Transitions from involuntary and other temporary work 1

Transcription

1 Transitions from involuntary and other temporary work 1 Merja Kauhanen* & Jouko Nätti** This version October 2011 (On progress - not to be quoted without authors permission) * Labour Institute for Economic Research, Helsinki ** University of Tampere, Finland 1 This paper is part of the project Between employment and unemployment. Involuntary part-time and temporary work in Nordic countries: extent, explanations, transitions and well-being outcomes funded by the Academy of Finland. We are grateful for this financial support. 1

2 1. Introduction Labour markets in Finland and in other advanced countries have experienced many changes during last decades. Globalisation and technological change have rapidly changed surroundings in which firms operate and labour markets function. These changes have also launched increasing demands for adaptability of firms and workers. Increasing adaptability has manifested itself also in enhanced use of flexible work arrangements such as temporary employment. The incidence of temporary employment has grown significantly across OECD countries during the last two decades (OECD, 2004). By using temporary works arrangements it is easier for firms to adjust the number of employees in response to fluctuations in demand. Temporary contracts may also provide needed flexibility for workers e.g. to combine work and studies. The other side of the coin is that increased use of temporary contracts has also lead to a growth in unwanted flexibility. With the increase in temporary employment the share of involuntary temporary workers has also grown in many countries (reference). By definition, these are temporary workers who work in temporary jobs because they could not find a permanent job. The growth of temporary employment has raised interest in the economic and social consequences of temporary work arrangements. As regards consequences of temporary workers an essential question is related to labour market mobility of temporary workers: do temporary jobs act a stepping stone to permanent employment or are they dead ends not helping workers to advance their working careers. There is a growing number of research on whether temporary work advances or hinders labour market career by looking at subsequent labour market outcomes (e.g. Dekker, 2001; Booth et al., 2002; Holmlund and Storrie, 2002; Kauhanen, 2002; D Addio and Rosholm, 2005; Gagliarducci, 2005, Guell and Petrongolo, 2007). In these studies both stepping-stone hypothesis, i.e. transitions to permanent jobs, and the segmentation theory hypothesis, i.e. transition to nonactivity or no transition at all, have been studied. The evidence on the effect of temporary jobs 2

3 on the probability of finding a permanent job is quite mixed, and the results vary by country and also by the time period investigated. However, there is still surprisingly little research where difference is made between involuntary and other temporary workers. In other words, does unwanted flexibility matter for the outcomes? One might expect that there could be difference in how involuntary and other temporary work advances and hinders labour market career if there exists less favourable treatment of involuntary temporary workers. There is evidence from Finland that this might be the case. According to Kauhanen and Nätti (2011) involuntary temporary workers access to job place training and possibilities for skills development and are worse compared to other temporary workers which might also have consequences on their later labour market careers even when one controls the impact of background characteristics. The purpose of this paper is to investigate to what extent involuntary temporary work acts as a stepping stone or a dead end for a labour market career, whether labour market transitions are different between involuntary and other temporary workers, and how much differences we find among different groups of involuntary workers. In other words, we are interested in the consequences of involuntariness, i.e. the temporary job not being person s optimal choice. In particular, we study this question by investigating hazard rates from involuntary temporary job and other temporary jobs to permanent employment (the stepping stone) and to unemployment and outside labour market (i.e. nonemployment) in a duration analysis framework. A novel feature of our paper is that we focus on studying transitions from both involuntary temporary work and other temporary work in order to see whether the unwanted flexibility matters for the outcomes. To our knowledge there exists only one other paper by Hernanz et al that has studied transitions by making difference between involuntary and other temporary workers with Spanish and German data but their analysis is not in a duration analysis framework. Their results 3

4 indicate no statistically significant difference between involuntary and other temporary employees in the likelihood of getting a permanent job. Studying the consequences from unwanted flexibility is an important question also from a policy perspective. We have witnessed an increase of flexibility in the labour market that has also lead to an increase in unwanted flexibility, i.e. rise in the number of involuntary temporary work. In this analysis our focus is on one country only, Finland. Finland is an interesting case to study the consequences of involuntary temporary work because the majority of temporary workers in Finland are involuntary, i.e. their reason to work in a fixed-term job is that they have been unable to find a permanent job. The involuntariness of temporary jobs is highlighted among women in Finland: nearly 70 percent of female temporary workers are involuntary temporary workers (see Figure 1). In addition, over 50 percent of new employment relations in Finland are fixed-term and the share of temporary employment is above the EU average (the share was for a long period among the highest in the EU15 countries) % Figure 1. Involuntary temporary workers in Finland, share (%) (Statistics Finland, LFS) Both genders Men Women

5 The rest of the paper is organized as follows. Section 2.1 introduces the data and provides descriptive evidence on the transitions from involuntary and other temporary jobs to other labour market states, i.e. permanent employment, unemployment and outside labour force. Section 2.2 provides description of the statistical model employed in the empirical analyses. Section 3 presents results from statistical analyses and, finally, section 4 summarizes and discusses the main findings of the paper. 2. Data and methodology 2.1 Data The empirical analyses are based on individual-level panel data from the Finnish labour force survey (FLFS) which is representative of the Finnish working age population aged between 15 and 74. Our data cover years from 2004 to year The FLFS data collection is based on a random sample drawn twice a year from the Statistics Finland population database. The design of the FLFS is a complex rotating panel where each individual in the sample is interviewed five times during fifteen months period. The monthly sample consists of five rotation groups each of which contains 2400 individuals. The size of monthly sample is altogether around 12,000 persons. The selected sample was subsequently interviewed three, six, twelve and fifteen months after entry to the sample. Panel characteristic of the data and the detailed information about the labour force status of the individual in the data makes it possible to study transition dynamics from temporary employment to other labour markets. Temporary employment in this data is defined as a fixed-term work. As the data also includes information on the reason of temporary job we are able to distinguish involuntary temporary employment from other temporary work. Involuntary temporary work refers to cases where individuals work in temporary jobs because they could not find a permanent job. The group 5

6 other ( voluntary ) temporary employment is a residual and includes all other types of temporary employment. We use a subsample of only those individuals who entered the survey between years 2004 and 2007, completed all five interviews 2, and informed to work in a temporary contract (either as involuntary or other temporary worker) during the first interview time, i.e. at the beginning of our observation period. In addition, we leave those temporary workers outside our analysis whose main activity during the first interview time was student 3. The size of the sample constructed in this way is 5991 individuals and xx spells. Table 1 shows descriptive statistics on changes of labour market statuses of involuntary temporary and other temporary workers after 3 months and 15 months. After 3 months majority of involuntary and other temporary workers (75 % and 63 %) are still in temporary jobs and even after 15 months almost half of involuntary temporary workers (45.8 %) and nearly 40 percent of other temporary workers hold temporary jobs showing considerable persistence. It seems to be the case that involuntary workers jobs end more often into unemployment, but slightly less into transition to permanent jobs and considerably less into transition out of labour force. As regards other temporary workers their temporary jobs end more often in transition outside labour force and less often in unemployment compared to the group of involuntary workers on average. 2 Note that those individuals who entered the survey during 2007 the observation period extends to year It is very common for students in Finland to have a job even if their main activity is studying. Working students are mainly working in temporary jobs voluntarily and their aim is to earn money while studying and they do not seek a permanent job at that stage. As they mainly are voluntary temporary workers having them included might distort results. 6

7 Table 1. Transitions from involuntary temporary and other temporary work to other labour market states after 3 months (second interview time) and after 15 months (fifth interview time) Labour market status at t + 3 months TE PE Unemployment Outside labour market Labour market status at: Involuntary temporary employment Other temporary employment Labour market status at t + 15 months TE PE Unemployment Outside labour market Labour market status at: Involuntary temporary employment Other temporary employment TE= temporary employment, PE=permanent employment Figure 2 shows proportion of involuntary and other temporary jobs converted to permanent jobs at the last observation time during years These conversion rates are slightly higher for other temporary workers compared to involuntary temporary workers during this period. It is noteworthy that the conversion rate is noticeably higher in 2008 which indicates the good economic situation in Finland in 2008 before the economic crisis. 7

8 % 40 Figure 2. Share of involuntary and other temporary work converted to permanent job (%) Involuntary Other Descriptive analysis shows that conversion rates of involuntary and other temporary work to permanent work at the end of the observation period (t+15 months) differ by gender, age and education of temporary workers (see figure 3). For both genders the conversion rates are higher for other temporary workers compared to involuntary temporary workers. It is also noteworthy that men have a higher conversion rate for both groups of temporary workers. By age group conversion rates to permanent employment are lower for involuntary temporary workers compared to other temporary workers among those aged years and those aged over 44. In contrast, among year-old workers the conversion rate is higher for involuntary temporary workers compared to other temporary workers which might be an indication that this latter group of young workers even do not want to get a permanent job at this stage of the lives. By education group other temporary workers with tertiary education face the highest conversion rate. As for the other transitions from involuntary and other temporary work to unemployment the descriptive analysis of our data shows that transition to unemployment is more common among 8

9 involuntary temporary workers compared to other temporary workers irrespective of the gender, age and education (see figure 3a). Highest transition to unemployment is faced by involuntary temporary workers who are over 44 and those with only primary education. Concerning transitions outside labour force they are more common among other temporary workers, but the differences are almost nonexistent among those aged between 25 and 44 years and among workers with tertiary education % Figure 3. Conversion rate to permanent employment (after t+15 months), % Involuntary Other Female Male y y. 45+y. Primary Secondary Tertiary 9

10 % Figure 4. Transitions to nonemployment from involuntary and other temporary work 4a. Labour market status unemployed, t+15 months, % Involuntary Other Female Male y y. 45+y. Primary Secondary Tertiary % 4b. Labour market status outside labour force, t + 15 months, % Involuntary Other Female Male y y. 45+y. Primary Secondary Tertiary 10

11 In the empirical analyses of transitions from involuntary and other temporary work we control for a large set of background characteristics related to individual characteristics such as gender, age, nationality and marital status, and job characteristics such as industry and sector (see Table 1 for descriptive statistics of the data). Table 1 shows that women outnumber men in the group of involuntary temporary workers whereas the share of women and men are much more even among other temporary workers. The age distribution also differs between these two groups: involuntary temporary workers are on average older compared to other temporary workers. By education involuntary temporary workers have more often tertiary education, they work more often in upper-collar jobs and in public sector compared to other temporary workers. Expectedly, they also look for a new job more often compared to other temporary workers. There also exist some differences in the timing of transitions from involuntary and other temporary jobs, especially to permanent jobs. The proportion of moving to permanent jobs within first six months is considerably higher for other temporary workers compared to involuntary temporary workers. Of those other temporary workers who succeed in moving from temporary job to permanent one nearly 70 percent has made this transition within one year after the start of temporary contract, whereas the corresponding share for involuntary workers is around 15 percent points lower. On average transition to permanent employment from involuntary temporary jobs lasts longer (22.8 vs months). But the average duration before moving to nonemployment (i.e. either unemployment or outside labour force) is also higher for involuntary temporary workers. 11

12 Table 2. Transitions distribution by duration and destination, % Transitions within, %: Transitions from involuntary temporary jobs to: Transitions from other temporary jobs to: PE NE PE NE 1-6 months months months months over 24 months Average duration in months TE= temporary employment, PE=permanent employment 2.2 Methodology We investigate transitions from involuntary temporary employment and other temporary employment to permanent employment (the stepping stone hypothesis), unemployment and outside labour force (the segmentation theory hypothesis) in a discrete-time hazard model framework (Prentice and Gloecker, 1978) as our data is interval-censored. As we study transitions to one of several states (three states) we employ an independent competing risks model (see Jenkins 1995). We make the assumption that transitions can only occur at the boundaries of the intervals (see Narendrenathan and Stewart, 1993). With this assumption the likelihood contribution partitions into a product of terms, each of which is a function of a single destination-specific hazard only, and we are able to estimate the overall independent competing risk model by estimating separate destination-specific models treating other destinations as censored observations (Jenkins, 2000). We estimate a discrete-time hazard model (complementary log log model) where the hazard denotes the probability of a spell of involuntary or other temporary spell being completed by time t+1 given that it was still continuing at time t. The discrete-time or grouped hazard is given by 12

13 ( β γ ) hi ( t / xi ) = 1 exp exp xi ) + ( t) where t+ 1 γ t = λ t ( ) ( u) du is the integrated baseline hazard and X i is the vector of covariates which include both individual-specific characteristics (e.g. gender, age, education and socioeconomic status) and individual s job-specific characteristics (e.g. industry and sector) and β is the vector of coefficients associated with the X. We estimate the above discrete-time model semi-parametrically without restrictions on the baseline hazard (see Meyer 1990). The contribution of the ith individual to the log-likelihood is d 1 d 1 i i { [ ]} L = c ln h ( d / x ) + ln[1 ( h ( t / x )] = c ln 1 exp exp( x β) γ( d ) exp( x β) γ( t) i i i i i i i i i i i t= 1 t= 1 where d i is an indicator variable equal to 1 if the temporary spell ends and 0 otherwise. (Narendranathan and Stewart, 1993). The temporary employment spells that we observe are already in progress when the observation period begins (delayed entry). With delayed entry we need to condition on the length of the temporary employment at the first interview date using the information on the elapsed duration of the current contract by that time. If the duration of the current contract is j at the first interview date and the duration before exiting during the observation period k, then the total duration is d 4. The contribution of the ith individual to the log-likelihood becomes (reference here) j + k 1 j + k 1 i i i i { [ ]} L = c ln h ( j + k x ) + ln[1 ( h ( t / x )] = c ln 1 exp exp( x β) γ ( j + k ) exp( x β) γ( t) i i i i i i i i i i i i i t= ji + 1 t= ji

14 The literature suggests several effects if unobserved heterogeneity is ignored: (i) the model might over-estimate the negative duration dependence in the hazard (i.e. under-estimate the degree of positive duration-dependence), (ii) the proportionate response of the hazard rate to a change in regressor k is no longer constant, but declines with age, (iii) one gets an under-estimate of the true proportionate response of the hazard to a change in a regressor k from the non-frailty model (Jenkins, 2005). We therefore also estimate models where unobserved heterogeneity is taken into account in the models and test with likelihood ratio test whether unobserved heterogeneity is relevant or not. 4. Results (to be appended ) We report the hazard model results of the determinants of transitions from involuntary and other temporary employment to permanent employment, unemployment and outside labour force in tables 4-5 in the appendix. Rather than reporting the estimated coefficients, we present the results as hazard ratios from the underlying continuous model (exponentiated coefficients). As regards the transition from involuntary temporary work to permanent job our results (table 4) suggest that of the personal factors gender, age and educational level play a statistically significant role in transition into permanent job. During women had 15 percent lower hazard of moving into permanent job compared to men from involuntary temporary jobs. Those aged 45 and older had more difficulties in turning involuntary temporary job into permanent one compared to those aged between 35 and 44 years (the hazard was almost 20 percent lower). In turn younger involuntary temporary workers seem to do better than older ones: year-olds had 21 percent higher probability of moving to permanent job compared to year-olds. As for the impact of educational level our results suggest that involuntary temporary workers with tertiary education 4 14

15 have a clearly higher hazard of getting a permanent job compared to those involuntary temporary workers who have primary education. Between individuals with primary and secondary education there is no statistically significant difference. The industry in which involuntary temporary workers work also seem to matter for the stepping stone effect. Compared to manufacturing industry those temporary workers who worked in transportation and in private service sectors such as trade and business activities had a clearly higher hazard of getting a permanent job during our inspection period. In contrast, involuntary temporary workers in education faced a 33 percent lower hazard of getting permanent job. In Finland temporary jobs are more frequently used in the public sector, i.e. in municipals and in state jobs. It also seems to be the case that it is more difficult for involuntary temporary workers to converse temporary job into permanent one in the public sector. The results suggest that the hazard is considerably lower for those involuntary temporary workers in state jobs compared to the private sector. A non-negligible share of temporary workers also work in part-time jobs at the same time. In the estimations we also controlled for the impact of working part-time for the transitions into permanent job, but we did not find a significant role for the part-time nature of the involuntary temporary job. Time fixed-effects imply a slight increasing trend in the proportion of temporary jobs leading to permanent jobs. Looking at the corresponding results for other temporary workers we can detect some similarities, but also interesting differences as regards the impact of several individual characteristics (table 5). Similarly, the stepping stone impact is lower for women working in other temporary jobs as well. Women working in other temporary jobs had around 20 percent lower hazard of moving from other 15

16 temporary job to a permanent job. An interesting result is that we do not find any significant differences in transition rates to permanent jobs by age groups. The impact of tertiary level education on the conversion of temporary job to permanent job is noticeably larger from other temporary job compared to involuntary temporary job. By sector, other temporary workers in state jobs have a half as small hazard of moving to permanent job compared to private sector. We no longer find a statistically significant difference between municipal sector and private sector. As for time fixed-effects we detect a slightly decreasing trend in the conversion rates. Transitions to unemployment and outside labour force can be regarded as transitions to nonwanted direction and be interpreted as giving evidence of the segmentation hypothesis. The results related to transitions to unemployment from involuntary temporary job show that women s hazard of ending up in unemployment from involuntary temporary job is 20 percent smaller compared to men when the impact of other factors is controlled for. By age group older involuntary temporary workers have around 33 percent higher risk compared to year-old workers whereas the youngest age groups face a lower hazard compared to this age group. Again high education seems to protect involuntary temporary workers from unemployment compared to the workers with low education the hazard being around 40 percent smaller. Our results related to transitions to nonwanted direction are in accordance with the results by D Addio and Rosholm (2004) who found that when temporary jobs are held by vulnerable individuals they do nit improve their career paths and may lead to instability and possibly exclusion. We find an even larger difference between women and men in transitions to unemployment from other temporary work. In this case women have a 40 percent larger probability of ending up in 16

17 unemployment from other temporary work compared to men. Younger age groups have a statistically significantly smaller hazard of moving to unemployment also from other temporary work whereas we do not find statistically significant differences by educational level or sector. The probability of moving outside labour force is distinctly larger for women than for men from both involuntary and voluntary temporary work. One explanation for this might be that for women working in temporary jobs and facing a threat of unemployment moving outside labour force and taking care of children at home is more prominent alternative to unemployment than for men (add reference). By age group we find that the youngest age group has 67 percent higher probability of moving outside labour market from involuntary work compared to year-olds. As expected the highly educated are less likely to move outside labour force from involuntary temporary jobs. compared to those involuntary temporary workers with low education. Hazards for public sector temporary workers of moving outside labour force are also lower compared to private sector workers. As regards the corresponding results for other temporary workers we find that year-olds have over three times higher hazard of moving outside labour force compared to the comparison age group. Our data inspections (not presented in this paper) suggest that the main reason for these young temporary workers to move outside labour force is that they begin to study. Similarly as in the case of involuntary temporary workers the highest educated have a clearly lower hazard of moving outside labour force. The results taking into account unobserved heterogeneity did not differ much from the ones presented here and are not reported separately in this version of the paper. Sensitivity analyses (to be added here) 17

18 5. Conclusions In this paper our focus was to investigate to what extent involuntary temporary work acts as a stepping stone or a dead end for a labour market career, whether labour market transitions are different between involuntary and other temporary workers, and how much differences we find among different groups of involuntary workers and other temporary workers. Our results suggest that the transition patterns differ between involuntary and other temporary workers to some extent. According to our descriptive analysis involuntary temporary workers are more trapped in temporary jobs. They also move more often to unemployment from temporary jobs. In turn, other temporary workers move more frequently outside labour force. Our preliminary results suggest that involuntary temporary work is different experience, i.e. has different impacts on labour market transitions, for different groups of workers. Involuntary temporary work seems to work better as a bridge to permanent employment for men than for women, for young workers compared to older ones, and for highly educated compared to those with low education. It is also more difficult to convert an involuntary temporary job to a permanent one in the public sector in Finland. The results also imply that for older workers and workers with low education involuntary temporary jobs are more dead ends by nature, as for these groups temporary jobs act less as a stepping stone and lead more often to a negative career development in the labour market, such as transitions to unemployment and outside labour force. From policy perspective it would be important to pay special attention to these groups of involuntary temporary workers for whom temporary jobs seem to be dead ends. One policy measure would be to provide equal opportunities for skills development for all workers irrespective of the job type. 18

19 As regards transitions from other temporary work our results suggest that also other temporary work seems to work better as a bridge to permanent employment for men and the highly educated workers, but by age group statistically significant differences are not detected. Again young workers and women have a higher hazard of moving from other temporary work outside labour force. References Booth, A.L., Francesconi, M., Frank, J., (2002), Temporary Jobs: Stepping Stones or Dead Ends? Economic Journal, vol. 112(4809, F189 F213. D Addio, A. and Rosholm, M. (2004), Exits from temporary jobs in Europe: A competing risks analysis, Labour. Dekker, R. (2001), A Phase They are Going Through: Transitions from Non-regular to Regular Jobs in Germany, the Netherlands and Great Britain, Tilburg University, mimeo. Gagliarducci, S. (2005), The Dynamics of Repeated Temporary Jobs. Labour Economics, vol. 12(4), Guell, M. and Petrongolo, B (2007), How Binding Are Legal Limits? Transitions from Temporary to Permanent Work in Spain. Labour Economics 14, Hernanz, V., Origo, F., Samek, M. and Toharia, l. (2002) Dreaming of a stable job: the transitions of temporary workers in Italy and Spain, a mimeo. Holmlund, B, and Storrie, D. (2002), Temporary Work in Turbulent Times: the Swedish Experience. Economic Journal, vol. 112(480), F245-F269 Kauhanen, M. (2002), Määräaikaiset työsuhteet ja toimeentulon riskit, (Temporary employment contracts and risks of inadequate income). Sosiaali- ja terveysturvan tutkimuksia 69. Helsinki: Kela. Kauhanen, M. and Nätti, J. (2011), Involuntary temporary and part-time work, job quality and wellbeing at work. Labour Institute for Economic Research Discussion Papers 272. Meyer B. (1990), Unemployment Insurance and Unemployment Spells. Econometrica 58: Narendranathan, W., Stewart, M. (1993), Modelling the Probability of Leaving Unemployment: Competing Risks Models with Flexible Base-line Hazards. Applied Statistics 42,

20 OECD (2004), Employment Outlook, OECD: Paris. Appendix. Table 1. Characteristics of involuntary and other temporary workers (at the first interview time) % Involuntary temporary Other temporary Female Age (average) Age years Age years Age years Age 45+ years Married Primary education Secondary education Tertiary education Blue-collar worker Lower white-collar worker Upper white-collar worker Services (all) Private sector Public sector Part-time

21 Table 3. Maximum likelihood estimates (hazard ratios) of the transitions from involuntary temporary work to permanent employment, unemployment and outside labour force to: Variables Permanent employment Female 0.850** (0.059) Age years 1.215* (0.113) Age years (0.070) Age 45+ years 0.785** (0.072) Married (0.068) Secondary education (0.105) Tertiary education 1.245** (0.135) Agriculture (0.258) Construction 1.869*** (0.263) Transportation 1.666*** (0.223) Trade 1.785*** (0.222) Hotels and restaurants 1.489** (0.263) Business activities 1.438*** (0.160) Education 0.631*** (0.082) Health and social work (0.108) Municipal 0.732*** (0.063) State 0.312*** (0.046) Part-time (0.097) Year (0.137) Year (0.135) Year ** (0.151) Year *** (0.270) Baseline hazard variables to be added here Unemployment Outside labour force 0.782*** (0.060) 1.700*** (0.144) 0.628*** 1.672*** (0.075) (0.218) 0.569*** 1.272** (0.057) (0.149) 1.329*** (0.118) (0.149) 0.666*** 1.209** (0.049) (0.102) ** (0.091) (0.082) 0.611*** 0.618*** (0.0684) (0.075) 1.509** 0.196*** (0.296) (0.080) *** (0.136) (0.050) *** (0.175) (0.028) 0.699* 0.096*** (0.131) (0.027) *** (0.237) (0.039) 1.218* 0.153*** (0.150) (0.036) * (0.118) (0.130) *** (0.097) (0.042) 1.182* 0.018*** (0.109) (0.006) *** (0.118) (0.012) (0.1111) (0.106) (0.115) 0.365*** (0.079) (0.1111) (0.106) (0.115) 0.365*** (0.079) 21

22 N of obs Log likelihood Note: Standard errors are in parenthesis.***: difference significant at 1 % level, **: difference significant at 5% level,*: difference significant at 10 % level. Table 4. Maximum likelihood estimates (hazard ratios) of the transitions from other temporary work (hazard ratios) to permanent employment, unemployment and outside labour force to: Permanent employment Unemployment Outside labour force Variables Female 0.831* (0.096) Age years (0.146) Age years (0.140) Age 45+ years (0.157) Married (0.127) Secondary education (0.163) Tertiary education 1.490** (0.262) Agriculture (0.317) Construction (0.313) Transportation 1.679** (0.404) Trade 1.649*** (0.313) Hotels and restaurants (0.486) Business activities 1.653*** (0.283) Education (0.177) Health and social work (0.217) Municipal (0.151) State 0.399*** (0.102) Part-time (0.162) Year (0.161) Year (0.185) 1.449** (0.260) 1.454*** (0.148) 0.640* 3.422*** (0.160) (0.803) 0.668* 1.551* (0.165) (0.377) *** (0.237) (0.505) 0.460*** (0.104) (0.173) * (0.241) (0.095) *** (0.185) (0.083) *** (0.424) (0.091) 1.897* 0.274*** (0.614) (0.081) *** (0.268) (0.049) 1.646* 0.102*** (0.440) (0.033) 1.897* 0.118*** (0.703 (0.053) *** (0.313) (0.043) * (0.393) (0.094) *** (0.309) (0.100) *** (0.289) (0.025) (0.276) (0.226) (0.237) (0.132) (0.144) 22

23 Year 2007 Year (0.170) (0.270) (0.161) 0.451** (0.172) (0.149) 0.265** (0.074) Baseline hazard variables to be added here N of obs Log likelihood Note: Standard errors are in parenthesis.***: difference significant at 1 % level, **: difference significant at 5% level,*: difference significant at 10 % level. 23