 4,616,500 21,365,350 154 244 269 NO Total 16,748,850 5,898,700 22,647,550 182 2010 YES 16,369,000 4,559,150 20,928,150 158 225 1,282,200 1,282,200 285 NO Total 16,369,000 6,228,925 22,597,925 193 227 1,669,775 1,669,775 287 Source: Own calculations with the MCVL data First, during the years of economic expansion and declining unemployment rates (2004–2007), the number of individuals who were recipients at some point each year was above three million Since the average number of recipients during a year—stock—was less than half this figure, it underlines the strong rotation of entry into and exit from the unemployment compensation system (UCS) during that period The effects of the economic crisis are clear in the following years Second, a wide majority (84 % in 2004–2008, 78 % in 2009 and 73 % in 2010) of the recipients along the year of observation also worked in the year: this group received benefits and worked, on average, during some months and months, respectively, in the period 2004–2008, and some months and months, respectively, in 2009–2010 Third, the proportion of individuals who only received benefits without contributing as employees in the year over the total number of recipients was stable around 16 % during the expansion (less than 600,000), increasing to 22 % in 2009 and 27 % in 2010 (1.3 and 1.7 million, respectively) This implies that around 2.5 % of all labour-market participants in the year in 2004– 2007 and 7.4 % in 2010 were receiving benefits without contributing as wage and salary workers in the year.13 The mean duration of the spells of unemployment benefits of this group remained quite stable around 9.5 months (280 days) during all periods About 25 % of these individuals received the benefits during 6–7 months or less and about 25 % of the whole year (data not shown in the table) These figures, interesting enough, have the limitation that they include the spells of unemployment in progress at the beginning of the year but only since then (so their durations prior to the 1st of January are not considered) and the ones that are in progress at the end of the year but only until then (so their durations after the 31st of December are not considered) Therefore the previous information does not give us the complete duration of the spells of unemployment (T indicator) In order to get it, we have to use the “longitudinal” perspective; that is, selecting the spells of unemployment benefits starting in a given period and following them up until their termination using the MCVL This is what we in the next section 14.4.3 The “Longitudinal” Perspective We select a sample of individuals starting their spells of unemployment benefits in a given year (2005 and 2008) and we follow them up through time (3 years) This will allow us to know how many people exhibit only one incidence of entry into the UCS within the window of observation and how many exhibit several incidences, with successive spells of benefits (they may be consecutive; for instance, a UI benefit which runs out followed by a UA benefit) and of employment In addition, in order to make the sample homogeneous, we select recipients aged between 16 and 64 (in the year of the first incidence) receiving UI benefits due to the ending of a labour relationship (because of a layoff, end of temporary contract, etc.) or UA benefits Table 14.6 provides the mean and the distribution of unemployment benefits durations for spells starting in 2005 and 2008 (in days) The first and second grand columns refer to equal periods of observation characterized by distinct economic conditions: the boom years of 2005–2007 and the recessive years of 2008–2010 Table 14.6 Means and distributions of duration (T indicator, in days) of benefits starting in 2005 and 2008: spells and individuals (Spain, 2004–2010) Benefits starting in 2005 (2005–2007 period) Spells Individuals Benefits starting in 2008 (2008–2010 period) Spells Individuals 125 200 141 248 1 4 13 14 10 22 25 25 30 52 30 61 50 87 120 95 141 75 181 239 183 347 90 282 532 334 715 95 395 729 485 815 99 729 1,096 729 989 Average number of spells – 2.6 (2) – 3.4 (3) Average number of incidences – 2.1 (1) – 2.4 (2) Average incidence duration (days) – 141 (78) – 162 (84) Accumulated average duration of incidences – (days) 296 (210) – 392 (334) Observations 2,208,400 5,060,250 3,421,550 First incidence Mean duration (days) Percentiles (days) All incidences 3,098,175 Source: Own calculations with the MCVL data In the table we distinguish two types of duration of covered unemployment: one is computed from the information on spells and the other from the information on individuals The information shown in the columns labelled “spells” refers to the duration of the spells of covered unemployment, while that shown in the columns labelled “individuals” refers to the effective duration of covered unemployment of persons (corresponding to the first incidence of unemployment within the year) The number of spells starting in any moment of 2005 (2008) was 3,098,175 (5,060,250), which corresponds to 208 400 (3,421,550) different persons When we compare the mean, the median and the distribution of durations by type of scenario— spells and individuals—what emerge is the existence of huge differences and that there have been important changes due to the alteration of the economic and labour market context The mean (and median) duration of the spells of unemployment benefits initiated in a given year (first incidence in the UCS) is lower before the crisis than during the crisis (125 days vs 141 days [mean] and 87 days vs 95 days [median]) Nevertheless, these figures are much lower than the effective duration of covered unemployment of individuals: a mean of 200 days and 248 days, respectively, and a median of 120 days and 141 days, respectively This information reveals that any analysis of the duration of covered unemployment focusing strictly on spells (not individuals) underestimates the “true” duration These differences come from the fact that the durations of individuals are constructed adding up the information of consecutive spells of unemployment benefits These differences are emphasized when we focus on long-term unemployment Ten per cent of the individuals starting a benefit in 2005 and followed up until the end of 2007 remain in covered unemployment for more than 532 days, while the same proportion exhibit a duration of more than 282 days when we consider the information on spells The effect of the economic crisis is evident, since these durations rise: 10 % of the individuals (spells) starting a benefit in 2008 and followed up until the end of 2010 remain in covered unemployment for more than 715 days (334 days).14 Given the evidence shown so far, in what follows we only use information on individuals, which allows us to exploit the longitudinal information of each person and compute the number of incidences per person in covered unemployment in a given period, the duration of each incidence and the total accumulated duration When a spell of unemployment benefit starts, the origin may be previous employment (due to either the ending of a fixed-term contract or a layoff) or previous benefit (after its exhaustion) Table 14.7 makes this distinction, showing the mean and the distribution of durations of covered unemployment of individuals using the information of the spells initiated in 2005 and 2008 by type of origin Nearly 90 % of the people starting a spell of benefits in either of the years come from employment Table 14.7 Means and distributions of duration (T indicator, in days) of benefits starting in 2005 and 2008 by types of benefit origin (either employment or previous benefit): effective duration of covered unemployment of individuals (Spain, 2004–2010) Benefits starting in 2005 (2005–2007 period) Benefits starting in 2008 (2008–2010 period) Previous benefit After a job Previous benefit After a job Mean duration (days) 332 180 372 232 Percentile (days) 4 18 13 16 14 10 25 32 89 21 49 31 94 24 59 50 192 116 280 125 75 457 201 630 305 90 774 449 876 668 95 1,096 669 946 783 99 1,216 973 1,058 958 Observations 291,625 1,916,775 398,725 3,022,825 Total 2,208,400 3,421,550 Source: Own calculations with the MCVL data As can be seen, the effective duration differs depending on whether a person starts a spell of unemployment benefits after a job or after (the exhaustion of) a previous benefit In the previous table, we found that the mean duration of covered unemployment of recipients was 200 days in 2005 and 248 days in 2008 The mean duration is 180 days and 232 days, respectively, when the receipt of benefits starts after the ending of a job, and 332 days and 372 days, respectively, when it comes after a previous benefit Therefore the effective duration of covered unemployment of individuals differs greatly, as expected, according to the reason for the beginning of the benefit, being shorter when it starts after the ending of a job Finally, as commented on in the introduction to this section, another feature of the duration of unemployment consists of considering that the time the individual remains in unemployment is “completely covered” by a benefit (if the individual exits to a job before or at the moment of the exhaustion of a benefit) or “partially covered” (if the individual exits to a job after the exhaustion of a benefit) In this sense, it is possible to distinguish between the duration of covered unemployment, which measures the time receiving benefits until an individual finds a job if this fact occurs before or at the moment of exhaustion (or even the receipt finishes without knowing anymore about the individual—assuming, therefore, that the spell is censored), and the complete duration of joblessness, which sums the duration of covered unemployment plus the time since the individual exhausts a benefit until he/she finds a job Table 14.8 contains the information on the complete duration of joblessness (for unemployment recipients) by types of benefit origin (either employment or other benefit) Combining this information with that of the previous table, and focusing on the individuals coming from a job, we find that, in the period prior to the crisis, the average complete duration of covered unemployment of individuals after a job was months (180 days) and the average complete duration of joblessness was months (210 days) With the onset of the crisis, the mean duration of covered unemployment increased to 7.7 months (232 days) and that of joblessness to 8.6 months (259 days) Table 14.8 Means and distributions of duration (T indicator, in days) of benefits starting in 2005 and 2008 by types of benefit origin (either employment or previous benefit): complete duration of joblessness of individuals (Spain, 2004–2010) Starts in 2005 (2005–2007 period) Starts in 2008 (2008–2010 period) Previous benefit After a job Previous benefit After a job Mean duration (days) 385 210 407 259 Percentile (days) 6 23 16 21 17 10 42 25 40 27 25 111 55 122 65 50 287 125 336 164 75 563 265 670 372 90 905 548 889 709 95 1,096 729 956 800 99 1,216 981 1,058 968 Observations 291,625 1,916,775 398,725 3,022,825 Total 2,208,400 3,421,550 Source: Own calculations with the MCVL data The distribution of durations also provides some interesting information Twenty-five percent of the individuals starting a benefit in 2005 end a period of joblessness in 55 days or less and 50 % in 125 days or less The durations corresponding to these proportions are longer when the benefit starts in 2008 (65 days and 164 days, respectively) The largest differences in duration seem to be concentrated on the top of the distribution: 10 % of the complete periods of joblessness starting in 2005 exhibit durations longer than 548 days, while for those initiated in 2008 this proportion corresponds to durations longer than 709 days In fact, around 25 % are periods of long-term joblessness (1 year or more) in 2008, being this proportion lower in 2005 14.5 Conclusions The aim of this article has been to shed some light into the measurement of the average duration of a spell of unemployment, providing an in-depth analysis of duration statistics based on cross-sectional information coming from the LFS and comparing them with others based on longitudinal data coming from administrative sources The case study has been the Spanish labour market The same analysis might be carried out for other countries which dispose of LFS microdata and longitudinal databases in order to carry out a similar study to the one we offer here Our results suggest that the average duration of unemployment as provided by the OECD statistics based on LFS grouped data (21.4 months in 2001, 13.3 months in 2005 and 14.8 months in 2010) is largely overestimated in the case of Spain This indicator, which refers to average “unemployment tenure” or elapsed—incomplete–duration of spells (Di), is even longer when one uses the individual responses included in the LFS microdata The existence of a relatively small number of individuals (5–14 %, depending on the year and the criterion), giving answers with an extremely high number of months, makes the distribution of durations have a long queue to the right, thereby affecting the mean Moreover, due to interruption bias, the average complete duration of spells of unemployment in progress (Dc) should be larger than the incomplete duration (Di) However, length bias also plays a part, since short-term unemployment spells are underrepresented (or not captured) in the LFS sample and the probability of capturing long periods is larger than the probability of capturing short periods, especially in a labour market with high worker turnover as the Spanish one This feature makes us think that the average complete duration of spells starting over a given period (T) is lower than the incomplete duration In fact, we are able to calculate the average complete duration of joblessness for unemployment recipients with longitudinal data (the Spanish MCVL) at individual level (what we have called the effective duration of covered unemployment of individuals, taking into account the time from the exhaustion of benefits until they eventually find a job) The result is that effective duration was months, if the individuals started the receipt of a benefit (after the ending of a job) in 2005, and 8.6 months, if they started in 2008 What is more relevant, the share of long-term unemployment is much lower than that supplied by conventional data based on the LFS The differences reported in the article among the measures for unemployment durations (in particular, the difference between Di and T), caused by the existence of many short spells of unemployment not captured appropriately by the LFS, constitute a relevant result, which can be used to inform the election of the more convenient database for analysing unemployment durations and to help the interpretation of the corresponding measures We interpret the comparison of all these pieces of evidence as indicating that the information on duration based on LFS data provides a misleading guide, in particular in countries where labour turnover is large (as is the case for the Spanish labour market) Therefore it should not be used (at least, not alone) to inform policymakers’ decisions and economists’ theoretical works Our results call for the use of aggregate unemployment measures incorporating the time dimension (as the ones developed in Sengupta 2009, and Shorrocks 2009a, b, for instance), although these have seldom been used in empirical analysis so far These measures may be helpful by providing more information for a better understanding of the nature of the variations in the unemployment rate along time but also of differences across countries (Gradín et al 2012) Acknowledgment José M Arranz acknowledges financial support from the Ramón Areces Foundation and Carlos García-Serrano acknowledges financial support from the Ministry of Science and Innovation (National Plan, ECO2010-19963) The authors wish to thank Spanish Social Security for providing the data for this research Obviously, the opinions and analyses are the responsibility of the authors The usual disclaimer applies References Akerlof G, Main B (1980) Unemployment spells and unemployment experience Am Econ Rev 70:885–893 Akerlof G, Main B (1981) An experience-weighted measure of employment and unemployment durations Am Econ Rev 71(5):1003– 1011 Alba A (1999) Explaining the transitions out of unemployment in Spain: the effect of unemployment insurance Appl Econ 31:183–193 [CrossRef] Alba A, Arranz JM, Moz-Bullón F (2007) Exits from unemployment: recall or new job? Labour Econ 14:788–810 [CrossRef] Alba A, Arranz JM, Moz-Bullón F (2012) Re-employment probabilities of unemployment benefit recipients Appl Econ 44:3645–3664 [CrossRef] Argimón I, González CI (2006) La Muestra Continua de Vidas Laborales de la Seguridad Social Boletín Económico del Banco de Espa, May, pp 40–53 Arranz JM, García-Serrano C (2011) Are the MCVL tax data useful? Ideas for mining Hacienda Pública Espola 199(4):151–186 Arranz JM, García-Serrano C (2014) The interplay of the unemployment compensation system, fixed-term contracts and rehirings Int J Manpower (forthcoming) Arranz JM, García-Serrano C, Hernanz V (2013) How we pursue ‘labourmetrics’? An application using the MCVL Revista Estadística Espola 55(181):231–254 Baker GM, Trivedi PK (1985) Estimation of unemployment duration from grouped data: a comparative study J Labor Econ 3(2):153–174 [CrossRef] Borooah VK (2002) A duration-sensitive measure of the unemployment rate: theory and application Labour 16(3):453–468 [CrossRef] Clark KB, Summers LH (1979) Labor market dynamics and unemployment: a reconsideration Brookings Pap Econ Act 1:13–72 [CrossRef] Corak M, Heisz A (1996) Alternative measures of the average duration of unemployment Rev Income Wealth 42(1):63–74 [CrossRef] Fallick B, Ryu K (2007) The recall and new job search of laid-off workers: a bivariate proportional hazard model with unobserved heterogeneity Rev Econ Stat 89(2):313–323 [CrossRef] Flinn C, Heckman J (1982) New methods for analyzing structural models of labor force dynamics J Econ 18:115–168 [CrossRef] García-Pérez JI (2008) La Muestra Continua de Vidas Laborales (MCVL): una guía de uso para el análisis de transiciones Revista de Economía Aplicada 16(E-1):5–28 García-Serrano C, Malo MA (2013) Beyond the contract type segmentation in Spain Contract case study on labour market segmentation Employment working paper 143, Employment Sector, International Labour Organization Gradín C, Cantó O, del Río C (2012) Unemployment and spell duration during the Great Recession in the EU Working paper series WP 12/05, Universidade de Vigo Ham J, Rea SA Jr (1987) Unemployment insurance and male unemployment duration in Canada J Labor Econ 5:325–353 [CrossRef] Jansson F (2002) Rehires and unemployment duration in the Swedish labour market—New evidence of temporary layoffs Labour 16(2):311–345 [CrossRef] Jenkins SP, García-Serrano C (2004) The relationship between unemployment benefits and re-employment probabilities: evidence from Spain Oxford Bull Econ Stat 66(2):239–260 [CrossRef] Jensen P, Svarer M (2003) Short- and long-term unemployment: how temporary layoffs affect this distinction? Empir Econ 28:23–44 [CrossRef] Kaitz HB (1970) Analyzing the length of spells of unemployment Mon Labor Rev 93:11–20 Katz LF, Meyer BD (1990a) The impact of the potential duration of unemployment benefits on the duration of unemployment J Public Econ 41:45–72 [CrossRef] Katz LF, Meyer BD (1990b) Unemployment insurance, recall expectations and unemployment outcomes Q J Econ 105(4):973–1002 [CrossRef] Kiefer NM (1988) Economic duration data and hazard functions J Econ Lit XXVI:646–679 Lapuerta, I (2010), “Claves para el trabajo la Muestra Continua de Vidas Laborales”, DemoSoc Working Paper, nº 2010-37, Universitat Pompeu Fabra Lindenboom M, Theeuwes J (1993) Search, benefits and entitlement Economica 60:327–346 [CrossRef] Mavromaras KG, Orme CD (2004) Temporary layoffs and split population models J Appl Econometr 19:49–67 [CrossRef] Mickelwright J, Nagy G (1999) Living standards and incentives in transition: the implications of UI exhaustion in Hungary J Public Econ 73:297–319 [CrossRef] Narendranathan W, Stewart M (1993) Modelling the probability of leaving unemployment: competing risks models with flexible base-line hazards Appl Stat 42:63–83 [CrossRef] Narendranathan W, Stewart M (1995) The determinants of individual unemployment durations in an era of high unemployment Econ J 105:321–332 [CrossRef] Nickell S (1979) Estimating the probability of leaving unemployment Econometrica 47:1249–1266 [CrossRef] Paul S (1992) An illfare approach to the measurement of unemployment Appl Econ 24:739–743 [CrossRef] Paul S (2001) A welfare loss measure of unemployment with an empirical illustration Manch Sch 69(2):1463–6786 [CrossRef] Riese M, Brunner JK (1998) Measuring the severity of unemployment J Econ 67(2):167–180 [CrossRef] Røed K, Nordberg M (2003) Temporary layoffs and the duration of unemployment Labour Econ 10:381–398 [CrossRef] Salant SW (1977) Search theory and duration data: a theory of sorts Q J Econ 91(1):39–57 [CrossRef] Sengupta M (2009) Unemployment duration and the measurement of unemployment J Econ Inequal 7(3):273–294 [CrossRef] Shorrocks A (2009a) Spell incidence, spell duration and the measurement of unemployment J Econ Inequal 7(3):295–310 [CrossRef] Shorrocks A (2009b) On the measurement of unemployment J Econ Inequal 7(3):311–327 [CrossRef] Stancanelli E (1999) Unemployment duration and the duration of entitlement to unemployment benefits: an empirical study for Britain Appl Econ 31:1043–1051 [CrossRef] Footnotes Based on LFS data, Eurostat estimates of the Spanish long-term unemployment rate are 10.1 % in 1996, 1.7 % in 2007 and 11.1 % in 2012, while the International Labour Organization provides the following figures regarding the proportion of long-term unemployed (in relation to total unemployment): 55.9 % in 1996, 27.6 % in 2007 and 45.1 % in 2010 For each person j, E[(Di)j] = (Dc)j And with a large population: Dc = 2Di This would be the case of Spain, where the share of temporary employment increased quickly during the 1980s to reach about one third of the wage and salary workers at the early-1990s (remaining at that level until the beginning of the current economic crisis) and the level of worker turnover has been high on the model of ‘flexibility at the margin’ hinged on a dual economic structure between jobs prone to stability and jobs prone to instability (García-Serrano and Malo 2013) Moreover, the combination of the extensive use of fixed-term contracts (more than half of all new signed contracts last less than months) and the fact that the unemployment compensation system is not experience rated (which imply that firms face lower costs for layoffs) seems to facilitate the succession of (short) spells of employment and unemployment benefits (and make rehirings more likely) (Alba et al 2007; Arranz and GarcíaSerrano 2013) See http://​stats.​oecd.​org/​ (Labour–Labour Force Statistics–Unemployment by duration–Average duration of unemployment) For the countries included in the corresponding table (not all the OECD countries), the average duration ranged from 2.2 months in Australia to 30.5 months in the Slovak Republic in 2012 (the figures were similar in 2005) The documentation of the OECD Statistics does not give any hint as to how they compute harmonized measures of unemployment duration See http://​www.​ine.​es/​ (Sociedad—Mercado laboral—Encuesta de Población Activa) We have excluded the unemployed who say that they have just found a job This is due to the way the Spanish Statistical Office gives the microdata files to researchers Before 2005, the responses to the question on the time looking for a job were grouped into the following categories: “less than months”; “from to less than months”; “from to less than 12 months”; “from 12 to less than 18 months”; “from 18 to less than 24 months”; “from to less than years”; and “4 years or more” The number of excluded “unemployed” would be, depending on the criteria, above 100,000 in 2001–2008 and around 150,000 in 2009– 2011 (first criterion) or about 200,000 in 2001–2008, around 300,000 in 2009–2010 and 400,000 in 2011 (second criterion) Argimón and González (2006) and García-Pérez (2008)—in Spanish—and Arranz and García-Serrano (2011)–in English–provide a good introduction to how the sample is used, while Lapuerta (2010)–in Spanish–and Arranz et al (2012)–in English–clearly set out the practical difficulties involved in handling the data 10 The MCVL is therefore only representative of the population related to the Social Security system in the year concerned, and is therefore not representative of the past: although it contains information on previous social security contributions by the individuals selected (dating back several years), it does not include past contributions by individuals who have died or who are no longer actively engaged in the labour market (see Arranz et al 2012, for an analysis of the impact of using data on a period prior to the years of reference on some key labour market variables) 11 The Spanish unemployment compensation system comprises two schemes: UI and UA UI is paid to employees (excluding civil servants, domestic workers and those without past work experience) who did not quit their job voluntarily, who can and want to work and who have paid a minimum number of contributions Length of entitlement depends on the number of months contributions are made Contributions for at least months over the last 72 months are required for eligibility since 1992 The duration of entitlement is equal to twice the modulus of the number of contribution months divided by 6, up to maximum of 24, that is, the potential entitlement periods are 4, 6, 8, 10, …, 24 The amount of UI paid is equal to a fraction of the average of the ‘regulatory base’ in the last months prior to unemployment, where the ‘regulatory base’ is the gross earnings used to calculate UI contributions UI payments decline with the unemployment spell: that fraction equals 70 % during months 1–6 of UI receipt and 60 % thereafter (50 % after a legal change in 2012) Payments are subject to a minimum amount equal to 75 % of the statutory minimum wage and to a maximum amount that varies with the number of children the unemployed person has UA is means tested and is available (depending on their characteristics) for those who exhaust the UI and those who are not eligible The UA benefit is a flat rate 12 The empirical literature has usually found that unemployment exit rates rise sharply when benefit exhaustion is imminent (Ham and Rea 1987; Katz and Meyer 1990a; Mickelwright and Nagy 1999), although it seems to be less pronounced in Spain (Jenkins and García-Serrano 2004) Furthermore, a spike in the hazard close to the expiry date of benefits is often more visible for new jobs than for recalls (Katz and Meyer 1990b; Jansson 2002; Mavromaras and Orme 2004) 13 We define the labour-market participants in a year as the sum of (a) the ones who have been contributing as wage and salary workers the whole year; (b) the ones who have been contributing as wage and salary workers part of the year and receiving benefits part of the year; and (c) the ones who have been receiving benefits part of or the whole year without contributing as wage and salary workers 14 We also observe the impact of the crisis on the average number of spells per person (from 2.6 to 3.4), on the average number of incidences (from 2.1 to 2.4), on the mean duration of the incidences (from 141 days to 162 days) and on the accumulated duration (from 296 days to 392 days) In sum, these results allow us to conclude that the economic and employment crisis has increased the average duration of covered unemployment and the re-incidence in unemployment in Spain ... Ángel Malo and Dario Sciulli (eds.), AIEL Series in Labour Economics, Disadvantaged Workers, 2014, Empirical Evidence and Labour Policies, DOI: 10.1007/978-3-319-04376-0, © Springer International... Disadvantaged Workers, 2014, Empirical Evidence and Labour Policies, DOI: 10.1007/978-3-319-04376-0_2, © Springer International Publishing Switzerland 2014 Disability and Work: Empirical Evidence. .. perspectives regarding various groups of disadvantaged workers This volume includes 13 chapters, mainly focusing on two groups of disadvantaged workers: disabled workers and young workers It also includes