Exit from short-term contract. A french empirical analysis,

Transcription

1 Ext from short-term contract. A french emprcal analyss, Mohamed Ben Halma To cte ths verson: Mohamed Ben Halma. Ext from short-term contract. A french emprcal analyss, Socété Canadenne de Scences Economques (SCSE), 2006, Montréal, Canada <halshs > HAL Id: halshs Submtted on 4 Apr 2007 HAL s a mult-dscplnary open access archve for the depost and dssemnaton of scentfc research documents, whether they are publshed or not. The documents may come from teachng and research nsttutons n France or abroad, or from publc or prvate research centers. L archve ouverte plurdscplnare HAL, est destnée au dépôt et à la dffuson de documents scentfques de nveau recherche, publés ou non, émanant des établssements d ensegnement et de recherche franças ou étrangers, des laboratores publcs ou prvés.

2 EXIT FROM SHORT-TERM CONTRACT A FRENCH EMPIRICAL ANALYSIS, Mohamed Al BEN HALIMA * Abstract : Ths paper tests the determnants of the transton out of short-term contracts by means of competng rs form of the sem-parametrc Cox proportonal hazard model. We use three labour maret states dstngushng between ext nto long-term contract, another short-term contract, and unemployment. For each type of ext, we defne a separate hazard functon that we ll call a type-specfc hazard. The estmates are carred out from a dataset recordng ndvdual labour maret hstores, the French Labour Force Survey (LFS) collected by the French Natonal Statstcal Insttute (INSEE). The competng rs model s estmated for all the ndvdual as well as separately for men and women. Our results show that, for men and women, the condtonal probablty of extng STC nto LTC decreases after 12 th months. Moreover, stayng n the same frm after a STC ncreases the chances of gettng stable jobs n the future. Keywords : Short-Term Contracts, Duraton model, Competng rs hazard model Classfcaton JEL : J41, J24, C10, C41 Résumé : Ce paper teste les détermnants de la transton des contrats de traval temporare (CTT) par un modèle à rsque concurrent, de type Cox à rsque proportonnel. Nous dstnguons tros statuts de transton sur le marché de traval: contrat à durée ndétermnée (CDI), un autre contrat de traval temporare et enfn le chômage. Pour chaque type de transton, nous allons défnr une foncton de rsque spécfque à chaque transton. Les estmatons sont réalsées à partr de l'enquête Emplo réalsée annuellement par l'insee. Le modèle à rsque concurrent a été estmé pour l ensemble des ndvdus ans séparément pour les hommes et les femmes. Les résultats montrent que, pour les hommes et les femmes, la probablté condtonnelle de la transton d un CTT vers un CDI décroît à partr du 12éme mos. D'alleurs, rester dans la même frme après un CTT augmente les chances d'obtenr un emplo stable dans le future. Mots-clés : Contrats de traval temporare, Modèle de durée, Modèle à rsque concurrent * GATE (Groupe d Analyse et de Théore Economque), UMR-CNRS n 5824, Unversty of Lyon 2. 93, chemn des Moulles - B.P ECULLY cedex (France) Correspondng author : benhalma@gate.cnrs.fr Phone +33(0) Fax +33(0) I would le to than Jean-Yves LESUEUR for hs support and hs advce, Alexandre KYCH, Davd GRAY and Claude MONTMARQUETTE for here constructve remars that were most helpful n achevng ths wor. 1

3 1. INTRODUCTION The type of labour contracts was a basc matter of dscusson durng the last decades, especally n the developed countres. Whereas, the short-term contracts had been the domnate pattern n ths major. Accordngly, the European advanced mportance n ths pont, and especally the France one, had pushed us to tae t as a real case of study n our research. Recent studes had focused on short-term contracts, whch consderably the preferable choce for the employers decson-tang, due to ts restrctons flexblty whch also could ncrease the economc moblty and reduce nequaltes between unemployers and employers and wouldn t create a dual labour maret, whch s descrbed n the lterature (Hunt, 2000; Maurn, 2000). However, employers prefer the short-term contracts n ther recrutment strateges due to a lot of reasons. Frst, costs reason, due to whch the temporary contracts are less costly on them wth respect to the permanent ones (Bentolla and Bertola, 1990; Bentolla and Sant-Paul, 1994; and Booth, 1997). Second, where as the short-term contracts have a hgh flexblty n fllng the job s vacances. Thrd, absorpton-tendency for the demand fluctuatons adjustment s costs. Fourth, reducng uncertanty n matchng value. Whereas, the matchng theory, that had been suggested by Jovanovc (1979), consttutes the model of reference ntegratng ths dmenson of the labour maret. Moreover, they added that t s necessary to nstall a mechansm mang t possble to produce an optmal parng n order to reach an effcent producton. In fact, the Frms current strateges are to ease ntally short-term contracts as a probatonary stage. Then, accordng to the applcant abltes and labour demand, long-term contracts would be proposed. Furthermore, as ponted out by Loh (1994), and Lazear (1995) probatonary perods may nduce self-selecton of those worers wth hgher ablty because they have a hgher probablty to obtan permanent contracts. Thus, we ll try focus n ths study short-terms contracts effects. We ll tae the msson of llustratng the steppng stone prncple or arrvng to a new methodology. Moreover, to what extent does a short-term contract (STC) ncrease the probablty of fndng a long-term contract (LTC)? Exstng theory suggests dfferent answer to ths queston. On the one hand, Booth et al. (2002b) examned whether temporary contracts deserve to be consdered as steppng stones. The authors confrm for ther Unted Kngdom sample that over a perod of 7 years, about 38% of all worers observed on short-term contracts move on to permanent employment after the STC expres wth hgher wages and frnge benefts. They also show that 2

4 hgh effort among temporary worers s postvely correlated wth the probablty of career advancement. On the other hand, the dead-end suggestons had been presented n the dscusson table, where Güell and Petrongolo (2001) were the most belevers on dead-end n studyng the duraton pattern of short-term contracts n Span and the determnants of ther converson nto permanent ones. D Addo and Rosholm (2005) study transtons out of temporary jobs usng the waves of the European Communty Household Panel. The authors confrm that very short contracts provde hgher chances of labour maret excluson especally for men. Whle, the emprcal wors avalable for Italy (Adam and Canzan, 1998), France (Abowd, Corbel and Kramarz, 1999), and Germany (Hagen, 2003) all ndcate that short-term contracts are steppng stones to permanent forms of employment rather than dead-end jobs. The structure of the artcle s organzed as follows. In the next secton we present the French legslaton of short-term contracts. In the secton 3, we hghlght the econometrc problems encountered to test the determnants of the ext of short-term contracts. Secton 4 descrbes the French Labor Force Survey, Enquête Emplo, collected by the French Natonal Statstcal Insttute (INSEE). Secton 5 shows a nonparametrc analyss of the hazard functons for men and women, nto long-term contract, short-term contract and unemployment by the method of Kaplan-Mer. Secton 6 shows the econometrc results from the estmate of the competng rs model for all the ndvdual as well as separately for men and women. Fnally, a synthess of the man results s gven as a concluson. 2. The French Regulaton of Short-Term Contracts French labor laws 1 allow frms to hre worers on two types of regular employment contracts: Indefnte-Term Contracts (Contrats à Durée Indétermnée, CDI) and Fxed-Term Contracts (Contrats à Durée Détermnée, CDD). The current archtecture of CDDs, ntroduced n 1979, dates bac to an agreement sgned n March Under ths agreement, CDDs can be offered by frms for only very precse reasons: CDD cannot be used to fll a job that would exst under normal and permanent busness condtons for a gven frm (Artcle L.122). 1 For more detals about French Labour Laws, see Abowd and Kramarz [2003] for an executve summary n englsh, and Lamy [1992] for an explanaton of the text of the law. 3

5 The use of fxed term employment contracts s covered by the detaled employment legslaton set out n the code du traval. Ths specfes the stuatons n whch employers may use fxed term contracts and places tght restrctons on ther duraton and renewal. The french government has made clear ts vew that employers should not mae excessve use of fxed term contracts, partcularly short-term contracts, whch have been cted as a ey factor n the development of socal excluson. Fxed term contracts can be used n fve dfferent stuatons: to replace absent employees; f a company grows temporarly and needs extra employees for a lmted perod of tme; for seasonal wor; to fll a vacant post untl a permanent employee starts the job; and n specfed jobs and sectors where permanent contracts are not approprate. A short-term contract must be n wrtng, otherwse t wll be judged to be a permanent contract. It must state the reason for a fxed term contract beng concluded, whch must correspond to the reasons outlned above. It must state the duraton of the contract, whch may be ether for a fxed length of tme or untl a specfed event happens (ncludng the completon of one or more tass). The contract may not be broen before ths fnshng pont unless both the employer and employee agree to t. Generally, successve fxed term contracts may last no longer than 18 months n total, and a fxed term contract cannot be renewed more than twce n ths tme. However, fxed term contracts concluded for urgent safety wor may not last longer than nne months. There must be a gap between successve contracts that s at least a thrd as long as the contract t follows, for example an employee on a nne month fxed term contract must be lad off for three months before beng hred on another fxed term contract by the same employer. Fxed term employees must receve the same legal, contractual and collectve rghts as comparable permanent employees. At the end of a fxed term contract, employees generally receve a payment worth at least 6% (10% snce January 2002) of ther total gross earnngs durng the perod covered by the contract and compensaton for any pad annual leave not taen. Although ther use s formally restrcted, CDDs are the most common method of hrng. For example, n 1990, 58% of all hres were through CDD, they were 68% n 1996 and 75% n 1999 (Coutrot [2000]). On the other hand, durng the 1990 s, more than 90% of the stoc of employees n prvate for-proft or sem-publc establshments were on CDIs. For those hred under CDD approxmately one n three s eventually converted to CDI (Abowd, Corbel and Kramarz [1999]). 4

6 3. The econometrc model In our analyss, we develop a competng rs form of the sem-parametrc Cox proportonal hazard model (see Cox and Oaes, 1988). Ths type of model allows controllng for all the standard determnants, as well as for duraton dependence, whch s the pure effect of tme spent n temporary employment on the probablty of movng to another state. For each type of ext, we are gong to defne a separate hazard functon that we ll call a typespecfc or cause-specfc hazard. The hazard model utlzes data on the sequence, tmng, and the state n whch an ndvdual occupes; the probablty of the type of ext s a functon of tme. Ths s referred to as the hazard rate or functon. The hazard rate s the dependent varable, and the relatonshp between ths rate and a varety of observed and unobserved varables s modelled. The hazard model has been wdely used n bomedcal and socal scence research, by Cox (1972), Kalbflesh and Prentce (1980), Hecman and Snger (1984), and Kefer (1988). We wll dstngush three labor maret states: long-term contract n prvate and publc establshments (state 1), short-term contract, ncludng fxed-term contract, ntermedares and seasonal worers (state 2) and unemployment (state 3). For each type of ext, we are gong to defne a separate hazard functon that we ll call a type-specfc or cause-specfc hazard. As before let T be a nonnegatve random varable denotng the length of short-term contract for ndvdual. In the competng rs model the choce set s specfed by a dscretestate space and the choce and swtchng between labor maret states are formalzed through transton rates, whle the tme spent on actvtes s estmated by a dstrbuton functon (see, Flnn and Hecman 1982, Hecman and Snger 1984). Let be a random varable denotng the type of ext that occurred to ndvdual. We now defne h ( t), the hazard for death type at tme t for ndvdual, as follows: h ( t) Pr = lm t 0 { t T t + t, J = j / T t} t, = 1,2,3 Thus, the condtonal probablty n ths equaton, s the probablty that ndvdual ext occurs between t and t + t, and the ext s of type, gven that the person had not already ded by tme t. The overall hazard of extng s just the sum of all the type-specfc hazards, that s, h ( t) = h ( t) 5

7 You can nterpret type-specfc hazards n much the same way as ordnary hazards. Ther metrc s the number of events per unt nterval of tme, except now the events are of a specfc type. Based on the type-specfc hazards we can also defne type-specfc survval functon: S t = 0 ( t) exp h ( u) du In order to estmate the transton rates of short-term contract and the effects of ndependent varables on these, we need to specfy the functonal form of the hazard model. The most wdely used specfcatons are the accelerated tme model and the Cox proportonal hazard model. Accordng to the Cox proportonal hazard model (Cox 1972), covarates have a multplcatve relatonshp wth the hazard functon (the proportonalty assumpton), causng them to shft the hazard functon up or downward. Therefore, covarates do not have a drect mpact on the duraton of an actvty. In contrast, the accelerated tme model (Cox and Oaes 1984) assumes that covarates act multplcatve on tme rather than on the hazard functon. Covarates, therefore, ncrease or decrease the tme spent on an actvty. As the current paper studes the mpact of explanatory varables on the ext of shortterm contract, Cox proportonal hazard model s used. The competng rs accelerated tme model s specfed as follows: where h ( t X ) = h ( t ) exp( x β ) t s the elapsed duraton of short-term contract for ndvdual, ( t X ) 0 h / denotes the hazard rate for ext condtonal on a vector of covarates, X; h ( ) denotes the baselne hazard functon measurng the effects of the tme passed snce the prevous labour maret state (STC) on the probablty movng to another state ; and x β denotes the covarates and parameter estmates for ext-specfc. The overall lelhood functon can be parttoned nto the product of the -specfc lelhoods. To see ths, suppose that there are n observatons at rs of falng n one of ways. The ndvdual contrbuton of the th ndvdual falng by event s gven by L = f ( t X, β ) S ( t X, β ) r r r r 0 t 6

8 where the subscrpt denotes the th event and r n the product term mples that the product s taen over the survvor tmes for all states except. The jont densty for the probablty densty functon and survvor functon for all actvtes are obtaned by ntegratng h ( t). The overall lelhood functon for the full sample s then gven by L = n r = 1 = 1 f ( t X, β ) S ( t X, β ) Ths parttonng s easer to see f we defne a censorng ndcator such that: when = 1 δ 1 f faled due to = 0 otherwse δ the observaton s observed falng due to rs ; when δ = 0, the observaton s rght-censored. Incorporatng δ nto the lelhood functon, the lelhood of the sampled duraton tmes may be expressed as L n r = = 1 = 1 f δ 1 δ ( t X, β ) S ( t X, β ) Importantly, for ths model, the assumpton must be made that the K rss are condtonally ndependent. Gven the covarates, the survval tmes for rs are ndependent of the survval tmes for rs r. 4. Data In ths nstance, we use a longtudnal mcro dataset recordng ndvdual labor maret hstores, the French Labor Force Survey (hereafter: LFS), Enquête Emplo, collected by the French Natonal Statstcal Insttute (INSEE). We consder the cohorts that were sampled from 1990 to The households n the cohorts are ntervewed n March of three consecutve years. The survey s representatve of the French populaton aged 15 and up. The samplng fracton s approxmately 1/300. Large samples of about 150,000 ndvduals aged 15 or more, n 75,000 households, can thus be ntervewed three tmes, n March of three subsequent years, about varous aspects of ther employment hstores. At each ntervew date, each ndvdual s ased to descrbe her actual labor maret state, and s also ased to provde some retrospectve nformaton on her labor maret hstory. Concentrate for example on ndvduals ntervewed n March At the frst ntervew, they are requested to retrace month by month, ther labor maret hstory between March 1999 and March At the second ntervew, they are request to recall (agan month by month) 7

9 the perod between March 2000 and March 2001, and at the thrd ntervew, the perod between March 2001 and March The followng standard nformaton s completed for each ntervew: sex, age, educaton, job tenure, ndustry, worng hours, prvate and publc sector. We had to dstngush ths a complete trajectores ( y y y ) t, t+ t+ 2 1,, where t s the date of the frst recorded observaton. For the estmaton of the emprcal model, the sample s restrcted to ndvduals respondng three consecutve waves and ndvduals who stll n the short-term contract at t, that means to households enterng the survey between 1991 and Households that are not ntervewed three tmes generate ncomplete labor maret hstores 2. Excludng such households from the cohort does not bas our results under the assumpton that non-partcpaton phenomenon (panel attrton) s exogenous. 5. An overvew of Short-Term Contract n French 5.1 The Aggregate Pcture In studyng the evoluton of the short-term contract proportons wth respect to the total French employment contracts (Fgure 1), a gradual ncrease for the beneft of the shortterm contracts had effectvely appeared between 1990 to 2000 (2.86% to 6.67%). Otherwse, the new mllennum were consderable the pc of the percentage of the short-term contract. Fgure 2 shows the share of short-term contract n total employment accordng to the varable of sex. Although the share of STCs n total employment ncreases for both men and women. Generally, the share of women n short-term contract s hgher than of men especally n the years between 1990 and to After 2000 the share of short-term contract contnued to declne margnally and n 2002 t reached 6.6% for women and 6.3% for men. The occurrence of ths type of employment have peaed n 2000 at 7.06% for women and 6.62% for men. 2 We excluded households that are only ntervewed once or twce. Ths can occur for two reasons. Frst, f a household moves to another address no attempt s made to fnd ths household s new address. The household that s lost s smply replaced by the INSEE by the new nhabtants occupyng the orgnal address. As a consequence, a household that have moved between the frst and the second ntervew date, s ntervewed only once. A household that have moved between the second and the thrd ntervew date, s ntervewed only twce. Second, even f a household remans at the same address durng the whole observaton perod, t can refuse to partcpate at the second and/or thrd ntervew date. Households that are not ntervewed three tmes generate ncomplete labour maret hstores. 8

10 Fgure 3 shows that the share of short-term contract n total employment s more frequent n some age groups than n others. We note that there s a greater proporton of shortterm contract among youth [15-24] wth respect to the old ages [50-59], whereas, the graph shows an ncrease n the gap dfference from (5.09% n 1990 to 8.53% n 2002). Thus, normally we can conclude that the short-term contract are also more related to the frst worng step for the employers way, whle the ncrease n the age-group gap could be related to the experence level. An analyss of the share of short-term contract n total employment by four sectors confrms the development of the short-term contract n partcular sectors of economy. Fgure 4 shows that the share of STC s s roughly ncreasng n all sectors durng the 90 s. Whle the bggest occupatons specfcally STC s are found among servce and constructon sectors. Snce 1993, the share of STC s n total employment ncreased sharply n servce sectors. Whle n 1993, 5.55% of STC s were actve n the servce sector, by 2000 ths percentage has ncreased to 11.73%. We should note, however, that the constructon and servces sector stll accounts for large part of all short-term contracts n the labour maret. In ths s related to the nd of the servces worng structure and the seasonable actvtes. By loong at the actvty rate for our sample of worers for the perod from the 1990 to 2002, tow facts are mmedately evdent (see table 1). The frst one s that worng as employee wth a long-term contract remans by far the most mportant form of employment but t contnues to decrease durng the perod. Ths decrease s reflected n the correspondng ncrease n the percentage of short-term contracts and unemployment, whle the percentage of those who are n educaton or job tranng remans roughly constant durng the same perod. The second fact s that worng as an employee wth a short-term contract remans by far the smallest form of employment but t contnues to ncrease by approxmately four percentage ponts. However, the share of short-term contract n total employment has ncreased from 3.95% n 1990 to 8.72% n When loong at the average annual growth n long-term and short-term contract concernng the perod from 1990 to 2002, fluctuatons n short-term contract over all perods are much more developed than for long-term contract. Table 2 shows one-year transton of short-term contract between the sx labour maret statuses n consderaton for the perod 1990 to It s nterestng to pont out when comparng the md-1990 to the begnnng of Durng the 90/91, around 35% of worers on a short-term contract are stll n the same status one year later, wth 38% of them managng to obtan a more stable form of employment but around 15% of them are not employed. 9

11 Approxmately 2% of these STC are stll n educaton or tranng one year later. Loong n more detals at the transton between 2001 and 2002, there s a hgh degree of persstence for people under short-term contract: around 57% of ndvduals are stll n the same stuaton of labor contract one year later, but only 22% of STC steppng-stones wth a long-term contract. Durng the perod from 1990 to 2002, the proporton of short-term contract that led to wth a long-term contract has decreased and at the same tme, the share of short-term contract that led wth a new STC has ncreased monotoncally. Table 3 shows the percentage of transton between dfferent economc actvty statuses dffer accordng to the personal characterstcs of the ndvduals. Loong at the gender dmenson, the share for men on short-term contracts to move to long-term contracts s around tow percentage ponts hgher than for women. Ths suggests that men havng a temporary job are more lely to obtan a stable employment, whle for women, ths s more lely to represent a perod of wor n a rather more ntermttent career path. Consderng the dfferent age group, t should be noted that the lowest degree of transton n long-term contracts (20.43%) s for those aged Snce the majorty of young people have no professonal experence on the labour maret, they accept even short-term contract (61.29%). The worers aged between [40-49], represent the hghest percentage n the transton towards the long-term contract. The analyss of transton by levels of qualfcaton hghlghts once more the mportance of educaton for labour maret outcomes. The percentage of temporary worers to move to long-term contract ncreases wth the level of qualfcaton, from 20.28% for the low-qualfed to 31.85% for the hgh qualfed. Also, regardless of the economc actvty statuses, the percentage that worers loosng or leavng ther job decreases wth level of qualfcatons of the ndvdual. 5.2 The hazard functon nto Long-term and Short-term contract: We have represented graphcally n Fgure 5, Fgure 6 and Fgure 7, the hazard functons for men and women, nto long-term contract, short-term contract and unemployment. We notce that both of them are non-monotonc. When consderng the condtonal probablty of extng STC nto LTC, we note that for men and women t decreases after 12 th months. The analyss of hazard rato confrms, men as well as for women, that short duraton of STC ncreases sgnfcantly ther chances to reach job stablty through LTC. 10

12 The hazard rate measurng the condtonal probablty of extng STC to another STC, reported n the Fgure 6 shows that women wth longer short-term contracts are more lely to get another STC. The hazard rates nto STC clearly confrm ths stuaton, by suggestng that temporary jobs have the same effects for men and women respectvely. Transton rate nto unemployment (Fgure 7) s decreasng after the frst year for both men and women. We should as the followng queston does the fact of obtanng a LTC after a STC related to the fact of stayng n the same frm n another word does the fact of movng to another frm ncrease or decrease the probablty of extng n a LTC? The hazard functon s decreasng whatever the type of frms of transton. One can compare hazard n long-term contract (Fgure 8) for the two types of frms (transton n the same frm; transton to another frm). The hazard functon nto LTC of the ndvduals wth transton n the same frm s always hgher that of the ndvduals wth transton to another frm. The probablty of extng to STC n the same frm (Fgure 9) s hgher than that of obtanng STC n another frm untl the 12 th month. But afterwards, two hazard functons are decreasng monotoncally. Ths can be explaned by the fact that STC s a probatonary perod n the frst year to test the ndvdual s slls. Then, accordng to the applcant abltes and labour demand, long-term contracts would be proposed. 6. Results of estmaton The estmaton results of the competng rs model out of short-term contracts confrm what has been observed by descrptve statstcs. Related to transtons nto LTC, we notce that the age, the human captal, the labour maret condtons (modelled here through the unemployment rate), beng occuped n some nd of elementary tass, sector of actvty, and fnally the prvate/publc type of employment, affect the hazard rate of men and women n the same way. 6.1 Results for the whole set of the ndvduals Table 4 report our results for the determnants of transtons probablty from shortterm contract to three economc actvty statuses for the perod 1990 to Frst, younger worers aged between relatve to the base of ndvduals aged between have a low probablty of fndng a long-term contract. Older worers aged between have also a hgher rs of excluson from the labour maret after a short-term 11

13 contract. Snce the majorty of young people have no professonal experence on the labour maret, they accept even precarous worng condtons. Ths result s confrmed by Booth, Francescon and Fran (2002a) on Englsh data. Men have a negatve and sgnfcant probablty of gettng another short-term contract or ext nto unemployment. The chances of havng an establshed poston under long-term contract are hgher for Europeans than for non-european people. The results show that worers wth low human captal wth techncal educaton hold more long-term contract than the worers wth frst cycle degree. The result found suggests that when short-term contract are held by ndvduals wth hgh slled (unversty degree), ths s not an effcent way to get promoted nto the labour maret, but a powerful tool to put them out of the labour maret. It s clear from the results assocated wth the ext nto unemployment that those holdng hgh slled postons have lower rss of facng excluson after the experence of a short-term contract. Occupatonal levels varables ndcate that executves or professonals have a postve and sgnfcant probablty to transtons nto long-term contract relatve to wormen. On the other hand, holdng a short-term contract wth ntermedary professons reduces the probablty of fndng a long-term contract. Ext from short-term contract vares accordng to the sze of frms. Compared to small frms, worers n frms wth 50 to 499 employees can also worsen the chances of gettng stable jobs n the future. Havng temporary jobs n the publc sector s assocated wth an ncreased rs of becomng non-employed. The negatve sgn assocated wth ths varable n the transton nto long-term contract ponts at the rs of such a practce. After a short-term contract n the publc sector (compared to prvate), worers are more lely to fall agan nto unemployment or short-term contract. Transtons probablty between dfferent economc actvty statuses dffer accordng to the experence n labour maret. The probablty of holdng a long-term contract s greater when ndvduals have a less experence n short term contract. On the other hand, people on short-term contract wth a professonal experence n the labour maret fnd dffcultes of havng an establshed poston and ther movng nto unemployment. The ntroducton of tme wor effects shows that full-tme worers have a postve and sgnfcant probablty to obtan an LTC compared to those n part-tme. 12

14 Consderng the type of short-term contract, t should be noted that the ndvduals under fxed-term contract have a hgher probablty of holdng long-term contracts. Therefore, ndvduals under fxed-term contract mprove ther chances of stablzaton n an employment poston compared to people wth seasonal or ntermedares contract. The local rate of unemployment tends to ncrease the probablty of transtons shortterm contract or unemployment. Persstence of unemployment obvously maes the recourse to ths stuaton of employment much more frequent. 6.2 Results for men and for women Table 5 and Table 6 report our results for the determnants of transton probablty from short-term contract to three economc actvty statuses for men and women ale. Concernng the age, for men as well as for women, young people occupyng an STC are more lely to obtan another STC. On the other hand, for older men of more than 50 years old, ther age s an obstacle that leads them to unemployment. The martal statute has the same sgnfcant and postve effect for men and women. Indeed, beng marred has an effect of ncreasng the probablty of havng a LTC. By examnng the results of the varable of the level of educaton, t should be noted that we have a sgnfcant dfference between men and women related to ther types of extng after a STC. Indeed, men wth hgher educaton have a postve and sgnfcant probablty to hold LTC. On the other hand, women havng the same level of educaton fnd dffcultes to obtan an LTC. Ths same phenomenon s observed for worers wth techncal educaton (havng a techncal currculum). The results show the latter hold more LTC whle women encounter more dffcultes n havng an establshed poston and ther movng to another STC. Women worng n large frms have a hgher probablty of fndng a long-term contract. On the other hand, men worng n large frms have also hgher rss of excluson from the labour maret after a short-term contract. Worng n the publc sector under a short-term contract has the same effect on men and women ale. Both sexes fnd dffcultes to move to an LTC. The tme of wor (full-tme / part-tme) doesn t have a sgnfcant effect on the types of extng from an STC for men. On the other hand, women havng full-tme jobs are more 13

15 lely to obtan an LTC. It should be noted too, that women havng a part tme job plays a sgnfcant and negatve effect on the probablty of movng to an LTC and doesn t ncrease the probablty of extng to another STC or to unemployment. As to the wages, they have the same postve and sgnfcant effect on the probablty of transton towards an LTC for both men and women. 6.3 Results for type of frm transton Table 7 and Table 8 report our results for the determnants of transton probablty from short-term contract to long-term contract or another short-term contract statuses by type of frm transton (n the same frm / n another frm). Men have a postve and sgnfcant probablty of steppng-stones to another longterm contract n the same frm. The chances of havng an establshed poston under long-term contract after a short-term contract n the same frm are hgher for marred than for executves or professonals worers. Transtons probablty between dfferent statuses dffer accordng to the type of shortterm contract n labour maret. The probablty of holdng a long-term contract s greater when ndvduals have a fxed-term contract and stayng n the same frm. On the other hand, people stayng n the same frm or movng to another frm, on ntermedares or seasonal contract fnd dffcultes of havng an establshed poston and ther extng nto short-term contract. 7. Concluson Ths paper test the determnants of the transton out of short-term contract. Moreover, does a short-term contract (STC) ncrease the probablty of fndng a long-term contract (LTC)? In our analyss, we develop a competng rs form of the sem-parametrc Cox proportonal hazard model Ths type of model allows controllng for all the standard determnants, as well as for duraton dependence, whch s the pure effect of tme spent n temporary employment on the probablty of movng to another state. We wll dstngush three labour extng: long-term contract n prvate and publc establshments (state 1), shortterm contract, ncludng fxed-term contract, ntermedares and seasonal worers (state 2) and unemployment (state 3). For each type of ext, we are gong to defne a separate hazard functon that we ll call a type-specfc or cause-specfc hazard. 14

16 The estmates are carred out from a longtudnal mcro dataset recordng ndvdual labour maret hstores, the French Labour Force Survey. We consder the cohorts that were sampled from 1990 to When consderng the hazard functons for extng nto LTC, we note that for men and women t decreases after 12 th months. The analyss of hazard rato confrms, men as well as for women, that short duraton of STC ncreases sgnfcantly ther chances to reach job stablty through LTC. One can compare hazard functon n long-term contract for the two types of frms (transton n the same frm; transton to another frm), we noted that the hazard functon nto LTC of the ndvduals wth transton n the same frm s always hgher that of the ndvduals wth transton to another frm. The competng rs model s estmated for all the ndvdual as well as separately for men and women. The estmaton results confrm what has been observed by descrptve statstcs. Related to transtons nto LTC, we notce that the age, the human captal, the labour maret condtons (modelled here through the unemployment rate), beng occuped n some nd of elementary tass, sector of actvty, and fnally the prvate/publc type of employment, affect the hazard rate of men and women n the same way. 15

17 Lterature Abowd, J.M. and F. Kramarz (2003), «The Costs of Hrng and Separatons», Labour Economcs, vol. 10, pp Abowd, John M., Patrc Corbel, et Francs Kramarz (1999) «he Entry and Ext of Worers and the Growth of Employment: An Analyss of French Establshments» Revew of Economcs and Statstcs, vol. 81, pp Adam and Canzan (1998), «Partal De-Regulaton: Fxed-Term Contracts n Italy and Span» CEP Dscusson Papers from Centre for Economc Performance. Bentolla and Sant-Paul 1994 Bentolla, S. and Sant-Paul, G. (1994), «A model of labour demand wth lnear adjustment costs», Labour Economcs, vol. 1, pp Bentolla, S. and Bertola, G. (1990) «Frng costs and labour demand: how bad s Euroscleross», Revew of Economc Studes, vol. 57, pp Booth, A.L. (1997) «An analyss of frng costs and ther mplcatons for unemployment polcy», n Unemployment Polcy, Cambrdge: Cambrdge Unversty Press, pp Booth, A.L., Dolado, J.J., Fran, J., (2002b) «Introducton: symposum on temporary wor» Economc Journal 112 (480), F181 F189. Booth, A.L., Francescon, M., Fran, J., (2002a) «Temporary jobs: steppng stones or dead ends?» Economc Journal 112 (480), F189 F213. Coutrot, T. (2000), «Les Facteurs de Recours Aux Contrats Temporares», DARES, Premères Synthèses, 2000, Cox, D.R. (1972), «Regresson models and lfe tables» Journal of Royal Statstcal Socety, vol 34, pp Cox, D.R. and Oaes, D. (1988) «Analyss of Survval Data» London: Chapman and Hall D Addo A.C. and Rosholm M. (2005), «Exts from temporary jobs n Europe: A competng rss analyss», Labour Economcs, Vol 12, pp Flnn, C. and J.J. Hecman (1982), «Models for the analyss of labour force dynamcs», n R. Baseman and G. Rhodes (eds), Advances n Econometrcs, Vol. 1, Greenwch: JAI. Güell, M. and Petrongolo, B. (2001), «Worer transtons from temporary to Permanent employment: the case of Span.», Centre for Economc Performance, LSE, 0438 Hagen, T., (2003) «Do fxed term contracts ncrease the long-term employment opportuntes of the unemployed?»zew Dscusson Paper, 1 56 (03-49, ZEW Mannhem / Germany). Hecman, J.J. and B. Snger (1984), «A Method of Mnmzng the Dstrbutonal Impact n Econometrc Models for Duraton Data», Econometrca, vol 52, pp Hunt et Jennfer (2000) «Frng Costs, Employment Fluctuatons and Average Employment: An Examnaton of Germany» Economca, London School of Economcs and Poltcal Scence, vol. 67(127), pp Jovanc B. (1979) «Job Matchng and the Theory of Turnover», Journal of Poltcal Economy, vol 87 n 5 p Kalbflesch, J.D., Prentce, R.L. (1980) «The Statstcal Analyss of Falure Tme Data» John Wley & Sons, New Yor Kefer, N. (1988) «Economc duraton data and hazard functons», Journal of Economc Lterature, vol 26, pp Lamy (1992), Lamy Socal, Drot Du Traval, Pars: Lamy S.A. edteur, Lazear E.P. (1995) «Hrng Rsy Worers» NBER Worng Paper, 5334 Loh, E. S. (1994), «Employment Probaton as a Sortng Mechansm», Industral and Labor Relatons Revew 47 (3), Maurn, E. (2000), The European Paradox : Do Flexble Contracts Create Rgd Labor Marets?, Insttut Natonal de la Statstque et des Etudes Economques, Papers

18 Fgure 1: The share of Short-Term Contract n total employment, Sources: French Labour Force Surveys, % STC Women % STC Men Fgure 2: The share of Short-Term Contract n total employment by sex, Sources: French Labour Force Surveys,

19 Fgure 3: The share of Short-Term Contract n total employment by age groups, Sources: French Labour Force Surveys, Agrculture Manufacturng Constructon Servces Fgure 4: The share of Short-Term Contract n total employment by sectors, Sources: French Labour Force Surveys,

20 Table 1: Man actvty status by year Long-Term Contracts Self-Employed Short-Term Contracts Unemployment Educaton / Tranng Out of Labour Force Sources: French Labour Force Surveys, Table 2: One-year transtons by man economc status 90/91 91/92 92/93 93/94 94/95 95/96 96/97 97/98 98/99 99/00 00/01 01/02 To Long-Term Contracts To Self-Employed To Short-Term Contracts To Unemployment To Educaton / Tranng To Out of Labour Force Sources: French Labour Force Surveys, Table 3: One year transtons by man economc status and by personal characterstcs Long-Term Contracts Short-Term Contracts Unemployment Educaton / Tranng Out of Labour Force Total 1-year transton (t+1) yaer transton (t+2) By Personal Characterstcs Male Female [16-24] [25-39] [40-49] [50-59] [60-64] Low qualfed Medum qualfed Hgh qualfed Sources: French Labour Force Surveys,

21 d u r a t o n W o m e n M e n Fgure 5: Transton to Long-Term Contract by genders D u r a t o n W o m e n M e n Fgure 6: Transton to Short-Term Contract by genders D u r a t o n W o m e n M e n Fgure 7: Transton to Unemployment by genders 20

22 D u r a t o n A n o th e r f r m S a m e f r m Fgure 8: Transton to Long-Term Contract by frm D u r a t o n A n o th e r f r m S a m e f r m Fgure 9: Transton to Short-Term Contract by frm 21

23 Table 4 : Transton from Short-Term Contract Cox Competng Rs Model Transton to Long-Term Contract Transton to Short-Term Contract Transton to Unemployment AGE ** (0.0563) *** (0.0408) (0.0789) AGE (0.0469) ** (0.0329) (0.0681) AGE (0.0788) (0.0525) ** (0.1045) AGE (0.2712) (0.1439) (0.2855) Men (0.0374) *** (0.0266) *** (0.0543) European *** (0.1362) (0.0493) ** (0.0840) Marred *** (0.0357) *** (0.0249) (0.0499) Unversty degree (0.0775) *** (0.0537) *** (0.1301) Bachelor degree *** (0.0629) *** (0.0473) ** (0.0969) Techncal Educaton ** (0.0420) *** (0.0305) (0.0559) General Educaton (0.0556) *** (0.0395) (0.0771) Executve or Professonal *** (0.0890) *** (0.0611) *** (0.1561) Intermedary Professon *** (0.0585) *** (0.0431) *** (0.0882) Employee *** (0.0462) *** (0.0340) *** (0.0642) Frm Sze2: [50, 99] ** (0.0625) (0.0472) (0.0867) Frm Sze3 : [100, 499] *** (0.0454) *** (0.0323) (0.0675) Frm Sze4 : [500, +] (0.0430) (0.0304) ** (0.0609) Publc sector *** (0.0421) ** (0.0987) *** (0.0689) Experence1 : [0, 1[ *** (0.1294) *** (0.0550) *** (0.1931) Experence2 : [1, 5[ (0.0883) ** (0.0626) (0.2348) Experence3 : [5, 10[ (0.1025) (0.0725) *** (0.1246) Full-Tme *** (0.1181) *** (0.0617) *** (0.1575) Partal-Tme 1: [30h, + [ / Wee (0.0951) ** (0.0756) ** (0.1245) Partal-Tme 2: [15, 29h [ / Wee (0.0971) (0.0619) (0.2838) Fxed-Term Contract *** (0.0462) (0.1616) (0.2980) Seasonal Contract (0.2433) *** (0.1671) (0.2841) Intermedares Contract (0.2251) *** (0.1614) (0.0521) Monthly Wage *** (0.0362) (0.0260) (0.0789) Rate of Unemployment *** (0.0122) *** (0.0089) *** (0.0175) Number of Observatons (Rs) (3922) (7689) (1932) Log-Lelhood Restrcted Log-Lelhood Wald ch-2 (30) Coeffcents are Cox Competng Rs model estmates: (***) sgnfcant at 1%, (**) sgnfcant at de 5%, (*) sgnfcant at 10%. Ref AGE40-49, Women, Frst Cycle Degree, Worman, Frm Sze1: [0, 49], Experence4 : [10, +[, Partal-Tme 3: [0, 15h [ / Wee. 22

24 Table 5: Men s transton from Short-Term Contract Cox Competng Rs Model Transton to Long-Term Contract Transton to Short-Term Contract Transton to Unemployment AGE (0.0855) *** (0.0625) ** (0.1171) AGE (0.0705) *** (0.0506) ** (0.1021) AGE (0.1197) ** (0.0810) * (0.1459) AGE (0.4528) (0.2300) *** (0.3678) European *** (0.1224) (0.0611) (0.1105) Marred *** (0.0533) *** (0.0381) *** (0.0760) Unversty degree ** (0.1215) ** (0.0881) * (0.2235) Bachelor degree *** (0.0940) * (0.0785) (0.1579) Techncal Educaton ** (0.0564) (0.0420) (0.0773) General Educaton (0.0827) (0.0606) (0.1132) Executve or Professonal *** (0.1283) (0.0910) ** (0.2242) Intermedary Professon *** (0.0788) *** ( *** (0.1212) Employee *** (0.0790) *** (0.0583) *** (0.1201) Frm Sze2: [50, 99] * (0.0650) (0.0698) (0.1289) Frm Sze3 : [100, 499] (0.0882) *** (0.0476) (0.0979) Frm Sze4 : [500, +] ** (0.0630) ( (0.0873) Publc sector *** (0.0708) *** (0.0477) *** (0.1166) Experence1 : [0, 1[ *** (0.1351) *** (0.0880) *** (0.2977) Experence2 : [1, 5[ (0.1350) *** (0.0857) ** (0.2993) Experence3 : [5, 10[ (0.1553) (0.0971) (0.3776) Full-Tme (0.1976) (0.1136) (0.2274) Partal-Tme 1: [30h, + [ / Wee (0.2574) (0.1497) (0.3096) Partal-Tme 2: [15, 29h [ / Wee (0.2159) (0.1279) (0.2533) Fxed-Term Contract (0.3042) (0.2119) (0.3995) Seasonal Contract (0.3304) *** (0.2213) (0.4287) Intermedares Contract (0.3038) *** (0.2121) (0.4001) Monthly Wage *** (0.0528) ** (0.0394) (0.0752) Rate of Unemployment *** (0.0176) *** (0.0132) *** (0.0253) Number of Observatons (Rs) 6410 (1944) 6410 (3532) 6410 (934) Log-Lelhood Restrcted Log-Lelhood Wald ch-2 (29) Coeffcents are Cox Competng Rs model estmates: (***) sgnfcant at 1%, (**) sgnfcant at de 5%, (*) sgnfcant at 10%. Ref AGE40-49, Women, Frst Cycle Degree, Worman, Frm Sze1: [0, 49], Experence4 : [10, +[, Partal-Tme 3: [0, 15h [ / Wee. 23

25 Table 6 : Women s transton from Short-Term Contract: Cox Competng Rs Model Transton to Long-Term Contract Transton to Short-Term Contract Transton to Unemployment AGE (0.0764) ** (0.0550) (0.1099) AGE ** (0.0642) (0.0439) ** (0.0924) AGE (0.1052) (0.0694) (0.1515) AGE (0.3406) (0.1857) (0.4579) European (0.1127) (0.0859) ** (0.1316) Marred ** (0.0491) (0.0336) (0.0692) Unversty degree ** (0.1031) (0.0688) *** (0.1620) Bachelor degree (0.0856) *** (0.0609) *** (0.1246) Techncal Educaton (0.0636) ** (0.0447) (0.0820) General Educaton (0.0761) *** (0.0530) *** (0.1057) Executve or Professonal (0.4809) *** (0.0864) *** (0.2225) Intermedary Professon ** (0.0913) *** (0.0635) *** (0.1297) Employee (0.0670) *** (0.0479) (0.0830) Frm Sze2: [50, 99] ** (0.0890) (0.0642) ** (0.1181) Frm Sze3 : [100, 499] *** (0.0637) *** (0.0441) (0.0937) Frm Sze4 : [500, +] *** (0.0591) (0.0416) *** (0.0854) Publc sector *** (0.0540) *** (0.0369) *** (0.0867) Experence1 : [0, 1[ *** (0.1191) *** (0.0751) *** (0.2512) Experence2 : [1, 5[ (0.1182) (0.0727) ** (0.2539) Experence3 : [5, 10[ (0.1371) *** (0.0823) (0.3011) Full-Tme ** (0.1149) *** (0.0764) *** (0.1576) Partal-Tme 1: [30h, + [ / Wee *** (0.1373) ** (0.0903) *** (0.1896) Partal-Tme 2: [15, 29h [ / Wee (0.1112) * (0.0725)) *** (0.1473) Fxed-Term Contract (0.3347) (0.2466) (0.3907) Seasonal Contract (0.3570) *** (0.2531) (0.4060) Intermedares Contract (0.3301) ** (0.2460) (0.3916) Monthly Wage *** (0.0499) (0.0356) (0.0734) Rate of Unemployment *** (0.0173) *** (0.0121) *** (0.0245) Number of Observatons (Rs) 7133 (1978) 7133 (4157) 7133 (998) Log-Lelhood Restrcted Log-Lelhood Wald ch-2 (29) Coeffcents are Cox Competng Rs model estmates: (***) sgnfcant at 1%, (**) sgnfcant at de 5%, (*) sgnfcant at 10%. Ref AGE40-49, Women, Frst Cycle Degree, Worman, Frm Sze1: [0, 49], Experence4 : [10, +[, Partal-Tme 3: [0, 15h [ / Wee. 24

26 Table 7 : Transton from Short-Term Contract to Long-Term Contract Transton to Long-Term Contract In the same frm To another frm AGE (0.0728) (0.0929) AGE (0.0639) (0.0698) AGE * (0.1202) (0.1054) AGE * (0.5811) (0.3110) Men ** (0.0498) *** (0.0584) European (0.0848) *** (0.1064) Marred *** (0.0467) (0.0557) Unversty degree (0.1047) (0.1180) Bachelor degree *** (0.0786) (0.1065) Techncal Educaton ** (0.0539) (0.0682) General Educaton (0.0731) (0.0872) Executve or Professonal *** (0.1232) *** (0.1330) Intermedary Professon ** (0.0738) *** (0.0965) Employee (0.0600) *** (0.0747) Frm Sze2: [50, 99] (0.0772) (0.1073) Frm Sze3: [100, 499] *** (0.0590) ** (0.0717) Frm Sze4: [500, +] (0.0580) *** (0.0645) Publc sector *** (0.0610) *** (0.0596) Experence1 : [0, 1[ (0.5818) *** (0.2079) Experence2 : [1, 5[ (0.5903) (0.0916) Experence3 : [5, 10[ (0.8258) (0.1042) Full-Tme ** (0.1647) * (0.1365) Partal-Tme 1: [30h, + [ / Wee (0.1728) (0.1360) Partal-Tme 2: [15, 29h [ / Wee (0.1417) (0.1354) Fxed-Term Contract ** (0.0546) (0.5336) Seasonal Contract (0.2984) ** (0.5451) Intermedares Contract (0.2638) (0.5318) Monthly Wage *** (0.0436) (0.0574) Rate of Unemployment *** (0.0158) ( ) Number of Observatons (Rs) 7102 (2313) 6441 (1609) Log-Lelhood Restrcted Log-Lelhood Wald ch-2 (30) Coeffcents are Cox Competng Rs model estmates: (***) sgnfcant at 1%, (**) sgnfcant at de 5%, (*) sgnfcant at 10%. Ref AGE40-49, Women, Frst Cycle Degree, Worman, Frm Sze1: [0, 49], Experence4 : [10, +[, Partal-Tme 3: [0, 15h [ / Wee. 25