Successful return to work during labor market liberalization: the case of Italian injured workers

Galizzi, Monica; Leombruni, Roberto; Pacelli, Lia Article Successful return to work during labor market liberalization: the case of Italian injured workers Journal for Labour Market Research Provided in Cooperation with: Institute for Employment Research (IAB) Suggested Citation: Galizzi, Monica; Leombruni, Roberto; Pacelli, Lia (2019) : Successful return to work during labor market liberalization: the case of Italian injured workers, Journal for Labour Market Research, ISSN 2510-5027, Springer, Heidelberg, Vol. 53, Iss. 9, pp. 1-24, https://doi.org/10.1186/s12651-019-0260-5 This Version is available at: https://hdl.handle.net/10419/216684 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Galizzietal. J Labour Market Res (2019) 53:9 https://doi.org/10.1186/s12651-019-0260-5 ORIGINAL ARTICLE Successful return towork duringlabor market liberalization: thecase ofItalian injured workers Monica Galizzi1* , Roberto Leombruni2,3 and Lia Pacelli2,3 Abstract We investigate the long term employment outcomes of Italian injured workers over a time period when the country introduced policy reforms that increased labor market flexibility but reduced job security. Using an employeremployee database matched with injury data, we observe that both before and after the reforms almost one-fourth of injured workers were no longer employed 3 years after their “first” return to work. We note a slight decrease in this share after the reforms (from 24 to 22%) while we find a decline in workers’ job security as measured by their probability of re-employment in permanent contracts. We use multinomial logit estimates to study how liberalization reforms were associated with a changing role of individual, firm, and injury characteristics in shaping long-term employment outcomes of injured workers after their recovery period. Heterogeneity analyses show that low wage employees, women, immigrants, and individuals who suffered a more severe injury were penalized more. Pre-injury individual characteristics became stronger predictors of long-term employment than firms’ characteristics. In particular, we find that the advantage provided by working in larger firms was significant before the liberalization reforms, but disappeared afterward, while the advantage provided by human capital became more relevant after the liberalization. Keywords: Occupational injuries, Return to work, Maximum medical improvement, Deregulation, Multinomial logit, Matched employer-employee data, Italy JEL Classification: J08, J28, J6 © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. 1 Introduction In Italy the cost of occupational injuries in 2007 was estimated to be equal to 2.63% of the national GDP. Onethird was attributable to safety investments and actions initiated by firms to prevent injuries. The remaining 27 billion euros were attributable to medical or indemnity costs incurred by national agencies, to production and adjustment costs sustained by employers, and to productivity losses or legal expenses faced by workers and their families (INAIL 2011). Across countries, all these expenses are known to be function of injury severity, and of the length of the time to recover and to return to work (RTW). Hence the extensive research to identify best practices that can speed the RTW process and the implementation of policies to facilitate it and enhance injured workers’ employment (Clayton etal. 2012; Barr etal. 2010). However, the meaning of successful RTW may vary across stakeholders (Young et al. 2005b; Leyshon and Shaw 2012). While for the insurance agency it may signify the end of disability payments, for employers it may indicate the time when productivity is fully restored. For workers it is likely to mean not only reentry to work but also ability to keep the job, the pre-injury wage, and to further advance in career (Young etal. 2005a). In this context, the seminal study by Butler etal. (1995) highlighted that the research focus on workers’ first RTW can produce a very misleading picture. It does not capture the real post-injury employment dynamics, as many injured workers fail to return to stable employment after their day of maximum medical improvement, and instead drop Open Access Journal for Labour Market Research *Correspondence: [email protected] 1 Department of Economics, University of Massachusetts Lowell, Lowell, MA 01854-2881, USA Full list of author information is available at the end of the article Page 2 of 24 Galizzietal. J Labour Market Res (2019) 53:9 off the labor force, or experience unemployment or new disability spells. Given such emerged research focus on injuries’ longer term employment outcomes, it is surprising that the literature has neglected the role played by existing labor market regulations. A handful of studies has examined whether differences in national disabilities policies (benefit generosity and eligibility requirements) contributes to different degrees of sustainable RTW (Anema etal. 2009; Barr et al. 2010; Collie et al. 2016; Burkhauser etal. 2016). However, to the best of our knowledge, to date no economic analysis has explored whether longer term employment patterns can also be associated with changes in national labor market regulations such as the ones introduced in several European countries since the 1990s to increase labor market flexibility. Our objective is to cast some light on such neglected topic. Liberalization reforms were introduced through interventions “at the margin” that facilitated and induced the creation of new more flexible contracts. However, the reforms had wider effects that changed the entire labor market functioning. They had a detrimental effect on workers’ bargaining power (Ciminelli etal. 2018) and played a role in exacerbating inequalities in the labor market (Barbieri and Cutuli 2016). They triggered demand side responses in terms of job openings and personnel recruitment which changed the context and the perspectives also of incumbent, permanent workers. Given such background, we want to investigate how such changed institutional environment framed the long term employment experience of injured workers. We research this in an institutional setting that differs greatly from the North American labor market that has been the object of most existing economic studies about injured workers’ RTW. In fact, in Italy injured workers are guaranteed “de facto” full wage compensation while off work, and this scenario changes the incentives to potentially speed the return to employment (Galizzi etal. 2016). We exploit a large administrative database on work and injury histories in Italy to describe how long term employment outcomes (up to 36months after the end of the recovery period) of injured workers evolved over 10 years when important labor market liberalization reforms were implemented (in 1998–2001). Such reforms aimed at introducing greater flexibility for employers and more employment opportunities for workers, but they were implemented in a heterogeneous way, affecting particularly those individuals who were already less protected before the reforms and bringing about or worsening a situation of great job insecurity for many sections of the Italian workforce (Berton etal. 2012; Barbieri and Scherer 2009). We examine whether such heterogeneous consequences were observable also among injured workers’ long term employment patterns, in particular in terms of differences in ability to maintain their pre-injury employment and contractual status. We investigate whether different outcomes characterized injured workers whose pre injury employment characteristics (human capital accumulation, firm’s size) were more likely to offer “protection” from the effects of the liberalization reforms. We also explore whether employment outcomes differed for the traditionally most vulnerable groups of workers (temporary workers, women and immigrants). To be more specific, we estimate the probability of different employment outcomes conditional on a rich set of controls, and compute the predicted probabilities of each outcome in years before, during and after the reform period. Since the policy change did not target a particular segment of the labor market, all workers saw their long term employment outcomes potentially affected by the reforms. Hence, we cannot identify a control group to assess the causal impact of the reforms on the levels of the estimated probabilities for workers in specific contractual groups.1 Then, our identification strategy relies on comparing the differences in the employment probabilities driven by several individual, firm, and injury characteristics, and studying whether the reforms were associated with a change in these differences across all workers. What we observe is that there were many probabilities trends that were in action before the reforms. Such trends were similar but separated (parallel) across the observed characteristics. However, we find meaningful departures from such parallel trends after the reforms. In particular, we find that the advantage provided by higher employment protection (EPL)-measured by firms’ size—in securing a successful outcome was significant before the liberalization reforms but is nonexistent afterward; while the advantage provided by human capital accumulation (HC)- measured by wages—increases in magnitude after the liberalization reforms. Our study is organized as follow: in Sect.2 we discuss the related literature, our research hypotheses about the “protective role” of some firms and individual attributes, and their testable implications; in Sect.3 we describe the Italian relevant institutional setting and its evolution over time with specific attention to labor market deregulation. Section4 addresses data, sample selection and methodological issues. Section5 presents our estimated results, robustness checks and comparisons with not-injured 1 At the same time, policy changes took place over several years and were concurrent with an economic slowdown (1999–2000), and with reforms in both the workers’ compensation and unemployment insurance systems (see section below). This precludes us from using also a before-after strategy (exploiting, for example, interrupted time series) to assess a clear-cut causal effect of the reforms on the probabilities of different employment outcomes. Page 3 of 24 Galizzietal. J Labour Market Res (2019) 53:9 workers. Finally, Sect. 6 includes a discussion of our results and our conclusions. 2 Related literature andtheoretical background Our aim is to describe how the long term RTW patterns of injured workers evolved over years characterized by labor market liberalization. Therefore, we build on two quite separate strands of economics literature, i.e. the one that has analyzed the conditions favoring a successful RTW and the one that has focused on the effects of labor market deregulation on the overall functioning of the labor market. 2.1 About RTW The RTW literature has shown that what seems a relatively simple chain of events (some workers get injured on the job, take time off work to heal, and return to work) disguises several complexities. Workers, firms, insurance agencies, government agencies, they all may face aligning or conflicting incentives in facilitating this process (Boden and Galizzi 2017). Workers’ demographics, firms’ attributes, injury characteristics and workers’ compensation rules about eligibility, generosity, and length of disability payment will also affect it (Anema etal. 2009; Barr etal. 2010; Collie etal. 2016). Regulations and norms about disabilities accommodation may play also a large role (Anema etal. 2009; Clayton etal. 2012; Gailey and Seabury 2010). Hence, the rich literature that over the last 30years has studied the different factors that may facilitate workers’ return to productive employment (Krause etal. 2001; Cullen 2018). However, workers’ RTW may not represent the end of the chain of events ignited by the incident. A first RTW may be followed by additional spells of employment or changes in employers, or by new occupational injuries, or lead to labor force separation (Butler etal. 1995; Krause etal. 2001; Bültmann etal. 2007; Côté etal. 2008; Vogel etal. 2011; Berecki-Gisolf etal. 2012; Galizzi 2013; Biering etal. 2013; Young 2014). All these additional developments are then one more time affected by the severity and degree of full recovery from the injury (Côté etal. 2008); by workers’ pre injury characteristics (Galizzi 2013); by firms’ ability to provide accommodation(Høgelund and Holm 2014), and potential retaliation against the injured worker (Strunin and Boden 2004). Given the focus of our research about the potential role played by EPL and HC, it is also important to note that these additional longer term employment outcomes can be affected by workers’ attachment to the job, earnings, and job status (Awang et al. 2016; Galizzi etal. 2016; Seing etal. 2015). For example, García- Serrano etal. (2010) show that highly flexible workers return to work earlier than others, ceteris paribus, possibly jeopardizing their health and future employability. Overall cyclical economic conditions and tightness of the labor market also play a role (Institute for Work & Health 2009). Therefore, it is quite plausible that after a first RTW, a worker’s rights and opportunities to remain employed and to have access to good jobs will be associated with the more general labor laws and regulations that characterize each country, and, within each country, different sectors or firms. 2.2 About deregulation Recent studies have shown that labor market deregulations introduced in several countries since the early 1990s to increase productivity and employment has reduced the share of labor income at the macroeconomic level (Ciminelli etal. 2018) and increased inequalities (Barbieri and Cutuli 2016). The literature also highlights that, when reducing employment protection legislation (EPL), workers’ mobility increases while the effect on unemployment is ambiguous (Bertola 1990). Giannelli et al. (2012) estimate that in Italy the duration of the first job spell of individuals entering the labor market decreased after the deregulation reforms (they study the period 1990–2000) and this effect was not counterbalanced by a higher probability of moving quickly to a new employer. Furthermore, they observe that “the share of workers with only one job spell within 3years decreases, while the share of those with three or more spells increases”. The port of entry effect—where temporary jobs lead to permanent employment-continued after the reforms but became less noticeable, as less than half of workers could move to a permanent contract after a series of temporary spells (Berton etal. 2011). Temporary jobs may end just substituting permanent ones (Kahn 2010). In the context of occupational injuries, temporary jobs and precarious employment have been found to be associated to higher occurrence of injuries (Amuedo-Dorantes 2002; Bender etal. 2012; Giraudo etal. 2016; Koranyi etal. 2018) and more severe injuries (Picchio and Van Ours 2017) in several countries, including Italy. Because the liberalization reforms increased the prevalence of job insecurity, it is reasonable to expect an overall association with changes in employment outcomes of occupational injuries. However, to the best of our knowledge, there are no studies that describe variations of injured workers’ subsequent employment and patterns of successful “first” RTW in the context of new labor market regulation. 2.3 Our hypotheses Our study contributes to such body of literature to test six different hypotheses among Italian injured workers. First, even in an institutional setting where injured workers enjoy more protection compared to the one examined in North American studies, a “first” RTW is Page 4 of 24 Galizzietal. J Labour Market Res (2019) 53:9 not a stable outcome and workers’ and firms’ characteristics affect such outcome: Hp0: Occupational injuries have a substantial limiting effect on long term employment also for workers who “first” RTW Hp1: The physical effect of injuries is a very important determinant of long term employment outcomes but workers’ and firms’ characteristics that are potentially associated with a higher likelihood of workplace accommodations play also a very large role Furthermore, as described in detail in the section below, labor market reforms introduced in Italy in the late 1990s affected firms differently depending on specific firms’ attributes such as size and unionization. Therefore, Hp2: The probability of keeping the pre-injury job after a first RTW was reduced for workers who were employed in firms where EPL decreased due to labor law reforms Second, we know that a worker’ high wage is likely to indicate her level of human capital, as well as productivity, effort, dedication to the job, and overall value to the firm’s production process. Therefore, higher wages might also imply a higher probability or staying in the same firm in the medium-long run after RTW, despite deregulation of the labor market that would allow the firm to dispose of the worker more easily. Therefore, we test the following hypothesis: Hp3: Compared to low wage workers, high wage workers were more likely to secure their pre-injury employment relationship in the medium-long run even if EPL was reduced On the other hand, workers who had suffered more serious injuries may be less employable, e.g. because of functional limitations due to the accident. It is an open empirical question whether deregulation of the labor market was associated with improvement or worsening of their condition. Hence we test: Hp4: The reduction in EPL was correlated with worst long-term employment outcomes for those employees who had suffered more serious injuries Finally, we know that women and immigrants are workers who compose the weaker segments of the labor market, e.g. they earn lower wages and are more often hired with more precarious contracts (Venturini and Villosio 2008; Olivetti and Petrongolo 2008). Hence, we test a fourth hypothesis: Hp5: The reduction in EPL was correlated with worst long-term employment outcomes for the more vulnerable employees such as injured women and immigrants To test these hypotheses we study a variety of long term employment outcomes: job-security (still holding the before-injury job), employment security (no more holding the before-injury job but still employed), a more precarious job (movements in and out of temporary/permanent contracts), unemployment, or new job related injuries. We study a 12years period (1994–2005) during which Italy introduced a set of labor market liberalization reforms that we illustrate below. 3 Institutional setting Italian injured workers hired with a permanent contract enjoy full job protection until the end of their recovery period. However, after their first RTW they are at risk of layoffs if no viable accommodation is found, or may quit if they cannot cope with job demands. For temporary workers the outcome is more uncertain if their contract expires before their day of maximum medical improvement. We aim at understanding whether changes in labor market regulations may be associated with modifications of long run employment outcomes across all these workers. In this paper, we focus on EPL reforms, defining a period “before” (1994–1997), a period “during” (1998– 2001) and a period “after” (2002–2005) such reforms. In the following section, we describe these reforms and the institutional settings, as well as some additional changes introduced in those years that affected both Workers’ Compensation rules and welfare provisions for the unemployed. 3.1 Employment protection legislation andliberalization reforms In the initial period that we study (1994–1997), the typical labor market contract for an Italian worker was a “permanent” one, i.e. a contract with no stated termination date and implying significant firing costs for the employer in case of not-consensual termination. Such firing costs included severance payments and compensation in case of unfair dismissal. They increased with firm size in correspondence of two legal thresholds, one at 15 and another one at 60 employees (Cavaletto and Pacelli 2014). In addition, unionization typically increases with firm size, making layoffs more conflictual, lengthier and costlier for larger employers. Until the late 1990s, the Page 5 of 24 Galizzietal. J Labour Market Res (2019) 53:9 only departures from the typical permanent contracts were apprenticeships and on-the-job-training contracts for youth. Temporary contracts were very limited and subject to strict constraints: each firm faced limits in the number of temporary employees it could hire and was required to prove the temporary nature of the occupation. Labor market liberalization reforms were concentrated between 1998 and 2001. Reforms started in 1998 (Law Decree No. 196/1997) and introduced and regulated new types of work (such as temporary agency work and quasidependent work2). They continued in 2000 (Law Decree No. 61/2000) deregulating part-time contracts. Finally, in 2001 we saw the full liberalization of temporary contracts (Law Decree No. 368/2001). Now they could last a few days or up to 36months, became renewable and faced no restrictions on their use. Afterward, only law Decree No. 276/2003 modified the EPL setting but did not introduce substantial novelties. Further and deeper reforms took place after 2008 but they are excluded from our analysis since their effects were largely affected by the macroeconomic recession. We should mention that the public sector did not undergo most of the reforms we discussed here. Instead, it faced a prolonged hiring freeze. Because of this difference, we exclude public employees from the current study. All these were reforms “at the margin”, decreasing EPL for new entrants on the labor market and for job movers, but not for incumbent workers who kept their open ended job. However, while small firms faced very low EPL even before the reforms were introduced, larger firms took advantage of them to decrease the average EPL of their workforce by expanding new hires with flexible contracts, and by enjoying stronger bargaining power (Ciminelli etal. 2018). The effect of the reforms became rapidly visible, as non-permanent contract workers in the private sector increased from 600,000 in the first quarter of 1998 to about 1.2 million in the first quarter of 2006 (Eurostat 2018). 3.2 Workers’ compensation In Italy, a public insurance system provides medical and disability benefits to all employees, both permanent and temporary workers, to all self-employed manual workers and to a part of non-manual self-employed workers. Those excluded from the public system—mostly selfemployed workers in the trade sector—have to resort to the private insurance sector (but they do not have the obligation to do so).3 The public system is managed by the National Workers’ Compensation Agency (INAIL), and is financed by firms through premiums that are proportional to payroll, increase with jobs risk, and are adjusted through experience rating. Workers who get injured are entitled to a recovery period, the length of which is established by a doctor who is certified to work for INAIL. They receive paid medical care directly or indirectly provided by INAIL, and disability benefits ranging from 60 to 75% of their earnings. However, a topup granted by employers according to collective agreements allows injured employees to earn a de-facto full wage replacement during their absence from work (Galizzi etal. 2016). The worker compensation system was the object of only minor adjustments in the last decades. The only relevant reform was delivered in 2000 (DLgs 38/2000). It introduced coverage also for incidents occurring on the way to or from work,4 a compensation for the so called “biological damage” (Rossi 2002), and modified injury severity thresholds that entitle injured workers to permanent disability (PPD) benefits. PPD is measured on a 0–100% scale and is set by the law in a very detailed way. Before the 2000 reform, workers with a PPD up to 10% (e.g. a fracture of the atlas without persistent neurological symptoms) did not receive any compensation after the completion of the healing period, while over the 10% threshold an annuity was paid compensating for long term income losses. The 2000 reform increased the generosity of the workers’ compensation PPD benefits: the threshold without compensation was lowered to 5% (e.g. a detectable scar, not visible on the face or neck); from 6 to 15% a lump sum compensating not the income losses but the biological damage is now paid according to the severity, the gender and the age of the worker5; above the 15% threshold the compensation is paid with an annuity, and pays for both the biological- and income loss.6 3.3 Unemployment benefits Up to 2005 the system assisting workers who lost their job was highly segmented. Workers laid off by larger (above 15 employees) manufacturing firms through collective bargaining (involving 5 or more employees) could enjoy generous “mobility benefits”: compensation up to 75% of their wage for up to 4years according to age and area of work. All other unemployed individuals were 2 Formally “self-employment” this is a de facto subordinate employment relationship. It is similar to a free-lance job and might be called also contractwork. 3 All self-employed workers are excluded from our analysis. 4 Such incidents in itinereare excluded from our analysis, as they were unobservable up to 2000. 5 For example, a fracture of the atlas with 10% PPD would grant a lump sum of 21,700 euro—about 1year of average salary—for a man under 20years of age. 6 Such annuity increases with the degree of PPD: from less than 2months of average salary to more than 1year of average salary for a PPD equal to 100%. Page 6 of 24 Galizzietal. J Labour Market Res (2019) 53:9 compensated by a system of unemployment benefits that was poorly endowed, provided only a 30% replacement rate for 6months (40% for up to 9months starting from 2001) and was subject to strict eligibility conditions on past employment7; overall, it was characterized by very low take-up rates and did not reach the minimum standards set by the 1952 ILO Social Security Convention until 2008 (Leombruni etal. 2012). 4 Study population andmethods 4.1 Data We use a database that combines individual employment histories from the Work Histories Italian Panel (WHIP) with injuries records from INAIL, the Italian National Workers’ Compensation Agency. The matched database is a 1:15 random sample of the population (about 1.5 million workers each year) covering the period 1994–2012, generating a unique source of information for the analysis of occupational injuries. WHIP’s reference population includes all Italian workers and pensioners. It excludes only public sector employees hired with an open-ended contract and high skilled professions (e.g., lawyers) who are compensated with different insurance funds. The dependent employment section of WHIP is a matched employer-employee database that includes start and end dates of each employment spell, as well as worker characteristics (age, sex, place of birth), job characteristics (temporary vs. permanent contract, full-time vs. part-time, occupation, location), labor market outcomes (the number of days and weeks worked, earnings and social security payments) and firm characteristics (size, opening and closing date, sector, location, monthly new hires and separations, average wages). INAIL data include a description of all injuries causing permanent or temporary disabilities across the whole country and with time off work longer than 3days. The data records a description of the injury event itself (when, how, where) and its consequences (nature of injury, part of body, length of temporary disability payment, and degree of permanent disability—if any). The INAIL dataset and the WHIP dependent employment section have been matched and this is the dataset on which we base our analysis (see Bena etal. 2012, and Galizzi etal. 2016, for further details).8 4.2 Sample selection For the purpose of this work, we select only employees who had a work incident between 1994 and 2005. As we observe outcomes up to 3years after their recovery period this brings us at the beginning of the economic crisis in 2008. After that year, the Italian macroeconomic environment was deeply affected by the consequences of the financial crisis, and became less comparable to the previous decade. We drop fatal events and injuries occurring on the way to or from work (as they were not compensated and therefore recorded before 2000), as well as those occurring in agriculture (where many are self-employed), and in education, health and personal services (which are mainly public sectors for which we do not have corresponding WHIP data). As in most studies concerning injured workers, a serious concern is the issue of underreporting of injuries (Boden and Ozonoff 2008; Picchio and Van Ours 2017), mainly from small establishments (Wuellner etal. 2016; Oleinick etal. 1995). Indeed, smaller firms have higher ability to under-report less severe injuries because governments’ health and safety controls are less frequently implemented among small firms. Furthermore, the additional insurance costs generated by an incident can be more significant for a small business and induce a larger incentive to underreport (evidence on Italy is provided in Galizzi etal. 2016). For our study, such concern is particularly relevant because it entails a potential bias in our estimations. As we described above, firm size modifies the degree of EPL, so that the estimated effect of the latter can be blurred by underreporting. Therefore, we exploit the information on the nature of incidents to study only injuries which usually require immediate treatments at a hospital (fractures, anatomic losses and removals of an alien corpus). By implementing this restriction we limit the likelihood of an underreporting bias because hospitals are required to report injuries to INAIL. Our final sample includes about 29,000 incidents for 27,442 workers. The sample composition does not change in a relevant way over the years we study with respect to 8 Our study did not require ethics review and approval. In fact, the data we use has been listed under the Italian National Statistical Program that includes data-collection projects that comply with the national regulation on all issues regarding the use of personal data. The data has been positively evaluated by the Italian authority that is responsible to guarantee and protect privacy. 7 Such conditions were: sector of activity (mostly manufacturing firms), at least 2years of past employment, and at least 1year of paid contributions over the last 2years. As a result, the take up rate was very low. Please notice that in case of acceptance for publication, we will not be able to directly provide the dataset we used because the data owner is theItalian Ministry of Health, and we had access to it according to a strict confidentiality agreement. The data can be accessed by any researcher establishing a specific research agreement with the Italian Ministry of Health, however. We are obviously ready to provide information to guide other researchers through the procedures for accessing the data on injuries. Please also notice that all the results and descriptive statistics only related to the labor market—e.g. work contracts or wages—can be easily replicated with publicly accessible datasets also based on INPS archives (e.g. LOSAI, distributed by the Italian Ministry of Welfare). The copies of the computer programs used to generate the results presented in the paper are available from the authors. Footnote 8 (continued) Page 7 of 24 Galizzietal. J Labour Market Res (2019) 53:9 the part of body affected and to the nature of injury, but for a 4 p.p. increase in the prevalence of fractures, from 48.9% before the reforms to 53% after them (see Table2). 4.3 Measures ofRTW outcomes The outcome we analyze is the work status of the person n months (up to 36months) after what we call their “first” RTW.9 We face no right censoring, as the database covers the period up to 2012. The outcomes we consider are: non-work, employed in the pre-injury firm, employed in a different firm with a permanent contract, employed in a different firm with a temporary contract,10 on leave because of a new injury. Unconditional probabilities (Table1) show that the most likely outcome was maintaining the pre-injury job. However, such probability was far below 100% across all three periods and decreases as time since recovery passes. Non-work was the second most likely outcome (Hp0). Regardless of time period, it affected almost one-fourth of injured workers and increased over time. 4.4 Statistical analysis To test Hp1 we estimate a standard multinomial logit model of the probability of different employment outcomes at different intervals (12 and 36months) after the “first” RTW, where yi takes values j = 0 for non-work, j = 1 if employed in the pre-injury firm, j = 2 if employed in a different firm with a permanent contract, j = 3 if employed in a different firm with a temporary contract, j = 4 if on leave because of a new incident (hence K = 5): (1) Prob yi=j=e b j X i 1+  K−1 k=0 ebkXi Table 1 Unconditional probabilities ofpostinjury long term employment outcomes Among those who were hired pre-injury with a permanent contract, only 1.7, 1.4 and 0.8% (in the three periods respectively) became temporary contract workers in the same firm—out of all possible outcomes. Among those who were hired pre-injury with a temporary contract in the same three periods, 18.2%, 15.2% and 12.6% became permanent contract workers in the same firm—out of all possible outcomes 1994–1997 Outcome aftern months sinceRTW 3months 6months 12months 36months Non work 15.0 14.7 16.9 25.4 Employed in the same firm 79.9 76.9 69.2 48.5 Employed in a different firm-permanent contract 3.7 6.2 10.6 19.5 Employed in a different firm-temporary contract 1.0 1.6 2.6 5.9 On leave due to a new_injury 0.4 0.6 0.8 0.7 Total 100.0 100.0 100.0 100.0 1998–2001 3months 6months 12 months 36months Non work 14.9 13.8 15.1 24.1 Employed in the same firm 78.1 75.1 67.2 46.8 Employed in a different firm-permanent contract 4.3 7.2 11.3 19.9 employed in a Different firm-temporary contract 2.2 3.4 5.7 8.5 On leave due to a new_injury 0.5 0.6 0.8 0.8 Total 100.0 100.0 100.0 100.0 2002–2005 3months 6months 12months 36 months Non work 15.4 14.1 15.2 23.0 Employed in the same firm 77.9 74.9 67.7 48.6 Employed in a different firm-permanent contract 3.6 5.7 9.5 17.0 Employed in a different firm-temporary contract 2.8 4.6 6.9 11.2 On leave due to a new_injury 0.3 0.8 0.7 0.2 Total 100.0 100.0 100.0 100.0 9 This is what is known as the day of “Maximum Medical Improvement” and is often used as a RTW measure in the literature. As we mentioned before,in the case of all Italian permanent contract this day coincides with the day of RTW because employees return to their previous job by law. However, Italian temporary contract workers can be unemployed upon return if their contract expires before the end of their healing period. In this latter case our data measure the employment outcome n months after the recovery period. We cannot identify these cases. This is unfortunate because injured workers in temporary contracts were a small portion of our injured population, but a portion that doubled (form 8 to 16%) over the 10years we study. 10 Fixed term, free-lance, training or apprentice contracts. Page 8 of 24 Galizzietal. J Labour Market Res (2019) 53:9 We then examine how probabilities changed during and after the labor market reforms for workers characterized by different pre-injury EPL and HC, i.e. we interact the covariates of interest with the three reform periods. We define a set of dummies D = {D1,D2,D3} signaling respectively periods “before” (1994–1997), “during” (1998– 2001) and “after” (2002–2005) market liberalization reforms. Not all covariates X are actually interacted with D(although most of them are, as explained in the empirical section); hence the estimated regressors in Eq.(1) can be further decomposed as bjXi=gjZi∗D+hjWi . Given our research hypotheses the covariates of main interest are the following. (i) A measure of EPL (log of the number of employees in the firm11) as firm size is highly related to firing costs and union protection to test Hp2. (ii) A measure of human capital (log of 1-year-lagged real daily wage) capturing individual productivity, status and labor market value to test Hp3. (iii) A measure of injury severity (log of days off work) to investigate whether more seriously injured employees experienced different employment outcomes before and after the reforms to test Hp4. Table 2 Summary statistics As explained in the text, we study only injuries which require immediate treatments at a hospital (fractures, anatomic losses and removals of an alien corpus) Period Statistic 1994–1997 1998–2001 2002–2005 Lagged real daily wage Mean 62.89 62.3 61.55 Wage above the mean wage by occupation in the firm Share 0.27 0.25 0.25 Past intermittent spells Mean 0.42 0.51 0.5 Number of employees in the firm Median 25.08 23.33 22.33 Growing firms Share 0.39 0.44 0.43 Shrinking firms Share 0.25 0.24 0.23 Firm age Mean 15.86 16.18 16.58 Excess turnover Mean 0.31 0.33 0.32 Past illness rate Mean 0.18 0.18 0.19 Years of experience plus firm tenure Mean 7.17 8.19 9.1 Female Share 0.08 0.09 0.09 Temporary contract Share 0.08 0.14 0.16 Part time Share 0.02 0.03 0.05 Migrants Share 0.1 0.17 0.26 Manual occupation Share 0.94 0.93 0.93 Worker’s age Mean 36.7 36.93 37.96 Unemployment rate Mean 7.65 6.5 5.44 North-west Share 0.31 0.33 0.33 North east Share 0.28 0.29 0.28 Center Share 0.17 0.18 0.17 South Share 0.24 0.21 0.22 Manufacturing Share 57.36 53.77 48.5 Construction Share 21.8 20.09 24.14 Services Share 20.85 26.14 27.36 Fracture Share 48.9 52.95 52.99 Anatomic loss Share 8.58 7.87 8.65 Alien corpus Share 42.53 39.18 38.36 Trunk Share 0.3 0.3 0.3 Head Share 25.1 32.5 33.1 Back Share 8.1 9.5 7.7 Lower extremities Share 19.2 19.9 21.2 Upper extremities Share 37.3 37.8 37.9 Days off work after the injury Mean 57.00 62.21 65.76 11 Head-count of individuals on the firm payroll, averaged over the calendar year. Page 15 of 24 Galizzietal. J Labour Market Res (2019) 53:9 pre-injury job were not different by wage quartile before the reforms, but afterward they are different (although differences are not strongly significant). This holds after 3years as well (Additional file1: Figure S7) and is consistent with our Hp3. Therefore, our first conclusion is that reforms21 were associated with changes in successful long term employment outcomes. Such outcomes became less linked with national employment protection rules, potential union protection, and firm’s accommodation (linked to firm size),to become more a function of individual’s human capital (as measured by wage). Such results confirm our first two hypotheses. Notice that—as discussed above— our estimates (see as an example Table4) control for (1) the individual propensity to move (number of pre-injury employment spells, interacted with periods), (2) total individual labor market experience (pre-injury days of employment, interacted with periods), and(3) firm’s propensity not to invest in long-term relationships with the workforce (excess firm turnover interacted with periods). All these regressors have the expected sign—(1) and (3) decrease the probability to stay, (2) increases it—although the estimated values tend to increase during and after the reform periods. 5.3 Severity oftheinjury When we focus on the role played by the severity of injury (captured bythe length of the spell off work), we find that more severe injuries were associated with lower conditional probability of keeping the pre injury job and higher conditional probability of non-work after 1year (Fig.5: “Conditional probability of outcomes after one year since RTW, by quartiles of number of days off after injury (9, 40, 80days)”). Surprisingly, none of the patters we study are correlated with the reform as lines remain pretty much parallel. The same outcome arises after 3years (not reported). It is important to recall that our analysis controls also for other measures of general health (age, the annual frequency of pre-injury sick leaves, nature of the injury, and degree of potential PPD caused by the injury). All these regressors were also interacted with periods and we did not find significant interactions. .7 .72.74.76.78 1 2 3 period Same Firm .11.12.13.14.15.16 1 2 3 period No Work .07.08.09.1 .11.12 1 2 3 period Other Permanent .01.02.03.04.05 1 2 3 period Other Temporary Fig. 4 Conditional probability of outcomes after 1 year since RTW, permanent contract workers only, by (real daily) wage quartiles (50, 60, 70 euro). 90% confidence intervals. Wage 50 euro: solid lines; Wage 60 euro: dash lines; Wage 70 euro: dotted lines 21 It is important to notice that, as discussed above a 2000 reform changed the compensation rules for PPD cases. The increased generosity of the system after 2000 may have increased the reservation wage and modified the labour supply of PPD workers, possibly reducing their labour force participation after the injury. As for the demand side, both before and after the 2000 reform employers remain entitled to legitimately lay off PPD workers only if they can prove that no accommodation can be provided within the firm. In general firms do not bear the immediate monetary costs of their PPD employees’ compensation, so we do not expect a distortionary effect of 2000 law on our results. However, to check our results against the 2000 change in the compensation system, we performed a robustness check of our main findings excluding PPD workers from the analysis (Additional file1: FiguresS2, S3). Results did not change. We thank an anonymous referee for suggesting this check. Page 16 of 24 Galizzietal. J Labour Market Res (2019) 53:9 Hence, our findings describe that the likelihood of different long-term employment outcomes by workers’ injury severity was not associated with the reforms. Workers who were more vulnerable because of their worst injury experience did not experience a change in their probability of successful RTW over the reforms period we study. This contradicts our Hp. 4. However, it is important to notice that one more time, we observe a striking increase in the overall likelihood of landing on a new temporary job over the reforms period. 5.4 Women andmigrants To test our fourth hypothesis, we analyze our main results separately in the case of women and immigrants only. It is important to recall that the share of immigrants in our sample increased from 10 to 17% and to 26% over the three periods while the share of women remained stable around 9%. Among women, temporary contracts increased more than in the whole sample,22 from 8%, to 19%, to 20% in the three periods considered; the same holds for immigrants, among which they increased from 8% to 16% to 19%. Notice however that the smaller sample size decreases the precision of the estimates with respect to the whole sample. We focus on the conditional probability of keeping the pre-injury job after 1year, for the sake of clarity and brevity. In Fig.6 (“Conditional probability of staying in the same firm at one year, by firm size (10, 100, 1000 employees) and by (real daily) wage quartiles (50, 60, 70 euro)”) we see that for both women and immigrants (regardless of contract type) the probability of remaining in the same firm flips after the reforms and becomes negatively related to firm size, showing a worsening of their long term employment outcomes during the reforms period compared to the whole population of injured workers (compare with Fig.1). For immigrants, this process started as soon as the reforms rolled in (period 2); for women, as the reforms were completed (period 3). Both women and immigrants show the same pattern of the full sample with respect to the protective effect of high wages. If, again, we exclude those working with a temporary contract at the time of the injury (Fig.7: “Conditional probability of staying in the same firm at one year, by firm size (10, 100, 1000 employees) and by (real daily) wage quartiles (50, 60, 70 euro). Permanent contract workers only”), we see that—as in the whole sample—the effect related to wages is unchanged. However, the protective effect of firm size for permanent contract workers tends .66.68.7 .72.74.76 1 2 3 period Same Firm .1 .12.14.16.18 1 2 3 period No Work .08.09.1 .11.12.13 12 3 period Other Permanent .02.03.04.05.06.07 1 2 3 period Other Temporary Fig. 5 Conditional probability of outcomes after 1 year since RTW, by quartiles of number of days of off after injury (9, 40, 80 days). 90% confidence intervals. Days off work 9: solid lines; Days off work 40: dash lines; Days off work 80: dotted lines 22 As shown in Table2, in the whole sample the share of temporary contracts increased from 9% to 14% to 16%. Page 17 of 24 Galizzietal. J Labour Market Res (2019) 53:9 .65.7 .75.8 .85 1 2 3 period Women, by firm size .5 .55.6 .65.7 1 2 3 period Not Native, by firm size .65.7 .75.8 1 2 3 period Women, by wage .55.6 .65.7 12 3 period Not Native, by wage Fig. 6 Conditional probability of staying in the same firm at 1 year, by firm size (10, 100, 1000 employees) and by (real daily) wage quartiles (50, 60, 70 euro). 90% confidence intervals. Firm size 10 employees: solid lines; firm size 100 employees: dash lines; firm size 1000 employees: dotted lines. Wage 50 euro: solid lines; Wage 60 euro: dash lines; Wage 70 euro: dotted lines .7 .75.8 .85 1 2 3 period Women perm, by firm size .55.6 .65.7 .75.8 1 2 3 period Not Native perm, by firm size .7 .75.8 .85 12 3 period Women perm, by wage .55.6 .65.7 .75 1 2 3 period Not Native perm, by wage Fig. 7 Conditional probability of staying in the same firm at 1 year, by firm size (10, 100, 1000 employees) and by (real daily) wage quartiles (50, 60, 70 euro). Permanent contract workers only 90% confidence intervals. Firm size 10 employees: solid lines; firm size 100 employees: dash lines; firm size 1000 employees: dotted lines. Wage 50 euro: solid lines; Wage 60 euro: dash lines; Wage 70 euro: dotted lines Page 18 of 24 Galizzietal. J Labour Market Res (2019) 53:9 to fade away, during the reforms for migrants and after the reforms for women. This association may suggest that larger firms took advantage of the liberalization reforms to reduce the share of injured women and migrants on their payroll with a temporary contract and also of those they had previously hired with a permanent contract. This supports our Hp. 5, i.e. worsening employment outcomes for injured women and immigrants during and after the introduction of the reforms.. Finally, with respect to the severity of the incident and the consequent length of the absence, results related to women and immigrants are unchanged with respect to the whole sample (not reported): one more time reforms were found to be unrelated to the subsequent probability of non-work for injuries of different severities. 5.5 Comparison ofinjured vs. non injured workers To better assess the implications of our findings, we should compare the observed employment outcomes for injured workers with the ones of Italian workers who did not experience an occupational incident. However, such direct comparison is not straightforward because we are focusing on a very specific subset of the population of workers (i.e. those more exposed to the risk of a work incident) and we study their employment experience after a quite relevant health shock (i.e. an injury that required immediate care). A counterfactual situation (i.e. workers exposed to a similar risk of a work injury but who did not experience it) cannot be easily found in the general working population, since a measure of occupational risk exposure is not reported in our data. Hence, the two comparison groups (injured vs not injured) are composed of potentially very different individuals according to their observable and unobservable characteristics. Based on what we can observe in our WHIP data,23 the general working population (in not-agricultural private firms) is composed by more women, fewer migrants, fewer manual workers, and higher wage earners with respect to our sample of workers employed in more risky occupations (described in Table2). The general working population is also exposed to the risk of a work incident for a shorter time: in fact, they have higher rates of part-time and less average work experience. All this would imply a quite different career path for our injured workers compared to non-injured employees regardless of the occurrence of a shock like a work incident. Hence, following Fadlon and Nielsen (2017), we look for a more appropriate comparison. Our comparison group is drawn from the same sample of injured workers used in the main analysis, but is selected only among those who will face a work accident later in time. Here we are assuming that they all come from the same population of workers at risk, and that the time of the accident is random. In Fadlon and Nielsen’s words (page 10) we “[…] compare households [workers in our case] with the same expectations over the distribution of future paths, but with different realizations, to isolate the unanticipated component of the shock […] exploiting the potential randomness of the timing of a severe […] health shock within a short period of time.” Those authors proceed to estimate a proper impact evaluation applying a difference in differences strategy. However, the aim of our research is different, because we do not focus on the outcomes per se as they did, but on whether the consequences of the actual occurrence of an incident are associated with changes in the institutional setting of the labor market. In particular, we are interested on how the probability of the different outcomes is affected by covariates like firm size or individual wages and how the effect of such regressors varied over the reform period. Therefore, we replicate the analysis as in the previous figures (e.g. Fig.1) and we compare the displayed patterns over the different reform periods. Finally, we use the estimated conditional probabilities to mimic a difference in differences comparison. To be more specific, we proceed as follows. First, we attribute to not-yet-injured workers a random date of “pretend injury” and a random duration of “pretend leave”, so that the distributions in the true and pretend injured samples are as similar as possible both in terms of time off work and tenure at the time of the incident (Additional file 1: Figures S10, S11). We impose that the “pretend date of injury” and “pretend day of “first” RTW” happen more than 1year before the calendar day of the actual injury, so that we always measure their outcomes before the actual injury occurs. Unfortunately, in this case the sample size for period 3 (after the reforms) becomes quite small, because of the right censoring of work incident data at December 2005; estimates are less reliable for that period, hence we must focus mainly on periods before and during the reforms. In terms of unconditional probabilities we calculate that our “not yet/pretend injured workers” face a much lower probability of no-work with respect to actual injured ones, and a higher probability of keeping their pre-injury job across the first two periods (Table5). This seems to suggest a clear vulnerability of injured workers when the incident actually occurs. Such general finding is confirmed when we replicate our multinomial logit analysis for the “not yet/pretend” injured group and plot again the conditional probabilities of the different outcomes in different periods (Fig.8: ““Not yet injured/pretend” injured workers:Conditional probabilities of staying in the same firm or not working 23 Results available under request. Page 19 of 24 Galizzietal. J Labour Market Res (2019) 53:9 one year after “pretend” RTW, by firm size (10, 100, 1000 employees)—first row—and by (real daily) wage quartiles (50, 60, 70 euro)—second row”; and Additional file1: FiguresS12, S13, S14, S15). We notice that over the reforms period (period 2) the “not yet/pretend” injured workers employed in small firms increase their probability to stay in the same firm one year after the imputed incident.24 This was not the case for the actual injured workers (back in Fig.1). In addition the estimated probabilities of not working 1year after the imputed incident day were much lower for workers in each firms size compared, again, to what was estimated for the real injured ones (Fig.1). “Not yet/pretend” injured workers who earned low, median and high wages also increased their probability to stay in the same firm and decreased their probability of nowork during the reforms. Instead, the probability to stay in the same firm was decreasing among actually injured low wage workers, and it stayed constant for median and high wage ones; among actually injured workers the probability of no-work decreased, but to a lesser extent among low wage workers (see Fig.3). Finally, we use the point estimates and confidence intervals of the conditional probabilities of staying in the same firm, as in Figs.1, 3, 8, and mimic a difference in differences comparison, without any possibility to verify the common trend assumption. However, we believe such assumption should hold because before period 1 both groups are “not yet injured”. In Table6 we note that actually injured workers suffer larger penalties in terms of the likelihood of remaining in the same firm 1year after a “first” RTW, especially if they were employed in small firms and earned lower pre injury wages. This suggests that reforms were associated with a worsening of the conditions of the actually injured workers compared to the not injured ones because their longer term employment prospects became more uncertain. 6 Discussion andconclusions The Italian labor market underwent a sequence of dramatic reforms between 1998 and 2001. Their goal was to increase labor market flexibility and, as a consequence, produce greater employment. New rules facilitated the creation of both more numerous and new types of flexible contracts. However, they ended up affecting the functioning of the whole labor market by eroding the bargaining power of all workers, including the ones hired with permanent contracts (Ciminelli etal. 2018). Our study describes the long term employment experience of injured workers during years when such reforms were introduced. Our data indicate that continuous work with the pre-injury employer remained the most common employment outcome for injured workers 1 year or 3years after a “first” RTW. However, this was true for only around 70% (after 1year) and 50% (after 3years) of injured workers and such percentages remained overall stable before, during, and after reform periods. The second most common outcome, non-work, became a bit less frequent. Before the reform 25% of injured workers were no longer employed 3years after their “first” RTW. After the reform this percentage decreased to 23%. More workers kept working but with a different employer and a temporary contract (from 6% before the reform to 11% after the reforms). We estimate a multinomial logit model that accounts for individual, firm and injury characteristics as determinant of the different employment outcomes we observe. Our results are consistent with what shown by North American studies (Butler etal. 1995) despite the fact that we are studying a system with stronger institutional protections for injured workers. We find that injury severity was a main determinant of long term employment outcomes. At the same time, our measures of human capital Table 5 Comparison of unconditional probabilities of outcomes for “true” injured vs. “not yet/pretend” injured workers1 year after“first” RTW True Pretend-not yetinjured Injury in period 1-before 1994–1997 Non work 16.9 9.0 Employed in the same firm 69.2 75.6 Employed in a different firm-permanent contract 10.6 10.9 Employed in a different firm-temporary contract 2.6 3.7 On leave due to a new_injury 0.8 0.9 Total 100.0 100.0 Injury in period 2-during 1998–2001 Non work 15.1 8.3 Employed in the same firm 67.2 70.3 Employed in a different firm-permanent contract 11.3 11.8 Employed in a different firm-temporary contract 5.7 8.7 On leave due to a new_injury 0.8 1.0 Total 100.0 100.0 Injury in period 3-after 2002–2005 Non work 15.2 7.9 Employed in the same firm 67.7 67.5 Employed in a different firm-permanent contract 9.5 12.1 Employed in a different firm-temporary contract 6.9 12.0 On leave due to a new_injury 0.7 0.5 Total 100.0 100.0 24 Results on only permanent contract workers disappear by firm size and are unchanged by wage, as in the sample of actual injured workers (see Additional file1). Page 20 of 24 Galizzietal. J Labour Market Res (2019) 53:9 accumulation (wages, relative wage status, labor market experience, employment stability) and of potential firms stability and ability to provide accommodations (stable employment, and size) were strong predictors of the most successful long term employment outcomes, i.e. ability to maintain employment with the preinjury employer. In addition, our estimations suggest that labor market liberalization reforms were associated with higher likelihood of maintaining employment after the injury and the “first” RTW. However, this happened at the expense of future job security: injured workers found more easily jobs with new employers, but fewer jobs with permanent contracts. From previous studies we know that reforms weakened the employment protection that Italian workers had historically enjoyed when hired by larger firms, who face higher firing costs, stronger union representation, and likely offer more accommodations to the injured. We investigate whether specific firm and individual attributes offered “protection” from the potential destabilizing effects of the reforms. Contrary to the common expectation about women’s weaker attachment to the labor force, we find that injured women were not more likely than men to stop working after a “first” RTW (as was already observed in Butler etal. 1995). We observe alsothat after the reforms, workers who were injured in larger establishments did not enjoy any longer a higher probability of long-term successful RTW (conditional on wages and the other controls), both in terms of their ability to keep their job with the pre-injury firms and in terms of losing employment (no work). Such loss of overall employment protection associated with firms’ size was driven by the most vulnerable workers: the ones who at the time of the injury were already hired with flexible contracts (males or females), or the ones who were immigrants, even if hired with a permanent contract. This suggests that the traditionally more vulnerable workers were also the ones for whom the disadvantage in post-injury long term employment outcomes were amplified during the reforms. Liberalization gave employers more leeway in terms of decreasing any long-term commitment toward these employees with further implications for social inequality (Gebel and Giesecke 2009; Barbieri and Cutuli 2016.) At the same time, our results confirm the importance of higher pre-injury wages (conditional on firm size and the other controls), in securing successful long-term employment. Previous research had already shown that higher wage workers are the more likely to return to work sooner after an incident (Galizzi et al. 2016). Given the same labor market experience, higher wage may proxy more successful investment in human capital .7 .75.8 .85 1 2 3 period Same Firm .04.06.08.1 1 2 3 period No Work .78.8 .82.84.86 1 2 3 period Same Firm .03.04.05.06.07.08 12 3 period No Work Fig. 8 “Not yet injured/pretend” injured workers: conditional probabilities of staying in the same firm or not working 1 year after “pretend” RTW, by firm size (10, 100, 1000 employees)—first row—and by (real daily) wage quartiles (50, 60, 70 euro)—second row—. 90% confidence intervals. First row—firm size 10 employees: solid lines; Firm size 100 employees: dotted lines; Firm size 1000 employees: dotted lines. Second row—Wage 50 euro: solid lines; Wage 60 euro: dash lines; Wage 70 euro: dotted lines Page 21 of 24 Galizzietal. J Labour Market Res (2019) 53:9 and, therefore, a higher productivity that employers do not want to lose. Or it may signal higher worker’s status within the company, another characteristic that may lead to higher likelihood of receiving accommodation by employers. The findings of our current study suggest that such factors are determinant also of successful employment outcomes over the long run despite the potential higher flexibility introduced by reforms. Our estimates show that only high wage workers kept being more likely to maintain their job with the pre-injury employer, or to avoid non work or a temporary contract if they changed employer one or 3years after their “first” RTW, even if they were women or immigrants. Finally, we find that workers with a higher degree of permanent disability were clearly more likely to stop working overtime. Workers who took long time to recover from the injury were also more likely to separate from the preinjury employer. However, we do not observe anychanges in such trends over the three periods we observe. Clearly, the worst the injury, the bigger the challenges across all the employment outcomes we studied. However, liberalization of the labor market was associated with neither worsened nor improved employment outcomes 1/3years after the day of maximum medical improvement across the different levels of injury severity. We acknowledge that our study has limitations. Because the reforms were introduced gradually and at times of additional institutional changes, we cannot implement a clear before-after policy evaluations design to provide clear evidence of a causal impact of the liberalization. Furthermore, our data is quite rich in terms of recorded workers’, firms’, and injuries’ characteristics, but it does not permit us to compare the experience of injured workers with the ones of similar workers employed, for example, in the same occupation and in the same firms. This limits our ability to establish clearly whether the reforms were associated with different outcomes for injured and non-injured employees. However, the comparison exercise we conducted using “not yet injured workers” as a comparison group suggests that during and after the reforms employment uncertainty may have increased more for injured workers than for the non-injured ones. This is clearly an area that should be further tested with different data sets. Our study represents also only a first step in our effort to describe how injured workers well-being may change over a period when labor market reforms are introduced. We assess our outcomes mainly in terms of job/employment security, but other dimensions of job quality such as wages, job title, and full vs. part time status are also very important dimensions of employment quality, and they deserve further analysis. Although our measures of long term employment outcomes enrich the ones used in previous studies by considering also the degree of job security attached to new employment contracts, we are aware that it remains an imperfect measure of potential success. Individuals differ in their needs, desires and aspirations. Therefore, a RTW on a job that is less secure but pays better or promises immediate or future promotions may be a preferable outcome to certain injured workers (Young 2014). Similarly, to certain individuals a temporary job that comes with more interesting duties, or with flexible time, could be more desirable than a more structured, permanent but less flexible job (Krause Table 6 Difference indifferences measure ofconditional probabilities ofoutcomes (remaining withsame employer) 1 year after“first” RTW Size Period T/C Margin 90% confidence interval Wage Period T/C Margin 90% confidence interval Small 2 T 0.689 0.680 0.698 Low 2 T 0.690 0.682 0.699 Small 1 T 0.687 0.677 0.696 Low 1 T 0.712 0.702 0.721 Small 2 C 0.756 0.749 0.763 Low 2 C 0.762 0.755 0.768 Small 1 C 0.731 0.726 0.737 Low 1 C 0.759 0.755 0.764 DID − 0.022 − 0.020 − 0.024 DID − 0.024 − 0.021 − 0.026 Medium 2 T 0.710 0.702 0.719 Median 2 T 0.702 0.694 0.709 Medium 1 T 0.740 0.730 0.749 Median 1 T 0.712 0.704 0.720 Medium 2 C 0.769 0.762 0.776 Median 2 C 0.768 0.761 0.774 Medium 1 C 0.794 0.789 0.800 Median 1 C 0.760 0.755 0.765 DID − 0.004 − 0.002 − 0.006 DID − 0.018 − 0.016 − 0.020 Large 2 T 0.727 0.712 0.742 High 2 T 0.714 0.705 0.724 Large 1 T 0.785 0.770 0.799 High 1 T 0.712 0.703 0.722 Large 2 C 0.781 0.769 0.793 High 2 C 0.774 0.767 0.782 Large 1 C 0.845 0.837 0.854 High 1 C 0.761 0.754 0.767 DID 0.006 0.009 0.004 DID − 0.012 − 0.010 − 0.013 Page 22 of 24 Galizzietal. J Labour Market Res (2019) 53:9 etal. 2001; Berecki-Gisolf etal. 2012; Young etal. 2005b; Leyshon and Shaw 2012). Finally, a RTW to a secure job may still be not successful if the worker keeps experiencing physical or mental health limitations (Bültmann etal. 2007). These are job dimensions and workers’ preferences that cannot be captured by our administrative data and will require survey or qualitative data. Despite such limitations, our study provides new evidence that policies aiming at improving RTW need to focus also on interventions that will affect a “successful” RTW, i.e. long term employment. Almost one-fourth of our Italian injured workers were no longer working 3years after a “first” RTW. This calls for incentives to induce firms to accommodate not only their most valuable (high wage, high status, long experience) employees but also the most “disposable” ones. In addition, we provide original evidence that the outcomes of occupational injuries do not happen in a vacuum, and are not only affected by workers’ compensation, disability, or human resource policies. They are correlated also with the more general rules that affect each country labor market. This implies that policies regulating the experience of injured workers need to be aware that their final effect is likely to depend on a much wider set of national or sectoral labor market regulations that will strengthen or weaken the “protective” effect of specific firms or individual attributes. At the same time, general labor market policies reforms cannot forget how they will end up framing the experiences of some of the less visible, but more vulnerable workers, such as the ones who carried already the burden of an occupational injury. Our findings show that labor market liberalization policies were associated with changes in outcomes for all injured workers, regardless of their pre-injury contractual status. During and after the reforms were introduced such workers found easier to find new—but less secure– employment. This has very important implications. Temporary contracts have been found to have a negative effect on productivity (Lisi and Malo 2017). Furthermore, there is evidence that more insecure jobs are associated with higher frequency and severity of injuries (Amuedo-Dorantes 2002; Guadalupe 2003; Fabiano etal. 2008; Bender etal. 2012; Picchio and Van Ours 2017) and, more generally, higher risk of poor health, of future sickness absences, and of larger use of health services (László etal. 2010). All these negative consequences are likely to be amplified for a person who already suffered an injury and ends up being reemployed in a more insecure job. Therefore, workers’ compensation policies and labor market liberalization policies need to be highly coordinated to avoid severe unintended consequences as higher costs caused by new lost production and medical expenses. Additional file Additional file1 This file includes robustness checks for our multinomial estimations; display of the conditional probabilities of outcomes 3 years after the “first” RTW; display of the probabilities of “no work” for women and immigrants 1 year after the “first” RTW; and additional comparisons of selected characteristics and outcomes between truly injured and “pretend/not yet injured” workers. Acknowledgements Not applicable. Authors’ contributions MG participated in the (i) conception and design of the work; (ii) interpretation of statistical results data for the work; (iii) drafting the work and revising it critically for important intellectual content; (iv) final approval of the version to be published; and (v) agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. RL participated in the (i) acquisition, preparation, and interpretation of data for the work; (ii) drafting the work and revising it critically for important intellectual content; (iii) final approval of the version to be published; and (iv) agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. LP participated in the (i) design of the work; (ii) econometric analysis, and interpretation of empirical results for the work; (ii) drafting the work and revising it critically for important intellectual content; (iii) final approval of the version to be published; and (iv) agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All authors read and approved the final manuscript. Funding This research was not supported by external funding. Availability of data and materials The data that support the findings of this study are available from Italian Ministry of Health but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. As explained in footnote 8 the authors are available to guide through the procedures for accessing and to make available the computer programs used to generate the results. Competing interests The authors declare that they have no competing interests. Author details 1 Department of Economics, University of Massachusetts Lowell, Lowell, MA 01854-2881, USA. 2 Department of Economics and Statistics, University of Torino, 10153 Turin, Italy. 3 Laboratorio R. Revelli, Turin, Italy. Received: 27 July 2018 Accepted: 29 May 2019 References Amuedo-Dorantes, C.: Work safety in the context of temporary employment: the Spanish experience. ILR Rev. 55(2), 262–285 (2002) Anema, J.R., Schellart, A.J.M., Cassidy, J.D., Loisel, P., Veerman, T.J., van der Beek, A.J.: Can cross country differences in return-to-work after chronic occupational back pain be explained? An exploratory analysis on disability policies in a six country cohort study. J. Occup. Rehabil. 19(4), 419–426 (2009) Awang, H., Shahabudin, S.M., Mansor, N.: Return-to-work program for injured workers: factors of successful return to employment. Asia Pac. J. Public Health 28(8), 694–702 (2016) Barbieri, P., Cutuli, G.: Employment protection, labour market dualism, and inequality in Europe. Eur. Sociol. Rev. 32, 501–516 (2016) Page 23 of 24 Galizzietal. J Labour Market Res (2019) 53:9 Barbieri, P., Scherer, S.: Labour market flexibilization and its consequences in Italy. Eur. Sociol. Rev. 25, 677–692 (2009) Barr, B., Clayton, S., Whitehead, M., et al.: To what extent have relaxed eligibility requirements and increased generosity of disability benefits acted as disincentives for employment? A systematic review of evidence from countries with well-developed welfare systems. J. Epidemiol. Community Health 64(12), 1106–1114 (2010) Bena, A., Leombruni, R., Giraudo, M., Costa, G.: A new Italian surveillance system for occupational injuries: characteristics and initial results.”. Am. J. Ind. Med. 55(7), 584–592 (2012) Bender, K.A., Green, C.P., Heywood, J.S.: Piece rates and workplace injury: Does survey evidence support Adam Smith? J. Popul. Econ. 25(2), 569–590 (2012) Berecki-Gisolf, J., Clay, F.J., Collie, A., McClure, R.J.: Predictors of sustained return to work after work-related injury or disease: insights from workers’ compensation claims records. J. Occup. Rehabil. 22(3), 283–291 (2012) Bertola, G.: Job security, employment and wages. Eur. Econ. Rev. 34(4), 851–879 (1990) Berton, F., Devicienti, F., Pacelli, L.: Are temporary jobs a port of entry into permanent employment?: Evidence from matched employer–employee. Int. J. Manpow. 32(8), 879–899 (2011) Berton, F., Richiardi, M., Sacchi, S.: The political economy of work security and flexibility: Italy in comparative perspective. Policy Press; distributed by University of Chicago Press, Bristol (2012) Biering, K., Hjøllund, N.H., Lund, T.: Methods in measuring return to work: a comparison of measures of return to work following treatment of coronary heart disease. J. Occup. Rehabil. 23(3), 400–405 (2013) Boden, L.I., Galizzi, M.: Blinded by Moral Hazard. Rutgers Univ. Law Rev. 69(3), 1213–1232 (2017) Boden, L.I., Ozonoff, A.: Capture-recapture estimates of nonfatal workplace injuries and illnesses. Ann. Epidemiol. 18(6), 500–506 (2008) Bültmann, U., Franche, R.-L., Hogg-Johnson, S., et al.: Health status, work limitations, and return-to-work trajectories in injured workers with musculoskeletal disorders. Qual. Life Res. 16(7), 1167–1178 (2007) Burkhauser, R.V., Daly, M.C., Ziebarth, N.R.: Protecting working-age people with disabilities: experiences of four industrialized nations. J Labour Mark. Res 49, 367–386 (2016) Butler, R.J., Johnson, W.G., Baldwin, M.L.: Managing work disability: why first return to work is not a measure of success. Ind. Labor Relat. Rev. 48(3), 452–469 (1995) Cavaletto, G., Pacelli, L.: Flessibilità del lavoro e formazione dei lavoratori. Il caso italiano. Polis, Ricerche e studi su società e politica in Italia 2, 251–286 (2014) Ciminelli, G., Duval, R., Furceti, D.: Employment Protection Deregulation and Labor Shares in Advanced Economies. International Monetary Fund. WP/18/186. https ://www.imf.org/en/Publi catio ns/WP/Issue s/2018/08/16/ Emplo yment -Prote ction -Dereg ulati on-and-Labor -Share s-in-Advan ced- Econo mies-46074 . (2018). Accessed 8 Jan 2019 Clayton, S., Barr, B., Nylen, L., et al.: Effectiveness of return-to-work interventions for disabled people: a systematic review of government initiatives focused on changing the behaviour of employers. Eur. J. Pub. Health 22(3), 434–439 (2012) Collie, A., Lane, T.J., Hassani-Mahmooei, B., Thompson, J., McLeod, C.: Does time off work after injury vary by jurisdiction? A comparative study of eight Australian workers’ compensation systems. BMJ Open 6(5), e010910 (2016) Côté, P., Baldwin, M.L., Johnson, W.G., Frank, J.W., Butler, R.J.: Patterns of sickleave and health outcomes in injured workers with back pain. Eur. Spine J. 17(4), 484–493 (2008) Cullen, K.L., Irvin, E., Collie, A., et al.: Effectiveness of workplace interventions in return-to-work for musculoskeletal, pain-related and mental health conditions: an update of the evidence and messages for practitioners. J. Occup. Rehabil. 28, 1–15 (2018) Eurostat. Database. European Commission. Available at: http://ec.europ a.eu/ euros tat/data/datab ase. (2018). Accessed May 7 2018 Fabiano, B., Currò, F., Reverberi, A.P., Pastorino, R.: A statistical study on temporary work and occupational accidents: specific risk factors and risk management strategies. Saf. Sci. 46(3), 535–544 (2008) Fadlon, I., Nielsen, TH.: Family labor supply responses to severe health shocks. (2017). NBER working paper 21352 Gailey, A.H., Seabury, S.A.: The impact of employment protection on workers disabled by workplace injuries. In: Kessler, D.P. (ed.) Regulation versus litigation: perspectives from economics and law: a National Bureau of Economic Research Conference Report. University of Chicago Press, Chicago and London (2010) Galizzi, M.: On the recurrence of occupational injuries and workers’ compensation claims. Health Econ. 22(5), 582–599 (2013) Galizzi, M., Leombruni, R., Pacelli, L., Bena, A.: Injured workers and their return to work. Beyond individual disability and economic incentives. Evid. Based HRM. 4(1), 2–29 (2016) García-Serrano, C., Hernanz, V., Toharia, L.: Mind the gap, please! The effect of temporary help agencies on the consequences of work accidents. J. Labor Res. 31(2), 162–182 (2010) Gebel, M., Giesecke, J.: Labour market flexibility and inequality: the changing risk patterns of temporary employment in West Germany. Zeitschrift für ArbeitsmarktForschung 42, 234–251 (2009) Giannelli, G.C., Jaenichen, U., Villosio, C.: have labor market reforms at the turn of the millennium changed the job and employment durations of new entrants? J. Labor Res. 33, 143 (2012) Giraudo, M., Bena, A., Leombruni, R., Costa, G.: Occupational injuries in times of labour market flexibility: the different stories of employment-secure and precarious workers. BMC Public Health. 16, 150 (2016) Guadalupe, M.: The hidden costs of fixed term contracts: the impact on work accidents. Labour Econ. 10(3), 339–357 (2003) Høgelund, J., Holm, A.: Worker adaptation and workplace accommodations after the onset of an illness. IZA J. Labor Policy 3, 17 (2014) INAIL- Istituto Nazionale per l’Assicurazione contro gli Infortuni sul Lavoro. “Il costo dei Danni da lavoro per l’Azienda Italia” http://www.inail .it/inter net_web/wcm/idc/group s/inter net/docum ents/docum ent/ucm_08627 6.pdf. (2011). Accessed 5 June 2015 Institute for Work & Health:. Workers’ compensation and the business cycle. March. https ://www.iwh.on.ca/briefi ngs/busin ess-cycle . (2009) Kahn, L.: Employment protection reforms, employment and the incidence of temporary jobs in Europe: 1996–2001. Labour Econ. 17(1), 1–15 (2010) Koranyi, I., Jonsson, J., Rönnblad, T., Stockfelt, L., Bodin, T.: Precarious employment and occupational accidents and injuries—a systematic review. Scand. J. Work Environ. Health 44(4), 341–350 (2018) Krause, N., Frank, J.W., Dasinger, L.K., Sullivan, T.J., Sinclair, S.J.: Determinants of duration of disability and return-to-work after work-related injury and illness: challenges for future research. Am. J. Ind. Med. 40(4), 464–484 (2001) László, K.D., Pikhart, H., Kopp, M.S., et al.: Job insecurity and health: a study of 16 European countries. Soc. Sci. Med. 70(6), 867–874 (2010) Leombruni, R., Paggiaro, A., Trivellato, U.: Per un pugno di euro. Storie di ordinaria disoccupazione. Politica Economica 28(1), 5–47 (2012) Leyshon, R., Shaw, L.: Using multiple stakeholders to define a successful return to work: a concept mapping approach. Work 41(4), 397–408 (2012) Lisi, D., Malo, M.A.: The impact of temporary employment on productivity. J. Labour Mark. Res 50, 91–112 (2017) Oleinick, A., Gluck, J.V., Guire, K.E.: Establishment size and risk of occupational injury. Am. J. Ind. Med. 28(1), 1–21 (1995) Olivetti, C., Petrongolo, B.: Unequal pay or unequal employment? A crosscountry analysis of gender gaps. J. Labor Econ. 26(4), 621–654 (2008) Picchio, M., Van Ours, J.C.: Temporary jobs and the severity of workplace accidents. J. Saf. Res. 61, 41–51 (2017) Rossi, P.: The new system of compensation of occupational accidents and diseases: features and details of the reform introducing the concept of biological damage. La Medicina del lavoro 93(2), 108–117 (2002) Seing, I., MacEachen, E., Ståhl, C., Ekberg, K.: Early-return-to-work in the context of an intensification of working life and changing employment relationships. J. Occup. Rehabil. 25(1), 74–85 (2015) Strunin, L., Boden, L.I.: The workers’ compensation system: worker friend or foe? Am. J. Ind. Med. 45(4), 338–345 (2004) Venturini, A., Villosio, C.: Labour-Market Assimilation of Foreign Workers in Italy. Oxford Rev. Econ. Policy 24(3), 517–541 (2008) Vogel, A.P., Barker, S.J., Young, A.E., Ruseckaite, R., Collie, A.: What is return to work? An investigation into the quantification of return to work. Int. Arch. Occup. Environ. Health 84(6), 675–682 (2011) Wuellner, S.E., Adams, D.A., Bonauto, D.K.: Unreported workers’ compensation claims to the BLS Survey of Occupational Injuries and Illnesses: establishment factors. Am. J. Ind. Med. 59(4), 274–289 (2016) Page 24 of 24 Galizzietal. J Labour Market Res (2019) 53:9 Young, A.E., Roessler, R.T., Wasiak, R., McPherson, K.M., van Poppel, M.N.M., Anema, J.R.: A developmental conceptualization of return to work. J. Occup. Rehabil. 15(4), 557–568 (2005a) Young, A.E., Wasiak, R., Roessler, R.T., McPherson, K.M., Anema, J.R., van Poppel, M.N.M.: Return-to-work outcomes following work disability: stakeholder motivations, interests and concerns. J. Occup. Rehabil. 15(4), 543–556 (2005b) Young, A.E.: An exploration of alternative methods for assessing return-to-work success following occupational injury. Disabil. Rehabil. 36(11), 914–924 (2014) Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.