A natural experiment on sick pay cuts, sickness absence, and labor costs

A natural experiment on sick pay cuts, sickness absence, and labor costs
of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institutionand sharing with colleagues.Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third partywebsites are prohibited.In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further informationregarding Elsevier’s archiving and manuscript policies areencouraged to visit:  Author's personal copy A natural experiment on sick pay cuts, sickness absence, and labor costs Nicolas R. Ziebarth a,b, ⁎ , Martin Karlsson c a German Institute for Economic Research (DIW) Berlin, Socio-Economic Panel Study (SOEP), Mohrenstraße 58, 10117 Berlin, Germany b Berlin University of Technology (TU Berlin), Germany c Technische Universität Darmstadt (TU Darmstadt), Marktplatz 15  —  Residenzschloss, 64283 Darmstadt, Germany a b s t r a c ta r t i c l e i n f o  Article history: Received 14 May 2009Received in revised form 5 September 2010Accepted 7 September 2010Available online 21 September 2010  JEL classi  fi cation: H51I18 J22 Keywords: Sickness absenceStatutory sick payLabor costsNatural experimentSOEP Thisstudyestimatesthereformeffectsofareductioninstatutorysickpaylevelsonsicknessabsencebehaviorand labor costs. German federal law reduced the legal obligation of German employers to provide 100%continued wage pay for upto six weeks per sickness episode. In1996 statutory sickpay was decreased to80%of foregone gross wages. Within the reform's target group  –  private sector employees  –  this measureincreased the proportion of employees having zero days of absence between 6 and 8%. Quantile regressionestimates indicate that employees with up to 5.5 annual absence days reduced their days of absence by about12%.Extendedanalysessuggestthatinindustriesthatenforcedthecut,behavioraleffectswereabouttwiceaslarge. We show that the direct labor cost savings effect stemming from the cut in replacement levels clearlyexceeds the indirect effect due to the decrease in absenteeism. Our calculations about the total decrease inlabor costs are very much in line with of  fi cial data which suggest that total employer-provided sick paydecreased by 6.7% or  € 1.7 billion per year.© 2010 Elsevier B.V. All rights reserved. 1. Introduction The relationship between unemployment bene fi ts and unemploy-ment duration has attracted labor economists' attention for decadesand provided material for countless numbers of publications. In lightof this, it is odd that comparably little research on the relationshipbetween sick leave bene fi ts and sickness absence exists, despite itsenormous relevance for labor supply, labor costs, labor productivity,population health, and the functioning of social insurance systems aswell as private insurance markets.While in Europe ownership of sickness absence insurance iswidespread and mostly universal, there is no comparable socialinsurance at the federal level in the U.S. At the federal level, the U.S.onlyhas DisabilityInsurance ( DI  ),whichcompensateswagelossesdueto work disability. As compared to the sickness insurance, theliterature on the DI is rich (cf. Bound, 1989; Gruber, 2000; Campolieti,2004; Chen and van der Klaauw, 2008; de Jong et al., forthcoming).However, the empirical  fi ndings on behavioral reactions towardsgenerosityexpansionsorcontractionsoftheDIaremixed.Inaddition,these results are unlikely to be transferable to the system of theEuropean sickness insurance since the DI is about permanent ratherthantemporarywithdrawalsfromthelabormarketandhencefocusesonlaborsupply atthe extensiveratherthan theintensive margin.TheU.S. also knows another social insurance which is run on a state-by-state basis: the  Workers' Compensation Insurance  ( WCI  ) solely coversincome losses due to work-related injury or sickness.Very few studies explicitly analyzed the impact of sick pay levelson absence rates. A handful of studies exploit legislative changes inthebene fi tlevels inSweden( JohanssonandPalme,1996,2002,2005;Henrekson and Persson, 2004; Pettersson-Lidbom and SkogmanThoursie, 2008). Two English studies provide some correlation-based evidence using 1970s era data from (Doherty, 1979; Fenn,1981). In addition, two papers analyze the impact of changes to WCIbene fi t levels in the U.S. (Curington, 1994; Meyer et al., 1995). All of  the aforementioned studies  fi nd that employees adapt their absencebehavior to increases and decreases in bene fi t levels. This  fi nding isreinforced by various other empirical studies which analyze furtherdeterminants of sickness absence behavior. Workplace conditions arerelevant (Dionne and Dostie, 2007) as are probation periods, employment protection, and economic upswings or downturns(Riphahn, 2004; Ichino and Riphahn, 2005; Askildsen et al., 2005).Average sickness absence days differ substantially across countries,ranging from 4 to 29 days per year and employee (see Fig. 1). Thissuggests that institutional arrangements and cultural in fl uences are of major importance and indicates that further explanation for the  Journal of Public Economics 94 (2010) 1108 – 1122 ⁎  Corresponding author. German Institute for Economic Research (DIW) Berlin,Socio-Economic Panel Study (SOEP), Mohrenstraße 58, 10117 Berlin, Germany. E-mail addresses: (N.R. Ziebarth), Karlsson).0047-2727/$  –  see front matter © 2010 Elsevier B.V. All rights reserved.doi:10.1016/j.jpubeco.2010.09.001 Contents lists available at ScienceDirect  Journal of Public Economics  journal homepage:  Author's personal copy signi fi cantdifferenceisrequired.Italsoreinforcesthepresumptionthatthere is huge potential for ef  fi ciency gains in the sickness absenceinsurance market.Depending on a country's institutional system, employers, privateinsurance companies, or social security systems provide sick pay. Inthe case of employer-provided sick pay, companies must bear theburden of labor costs in addition to productivity losses caused byabsences from the workplace.Under Germany's generous sick pay system, employers are legallyobliged to continue to pay employees their full wage for up to sixweekspersicknessepisode.Unlikeinmostothercountries,nobene fi tcap is applied. Nevertheless, as Fig. 1 demonstrates, Germany ispositioned in the middle region of the country ranking and somecross-country comparisons even place Germany below the interna-tional average in terms of sickness absence rates (Bonato andLusinyan, 2004). One explanation might lie in the anecdotal evidencethat Germans have a strong work ethic. Other explanations may be awell-functioning monitoring system or high unemployment rates.In 1996, the Kohl government decided to reduce the statutory sickpay level from 100 to 80% of foregone gross wages, effective fromOctober1,1996.Theintentionwastwofold:toreducemoralhazardinthe sickness absence insurance and to reduce labor costs in order tofoster employment creation. At that time, employers had sick leavepayments amounting to  € 28.2 billion per year, representing 1.5% of 1996 GDP (German Federal Statistical Of  fi ce, 1998). Germany waspositioned at the top of cross-country rankings comparing total laborcosts per hour. At the time there was a consensus among economiststhat the extraordinarily high labor costs were the main reason for thepersistently high unemployment rate in Germany.The main aim of this study is to estimate the causal impact of thecut in statutory sick pay on sickness absence and labor costs. Weexploit the exogenous variation in the absence costs by using adifference-in-differences methodology and longitudinal survey datafromtheGermanSocio-Economic Panel Study(SOEP).In the fi rst partoftheempiricalanalysis,weapplyanintention-to-treatapproachthatcompares those who were totally unaffected by the law – public sectoremployees, self-employed, and apprentices – to the target populationof the reform: private sector employees. We thereby estimate theactual overall reform effect rather than the potential effect had thereformbeenstrictlyappliedbyeverysinglecompany.The latteris notthe case since a) employers are always free to provide fringe bene fi tsontopofstatutoryregulationsandb)persistentmassdemonstrationsand strikes forced employers' representatives in some industries toagreetothecontinuationoftheoldsickpayschemeincollectivewageagreements.Thus, in the second part of the empirical analysis, we exploitdifferences in sick leave schemes across collective agreements of themost important industries. Our data comprise individual-level infor-mation on the employees' industries and on whether individuals werecoveredbycollectiveagreements.Inthesecondpart,wesolelyfocusonprivate sector employees, using those who were covered by collectiveagreements with 100% guaranteed sick pay as controls. Contrastingthem withemployees inindustries where sickpaywasunambiguouslycut as well as with employees that were not covered by collectiveagreements provides additional evidence on the reform's impact.This study makes a contribution to the literature on the topic inseveralways:weestimatethecausaleffectsofcutsinsickpaylevelsonsickness absence behavior using non-Swedish and uncensored data.Toidentify the causal reform effect we make use of three differentapproaches that rely on different subsamples that were unequallyaffectedbythereform.Bythismeans,theplausibilityandrobustnessof the identi fi ed effects is veri fi ed. While we provide evidence on theoverall reform effects in the  fi rst part of the analysis, we gain greaterinsight into the speci fi c mechanisms of the reform in the second part.Thanks to the panel structure of the data, we are able to take thesample composition into account. Most of the evaluation literaturestruggles with selection issues that often signi fi cantly hamper theanalysis. In this context, we can identify job changers and employeeswho potentially selected themselves into or out of the treatment.Thus, sorting is unlikely to be a major issue. Unlike studies thatestimate effects in certain regions or states, we use a representativesample of the most populous European country.This paper also contributes to the broader fi eld of literature on theinterdependencies between social insurance systems and laborsupply. Since reduced labor costs was one of the main objectives of the reform, we calculate employers' total labor cost savings androughly estimate the number of jobs which may have been created asa consequence of the reform.Finally, this study illustrates the pitfalls that policymakers facewhen planning to implement unpopular labor market reforms incountries with a strong tradition of collective bargaining and strongunions.Hadthepurposeofthereformbeenbettercommunicatedand Fig. 1.  Differences in annual absence days by OECD country.1109 N.R. Ziebarth, M. Karlsson / Journal of Public Economics 94 (2010) 1108 – 1122  Author's personal copy had the new law been applied one-to-one by all employers, ourcalculations suggest that twice as many jobs could have been createdas actually occurred.Section 2 outlines the institutional setting in Germany. Section 3provides more detail on the data. Section 4 discusses the empiricalestimationstrategy.Thisis followed by Section5 in which weprovidethe results of our empirical analyses. Section 6 outlines the study'sconclusions. 2. The German sick pay scheme and policy reform  2.1. The sick pay scheme and monitoring system Germany has one of the most generous sick pay schemes in theworld. Before the implementation of the new law, every employerwas legally obliged to continue usual wage payments for up to sixweekspersicknessepisode.Inotherwords,employershadtoprovide100%sick payfromthe fi rstdayofa periodof sicknesswithnobene fi tcaps. 1 Henceforth, we use the term short-term sick pay as a synonymfor employer-provided sick pay and short-term absenteeism as asynonym for periods of absence of less than six weeks.Inthecaseofillness,employeesareobligedtoimmediatelyinformtheir employer about both the sickness and expected duration. Fromthe fourth day of a sickness episode, a doctor's certi fi cate is requiredand is usually issued for up to one week, depending on the illness. If thesicknesslastsmorethansixcontinuousweeks,thedoctorneedstoissue a different certi fi cate. From the seventh week onwards, sick payisdisbursedbythesicknessfundandloweredto80%offoregonegrosswages for those who are insured under Statutory Health Insurance(SHI). 2 The monitoring system mainly consists of an institution called Medical Service of the SHI   (Medizinischer Dienst der Krankenversicher-ung).One of the srcinal objectives oftheMedical Serviceis to monitorsickness absence. German social legislation codi fi es that the SHI isobliged to call for the Medical Service and a medical opinion to clarifyany doubts about work absences. Such doubts may arise if the insuredperson is short-term absent with unusual frequencyor is regularly sickonMondaysorFridays.Similarly,ifdoctorscertifysicknesswithunusualfrequency,theSHImayaskforexpertadvice.Theemployeralsohastheright to call for the assistance of the Medical Service and expert advice.Expert advice is based on available medical documents, informationabout the workplace, and a statement which is requested from thepatient. If necessary, the Medical Service has the right to conduct aphysical examinationof the patient and to cut bene fi ts. 3 In 2007,about2000 full-time equivalent and independent doctors worked for themedical service and examined 1.7 million cases of absenteeism(Medizinischer Dienst der Krankenversicherung (MDK), 2008).  2.2. The policy reform In 1996, the total sum of employer-provided sick pay amounted toDM 55.3 billion ( € 28.2 billion) (German Federal Statistical Of  fi ce,1998) and it was claimed to contribute to persistently highunemployment rates by functioning like a tax on labor. Togetherwith speculation about a high degree of moral hazard in the generousGerman sick pay scheme, these considerations prompted the Germangovernment to pass Employment Promotion Act 4 which went intoeffect October 1, 1996.The law reduced the sick pay employees are entitled to claim from100 to 80% of gross wages during the  fi rst six weeks per sicknessepisode. The Self-employed were unaffected by the new law. Due topolitical considerations and the existence of other laws, both publicsector employees and apprentices were exempt from the reform. 5 Similarly unaffected were employees on sick leave due to workrelatedaccidents.Asanalternativetothecutinsickpay,fromthedatewhen the new law became effective, employees had the right toreduce their paid vacation by one day for every  fi ve days of sicknessrelated absence, thereby avoiding the sick pay cut.In addition to this law, which lowered short-term sick pay and isthe focus of this study, another law was passed on November 1, 1996,andbecameeffective from January1, 1997,onwards. 6 The second lawreduced sick pay from the seventh week onwards from 80 to 70% of forgonegrosswages.Theimpactofthislawonlong-termabsenteeismis analyzed elsewhere (Ziebarth, 2009). So as to not confuse theimpact of the cut in long-term sick pay with the impact of the cut inshort-term sick pay, it is important to analyze the effects of bothreformsseparately.Thisisalsonecessarysincethesubgroups affecteddiffered between the two reforms. 7 Before,andintheaftermathofthelaw'simplementation,thegeneralpublic and the unions put pressure on employers through massdemonstrations and strikes. Germany is the country of srcin of Bismarckian corporatism that has been adopted as a role model byseveralEuropeancountries.AnintegralpartofBismarckiancorporatismis the idea of social partnership between employers and unions alongwithautonomyinbargaining.Asaresult,unionsaretraditionallystrongin Germany, as is the degree of collective bargainingcoverage. In 1998,about 68% of all employees in West and 50% in East Germany werecovered by collective wage agreements (Hans Böckler Stiftung, 2009).Ongoing union pressure forced employers' associations in variousindustries to agree in collective agreements to provide sick payvoluntarilyontopofthestatutoryregulations.However,thesecollectiveagreements were only binding for  fi rms under collective bargainingcoverage. 8 Fig.2givesanoverviewofhowthereformworked(GermanFederal Statistical Of  fi ce, 1996, 2008; Hans Böckler Stiftung, 2009).Principally, one needs to differentiate between those occupationalgroupsthatwereintendedtobeaffectedbytheweakeningofstatutoryminimum standards and those that were not. The law applied to allemployees in the private sector (A), whereas apprentices, self-employed, and public sector employees (B) were exempt as explainedpreviously.Atthelowerlevel,oneneedstodifferentiatebetweenthoseprivate sector employees who were covered by collective agreements(A.2) and those who were not (A.1). For companies not covered bycollective bargaining, the decision on whether fringe bene fi ts areprovided on top of statutory standards is made at the company level.However, companies thatare not covered by collective agreements areusuallyunlikelytoprovideasubstantialamountoffringebene fi ts.Thus,itisverylikelythatthosetenmillionprivatesectoremployeeswhowerenot covered by a collective agreement experienced the cut in statutory 1 The entitlement is codi fi ed in the  Gesetz über die Zahlung des Arbeitsentgelts anFeiertagen und im Krankheitsfall (Entgeltfortzahlungsgesetz) , article 3, 4. 2 In principle, there is no limit on the frequency of sick leave spells. However, if employees fall sick again due to the same illness after an episode of six weeks, the lawexplicitly states that they are only again eligible for employer-provided sick pay if atleast six months have been passed between the two spells or twelve month have beenpassed since the beginning of the fi rst spell. This clause intends to avoid substitution of long-term spells by short-term spells. 3 The wording of the laws can be found in the Social Code Book V, article 275, para.1, 1a; article 276. 4 The  Arbeitsrechtliches Gesetz zur Förderung von Wachstum und Beschäftigung  (  Arbeitsrechtliches Beschäftigungsförderungsgesetz  ),  BGBl. I 1996 p. 1476  – 1479 , waspassed September 25, 1996. 5 In the case of apprentices, the  Berufsbildungsgesetz (BBiG)  prevented theapplication of the law. 6 This law is the  Gesetz zur Entlastung der Beiträge in der gesetzlichen Krankenver-sicherung   ( Beitragsentlastungsgesetz   —  BeitrEntlG ),  BGBl. I 1996 p. 1631 – 1633. 7 As an example, Puhani and Sonderhoff (2010) do not differentiate between bothreforms. In addition, they solely contrast group A.1 with group A.2 in Fig. 2. In doing sothey compare two groups that both include treated and non-treated employees. 8 Employers are free to leave collective agreements after the expiration of thecontract. Average contract terms were 16.8 months in 1997; they even increasedsteadily from 1994 to 1998 (Hans Böckler Stiftung, 2009). In addition, we have notfound any evidence that an unusual number of   fi rms left collective agreements afterthe reform.1110  N.R. Ziebarth, M. Karlsson / Journal of Public Economics 94 (2010) 1108 – 1122  Author's personal copy sick pay. Those who were covered by collective bargaining were eithernot affected or were affected by the cut in sick pay, depending on theindustry and theresult of collectivebargaining. In ourdata,we are ableto identify all four groups as displayed in Fig. 2. As shown later, weestimatethreedifferentempiricalmodelsthatcomparegroupsa.)AandB,b.)A.2.1andA.2.2,andc.)A.1andA.2.2toidentifythereformeffects.We call these three models Approach I, Approach II, and Approach III. 3. Data and variable de fi nitions Theempiricalspeci fi cationsmakeuseoftheGermanSocio-EconomicPanelStudy(SOEP).TheSOEPistheonlyavailablerepresentativedatasetforGermanythatincludesinformationonsicknessabsence.TheSOEPisalongitudinal annual household survey that has existed since 1984.Wagner et al. (2007) provide further information.For the empirical speci fi cations, we extract three pre- and two post-reform years from the survey, i.e., the waves from K (1994) to P (1999)thateachcontainssicknessabsenceinformationaboutthepreviousyear. 9 Wediscardtheyear1996inmostspeci fi cationsbecausethereformwentintoeffectOctober1,1996,andbecauseweonlyhaveabsenceinformationbased on calendar years. However, in one robustness check, we estimatetreatment effects separately for the years 1996, 1997, and 1998.We restrict our sample to those in the labor force who are eligiblefor sick pay (plus self-employed) and who are between 18 and65 years of age. 10 Work accident related sick leave was exemptedfrom the new regulations. Thus, we exclude all respondents whoneeded medical treatment due to a work accident in one of thesampling years. Besides short-term sick pay, long-term sick pay,which is disbursed from the seventh week onwards, was alsoeffectively reduced as of January 1997. Since we intend to isolatethe reform effects on short-term absenteeism, we discard allrespondents having had a long-term sickness spell of more than sixweeksin oneof thesamplingyears. 11 Inthe empiricalassessment, weprovide evidence suggesting that excluding the homogeneous, butspecial, group of the long-term sick poses no severe threat to theevaluation of the cut in short-term sick pay. We also discardindividuals with item non-response.  3.1. Endogenous and exogenous variables The SOEP is a rich dataset, particularly with respect to jobcharacteristics. Detailed questions about the type of job, the numberof years with the employer, the gross and net wage, etc. are sampled.Additionally, there are questions on sick leave behavior.Wegenerateourdependentvariablesfromthefollowingquestion: “ How many days off work did you have in 19XX because of illness?Please enter all days, not just those for which you had a doctor'scerti fi cate. ” a.)Nodaysb.)XXdaysintotal.ThegreatadvantageoftheSOEP and this question is that the  total  number of absent days isdocumented, not only those with a certi fi cate or those that arecompensated by a federal institution as it is the case with mostregister data. Particularly when the focus is on short-term absentee-ism, it is a big advantage to have such a total measure. However, thiscomes at the cost of not having detailed spell data.The  fi rst dependent variable measures the proportion of employ-ees with no absence days, i.e., the incidence of sickness absence. It isnamed  noabs  and has a one for employees with no sick leave days; allemployees with sick leave are coded as zeros.  Noabs  should not bevery prone to measurement errors.The second dependent variable measures the total number of absentdaysandiscalled daysabs .However,lookingatthedistributionof this variable, the potential issues of measurement errors,misreporting behavior, and outliers become quite obvious. Forexample, 0.2% (i.e., 30 respondents) of the sample indicated a totalnumber of absence days of more than 50, which is theoreticallypossible but, given that these respondents also denied an absencespell of more than six weeks, quite unlikely. While the evaluation of the reform effects should not be seriously distorted as long as thereform did not affect measurement errors, outliers and misreportingpotentially exacerbate standard errors and lead to imprecise esti-mates. Moreover, in the case of underreporting, with 0 b α  b 1 beingthe reported fraction of true days, our estimates would be downwardbiased by  α  .The entire set of explanatory variables can be found in theAppendix A and is categorized as follows: the fi rst group incorporatesvariables on personal characteristics, such as the dummies  female,immigrant  ,  East German ,  partner  ,  married ,  children ,  disabled ,  health good ,  health bad ,  no sports , and  age  ( age 2 ). The second group consistsof educational controls such as years of educational attainment orwhether the person was trained speci fi cally for their job. The lastgroup contains explanatory variables on job characteristics. Among Fig. 2.  Overview of treatment and control groups. 9 What is meant here is that we collect data from the years 1993 – 1998. Sincecurrent as well as retrospective information is sampled in every wave, we match theretrospective information which we are interested in with the current information of the relevant year as long as the respondent was interviewed in both years. If this wasnot the case, we use both types of information from the same interview and assumethat the current statements have not changed since the previous year. 10 Although marginally employed (employees who earn less than  €  400 per month)are eligible for sick pay and have been on a par with the full-time employed since June1, 1994, we do not include them since it is likely that marginally employed were notfully aware of their rights at that time and since anecdotal evidence suggests that asigni fi cant proportion of employers refused to provide this bene fi t. 11 The identi fi cation of these respondents is feasible since a question on whetherrespondents had such a long-term spell is annually asked. In Section 5.1, we again usethe whole sample to estimate the total labor cost savings for Germany. Likewise,respondents are asked every year whether they had a work accident and whether thisaccident required medical treatment.1111 N.R. Ziebarth, M. Karlsson / Journal of Public Economics 94 (2010) 1108 – 1122
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks