The economic and social determinants of employee behavior: evidence from insider econometric studies / von Konstantin Böddeker, M.Sc.

Fakul ä ü Wi scha swissenscha en Leh s uhl ü O ganisa ions-, Medien- und Spo ökonomie ___________________________________________________________________________________ THE ECONOMIC AND SOCIAL DETERMINANTS OF EMPLOYEE BEHAVIOR: EVIDENCE FROM INSIDER ECONOMETRIC STUDIES ___________________________________________________________________________________ De Fakul ä ü Wi scha swissenscha en de Uni e si ä Pade bo n zu E langung des akademischen G ades Dok o de Wi scha swissenscha en - Doc o e um poli ica um - o geleg e Disse a ion on Kons an in Böddeke , M.Sc. gebo en am 10.12.1985 in Pade bo n (2016) The da a used in his hesis a e p op ie a y and can, he e o e, no be made a ailable o o he esea che s. The esul s, opinions and conclusions o his disse a ion a e hose o he au ho and no necessa ily hose o he Volkswagen AG. Table o Con en III TABLE OF CONTENT LIST OF FIGURES ............................................................................................................ V LIST OF TABLES ............................................................................................................. VI LIST OF ABBREVIATIONS ........................................................................................ VIII 1 INTRODUCTION ........................................................................................................... 1 2 WORK TEAM DIVERSITY AND THE OPTIMAL COMPOSITION OF TEAMS: A META-ANALYTIC APPROACH .................................................... 11 2.1 In oduc ion ........................................................................................................... 11 2.2 Di e si y ................................................................................................................ 13 2.2.1 De ini ions o Di e si y ................................................................................. 14 2.2.2 Dimensions o Di e si y ................................................................................ 16 2.2.3 Measu es o Di e si y .................................................................................... 16 2.3 Theo e ical Backg ound o Di e si y Resea ch .................................................... 18 2.3.1 Social Ca ego iza ion and Simila i y-A ac ion Pa adigm ........................... 18 2.3.2 In o ma ion P ocessing Pe spec i e and Economic Theo y .......................... 21 2.3.3 Mode a o s o he Di e si y-Pe o mance Rela ionship ................................ 23 2.4 Me a-Analysis App oach ...................................................................................... 25 2.5 Resul s ................................................................................................................... 30 2.6 Conclusion and Implica ions ................................................................................ 36 3 EMPLOYEE ABSENTEEISM: DETERMINANTS IN THE INTERNATIONAL CONTEXT ................................................................................. 39 3.1 In oduc ion ........................................................................................................... 39 3.2 Theo e ical F amewo k ......................................................................................... 42 3.2.1 Social Pee In luences in Wo k Teams.......................................................... 45 3.2.2 Economic In luences / Incen i es .................................................................. 48 3.2.3 Wo ke Cha ac e is ics .................................................................................. 49 3.2.4 Wo king Condi ion ........................................................................................ 52 3.3 Da a Se and Desc ip i e S a is ics ....................................................................... 52 3.4 Es ima ion S a egy ............................................................................................... 57 3.5 Empi ical Resul s .................................................................................................. 61 3.6 Conclusion and Implica ions ................................................................................ 74 Table o Con en IV 4 ON THE ROAD AGAIN: CROWDING-OUT EFFECTS OF EXTRINSIC MOTIVATION IN COMMERCIAL TRUCKING.................................................... 79 4.1 In oduc ion ........................................................................................................... 79 4.2 Theo e ical F amewo k ......................................................................................... 82 4.3 Da a Se and Desc ip i e S a is ics ....................................................................... 88 4.4 Empi ical Models and Es ima ion Resul s ............................................................ 93 4.5 Conclusion .......................................................................................................... 106 5 BEHAVIORAL CONSEQUENCES OF THE TRANSITION FROM TEMPORARY TO PERMANENT EMPLOYMENT ............................................. 110 5.1 In oduc ion ......................................................................................................... 110 5.2 Li e a u e Re iew and Theo e ical F amewo k .................................................. 112 5.3 Da a Se and Desc ip i e S a is ics ..................................................................... 117 5.4 Empi ical Models and Es ima ion Resul s .......................................................... 121 5.5 Conclusion .......................................................................................................... 128 6 SUMMARY AND FUTURE OUTLOOK ................................................................. 131 APPENDIX ........................................................................................................................ IX REFERENCES ............................................................................................................ XXIX EIDESSTATTLICHE ERKLÄRUNG ......................................................................... LXI Lis o Figu es V LIST OF FIGURES Figu e 1.1: Di e si y as Sepa a ion, Va ie y and Dispa i y ................................................. 15 Figu e 3.1: S ee s and Rhodes (1978) P ocess Model o Employee Absence .................... 43 APPENDIX Figu e A.1 : Fo es Plo – Age Di e si y and O e all Pe o mance.................................... IX Figu e A.2: Fo es Plo – Gende Di e si y and O e all Pe o mance ................................ X Figu e A.3: Fo es Plo – Cul u e Di e si y and O e all Pe o mance ............................... XI Figu e A.4: Fo es Plo – Tenu e Di e si y and O e all Pe o mance .............................. XII Figu e A.5: Fo es Plo – Func ional Backg ound Di e si y and O e all Pe o mance... XIII Figu e A.6: Fo es Plo – Educa ional Backg ound Di e si y and O e all Pe o mance . XIV Figu e A.7: Fo es Plo – Educa ion Le el Di e si y and O e all Pe o mance ............... XV Figu e A.8: Fuel Consump ion wi h / wi hou Incen i es ................................................ XXI Figu e A.9: Ke nel Densi y Plo o Fuel Consump ion by T ip E alua ion ............... XXVIII Lis o Tables VI LIST OF TABLES Table 2.1: Fo mulas o he Calcula ion o He e ogenei y Measu es ................................. 17 Table 2.2: Rela ionship be ween Di e si y and O e all Team Pe o mance ...................... 31 Table 2.3: Team Size as Mode a o o he Pe o mance-Di e si y Rela ionship ................ 33 Table 2.4: Team Type as a Mode a o o he Pe o mance-Di e si y Rela ionship............ 34 Table 3.1: Summa y S a is ics – Da a Se Composi ion ...................................................... 54 Table 3.2: Summa y S a is ics – Means and s anda d de ia ions........................................ 55 Table 3.3: Fixed-E ec s In e ac ion Te ms on Absence Ra es (I) ...................................... 63 Table 3.4: Fixed-E ec s In e ac ion Te ms on Absence Spell Du a ion (II) ...................... 64 Table 3.5: Fixed-E ec s In e ac ion Te ms on Absence F equency (III) .......................... 65 Table 4.1: Summa y S a is ics - D i e Demog aphics ....................................................... 90 Table 4.2: Summa y S a is ics - D i ing S yle .................................................................... 90 Table 4.3: Summa y S a is ics - T a ic Ligh Pe o mance E alua ion Tool..................... 92 Table 4.4: Baseline Resul s o he Fixed-E ec s Es ima ions (Models 1 o 6) ................... 98 Table 4.5: Seemingly Un ela ed Reg ession (SUR) Resul s on D i ing Pa ame e s ........ 102 Table 4.6: Seemingly Un ela ed Reg ession (SUR) Resul s on E alua ion Pa ame e s ... 103 Table 4.7: Es ima ed P obabili ies o T ip E alua ion in bo h Incen i e En i onmen s . 104 Table 5.1: Summa y S a is ics – D i e Demog aphics .................................................... 118 Table 5.2: Summa y S a is ics – Dependen Va iables ..................................................... 119 Table 5.3: Impac o Con ac S a us on Sho -Te m Fuel Consump ion .......................... 124 Table 5.4: Impac o Con ac S a us on Long-Te m Fuel Consump ion .......................... 125 Table 5.5: Es ima ed P obabili ies o T ip E alua ion ..................................................... 127 Lis o Tables VII APPENDIX Table A.1: He e ogenei y Measu es Lis ed by F equency o Use ....................................... IX Table A.2: Di e si y-Pe o mance Rela ion by Pe o mance Measu e ............................ XVI Table A.3: Fixed-E ec s Es ima ions on Absence Ra e (IV) ....................................... XVIII Table A.4: Fixed-E ec s Es ima ions on Mean Absence Spell Du a ion (V).............. 19IXX Table A.5: Fixed-E ec s Es ima ions on Absence F equency (VI) .................................. XX Table A.6: Desc ip i e S a is ics wi h and wi hou Incen i es in P ac ice ....................... XIX Table A.7: Two-Way Ano a o Incen i es and D i e on Fuel Consump ion.............. XXIII Table A.8: Findings on Fuel Consump ion a e Ins alla ion o Incen i es .................. XXIII Table A.9: Findings on Fuel Consump ion a e Aboli ion o Incen i e ...................... XXIV Table A.10: O de ed Logi Es ima ion o T ip E alua ion .......................................... XXIV Table A.11: Baseline Resul s o Fixed-E ec s Es ima ion (incl. Bonus Regime) ......... XXV Table A.12: Baseline Resul s o Fixed-E ec s Es ima ion (incl. T end Va iable) ....... XXVI Table A.13: O de ed Logi Es ima ion o D i e Pe o mance .................................... XXVII Lis o Abb e ia ions VIII LIST OF ABBREVIATIONS CI Con idence In e al CV Coe icien o Va ia ion DWD Deu sche We e diens (Ge man Me eo ological Se ice) ES E ec Size ESP Spain EU Eu opean Union FE Fixed-E ec s (Reg ession) GDP G oss Domes ic P oduc GER Ge many GPS Global Posi ioning Sys em HR Human Resou ces HRM Human Resou ce Managemen ILI In luenza-like Illnesses KPI Key Pe o mance Indica o OECD O ganiza ion o Economic Co-Ope a ion and De elopmen OHC Occupa ional Heal h Ca e OLS O dina y Leas Squa es (Reg ession) R&D Resea ch and De elopmen RE Random-E ec s (Reg ession) ROA Re u n on Asse s ROE Re u n on Equi y ROS Re u n on Sales s.d. S anda d De ia ion SUR Seemingly Un ela ed Reg ession TMT Top Managemen Team UK Uni ed Kingdom US Uni ed S a es o Ame ica WHO Wo ld Heal h O ganiza ion In oduc ion 1 1 INTRODUCTION Since he 1980s he s and o pe sonnel economics has eme ged om labo economics in o de o add ess bo h he lack o economic pe spec i es in ea ly human esou ce manage- men esea ch as well as he g owing need o hands-on, business-o ien ed insigh s on hu- man esou ce managemen (HRM) (Lazea 2000a). Ini ially in luenced by in o ma ional economics and u he ex ended by ideas o igina ing om social psychology and o ganiza- ional sociology, pe sonnel economics is nowadays a well-es ablished and subs an ial pa o labo economics esea ch. Re iews by P ende gas (1999), Lazea (1999, 2000a) and Lazea and Oye (2013) p o ide a comp ehensi e o e iew o he his o ical de elopmen as well as he con empo a y deba e wi hin he ield o pe sonnel economics. Pe sonnel economics applies (labo ) economic me hods such as econome ics and game heo y o p o ide a de ailed unde s anding o he unc ioning o a i m and i s human e- sou ce p ac ices. Howe e , due o limi ed da a a ailabili y in he ea ly yea s, pe sonnel economics s a ed o wi h mos ly heo e ical conside a ions. La e con ibu ions combine heo y wi h empi ical analyses as hey a e buil on ich i m-based da a se s. In gene al, pe sonnel economic models and heo ies play by he adi ional ules o economics, in pa - icula he assump ions o a ional maximizing agen s, speci ic equilib ium analysis o la- bo and p oduc ma ke s and e iciency as a esul o ma ke equilib ia (Lazea 2000a, Lazea , Shaw 2007). Ne e heless, he main esea ch ields o pe sonnel economics a e no limi ed o he classic HRM issues such as compensa ion, u no e and incen i es. Ins ead, i s scope o esea ch is ex ended beyond adi ional economic ields, including opics such as g oup no ms, eamwo k se ings, o wo ke empowe men . In b ie , one can summa ize he main esea ch in e es s o pe sonnel economis s o comp ise incen i es, wo ke - i m ma ching, compensa ion, skill de elopmen and o ganiza ion o wo k (Lazea , Oye 2013). These opics sha e he common h ead in ha pe sonnel economics is all abou enhancing wo ke p oduc i i y (Lazea , Gibbs 2009). In con as o labo economics, pe sonnel economics does no appea policy-o ien ed: I ocuses exclusi ely on he wel a e wi hin an indi idual employmen ela ionship and no on he o e all social wel a e (Lazea , Oye 2013). In o he wo ds, pe sonnel economics aims a o e ing p ac ical implica ions o manage s on how o imp o e ope a ions and, hus, enhance he o e all p oduc i i y o a gi en i m. As a consequence, pe sonnel eco- nomics esea ch mainly ocuses on hose a iables ha manage s ha e an ac ual bea ing in. In oduc ion 8 hand, simila i y-a ac ion pa adigm and social ca ego iza ion heo y sugges g oup he e o- genei y o induce pe o mance and p oduc i i y o be in e io due o sub-g oup o ma ion (in-g oup s. ou -g oup) and wo k g oup con lic . As a consequence, homogeneous eams a e assumed o be p e e able (Taj el 1982, Taj el, Tu ne 1986, Tu ne 1987). On he o he hand, he in o ma ion p ocessing pe spec i e assumes he e ogeneous eams o inc ease pe o mance and p oduc i i y due o complemen a y cogni i e esou ces, expe iences and ne wo k ies. Hence, he e ogeneous eams a e expec ed o be ad an ageous in e ms o pe o mance and p oduc i i y (Cox, Blake 1991). The e is ex ensi e empi ical e idence o bo h app oaches. Applying me a-analy ic esea ch o a o al o 66 indi idual samples de i ed om 63 p ima y s udies on wo k g oup di e si y, his pape seeks o analyze he b oad li e a u e on di e si y in sea ch o common s a is ical pa e ns on he ad an ageous- ness o ei he homogeneous o he e ogeneous eams. Di e si y is conside ed along less ask- ela ed (age, gende , cul u e) and highly ask- ela ed ( enu e, unc ion, educa ional backg ound and educa ional le el) a ibu es. Fu he mo e, empi ical es s o po en ial mode a ing e ec s o eam ype ( op managemen eams, wo k eams, esea ch and de el- opmen eams and mixed eams) and eam size a e pe o med. O e all, indings con i m he ambi alen na u e o wo k g oup he e ogenei y. In suppo o social ca ego iza ion he- o y and simila i y-a ac ion pa adigm, he esul s sugges nega i e popula ion co ela ions o gende and age he e ogenei y on eam pe o mance. Likewise, educa ional backg ound di e si y is obse ed o be posi i ely ela ed o eam pe o mance, hus, con i ming a gu- men s s a ed by in o ma ion p ocessing pe spec i e. Two mode a ing e ec s a e obse ed. Fi s , inc easing eam size mode a es he in luence o eam di e si y on pe o mance nega- i ely o less ask- ela ed di e si y a ibu es and posi i ely o highly ask- ela ed a ib- u es. Second, op managemen eam ype and esea ch and de elopmen eam ype posi- i ely mode a e he di e si y-pe o mance ela ionship ega dless o he ask- ela edness o di e si y. Mo eo e , wo k eam ype nega i ely mode a es he di e si y-pe o mance ela- ionship o wo k eams ega dless o he ask- ela edness o di e si y a ibu es. Chap e h ee depic s join wo k wi h Be nd F ick in which a hi he o una ailable da a se co e ing 160 blue-colla wo k uni s om ou in e na ional manu ac u ing plan s o a la ge Eu opean au omobile manu ac u e is used o analyze social and economic de e minan s o employee absen eeism. In his chap e , he unde s anding o absen eeism is based on he S ee s and Rhodes (1978) p ocess model as well as on economic heo ies as p oposed by neoclassical labo supply models (e.g. Allen 1981, Dunn, Youngblood 1986) and e icien- In oduc ion 9 cy wage heo y (Shapi o, S igli z 1984). By means o a comp ehensi e e iew o exis ing empi ical indings on absence esea ch, a se o hi een po en ial de e minan s o absence is iden i ied ha s em om social pee in luences (wo k uni size, u no e , sha e o em- po a y co-wo ke , sha e o heal h-impai ed co-wo ke s), economic in luences / incen i es (sick pay egula ions, employmen p o ec ion laws, unemploymen , p ospe i y le el), wo ke cha ac e is ics (age, gende , enu e, acu e heal h) and wo king condi ions (shi sys em). Absence is ound o be a mul i ace ed phenomenon since only ew s a is ical pa - e ns hold ue o all ou plan s. Ins ead, co e indings sugges esul s o a y signi ican - ly by p oduc ion si e. Based on empi ical analyses as well as in e iews wi h on-si e ex- pe s (i.e. HR manage s, wo ke ep esen a i es, manage s and shop loo s a ), esul s sugges absence o inc ease wi h uni size and o e all uni u no e . Likewise, a high sha e o empo a y co-wo ke s inc eases g oup-le el absence among pe manen employees. In con as , he hypo hesized link be ween absence and he sha e o co-wo ke s ha su e any kind o pe manen o empo a y heal h-impai men could no be suppo ed. Addi ion- ally, posi i e ela ionships be ween absence and s ic employmen p o ec ion laws as well as a o able p ospe i y le els a e iden i ied. Findings wi h espec o he na ional (un)employmen si ua ion a e inconclusi e and esul s o na ional sickness bene i s legis- la ion ail o each s a is ical signi icance. Fu he mo e, g oup-le el absence is obse ed o inc ease wi h he sha e o emale employees in he eam. Mo eo e , bo h in olun a y and olun a y absence is obse ed o inc ease wi h age. Howe e , esul s o enu e as well as na ional in luenza ou b eaks a e inconclusi e. Wi h ega d o wo king condi ions, absence is obse ed o be highe in uni s ha ope a e in wo- and h ee-shi sys ems compa ed o hose ollowing he s anda d one-shi sys em. Employees’ beha io al esponses o changes in mone a y incen i e design a e discussed in chap e ou , which is join wo k wi h Be nd F ick. Buil on lee managemen panel da a o an in-house haule o a la ge Eu opean uck manu ac u e , his pape analyzes he pe - o mance o 37 comme cial uck d i e s on a o al o 6,326 indi idual ips wi hin a h ee- yea ime ame. Pe o mance is measu ed ia ou di e en se s o a iables ha a e ec- o ded by GPS-based on-boa d compu e s: o e all uel consump ion, d i ing beha io as e lec ed by se e al d i ing pa ame e s, ela i e pe o mance based on d i ing sco es and, e en ually, a compu ed o e all pe o mance e alua ion. The o ganiza ional se ing ollows a quasi-expe imen al app oach since he exis ing pe o mance bonus sys em had been abolished by he haule a e he i s wo yea s o he h ee-yea obse a ion pe iod. F om In oduc ion 10 heo y, wo po en ial beha io al esponses migh be expec ed om employees. Classic con ac heo y assumes human beings o di ec ly espond o ex insic incen i es in in- c easing e o as well as educing shi king beha io (e.g. Lazea 2000b, Gibbons 1998, Gibbons, Robe s 2013). Howe e , a compe ing heo e ical app oach doub s hese bene i s o ex insic (mone a y) incen i es a guing ha ex insic incen i es li e ally “buy o ” an employee’s in insic mo i a ion and, he e o e, esul in dec easing o e all pe o mance (e.g. Deci 1971, Bénabou, Ti ole 2003, 2006, Gneezy e al. 2011). Empi ical and expe i- men al e idence o bo h app oaches is discussed in he chap e . In e es ingly, indings con adic economic heo y in gi ing suppo o he c owding-ou app oach. T uck d i e s a e obse ed o display signi ican ly lowe e o and pe o mance when incen i ized ex- insically by means o a pe o mance bonus. Es ima ions e eal ha d i e s beha e less eco- iendly and less in acco dance wi h company s anda ds yielding an o e all lowe pe - o mance e alua ion. This esul implies ha d i e s’ in insic and social mo i es – e.g. image conce ns, en i onmen al belie es – a e c owded ou by mone a y ewa ds. As men ioned ea lie , chap e i e – again co-au ho ed by Be nd F ick – analyzes beha - io al esponses o employees who a e “p omo ed” om empo a y o pe manen con ac s. This chap e is based on he iden ical lee managemen panel da a used in chap e 4. Again, he o ganiza ional se ing ollows a quasi-expe imen al design since i obse es eigh comme cial uck d i e s on 1,299 ips whose con ac s a us changed om empo- a y o pe manen . Conside ing ad an ages o pe manen o e empo a y con ac s (e.g. be e wo king condi ions, highe pay and s ic e employmen p o ec ion), empo a y em- ployees may be assumed o be highly incen i ized o display high e o in o de o quali y o a pe manen con ac . Empi ical suppo o his a gumen is discussed wi hin he chap- e (e.g. B adley e al. 2012). Findings sugges abou hal o he d i e s o signi ican ly inc ease uel consump ion a e ha ing signed a pe manen con ac wi h a s ong ini ial e ec and a long decline back o “no mal” le els. This esul indica es hese d i e s o be- ha e as a ional chea e s (Nagin e al. 2002). The same holds ue o compu e ized pe - o mance e alua ions. Subsequen ly, chap e 6 summa izes he main esul s o his hesis and p o ides bo h gen- e al conclusions as well as sugges ions o u u e esea ch. Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 11 2 WORK TEAM DIVERSITY AND THE OPTIMAL COMPOSITION OF TEAMS: A META-ANALYTIC APPROACH 2.1 In oduc ion These days, o ganiza ions inc easingly implemen eamwo k o mee he demands o sus- ainable de elopmen s in economic, echnological and socie al en i onmen s (e.g. De ine e al. 1999). Gi en he g owing impo ance o eams, managemen schola s ha e eage ly aken on he s udy o eamwo k and i s in luence on business p ocesses as well as employ- ee pe o mance (e.g. Sunds om e al. 2000). As a consequence, eamwo k esea ch is nowadays es ablished as one o he mos ele an opics in managemen esea ch, a posi- ion anecdo ally desc ibed by Lazea (1999) who s a es ha “i is impossible o pick up a business publica ion hese days wi hou eading abou he wonde s o eamwo k” (Lazea 1999, p. C15). A he same ime, he composi ion o he global wo k o ce has changed in ha wo ke he - e ogenei y has inc eased d ama ically o e he las decades. Mul i ace ed social and demo- g aphic de elopmen s may be held accoun able o his phenomenon. Conside , o in- s ance, he cons an ly g owing p opo ion o women pa icipa ing in he labo ma ke (e.g. Ali, Kulik, Me z 2011). A he same ime, women inc easingly ad ance in o manage ial anks (e.g. Elsass, G a es 1997) and in o jobs o me ly known o be all-male p o essions such as medicine o law (e.g. No g en 2010). Fu he mo e, he indi idual wo king li e span has p olonged since employees en e he labo ma ke a a younge age and s ay much longe be o e becoming eligible o e i emen (e.g. G und, Wes e gaa d-Nielsen 2008). E en ually, cul u al and e hnic backg ound he e ogenei y is e e mo e inc easing in line wi h global mig a ion and human capi al mobili y (e.g. A uc e al. 2015). So a , he e a e no clea indica ions ha he ongoing social and demog aphic ends will slow down o e en e e se in he nea u u e (e.g. Dead ick, S one 2009, Tsui, Gu ek 1999). Gi en he de elopmen owa ds a mo e he e ogeneous wo k o ce, i is o g owing im- po ance o unde s and how di e si y issues a ec p ocesses and p oduc i i y wi hin o - ganiza ions and eams. Acco ding o Ba sade e al. (2000), esea ch on he “cos s and ben- e i s o di e si y in he wo kplace has been going on a a igo ous pace o e he las wo decades and mo e” (Ba sade e al. 2000, p. 802). As a consequence, he amoun o s udies on di e si y app oxima ely doubles e e y i e yea s, o ins ance schola s published 134 Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 12 s udies in 2003 compa ed o only 19 s udies back in 1988 (Ha ison, Klein 2007). Howe - e , he exis ing esea ch on di e si y in he wo kplace is con using and disappoin ing in a way ha “cumula i e indings abou he consequences o wi hin-uni di e ences ha e been weak, inconsis en , o bo h” (Ha ison, Klein 2007, p. 1199). Di e si y esea ch is mainly guided by wo opposing esea ch adi ions ha clea ly demons a e he ambi alen na u e o di e si y ha is o en conside ed a “double-edged swo d” (Milliken, Ma ins 1996, p. 403). On he one hand, a subs an ial body o he li e a- u e s a es ha wi hin eams he e ogenei y may induce social ca ego iza ion ha in u n implies sub-g oup o ma ion (in-g oup s. ou -g oup) and eam con lic and, as a conse- quence, esul in lowe eam pe o mance. Hence, homogenei y in eams is a gued o be p e e able o he e ogenei y wi h ega d o eam pe o mance. The mos well-known con i- bu ions o his concep a e exp essed in simila i y-a ac ion pa adigm as well as social ca ego iza ion heo y (e.g. Taj el 1982, Taj el, Tu ne 1986, Tu ne 1987). On he o he hand, schola s suppose ha eam p oduc ion may bene i om he e ogeneous cogni i e abili ies, expe iences, skills and ne wo k ies o eam membe s. These may be comple- men s and, he e o e, cons i u e a b oad se o esou ces a ailable o he eam. As a esul , he e ogenei y in eams is easoned o be ad an ageous o e homogenei y ega ding eam pe o mance. These conside a ions a e summa ized in in o ma ion p ocessing pe spec i e (e.g. Ho man, Maie 1961, Cox, Blake 1991) and economic heo y alike (e.g. Lazea 1999). In his pape , we conside bo h heo e ical app oaches o de i e ou main esea ch hypo h- eses on he di e si y-pe o mance ela ionship wi hin o ganiza ional wo k g oups. We aim a con ibu ing new insigh s o he wide ield o di e si y esea ch by means o a comp e- hensi e me a-analysis. Mo eo e , we es o wo po en ial mode a ing e ec s on he di- e si y-pe o mance ela ionship – eam size and eam ype. Bo h ha e been iden i ied as being in luen ial by p e ious esea ch (e.g. S ewa 2006, Ho wi z, Ho wi z 2007, De ine 2002). Fo his pu pose, we quan i a i ely summa ize he las wen y yea s o empi ical esea ch on di e si y in o ganiza ional wo k eams ( om 1994 o 2014). Based on a ho - ough li e a u e e iew, we iden i y 242 empi ical s udies on he link be ween di e si y and pe o mance in o ganiza ional eamwo k se ings. Following he adi ion o p e ious me- a-analyses on wo kg oup di e si y (e.g. Bell 2007, Bell e al. 2011, De D eu, Weinga 2003, Ho wi z, Ho wi z 2007, S ewa 2006, Webbe , Donahue 2001), we apply s ic in- Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 13 clusion c i e ia o manage he as amoun o li e a u e. E en ually, we include a o al o 66 indi idual samples om 63 s udies in o he me a-analysis. O e all, ou indings con i m he ambi alen na u e o di e si y in eams. We iden i ied he impac o di e si y on eam pe o mance o depend p ima ily on he di e si y a ibu e being ei he highly o less ask- ela ed. On he one hand, we obse e a nega i e ela ion- ship be ween eam pe o mance and he e ogenei y along less ask- ela ed a ibu es (gende and age). These indings la gely suppo social ca ego y heo y and simila i y-a ac ion pa adigm. On he o he hand, we obse e he e ogenei y along highly ask- ela ed a ibu es (educa ional backg ound) o be posi i ely linked o eam pe o mance. This inding la gely suppo s economic assump ions and in o ma ion p ocessing pe spec i e. Fu he mo e, we obse e bo h eam size and eam ype o mode a e he di e si y-pe o mance ela ionship. Wi h inc easing eam size he in luence o di e si y on pe o mance dec eases o less ask- ela ed a ibu es and inc eases o highly ask- ela ed a ibu es. Conce ning eam ype, we obse e a posi i e mode a ing e ec o op managemen eams (TMT) as well as esea ch and de elopmen (R&D) eams on almos all di e si y a ibu es i espec i e o he ask- ela edness. In con as o his, wo k eams nega i ely mode a e all di e si y- pe o mance ela ions. All esul s will be discussed and in e p e ed in de ail h oughou his s udy. This pape p oceeds as ollows. Sec ion wo will discuss he di e si y cons uc , i s dimen- sions, he mos common he e ogenei y measu es as well as a de ini ion o di e si y as un- de s ood in his pape . Subsequen ly, sec ion h ee will p esen he wo compe ing heo e i- cal concep s and de i e he main esea ch hypo heses o his pape . The me a-analy ic p o- cedu e applied in his s udy will be desc ibed in sec ion ou . E en ually, sec ion i e p e- sen s he esul s o he me a-analysis while sec ion six concludes and o e s implica ions o p ac ical use and u u e esea ch. 2.2 Di e si y Despi e he ele ance o di e si y in he ecen academic discou se, he e is a lack o bo h a common unde s anding and a unique de ini ion o he unde lying cons uc (e.g. Guzzo, Dickson 1996). This becomes e en mo e appa en conside ing he high numbe o syno- nyms ha a e used in e changeable o he umb ella e m ‘di e si y’ by schola s: “’dispe - sion’, ‘inequali y’, ‘wi hin-g oup a iabili y’, ‘(dis)ag eemen ’, ‘consensus’, ‘he e ogenei- y’, ‘homogenei y’, ‘de ia ion’, ‘di e ence’, ‘dis ance’, ‘ ela ional demog aphy’, Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 14 ‘sha edness’ and mo e” (Ha ison, Sin 2006, p. 195). Since each sepa a e de ini ion is linked wi h a sligh ly di e en unde s anding and me ic, he lack o a common di e si y concep seems o be e en mo e obs uc i e. Hence, any compa ison ac oss s udies is com- plica ed and needs pa icula a en ion due o he ac ha esea che s do no necessa ily mean he same phenomenon when e e ing o di e si y. In gene al, h ee main inconsis encies ha e been iden i ied in di e si y esea ch: he lack o a unique cons i u i e de ini ion, he high a ie y wi hin di e si y a iables and he incon- sis en use o di e si y measu es (Ha ison, Sin 2006). This pa ag aph add esses hese in- consis encies, i s by gi ing a clea de ini ion o di e si y as unde s ood h oughou his pape , and second by discussing hose measu es ha a e mos commonly used o assess wo k g oup he e ogenei y. 2.2.1 De ini ions o Di e si y In many cases, s udies on di e si y ocus exclusi ely on how di e si y a ec s p ocesses and ou comes o he uni s unde obse a ion. Howe e , a con o e sial examina ion o he di e si y cons uc i sel is seldom ound in he li e a u e. Thus, i comes as no su p ise ha he e is no join ag eemen on how di e si y can be de ined. Ins ead, li e a u e p o ides a ple ho a o di e gen a emp s ha somehow dilu e he unde s anding o di e si y (Ha i- son, Sin, 2006). A pionee ing de ini ion o di e si y in i s psychological, sociological and economic sense has been p oposed in a seminal wo k by Blau (1977). Acco ding o his unde s anding, he - e ogenei y on a gi en a ibu e depends on i s dis ibu ion: “ he la ge he numbe o g oups and he mo e e enly dis ibu ed he popula ion is di ided among hem, he g ea e is he e ogenei y” (Blau, 1977, p. 9). Howe e , his de ini ion equi es a iables o allow o ank-o de ing and, hus, is s ic ly associa ed wi h s a us o powe hie a chy. Schola s e e since ha e b oadened his na ow de ini ion by p o iding mo e ex ensi e di e si y concep s (Ha ison, Sin, 2006). The undamen al ques ion emains he same: how is max- imum di e si y de ined. While he idea o minimum di e si y simply e lec s pe ec ho- mogenei y o a g oup in e ms o a gi en a ibu e, maximum di e si y seems o ha e a wo- olded dimension. On he one hand, le us assume a g oup wi h all membe s ha ing di e en cha ac e is ics along a pa icula a ibu e. This g oup may be assessed o each he maximum deg ee o di e si y. On he o he hand, conside a g oup ha can be di ided in o wo subg oups ha a e clea ly sepa a ed along a ew bu ex emely opposing cha ac- Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 15 e is ics. This g oup may be named maximal di e se as well. Mo eo e , one e en migh assume maximum di e si y o appea in g oups wi h one membe clea ly su passing he o he s along a gi en cha ac e is ic. Howe e , di e si y li e a u e has no pu much a en ion on he cla i ica ion o hese ques ions so a (Ha ison, Klein 2007). In his me a-analysis, we unde s and di e si y acco ding o a de ini ion ecen ly p oposed by Ha ison and Klein (2007). Add essing he ambigui y o exis ing ideas on di e si y, hey dis inguish be ween h ee ypes o g oup he e ogenei y: sepa a ion, a ie y and dispa - i y (see Figu e 1.1). Fi s , di e si y as sepa a ion exp esses ho izon al di e ences in posi- ion wi hin a g oup along a ibu es such as a i udes, alues o opinions. Second, di e si y as a ie y desc ibes di e gen ca ego ies among eam membe s on a gi en a ibu e such as in o ma ion o educa ional backg ound. Thi d, di e si y as dispa i y assesses e ical di - e ences wi hin a g oup on alued social asse s such as pay o hie a chy. In gene al, he au ho s ega d di e si y as “ he dis ibu ion o di e ences among he membe s o a uni wi h espec o a common a ibu e, X” (Ha ison, Klein 2007, p. 1200). Figu e 2.1: Di e si y as Sepa a ion, Va ie y and Dispa i y Sou ce: Ha ison, Klein (2007), p. 1202. Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 16 2.2.2 Dimensions o Di e si y I espec i e o i s ype, di e si y is assumed o be a ibu e-speci ic. Pu di e en ly, a eam is no di e se pe se, bu only ega ding one o mo e clea ly speci ied eam membe cha - ac e is ics (Ha ison, Klein, 2007). Any human being is a unique combina ion o ea u es such as age, gende , cul u al backg ound, educa ion, o ganiza ional enu e e c. In esponse o he la ge numbe o po en ial di e si y a ibu es, schola s ypically apply clus e ing in o de o inc ease s uc u e and manageabili y o da a. The mos undamen al clus e may be a dis inc ion be ween easily obse able a ibu es o su ace le el di e si y (age, sex, and ace) and less isible a ibu es o deep le el di e si y ( enu e, educa ion and unc ion) (e.g. Ha ison, P ice, Bell 1998, Milliken, Ma ins 1996). Ano he commonly used so ing o di e si y a ibu es has been used by Polze , Mil on and Swann (2002) who dis inguish be ween demog aphic (age, sex, ace, ci izenship) and unc- ional ea u es (p e ious deg ee, p e ious job unc ion, MBA concen a ion). A sligh ly di e en logic is p oposed by Pelled (1996) as well as Pelled, Eisenha d and Xin (1999): They classi y a ibu es o be ei he ask- ela ed (educa ion, o ganiza ional enu e, g oup enu e, unc ional backg ound) o non- ask ela ed (age, gende , ace). In his s udy, we ollow hei ecommenda ion by di e en ia ing be ween less ask- ela ed and highly ask- ela ed di e si y a ibu es. 2.2.3 Measu es o Di e si y Usually, eam di e si y is assessed based on s anda dized he e ogenei y measu es a he g oup-le el. The ou mos equen ly used measu es include he simple wi hin-g oup s anda d de ia ion (s.d.), Blau’s (1977) index o he e ogenei y, Allison’s (1978) coe icien o a ia ion (CV) and Teachman’s (1980) index o he e ogenei y. Wi hin-g oup s.d. is one o he basic s a is ical measu es o assess wi hin-g oup a ia ion in he dis ibu ion o a gi en a ibu e. I eaches i s maximum in dis ibu ions ha a e bi- modal, i.e. he sample spli in hal on a gi en a ibu e (Ha ison, Sin 2006). The Blau (1977) index is one o he mos widely used measu es o assess di e si y o ca ego ical a iables such as gende o age (e.g. Timme man 2000). Maximum he e ogenei y is eached when he p opo ions o all possible ca ego ies a e equally dis ibu ed. The e o e, he Blau index minimizes wi h o al homogenei y and app oaches one wi h inc easing he - e ogenei y (Ha ison, Sin 2006). Teachman’s (1980) index may ange om ze o o posi i e in ini y depending on he numbe o ca ego ies o any (ca ego ical) a iable. I is equal o Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 17 ze o in si ua ions whe e all eam membe s belong o he same social ca ego y and inc eases wi h he numbe o di e en ca ego ies in he eam (Ha ison, Sin 2006). Allison’s CV (Allison 1978) is p obably he mos commonly used measu e o assess wi hin-g oup di e - si y (e.g. Williams, O’Reilly 1998). I is calcula ed by di iding he wi hin-g oup s anda d de ia ion by he g oup mean. In con as o o he di e si y measu es, CV does no maxim- ize when he a ie y o ca ego ies in he g oup inc eases bu , likewise o s.d., g ows wi h he magni ude o con as s in he dis ibu ion o a gi en a ibu e. I eaches maximum al- ues in eams wi h only one membe being di e en o he o he s (Ha ison, Sin 2006). The o mulas o calcula ing hese he e ogenei y measu es a e p esen ed in Table 2.1. Howe - e , all measu es sha e a join concep ual weakness as hey a e all p one o small eam bias (Bedeian, Mossholde 2000) as well as dis o ions a ising om di e gen eam sizes in only one sample (Biemann, Ke ney 2010). Table 2.1: Fo mulas o he Calcula ion o He e ogenei y Measu es He e ogenei y measu e Fo mula Allison’s (1978) Coe icien o Va ia ion 𝐶𝑉= √∑(𝑥𝑖− 𝑥)2 𝑁 𝑁 𝑖=1 𝑥 Blau’s (1977) Index 𝐵=1 − ∑𝑝𝑘2 𝑁 𝑘=1 S anda d De ia ion 𝑆𝐷= √∑(𝑥𝑖− 𝑥)2 𝑁 𝑁 𝑖=1 Teachman’s (1980) Index 𝑇= −(∑𝑝𝑘ln (𝑝𝑘)) 𝑁 𝑘=1 No e. N = o al numbe o eams; xi = indi idual le el cha ac e is ics; x = mean cha ac e is ics o he eam; pk = p opo ion o eam membe s in he k h ca ego y. Apa om he g oup-le el he e ogenei y measu es p esen ed abo e, he e a e o he ways o concep ualize he e ogenei y. Fi s , ela ional di e si y desc ibes an indi idual’s (dis)simila i y wi h his eam ma es and, hus, measu es di e si y a he indi idual le el (e.g. O’Reilly, Caldwell, Ba ne 1989, Tsui, Egan, O’Reilly 1992, Ha ison, P ice, Bell Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 24 e is ics such as age, gende o cul u al backg ound o nega i ely mode a e he link be- ween g oup di e si y and pe o mance. Findings by S ewa (2006) suppo his a gumen . He con i ms he posi i e mode a ing e ec o op-managemen eams and p ojec eams while obse ing a nega i e e ec o p oduc ion eams. Being awa e o he cogni i e na- u e o eam p ocesses in TMT and p ojec eams in con as o he epe i i e and s anda d- ized wo k o p oduc ion eams, hese esul s come as no su p ise. Based on hese conside - a ions, we p edic : H2a: Team size nega i ely mode a es he ela ionship be ween di e si y and eam pe o mance o less ask- ela ed demog aphics. H2b: Team size posi i ely mode a es he ela ionship be ween di e si y and eam pe o mance o highly ask- ela ed demog aphics. As discussed abo e, eam size migh mode a e he di e si y-pe o mance ela ionship simply o s a is ical easons since he mos commonly used he e ogenei y measu es (e.g. Blau index o he e ogenei y, coe icien o a ia ion, Teachman index) a e p one o biases a ising ei he om small eam size (Bedeian, Mossholde 2000) o om eam size a ying wi hin he sample (Biemann, Ke ney 2010). The necessi y o es o s a is ical a e ac s in he di e si y-pe o mance ela ionship a i ms us in assessing eam size as a po en ial mode a o o he di e si y pe o mance ela ionship. Team ype I is a well-es ablished ac in he eam li e a u e ha eam pe o mance and he e iciency o eamwo k is highly dependen on he ype o he eam (e.g. Cohen, Bailey 1997, Thyle- o s, Pe sson, Hells öm 2005). A he same ime, he e is b oad e idence ha eam ype mode a es he di e si y-pe o mance ela ionship (e.g. Ho wi z 2005, S ewa 2006). Team esea ch applies mul iple ways o dis inguish be ween eam ypes. Fo ins ance, Joshi and Roh (2009) di e en ia e eams along he longe i y o coope a ion. They assume isible di e si y (e.g. age, gende ) o be o less signi icance in eams designed o long- e m coop- e a ion (such as TMT o wo k g oups deeply es ablished in he o ganiza ion), while in isi- ble di e si y (e.g. alue, pe sonali y) is posi i ely linked o leng h o coope a ion (Ha ison e al. 2002). Ano he s and o li e a u e dis inguishes eams along he asks hey pe o m (Sunds om e al. 2000). In his s udy, we ollow a eam ypology p oposed by De ine (2002). Fi s , in ellec ual wo k eams (such as TMTs) a e assigned complex asks in o en unce ain si ua ions and Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 25 a e unning non-s anda dized p ocesses. Second, design eams (such as R&D) equi e c ea- i i y and echnical inno a ion o ul ill hei asks. Thi d, physical wo k eams (such as p oduc ion eams, pe o mance eams o se ice eams) simply ollow s anda dized asks. We aim a es ing o mode a ing e ec s along hese h ee di e gen eam ypes. In gene al, TMTs and R&D eams a e assumed o be he e ogeneous on ask- ela ed a ibu es (i.e. unc ional and educa ional backg ound) as he espec i e asks equi e a b oad base o a - ying knowledge and expe ience. A he same ime, hese eams a e likely o be homogene- ous on non- ask ela ed cha ac e is ics (i.e. age, gende , cul u al backg ound). In con as o his, physical wo k eams a e ound o be mo e he e ogeneous in hei composi ion (e.g. Ho wi z, Ho wi z 2007, De ine 2002). As a consequence o hese conside a ions, we de- i e wo hypo heses on he impac o less ask- ela ed as well as highly ask- ela ed he e o- genei y on eam pe o mance: H3a: The p edic ed nega i e ela ionship be ween less ask- ela ed he e ogenei y and eam pe o mance will be weake in R&D eams and TMT compa ed o o he eam ypes. H3b: The p edic ed posi i e ela ionship be ween highly ask- ela ed he e ogenei y and eam pe o mance will be s onge in R&D eams and TMT compa ed o o he eam ypes. 2.4 Me a-Analysis App oach In he ollowing, we use he me a-analysis app oach o es o he abo e s a ed hypo heses on he di e si y-pe o mance ela ionship as well as o examine po en ial mode a ing e - ec s o eam size and eam ype. Li e a u e e iew A ho ough li e a u e e iew was conduc ed in o de o iden i y s udies on wo kg oup di- e si y. We ocus on esea ch published be ween 1994 and 2014 since we aim a quan i a- i ely summa izing he las 20 yea s o di e si y esea ch wi hou eplica ing ea lie me a- analyses ha ha e al eady comp ehensi ely e iewed ea lie wo ks (e.g. Bowe s, Pha me , Salas 2000, Webbe , Donahue 2001). We loca ed ele an s udies on wo kg oup di e si y o po en ial inclusion h ough se e al sea ch s a egies. Fi s , we e iewed ecen me a- analyses and na a i e e iews by Bowe , Pha me , Salas (2000), Williams, O’Reilly (1998), Ho wi z, Ho wi z (2007), S ewa (2006), Webbe , Donahue (2001), Bell (2007), Bell e al. (2011), De D eu, Weinga (2003) and De Chu ch, Mesme -Magnus (2010). Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 26 Second, we elec onically sea ched he online da abases Scopus, Business Sou ce Comple e and PsycINFO wi h combina ions o he ollowing keywo ds: eam, g oup, di e si y, he - e ogenei y, homogenei y, di e se, ( eam) composi ion, demog aphy, demog aphic di e si- y, age, gende , emale, male, enu e, educa ion, unc ion, cul u e, ace, e hnici y, pe o - mance, ou come, e ec i eness and op-managemen - eam. Thi d, he da abase sea ch was supplemen ed wi h a manual sea ch o in-p ess a icles in bo h op- ie ed jou nals and jou nals wi h a pa icula ocus on g oup esea ch including Adminis a i e Science Qua - e ly, O ganiza ion Science, Jou nal o Applied Psychology, Academy o Managemen Jou nal, Academy o Managemen Re iew, Pe sonnel Psychology, G oup and O ganiza- ional S udies/Managemen , O ganiza ional Beha io and Human Decision P ocesses, Jou nal o O ganiza ional Beha io , Jou nal o Managemen , Small G oup Resea ch, G oup Dynamics and Human Rela ions. We only conside pee - e iewed, English- speaking jou nal a icles, hus excluding disse a ions, con e ence pape s, manusc ip s o wo king pape s. We a e awa e o he isk ha his app oach migh pu ou esul s in o ques ion due o in o ma ion loss as a consequence o publica ion bias. Ye , we p oceed as desc ibed o be able o deal wi h he ga bage-in-ga bage-ou issue (e.g. Egge , Smi h, S e ne 2001) ha seems o pa icula impo ance o us in he wide and complex ield o di e si y esea ch. Howe e , since published indings ea u e posi i e as well as nega i e ela ionships and do also epo non-signi ican esul s a some ins ances, we eel con iden o ca ch he subs an ial con ibu ions wi hou losing ele an in o ma ion. S udy abs ac s we e e iewed o sui able esea ch on he ela ionship be ween demog aphic di e si y a iables and bo h eam pe o mance and c ea i i y. This li e a u e sea ch yields a o al o 242 s udies ha we e u he examined o po en ial inclusion in he me a-analysis. In o de o be included, a s udy has o mee se e al inclusion c i e ia. Fi s , we ocus ex- clusi ely on empi ical esea ch ha s udies a leas one o he hypo hesized ela ionships be ween demog aphic di e si y a iables and wo kg oup p oduc i i y. In o he wo ds, we exclude s udies ha obse e ou comes such as c ea i i y, inno a i eness, sa is ac ion o g oup cohesion. Addi ionally, we e ain om including agg ega ed measu es o di e si y such as demog aphic di e si y, which is o en used o combine age, gende and cul u al backg ound/e hnici y in only one a iable. Mo eo e , we do no conside s udies ha ob- se e psychological di e si y a iables such as cogni i e abili y o pe sonali y. Second, we only include s udies ha unde s ood wo k g oups as eams ha mee he ollowing c i e ia: Team membe s need o sha e a common goal, ha e simila o e en iden ical wo king asks Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 27 and be wo king in e dependen ly. Mo eo e , eams should be embedded in an o e all so- cial sys em such as an o ganiza ion and be ega ded as a eam by i s membe s as well as by ex e nal pe sons (Guzzo, Dickson 1996). Thi d, since mixed le els o analysis a e inap- p op ia e o calcula ing sample-weigh ed e ec s (Beal e al. 2003), we ocus on s udies which assess di e si y on he eam-le el and d op hose s udies ha epo indings on he indi idual le el. Fu he mo e, we ollow sugges ions by Ka zenbach and Smi h (2005) and limi he maximum eam size o wen y- i e pe sons. Membe s o la ge eams a e usually no subjec o ask in e dependency and do no adequa ely in e ac as a eam, ye bo h ac- o s a e necessa y o s udy di e si y e ec s on eam pe o mance. Minimum eam size is wo pe sons. Fou h, we conside only s udies ha assess di e si y based on objec i e measu es such as wi hin-g oup s.d., Blau’s (1977) index o he e ogenei y, Allison’s (1978) CV and Teachman’s (1980) index o he e ogenei y. Subjec i e measu es o g oup he e o- genei y a e mos likely biased due o indi idually di e ing pe cep ions o cu en g oup se ings o cul u al a ia ions and a e, he e o e, no included in his me a-analysis. Addi- ionally, we exclude s udies ha unde s and di e si y ei he as ela ional di e si y o as aul lines since bo h app oaches a e no compa able o he di e si y concep applied he e. Fi h, s udies ha e o s em om eal-li e o ganiza ional se ings. This excludes expe i- men al esea ch, case s udies and simula ion esea ch design. Mo eo e , we concen a e on eams in wo king en i onmen s. The e o e, we exclude s udies ha a e based on s uden wo k g oups, spo eams o o he non- ask ela ed o ganiza ional se ings. We do so in o de o gua an ee high p ac ical ele ance o implica ions d awn om his me a-analysis. E en ually, we exclude s udies ha we asses o su e me hodological weaknesses and/o a e no compa able o he o e all se o s udies. Applying hese s ic inclusion c i e ia is necessa y o gua an ee be ween-s udy compa abili y. A o al o 179 s udies ou o 242 ini- ially iden i ied s udies ailed o mee he inclusion c i e ia. This lea es a inal numbe o 63 s udies o be included in ou analysis. Two s udies con ibu e wo o mo e samples. As a esul , we yield a inal da a se including 66 samples wi h in o ma ion on a o al o 12,478 eams. Di e si y a iables The independen a iables in es iga ed in his me a-analysis include he demog aphic ea- u es o age, gende , cul u al/e hnical backg ound, o ganiza ional enu e, unc ional and educa ional backg ound as well as educa ion le el. To mee he inclusion c i e ia, di e si y a iables ha e o be agg ega ed o he eam-le el by means o an app op ia e measu e o Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 28 g oup-le el he e ogenei y, e.g. wi hin-g oup s.d., he s anda d measu es Blau’s (1977) in- dex o he e ogenei y, Allison’s (1978) CV and Teachman’s (1980) index o he e ogenei y. S ill, we allow o o he me hods o assess di e si y objec i ely – such as he He indal index, he Gini index o he simple pe cen age sha e – as long as i s choice seems easona- ble and compa abili y wi h o he measu es is gi en. Table A.1 in he appendix lis s he equency dis ibu ions o di e si y measu es used in p ima y s udies. Following Pelled (1996) in clus e ing di e si y a ibu es along he dimension o ask- ela edness, we ca ego ize age, gende and cul u al backg ound/e hnici y ( e e ed o as cul u e in he ollowing) as less ask- ela ed, while we a e o ganiza ional enu e, unc ion- al and educa ional backg ound as well as educa ion le el o be highly ask- ela ed. While he me ic o age, gende , cul u e and o ganiza ional enu e di e si y is qui e sel - explana o y and only a ies sligh ly be ween p ima y s udies (e.g. age and enu e measu ed con inuously o in ca ego ies), h ee di e si y a iables need u he conside a ion. Func- ional di e si y is unde s ood as unc ional expe ience gained du ing he ca ee , he domi- nan unc ion, o ganiza ional oles and he p ima y p o essional o ien a ion. Educa ional backg ound di e si y is concep ualized as he majo le el o he specializa ion o s udies. Educa ion le el di e si y is assessed ei he as he numbe o yea s o ( o mal) educa ion o as he highes educa ional le el achie ed. Ou come a iables The dependen ou come a iables include di e gen ypes o eam pe o mance assessed along ei he objec i e o subjec i e pe o mance measu es. Objec i e pe o mance measu es include inancial key igu es such as e u n on equi y (ROE), e u n on asse s (ROA) and e u n on sales (ROS) as well as measu es o eam e ec i eness such as p oduc i i y, p o i abili y and goal achie emen . Subjec i e pe o mance measu es a e based ei he on a eam’s sel -assessed pe o mance app aisals o on ex e nal a ings om manage s, supe iso s o o he pe sons ou side he eams. Mode a o a iables E en ually, in o de o es o po en ial mode a ing e ec s in he di e si y-pe o mance ela ionship we include in o ma ion on a e age eam size and eam ype. We include mean eam size when epo ed in he ini ial s udy – usually measu ed as simple headcoun o eam membe s. S udies ha do no epo a e age eam size a e excluded om he mode a- o analysis. Rega ding eam ype, we clus e eams along De ine’s (2002) ypology ei he Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 29 o be TMT, wo k eam, R&D eam o mixed eams. While TMT (execu i e) and R&D eams (design) ma ch De ine’s (2002) in ellec ual wo k eam clus e , wo k eams e e o he physical wo k eam clus e . We unde s and “wo k eams” o include hose eams in o- duced as wo k eams in he ini ial s udies, bu also eams ha con inuously wo k oge he on he same pe sis ing asks, e.g. eams labeled as school s a , sales eams o b anch eams. S udies on wo k eams ha a e engaged in speci ic ypes o coope a ion (e.g. i ual eams) o subjec o ask p o iles ha change o e ime (e.g. en u e o p ojec eams) a e classi ied as “mixed eams”. The same is ue o s udies which include di e gen eam ypes wi hou epo ing esul s sepa a ely. Again, s udies ha do no p o ide he app op i- a e in o ma ion on eam ype a e excluded om mode a o analysis a his ins ance. Me a-analy ic p ocedu e Fo his me a-analysis, we chose he p oduc -momen co ela ion coe icien ( ) o be he p ima y e ec size index, because we exclusi ely ocus on obse a ional esea ch and do no include expe imen al indings. Howe e , p oduc -momen co ela ion coe icien s may en ail se ious s a is ical di icul ies such as p oblema ic s anda d e o o mula ion (Lipsey, Wilson 2001) and excessi e Type I e o a es (Alexande , Scozza o, B odkin 1989). In esponse o hese undesi able s a is ical p ope ies, we ollow sugges ions by Lipsey and Wilson (2001) as well as Hedges and Olkin (1985) and co ec co ela ion coe icien s by Fishe ’s Z - ans o ma ion. Wi h ega d o model speci ica ion, we chose andom e ec models as in oduced by Hedges and Olkin (1985) since his app oach allows he e ogenei y in he e ec size o a ise om wo sou ces, subjec le el sampling e o and a iabili y o e ec s andomly dis ibu ed along s udies (Lipsey, Wilson 2001). To es o he deg ee o p ecision o he mean e ec size es ima es, we calcula ed he 95% con idence in e al (CI) a ound he popula ion co ela ion. A CI ha does no include ze o indica es he mean e ec size o be signi ican (Lipsey, Wilson 2001, Whi ene 1990). I he 95% CI does no include ze o, we ollow Lipsey and Wilson (2001) in calcula ing z- es s o check whe he he mean popula- ion e ec size is signi ican a p≤α (e.g. 10%-signi icance). In o de o es o he homo- genei y o he e ec size, we apply chi-squa e dis ibu ed Q s a is ics (Hedges, Olkin 1985). In gene al, homogenei y o e ec sizes would imply ha indi idual e ec sizes di - e om he es ima ed popula ion mean e ec size only by sampling e o (Lipsey, Wilson Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 30 2001). 9 In o he wo ds, ejec ing Q s a is ics would deno e he e ogeneous e ec size dis i- bu ions wi h wo main implica ions. Fi s , he e ogenei y a o s he use o andom e ec s a he han ixed e ec s. Since we obse e he majo i y o Q s a is ics o con i m he e oge- nei y in ou da a, ou ini ial choice o andom e ec models is s a is ically suppo ed. Sec- ond, he e ogenei y sugges s ha andomly dis ibu ed a iables may mode a e he a iance in indi idual e ec sizes. As a consequence, we es o he p oposed mode a ing e ec s o eam size and eam ype by applying he De Simonian and Lai d (1986) andom-e ec s me a- eg ession me hods app oach wi h Knapp and Ha ung (2003) es s o e ec es i- ma es a iance. QE is he co esponding esidual he e ogenei y s a is ic ha indica es mod- el signi icance (Lipsey, Wilson 2001). 2.5 Resul s We s a he discussion o ou indings by examining he ela ionship be ween di e si y a iables and eam pe o mance as hypo hesized in H1a o H1g (Table 2.2). Subsequen ly, we p esen ou indings wi h espec o he mode a ing e ec s o a e age eam size (H2a, H2b, Table 2.3) as well as eam ype (H3a, H3b, Table 2.4). Fo each analysis, we epo he numbe o co ela ions om independen samples included in he me a-analysis (k), he o al numbe o eams (N), he e ec size gi en as he co ec ed popula ion co ela ion ( z) (sample-size weigh ed h ough Fishe ’s Z- ans o ma ion), he s anda d e o o he co - ec ed popula ion co ela ion ( z(se)), he lowe and uppe bounds o he 95% con idence in e al (95% CI), Coch an’s Q as homogenei y s a is ics (Q) wi h he espec i e deg ee o eedom (d ), and e en ually he I2 s a is ic which p esen s he pe cen age o be ween-s udy a ia ion a ibu able o he e ogenei y (I2). Main esul s on he di e si y-pe o mance ela ionships (hypo heses H1a o H1g) a e e- po ed in Table 2.2. Fo es plo s ha illus a e he co ela ions a s udy con ibu es o he me a-analysis a e p esen ed o each di e si y a iable in Figu es A.1 o A.7 in he appen- dix. In suppo o Hypo hesis H1a, we obse e a nega i e e ec size ega ding he link be ween age he e ogenei y and eam ou come ( z = -0.052, 95% CI: -0.109 | 0.005). Since ze o is included in he 95% CI, we calcula e he z- es o e ec size (ES) which allows o 9 As an al e na i e o Q-s a is ics we calcula e he I2 measu e ha gi es he p opo ion o o al a ia ion in mean e ec size a ibu able o be ween-s udy he e ogenei y (Higgins, Thompson 2002). I is calcula ed as 𝐼²=100% 𝑥 𝑄−𝑑𝑓 𝑄. As ule o humb, one can say ha an I² alue a ound 25% indica es low he e oge- nei y, while alues a ound 50% and 75% can be ansla ed o mode a e and high le els o he e ogenei y, espec i ely (Higgins e al. 2003). Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 31 ejec he null-hypo hesis (ES=0) and, hus, con i m H1a a he 10%-le el o signi icance. In o he wo ds, he mo e di e se eam membe s a e wi h ega d o age he lowe is o e all g oup p oduc i i y. Hypo hesis H1b p edic ed a nega i e ela ionship be ween gende he - e ogenei y and eam pe o mance. Ou indings con i m he expec ed link ( z =-0.036, 95% CI: -0.076 | 0.004). Again, z- es s o e ec size (ES) allow o con i ma ion o H1b (a he 10%-le el). This indica es ha gende he e ogenei y in eams is de imen al o eam pe - o mance. Rega ding he ela ion be ween cul u al di e si y and eam pe o mance, we do no ind any suppo o he expec ed nega i e ela ionship (H1c). Cul u al di e si y does no a ec eam p oduc i i y signi ican ly. In summa izing indings on he less ask- ela ed di e si y a iables, esul s o bo h age and gende p o ide suppo o social ca ego iza- ion heo y as well as he simila i y-a ac ion pa adigm since di e si y along hese a ib- u es is iden i ied o a ec o e all eam pe o mance nega i ely. Table 2.2: Rela ionship be ween Di e si y and O e all Team Pe o mance Di e si y k z z(se) 95% CI Q d I2 Lowe Uppe Age 25 -0.052* 0.029 -0.109 0.005 88.79*** 24 73.00% Gende 30 -0.036* 0.021 -0.076 0.004 57.69*** 29 49.70% Cul u e 20 0.003 0.033 -0.061 0.067 105.98*** 19 82.10% Func ion 27 0.041 0.039 -0.035 0.117 369.37*** 26 93.00% Tenu e 25 0.007 0.031 -0.054 0.069 178.85*** 24 86.60% Educa ion Backg ound 10 0.064* 0.038 -0.011 0.138 48.85*** 9 81.60% Le el 12 -0.006 0.052 -0.108 0.097 43.23*** 11 74.60% No e. k = o al numbe o co ela ion coe icien s me a-analyzed; z = co ec ed popula ion co ela- ion (sample-size weigh ed based e ec size (ES) on Fishe ’s z ans o med co ela ion coe icien s wi h signi icance es o ES=0; z = s anda d e o o he co ec ed popula ion co ela ion; 95% CI = lowe and uppe bound o he 95% con idence in e al; Q = homogenei y s a is ics; d = deg ee o eedom; I2 = pe cen age o be ween-s udy a ia ion due o he e ogenei y. ***p<.01; **p<.05; *p<.1 Wi h ega d o he expec ed posi i e ela ionship o he e ogenei y in highly ask- ela ed a ibu es on eam p oduc i i y, we only ind suppo o educa ional backg ound di e si y as s a ed in H1 ( z = 0.064, 95% CI: -0.011 | 0.138). A signi ican z- es allows us o con- i m H1 a he 10%-le el o signi icance. This inding sugges s eams o pe o m signi i- can ly be e when d awing on a b oad se o di e se educa ional backg ounds. Howe e , he e is no suppo o he p edic ed posi i e ela ionship be ween eam pe o mance and o ganiza ional enu e, unc ional and educa ion le el he e ogenei y (H1d, H1e and H1g). Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 32 The e o e, o e all suppo o he in o ma ion p ocessing pe spec i e is weak and only based on indings wi h espec o educa ional backg ound he e ogenei y. In o de o in es iga e he ela ionship be ween eam he e ogenei y and eam pe o mance in mo e dep h, we sepa a e p ima y s udies depending on he way hey assess eam pe - o mance (Table A.2 in he appendix). While he e a e no ou comes o he han hose dis- cussed abo e in he ields o age, gende , cul u al, unc ional and educa ional backg ound di e si y, we gain addi ional insigh s o enu e di e si y as well as educa ion le el di e si- y. In pa ial suppo o Hypo hesis H1d, we obse e enu e di e si y o be posi i ely ela - ed o eam pe o mance a leas when pe o mance measu es a e based on subjec i e ex e - nal a ings, e.g. by supe iso s ( z = 0.095, 95% CI: 0.021 | 0.106). Con a y o ou expec- a ions as exp essed in hypo hesis H1g, we ound a nega i e ela ionship be ween educa- ional le el di e si y and pe o mance measu ed on ou side a ings (-0.132, 95% CI: - 0.267 | 0.002). All in all, we a e able o con i m he expec ed nega i e link be ween eam pe o mance and bo h age di e si y (H1a) and gende di e si y (H1b). Fu he mo e, we ind suppo o he hypo hesized posi i e ela ionship be ween educa ional backg ound di e si y (H1 ) and eam pe o mance. Addi ionally, we can p o ide pa ial suppo o he expec ed posi i e ela ion be ween eam pe o mance and enu e di e si y (H1d). A he same ime, we ha e o pa ially ejec H1g as we obse e a nega i e ela ionship be ween educa ion le el and pe o mance a ed by ou side s. Subsequen o he discussion o he main esul s o he me a-analyzed di e si y-pe o mance ela ionship, he second se o analyses cen e s on he po en ial mode a ing e ec s o a e age eam size (as s a ed in H2a and H2b) and eam ype (as s a ed in H3a and H3b). Team size In o de o es o he mode a ing e ec o eam size on he di e si y-pe o mance ela- ionship, we applied andom-e ec s me a- eg ession. Resul s a e epo ed in Table 2.3. As s a ed in H2a, we expec eam size o nega i ely mode a e he link be ween he e ogenei y in he less ask- ela ed demog aphic a ibu es o age, gende and cul u e. Me a- eg ession indings la gely suppo eam size o be he p edic ed nega i e mode a o . In pa icula , he co ela ion coe icien s o he di e si y-pe o mance ela ionship educe wi h each addi- ional uni membe ( o age by -0.001, o gende by -0.013 and o cul u e by -0.021). Highly signi ican alues o QE con i m he signi icance o he eg ession models. These Wo k Team Di e si y and he Op imal Composi ion o Teams: A Me a-Analy ic App oach 33 indings may sugges ha coo dina ion and communica ion among eam membe s a e complica ed wi h inc easing eam size. Mo eo e , social ca ego iza ion and con lic may be mo e p onounced in la ge eams. As p edic ed in H2b, we obse e posi i e mode a ing e ec s on he di e si y-pe o mance ela ionship o he ask- ela ed demog aphic a ib- u es enu e (0.003), unc ional (0.013) and educa ional backg ound (0.063) as well as edu- ca ion le el (0.016). These indings sugges ha addi ional eam membe s may con ibu e addi ional (complemen a y) knowledge, skills and expe ience. This seems o be bene icial o g oup pe o mance a leas o ask- ela ed he e ogenei y. O e all, we a e able o con- i m hypo heses H2a and H2b. Table 2.3: Team Size as Mode a o o he Pe o mance-Di e si y Rela ionship Di e si y k z z(se) 95% CI QE d I2 Lowe Uppe Age 19 -0.001 0.008 -0.019 0.017 62.31*** 17 72.72% Gende 21 -0.013 0.005 -0.024 -0.001 35.54** 19 46.54% Cul u e 11 -0.021 0.013 -0.051 0.008 48.11*** 9 81.29% Func ion 15 0.013 0.016 -0.021 0.048 79.24*** 13 83.59% Tenu e 14 0.003 0.022 -0.045 0.051 56.19*** 12 78.65% Educa ion Backg ound 6 0.063 0.049 -0.074 0.2 16.03*** 4 75.05% Le el 11 0.016 0.026 -0.042 0.073 32.86*** 9 72.61% No e. k = o al numbe o co ela ion coe icien s me a-analyzed; z = co ec ed popula ion co ela- ion (sample-size weigh ed based ES on Fishe ’s z ans o med co ela ion coe icien s wi h signi - icance es o ES=0; z = s anda d e o o he co ec ed popula ion co ela ion; 95% CI = lowe and uppe bound o he 95% con idence in e al; QE = homogenei y s a is ics; d = deg ee o ee- dom; I2 = pe cen age o be ween-s udy a ia ion due o he e ogenei y. ***p<.01; **p<.05; *p<.1 Team ype Likewise o eam size, we use me a- eg ession o es o mode a ing e ec s o eam ype on he di e si y-pe o mance ela ionship. Resul s a e epo ed in Table 2.4. In hypo hesis H3a we expec ed he nega i e ela ionship be ween less ask- ela ed he e ogenei y and eam pe o mance o be weake in TMT and R&D eams compa ed o o he eam ypes. Fo all h ee less ask- ela ed di e si y a ibu es (age, gende and cul u e), we obse e he p edic ed nega i e di e si y-pe o mance ela ionship o wo k eams and mixed eams. Howe e , o ou g ea su p ise, esul s o TMT and R&D eams do no show he expec ed sligh ly weake nega i e ela ionship bu , ins ead, e en sugges a posi i e ela ionship o age, gende and cul u e wi h eam pe o mance. In o he wo ds, o TMT and R&D eams Employee Absen eeism: De e minan s in he In e na ional Con ex 40 o e all wo king li e span (e.g. G und, Wes e gaa d-Nielsen 2008). On his accoun he e is a g ea need o educing exis ing wo kplace bu dens ha migh p o oke s ess o illnesses in o de o sus ain employees’ heal h in he long un. De ailed knowledge on he de e mi- nan s o absence may help ackling hese issues. In gene al, absence can be de ined as non-a endance a wo k when scheduled o (e.g. K is- ensen e al. 2006). This idea allows o dis inguish be ween in olun a y (e.g. sickness) o olun a y (e.g. shi king) absence beha io (e.g. Johns 1997, Sagie 1998). In lack o a unique heo y on absence, se e al a emp s o concep ualize absen eeism ha e eme ged o e he yea s wi h he well-known p ocess model o S ee s and Rhodes (1978) being he mos commonly ci ed. In hei adi ion, absence is usually unde s ood as being subjec o a ious pe sonal, o ganiza ional, social and en i onmen al a iables ha in luence bo h mo i a ion and abili y o a end wo k. Much o he (pe sonnel) economics wo k on absen- eeism ela es o ei he neoclassical labo supply models (Allen 1981a/b, Dunn, Youngblood 1986) o e iciency wage heo y (Shapi o, S igli z 1984). Theo y is backed up by expe imen al and empi ical esul s on demog aphics (e.g. Voss, Flode us, Dide ichsen 2001; Ba mby, E colani, T eble 2002), g oup absence no ms (e.g. B adley, G een, Lee es 2007; Bambe ge , Bi on 2007), wo k sa is ac ion (e.g. K is ensen e al. 2006), wo king condi ions (e.g. Dionne, Dos ie 2007), na ional sickness bene i egula ions (e.g. F ick, Malo 2008; Johansson, Palme 2005), employmen p o ec ion (e.g. Riphahn, Thalmaie 2001; Ichino, Riphahn 2005) as well as economic en i onmen and unemploymen (e.g. Leigh 1985) – ye a some ins ances wi h inconsis en indings. In his a icle, we empi ically in es iga e why absence igu es o blue-colla employees a y conside ably be ween di e en p oduc ion si es o a la ge co po a ion al hough em- ployees ace simila ci cums ances in e ms o wo king condi ions, manu ac u ing p ocess- es and inal p oduc a each acili y. In pa icula , we a e in e es ed in he ole o economic and social de e minan s in de e mining employee absence. Mo eo e , we examine he in- luence o wo ke cha ac e is ics and wo king condi ions on absence. Using a hi he o un- a ailable da a se co e ing 160 blue-colla wo k uni s a ou in e na ional p oduc ion si es o a la ge Eu opean au omobile manu ac u e , we o e a comp ehensi e analysis o po en- ial de e minan s o absen eeism including:  social pee in luences ( eam size, eam u no e , he sha e o empo a y em- ployed wo ke s and he sha e o wo ke s su e ing heal h impai men s) Employee Absen eeism: De e minan s in he In e na ional Con ex 41  economic in luences / incen i es (sick pay egula ions, employmen p o ec ion, (un)employmen and p ospe i y le el)  employee cha ac e is ics (age, gende , enu e and heal h)  wo king condi ions (shi sys em) Following he da a analyses, we conduc ed a o al o eigh een in e iews wi h on-si e ex- pe s o u he in e p e ou empi ical indings. Expe s include HR manage s, line manag- e s, wo ke ep esen a i es and shop loo s a . The esul s p esen ed in his pape indica e ha absence a he espec i e co po a ion is a mul i ace ed phenomenon ha only shows ew s a is ical pa e ns ha a e alid a all ou plan s unde obse a ion. Ins ead, all de e minan s examined a e di e en ly a ec ing em- ployee absence a each o he ou p oduc ion si es – de ails a e discussed h oughou chap- e 3.5. In a nu shell, we obse e a posi i e link be ween absence and social pee in luences such as uni size and uni u no e . Fu he mo e, a high sha e o empo a y wo ke s in- c eases absence o pe manen ly employed wo ke s while we do no ind any absence e ec o he sha e o wo ke s su e ing heal h impai men s. The esul s ega ding economic in- luence sugges s ic employmen p o ec ion laws as well as a a o able p ospe i y le el o signi ican ly inc ease absence, whe eas indings o he na ional (un)employmen si ua ion emain inconclusi e. Resul s o na ional sickness bene i s legisla ion ail o each s a is i- cal signi icance. Wi h ega d o wo ke cha ac e is ics, we obse e uni absence o inc ease wi h he sha e o emale employees. Addi ionally, we de ec a posi i e link be ween age and bo h in olun a y as well as olun a y absence. E idence on he ela ion be ween ab- sence and enu e as well as indi idual heal h is somewha inconclusi e. E en ually, esul s on wo king condi ions imply ha absence is highe in a wo-shi and h ee-shi sys ems compa ed o he s anda d one-(day-)shi sys em. In chap e 3.6, p ac ical implica ions o a endance managemen a e conside ed based on he empi ical indings and insigh s gained du ing expe in e iews. Ou s udy con ibu es o he exis ing li e a u e in se e al ega ds. Fi s , we add o he g owing esea ch on de e minan s o absen eeism by o e ing new pe spec i es om inside an o ganiza ion. In pa icula , we bene i om unique da a ha allows o comp ehensi e empi ical es ing. Second, we appea o be among he i s esea che s who a e able o wo k wi h in e na ional da a om wi hin only one o ganiza ion. Since all da a is epo ed on company s anda ds, we can neglec o ganiza ional and in e na ional di e ences in e- Employee Absen eeism: De e minan s in he In e na ional Con ex 42 co ding me hodology ha usually limi in e na ional compa ison (e.g. Eu opean Founda- ion o he Imp o emen o Li ing and Wo king Condi ions 2010). Thi d, ou da a allows us o examine h ee absence measu es. Thus, we a e able o empi ically p oxy o olun- a y s. in olun a y absence. Fu he mo e, by using uni -le el absence da a we add ess a weakness o p e ious absence esea ch ha has ocused almos exclusi ely on indi idual- le el absen eeism and only accoun o uni -le el e ec s in pa ches (e.g. Ren sch, S eel 2003). Howe e , since absence is usually modeled as a social phenomenon (e.g. S ee s, Rhodes 1978, Kaise 1998) i should, by de ini ion, be in luenced by uni -le el pee s. Fi- nally, ou s udy adds o he g owing inside econome ics app oach ha has been i s in- oduced o pe sonnel economics in seminal pape s o Ichniowski, Shaw and P ennushi (1997) as well as Lazea (2000b). The emainde o his pape is o ganized as ollows. The upcoming sec ion p o ides bo h a discussion o he absen eeism concep as unde s ood in his pape as well as a e iew o he exis ing absence esea ch. Sec ion h ee will p esen he da a se and he peculia i ies o he o ganiza ional se ing unde obse a ion. Sec ion ou discusses he es ima ion s a egy while sec ion i e p esen s he empi ical esul s. E en ually, sec ion six concludes and o - e s implica ions o bo h p ac i ione s and u u e academic esea ch. 3.2 Theo e ical F amewo k Resea ch on absen eeism is mul idisciplina y. I mainly o igina es om managemen li - e a u e bu mo e ecen ly a ouse in e es in (labo ) economics and social psychology (Kai- se 1998). Howe e , despi e a subs an ial body o expe imen al and empi ical e idence on absen eeism, a comp ehensi e unde s anding o he absence concep and i s unc ioning is s ill missing. So a , esea ch has me ely ag eed on a common de ini ion o employee ab- sence ha includes all ypes o non-a endance a wo k when o iginally scheduled o, ye excluding scheduled absence ag eed on wi h he employe such as holidays o lex ime lea es (e.g. K is ensen e al. 2006). In o he wo ds, absen eeism is usually assigned o sel - ce i ied o medically-ce i ied sickness absence and ecognized as such by he employe (e.g. Whi ake 2001). In his s udy, we ollow his de ini ion o absence. A “concep ual b eak h ough” (Kaise 1998, p. 81) in de ining a heo e ical amewo k o psychological absence esea ch is he p ocess model in oduced in a now seminal a icle by S ee s and Rhodes (1978). In hei model, absence is unde s ood as an indi idual decision Employee Absen eeism: De e minan s in he In e na ional Con ex 43 based on he abili y and he mo i a ion o come o wo k (see Figu e 3.1). In hei model, he au ho s add ess weaknesses o p e ious con ibu ions in ques ioning he assump ions o job sa is ac ion being he p ima y cause o absence (a comp ehensi e o e iew is gi en by Kaise (1998)). In esponse, S ee s and Rhodes (1978) model absence o be depending on bo h he mo i a ion as well as he abili y o a end wo k which in u n bo h a e di ec ly in luenced by pe sonal cha ac e is ics (e.g. age, gende , enu e). Addi ionally, he mo i a- ion o a end is u he subjec o he indi idual job si ua ion (e.g. job scope, job le el, wo k g oup size) as well as a ious a endance p essu es (e.g. economic condi ions, wo k g oup no ms). The au ho s assume he abili y o a end (e.g. sickness, amily esponsibili- ies) o mode a e he e ec o a endance mo i a ion on absence. La e , he au ho s modi- ied hei i s model in o an ad anced diagnos ic model by including economic and social psychological ac o s ha ha e no been conside ed in hei ini ial model (Rhodes 1990). 13 Figu e 3.1: S ee s and Rhodes (1978) P ocess Model o Employee Absence Sou ce: Illus a ion o S ee s and Rhodes p ocess model (S ee s, Rhodes 1978, p.393). In economic esea ch, he unde s anding o absence di e s om he psychological con- cep s. Much o he economic absence esea ch ela es o wo main heo ies. Fi s , neoclas- sical labo supply models assume absence o occu om indi idual day- o-day labo - 13 Two al e na i e heo ies o absen eeism a e discussed in he psychological absence li e a u e. Fi s , he wi hd awal model which assumes absence o be wi hd awal om un o una e wo k ci cums ances (e.g. Hanisch, Hulin 1990, Po e , S ee s 1973). In o he wo ds, absence may p o ide employees wi h s ess- elie and allows hem o e u n o wo k mo e p oduc i e (e.g. Bachle 1995). Second, he social- psychological app oach ha emphasizes he impo ance o absence no ms wi hin g oups (e.g. Kaise 1998, Ren sch, S eel 2003). Employee Absen eeism: De e minan s in he In e na ional Con ex 44 leisu e choice decisions (e.g. Allen 1981a/b, B own, Sessions 1996, Dunn, Youngblood 1986). The idea is simple: he indi idual u ili y is maximized when he ma ginal a e o subs i u ion be ween income and leisu e equals he wage o e ed by he employe . Howe - e , due o impe ec labo ma ke s and job sea ch cos s, one can assume indi iduals o ac- cep a job o e e en i his cons ain is no ul illed a he con ac ed numbe o wo k hou s. I a a gi en wage he con ac ed numbe exceeds he desi ed numbe o wo k hou s, he indi idual has an incen i e o be absen om wo k (Allen 1981a/b). Ye , his only holds ue as long as he u ili y gains o being absen exceed he indi idual cos s o absen- eeism. These cos s may include he gap be ween egula pay and sick pay. In o he wo ds, employees may use absence o maximize hei indi idual u ili y (Dunn, Youngblood 1986). The e o e, absen eeism can be seen as “a desi able nonpecunia y elemen o he compensa ion package” (Allen 1981b, p. 207). Second, e iciency wage heo y as dis- cussed by Shapi o and S igli z (1984) assumes absence o a ise om mo al haza d and shi king. The e o e, absence may se e as a measu e o indi idual e o le el choice deci- sions and wo ke p oduc i i y and is widely used as such in he esea ch (e.g. Ichniowski, Shaw 2013). The Shapi o-S igli z (1984) model is based on he assump ion ha he alue o a p esen s a e (i.e. an employmen ela ion) consis s o he ne cu en e u n as well as he expec ed u u e e u n. By being absen om wo k an employee isks dismissal and, as a consequence, he loss o he li e ime u ili y associa ed wi h he employmen ela ion. Bo h neoclassical labo supply and e iciency wage heo y sha e he assump ion ha ab- sence, a leas pa ially, is an employee’s indi idual decision unde some cons ain s im- posed by he employe . Hence, absence needs o be conside ed appea ing ei he in olun- a y (e.g. as consequence o sickness o inju y) o olun a y (e.g. o maximize indi idual u ili y). While in olun a y absence spells a e no expec ed o be in luenced by mo i a ion, olun a y absence can be modeled as u ili y maximiza ion o shi king (e.g. Ba mby, Ses- sions, T eble 1994, B own, Sessions 1996, Chadwick-Jones, B own, Nicholson 1973, Sagie 1998). Mo eo e , in he adi ion o S ee s and Rhodes (1978) absence can be in e - p e ed as a social phenomenon wi h absence decisions ne e made solely by he indi idual, bu ins ead being subjec o ce ain cons ain s imposed by pee s, subo dina es, supe io s and he o ganiza ion’s o e all absence cul u e (e.g. Chadwick-Jones, Nicholson, B own 1982; Ren sch, S eel 2003). Ne e heless, a g ea pa o absence esea ch ocus exclusi e- ly on indi idual le el absence (e.g. Kaise 1998, Ren sch, S eel 2003). Employee Absen eeism: De e minan s in he In e na ional Con ex 45 Conside ing a po en ial so ing o de e minan s o absence a iance we a e loosely inspi ed by Kaise (1998). He sugges s absen eeism o be de e mined by pe sonal cha ac e is ics and indi idual esponses o in luences s emming ei he om he wo k o non-wo k en i- onmen . Adding ou own eading o he li e a u e, we iden i y ou ields o po en ial de- e minan s o employee absen eeism: social pee in luences in eams, economic in luences / incen i es, wo ke cha ac e is ics and wo king condi ions. In wha ollows, we b ie ly discuss he de e minan s o absence along his classi ica ion in o de o de i e ou main esea ch hypo heses. 14 3.2.1 Social Pee In luences in Wo k Teams Tu no e I has o en been poin ed ou in absen eeism esea ch ha absence a iance is small wi hin and la ge be ween g oups (e.g. Xie, Johns 2000). Usually, his phenomenon is a ibu ed o social absence no ms ha exis ei he among g oups, o ganiza ions o cul u al sphe es (e.g. Bambe ge , Bi on 2007, Ren sch, S eel 2003). Absence no ms as unde s ood in his pape can be de ined as “se o absence- ela ed belie s, alues and beha io al pa e ns ha a e sha ed among membe s o a wo k g oup o o ganiza ional uni ” (Gella ly, Luchak 1998, p. 1086). Al hough he e is comp ehensi e empi ical e idence on he link be ween pee absence and indi idual absence (e.g. Ma occhio 1994, de Paola 2010, Ichino, Maggi 2000, Rosenbla , Shapi a-Lishchinsky, Shi om 2010), li le is known on he social mechanisms behind his ela ionship (Johns 1997). 15 Since eam membe s a e a ional u ili y- maximize s, we assume absence no ms o be mos likely in a o o he employees and, hus, inc ease absence. Now, conside a eam whose membe s sha e a common belie abou he legi imacy o absence. Wi hin his eam, he en o ceabili y o he sha ed absence no m is subjec o changes in he g oup composi ion since new wo ke s a e no awa e o any exis ing absence no m o migh no ag ee on pa icipa ing (e.g. Mine s e al. 1995). As a consequence, eam-le el absence can be assumed o dec ease wi h eam u no e . This a gumen leads us o hypo hesis H1: H1: “Absence on he uni -le el dec eases wi h u no e .” 14 We limi ou e iew o he li e a u e o hose de e minan s ha will ac ually be su eyed wi hin his s udy. Fo a comp ehensi e o e iew on absence esea ch see, o ins ance, e iew a icles and me a- s udies by Beems e boe e al. (2009), Duij s e al. (2007) as well as Fa ell and S amm (1988). 15 One possible a gumen is buil on sel -ca ego iza ion (Tu ne 1987) and social iden i ica ion heo y (Taij el 1982) since indi iduals p e e wo king wi h simila pee s and, hus, adop a pa icula pee be- ha io . Employee Absen eeism: De e minan s in he In e na ional Con ex 46 Uni size Re e ing o neoclassic labo supply models as well as e iciency wage heo y, one can assume ha some wo ke s s ay away om wo k wi hou being genuinely sick. Gi en his assump ion, esea che s o en model absen eeism as a po en ial indica o o shi king (e.g. Ba mby, O me, T eble 1995, Riphahn 2004, B adley, G een, Lee es 2007). Economic heo y sugges s ha he sha e o shi king employees usually depends on he p obabili y o being de ec ed as well as he po en ial consequences once being caugh (e.g. Alchian, Demse z 1972). In his con ex , he e ec o g oup size on shi king is s aigh o wa d: Shi king can be de e ed by app op ia e ac ions ei he by employe s (e.g. o ganized moni- o ing, incen i e design) o by co-wo ke s (e.g. mu ual moni o ing, social sanc ions). 16 Howe e , any o hese ac ions will be ha de o en o ce wi h inc easing g oup size. Fo ins ance, he abili y o mu ually moni o co-wo ke beha io is ge ing mo e di icul wi h each addi ional eam membe and, as a esul , he e ec i eness o social sanc ions ha discipline wo ke s who a e caugh shi king declines. Likewise, moni o ing by supe iso s is ge ing mo e complica ed and/o cos ly wi h inc easing eam size (Ba on, K eps 1999). Fu he mo e, om Bandie a, Ba ankay and Rasul (2005, 2009) we know abou he im- po ance o social ies and amilia i y among eam membe s in de e mining p oduc i i y. The magni ude o wo ke amilia i y, howe e , can be assumed o dec ease wi h eam size. Ul ima ely, one can p edic shi king o be acili a ed in la ge eams due o educed moni- o ing and less amilia i y. These p edic ions a e exp essed in hypo hesis H2: H2: “Absence on he uni -le el inc eases wi h eam size.” Tempo a y wo ke s I is widely ecognized in economic esea ch ha empo a y wo ke s a e less absen han pe manen wo ke s. Ye , as soon as aken on a pe manen con ac hey signi ican ly in- c ease hei absence beha io (e.g. B adley, G een, Lee es 2007, Engelland , Riphahn 2005). This phenomenon is mainly a ibu able o incen i izing con ac cha ac e is ics since empo a y agen s seek o quali y hemsel es o pe manen con ac s ha a e usually associa ed wi h be e wo king condi ions (e.g. A onsson 1999, Paoli, Me llié 2001) and highe pay (e.g. Me ens, Gash, McGinni y 2007). Howe e , i emains an open ques ion how empo a y agen s in luence hei pe manen ly employed co-wo ke s wi h ega d o absence beha io . Two po en ial a gumen s may be conside ed. Fi s , assume ha pe ma- nen ly employed wo ke s a e awa e o he incen i izing na u e o empo a y con ac s. 16 On he impo ance o moni o ing o educe shi king in eam se ings ecall Holms öm (1982). Employee Absen eeism: De e minan s in he In e na ional Con ex 47 Being a ional u ili y-maximize s, pe manen agen s migh allow hemsel es a mo e “gen- e ous” absence beha io as hey suppose hei empo a y co-wo ke s o be incen i ized o ill in he gap le by hei own absence. This implies ha a high sha e o empo a y wo k- e s inc eases he absence o he pe manen ly employed co-wo ke s due o shi king. Sec- ond, as men ioned abo e, empo a y employmen is usually associa ed wi h wo se wo king condi ions. Suppose, o ins ance, ha empo a y agen s a e appoin ed o he mos bu den- some wo king condi ions, e.g. by being excluded om job o a ion. In u n, his would allow pe manen ly employed wo ke s o s ay on less demanding wo k asks and su e less wo k- ela ed s ains. Following his a gumen , a high sha e o empo a y wo ke s migh dec ease pe manen wo ke s’ absence. Since we know ha empo a y agen s a e appoin ed on he same asks and mee iden ical wo king condi ions han pe manen wo ke s a he obse ed company, we belie e he la e a gumen o be less con incing in ou s udy de- sign. The e o e, we base ou hough s on shi king beha io and hypo hesize: H3: “Absence o pe manen ly employed eam membe s inc eases wi h he sha e o empo a y eam membe s.” Wo ke s wi h heal h limi a ions I is common p ac ice a he s udied o ganiza ion ha wo ke s su e ing om pa icula ce i ied heal h limi a ions migh be subjec o empo a y o pe manen heal h impai men s such as he p ohibi ion o ca y hea y loads o wo k o e head. As a esul , hese wo ke s a e exemp om pe o ming asks ha would u he de e io a e hei heal h condi ion. Ye , i emains an open issue i he p esence o employees wi h heal h impai men s a ec s o e all uni absen eeism. This ques ion is mainly based on wo conside a ions. Fi s , i seems o be qui e a s aigh o wa d a gumen ha heal h- es ic ed employees a e mo e p one o absence due o hei physical (o men al) condi ion. Medical esea ch la gely em- phasized he g ea signi icance o ch onical and p e-exis ing heal h impai men s in de e - mining absence (e.g. Dewa, Lin 2000, Kessle e al. 2001). Second, he easie jobs in he eam’s ope a ions may o en be occupied by wo ke s wi h heal h impai men s and, as a consequence, a e excluded om uni -in e n job o a ion. In his si ua ion, ini ially heal hy employees migh ace addi ional bu dens as hey need o s a hose wo kplaces ha hei eam ma es wi h heal h impai men s canno co e . These highe bu dens may inc ease hei own absence. The e o e, we expec a posi i e ela ionship be ween absence and he sha e o eam membe s wi h heal h-impai men s. Employee Absen eeism: De e minan s in he In e na ional Con ex 48 H4: “Absence on he g oup-le el inc eases wi h he sha e o eam membe s wi h heal h impai men s.” 3.2.2 Economic In luences / Incen i es Sickness bene i s I is a s ylized ac in economics ha he ins i u ional amewo k de e mines employees’ absence beha io . Ea ly e idence is gi en by Buzza d and Shaw (1952) who s a e ha ab- sence inc ease wi h highe sick pay eplacemen a es as he cos s o absence dec ease o employees. A co esponding labo supply model is o e ed by B own and Sessions (1996). They a gue ha highe sickness bene i s aise incen i es o wo ke s o be absen om wo k. These assump ions a e suppo ed by F ick and Malo (2008). Based on he Eu opean Su ey o Wo king Condi ions hey obse ed a signi ican inc ease in absen eeism wi h mo e gene ous sickness bene i s wi hin he EU-14. O e he yea s, se e al EU go e n- men s ini ia ed changes o na ional sickness legisla ion in o de o add ess he bu dens o sick pay o social secu i y sys ems and employe s. Absence esea ch o en bene i s om hese legisla i e changes ha allow o na u al expe imen s, e.g. Johansson and Palme (2002, 2005) as well as Voss, Flode us and Dide ichsen (2001) o Sweden and Puhani and Sonde ho (2010) as well as Zieba h and Ka lsson (2010, 2013) o Ge many. These s udies p o ide b oad suppo o he posi i e link be ween s a u o y sickness bene i egu- la ions and employee absence. We, he e o e, p edic : H5: “Absence will inc ease wi h mo e gene ous sickness bene i egula ions.” Employmen p o ec ion The same line o a gumen holds ue o employmen p o ec ion. Focusing on a legisla i e change in Sweden, Olsson (2009) obse es ha a less a o able labo p o ec ion law in e ms o senio i y dismissal p o ec ion signi ican ly dec eases absence beha io in o gani- za ions ha a e a ec ed by he change. Based on Ge man da a, Riphahn and Thalmaie (2001) ind ha absen eeism inc eases a e p oba ion pe iods end and manda o y labo p o ec ion se s in. Fu he e idence is gi en by Ichino and Riphahn (2005) who obse e he same beha io o I alian bank employees. B adley, G een and Lee es (2012) epo a signi ican aise in absen eeism when empo a y wo ke s a e aken on pe manen ly. They a gue ha he low employmen p o ec ion associa ed wi h empo a y employmen incen i - izes wo ke s o show high e o in e ms o low absence. As soon as labo p o ec ion ap- Employee Absen eeism: De e minan s in he In e na ional Con ex 49 plies o hei employmen ela ion, he o me ly empo a y agen s educe e o and no longe e ain om shi king. This a gumen leads o hypo hesis H6: H6: “Absence will inc ease wi h s ic employmen p o ec ion legisla ion.” Unemploymen and p ospe i y le el In addi ion o he legisla i e amewo k, he economic en i onmen and unemploymen a es a e o en linked o absence beha io beginning wi h a seminal pape by Leigh (1985). Using da a om he ime o he US ecession in he 1970s, he obse ed employees o e- ain om absen eeism when unemploymen is high as hey assume employe s o i s lay o wo ke s who ha e p o en absence-p one. Askildsen, B a be g and Nilsen (2005) epo simila indings o he No wegian labo ma ke : An inc ease in local unemploymen a es signi ican ly dec eases absence. This esul e en holds when con olling o wo k o ce composi ion which is o en claimed o be esponsible o cyclical a ia ion in absen eeism. Knu sson and Goine (1998) obse e he link be ween absence and unemploymen o Swedish males. Wi h ega d o he economic si ua ion, Vi anen e al. (2005) obse ed o Finish public sec o employees ha a cons an ly poo local economy has a dec easing im- pac on sel -ce i ied sickness. Audas, Godda d (2001) con i m he link be ween business cycle and absen eeism o he US. Thus, we p edic : H7: “Absence will dec ease wi h inc easing unemploymen .” H8: “Absence will inc ease wi h p ospe i y le el.” 3.2.3 Wo ke Cha ac e is ics Non-wo k ela ed wo ke cha ac e is ics (gende , age) The link be ween non-wo k ela ed wo ke cha ac e is ics (i.e. gende and age) and em- ployee absen eeism is well es ablished. While o gende i is a s ylized ac ha emales exhibi highe sickness absence han males (e.g. Ba mby, E colani, T eble 2002, Mas e- kaasa 2000), esul s o age a e less conclusi e (e.g. Rhodes 1983). Acco ding o Bekke , Ru e and an Rijswijk (2009), he e a e h ee main easons o gen- de di e ences in absen eeism. Fi s , due o physical di e ences women su e mo e om ep oduc ion- ela ed heal h issues such as p egnancy and mens ua ion (e.g. Alexande son e al. 1996, Sydsjö, Sydsjö, Alexande son 2001). Second, emales ace o he daily obliga- ions han males such as he double bu den o job and amily since child ca e is adi ional- ly mo e associa ed wi h women (e.g. Åke lind e al. 1996, B a be g, Dahl, Risa 2002, Employee Absen eeism: De e minan s in he In e na ional Con ex 56 Ou s udy design has nume ous ad an ages in add essing issues aised by p e ious e- sea ch on absen eeism. Fi s , s a ing wi h ea ly wo k by Chadwick-Jones, Nicholson and B own (1982) absence is no solely modeled as an indi idual phenomenon bu as being subjec o social pee in luences. Gi en ha wi hin-uni absence a ies only li le whe eas be ween-uni absence a ies widely one can assume uni -le el in luences o play an im- po an ole in de e mining absence (e.g. Ren sch, S eel 2003, Ha ison, Ma occhino 1998). Thus, i seems su p ising ha ela i ely li le wo k is done on absen eeism a he g oup le el (e.g. Ren sch, S eel 2003) wi h only ew excep ions (e.g. Xie, Johns 2000, K is ensen e al. 2006). We add ess his gap in absence esea ch by o e ing new insigh s on sickness absence a he eam le el. Second, absence esea ch mainly elies on sel - epo ed sickness da a, e.g. ga he ed ia ques ionnai es, since egis e da a is o en una ail- able o esea che s. Howe e , he esponse sensi i i y o sel - epo ed da a in compa ison o egis e da a a ies om high (e.g. 91% (Voss e al. 2008), 88% (Bu do , Pos , B ug- geling 1996)), o medium (e.g. 82% (Aguis e al. 1994)) and low (e.g. 55% ( an Poppel e al. 2002)). In line wi h he la e , Johns (1994) inds employees o be absen ac ually wice as much as sel - epo ed. 21 Since we a e able o use egis e da a om company eco ds we can neglec conce ns abou he eliabili y and alidi y o ou da a. Thi d, since ou da a se includes ou p oduc ion plan s om Ge many, Spain and he UK we a e able o con ol o a ying ins i u ional amewo ks ac oss hese coun ies. As discussed ea lie , egula- ions o sickness bene i s and employmen p o ec ion a e ound o signi ican ly de e mine absence beha io in he in e na ional con ex (e.g. F ick, Malo 2008, Riphahn, Thalmaie 2001). Howe e , in e na ional compa abili y o absence s a is ics usually p o es di icul due o widely a ying epo ing s anda ds o speci ic cases such as amily ca e o ma e ni- y (Eu opean Founda ion o he Imp o emen o Li ing and Wo king Condi ions 2010). Fo his s udy, we can dispel any doub s abou in e na ional compa abili y since all ab- sence da a is eco ded based on company s anda ds. Finally, we a e able o apply mul iple measu es o absence as dependen a iables o accoun o di e en aspec s o absen eeism such as olun a y and in olun a y absence. All absence measu es used in his s udy a e discussed in de ail in he ollowing sec ion. 21 Mo eo e , accu acy o sel - epo ed sickness absence is ound o dec ease wi h ime (Se e ens e al. 2000) and wi h inc easing numbe o sickness days (Fe ie e al. 2005, Johns 1994). Employee Absen eeism: De e minan s in he In e na ional Con ex 57 3.4 Es ima ion S a egy Ou empi ical analyses a e based on h ee di e en absence measu es o conside absen ee- ism om a ious pe spec i es. We use absence KPIs epo ed on company eco ding s anda ds. The mos impo an absence igu e is a uni ’s mon hly absence a e. I is de ined as he pe cen age o ime o iginally scheduled o wo k wi hin he uni missed due o sel - ce i ied and medically-ce i ied sickness. All company-in e n discussion on absence is en i ely cen e ed on his measu e. Absence a es a y ac oss uni s and loca ions, he o e - all s a is ics a e p esen ed in Table 3.2. Mo eo e , we ollow Hensing e al. (1998) 22 in cons uc ing u he absence measu es since wo king wi h egis e da a does no allow o a clea sepa a ion be ween in olun a y and olun a y absence. We, he e o e, apply wo mon hly p oxies o bo h absence ypes p oposed among o he s by Sagie (1998). Fi s , one can assume in olun a y absence spells o be usually longe since sicknesses and inju ies a e seldom cu ed comple ely a e one o wo days. Hence, we calcula e he mean du a ion o sick lea es o a eam ounded o in ege alues (Hensing e al. 1998). Second, olun a y absence is widely ecognized as sho -las ing bu mo e equen . Tha is why equency measu es se e well as p oxy o olun a y absence (e.g. Chadwick-Jones e al. 1971). We ollow Hensing e al. (1998) in calcula ing a equency measu e by di iding he numbe o absence spells o a gi en eam by he numbe o pe sons in ha pa icula uni . S ill, a de- ini i e sepa a ion be ween in olun a y and olun a y absence emains unclea since bo h p oxies migh include he o he absence beha io as well. In gene al, ou es ima ions a e based on he ollowing models wi h ABS ep esen ing he h ee absence measu es discussed abo e as dependen a iables: ABS = α + βPEER + γ ECON + δWORK + φ COND + 𝜗CONT + ε whe e α is he cons an , PEER is a ec o o social pee in luences on he eam le el, ECON is a ec o o economic in luences / incen i es, WORK is a ec o o wo ke cha ac- e is ics, COND is a ec o desc ibing wo king condi ions and CONT is a ec o o u he con ols, β, γ, δ, φ and ϑ a e he es ima ed coe icien s and ε is he e o e m. All a iables a e discussed in mo e de ail below. 22 Hensing e al. (1998) p opose equency, leng h, incidence a e, cumula i e incidences and du a ion o sick-lea e as measu es o absen eeism. Al hough we a e no able o cons uc all measu es o ou sample due o una ailable o missing da a, we use hei equency and du a ion measu es. Employee Absen eeism: De e minan s in he In e na ional Con ex 58 Social pee in luences (PEER) We model social in luences o wo ke s’ pee s o s udy hei impac on employee absence. One impo an pee in luence is ocusing on g oup absence no ms (e.g. Bambe ge , Bi on 2007, Ren sch, S eel 2003). These no ms migh change wi h u no e (Mine s e al. 1995). Wi h ega d o hypo hesis H1 we p oxy o u no e by adding all new a i als o and all exi s om he eam in o de o c ea e one o e all u no e a iable. The idea is simple: we use his qui e in ui i e measu e o a iance in eam composi ion since absence no ms should be in luenced by bo h new and lea ing wo ke s alike. In addi ion, we add ess e- sea ch on shi king in eam se ings (e.g. Dunn, Youngblood 1986) by s a ing ha wi h in- c easing eam size absence should be mo e p onounced due o acili a ed shi king op ions (H2). To inco po a e his idea in ou es ima ions we include mon hly eam size o all uni s. We measu e eam size as he simple headcoun o eam membe s in a gi en mon h. As exp essed by hypo hesis H3, he sha e o empo a y wo ke s migh inc ease he absence beha io o pe manen wo ke s. In o de o con ol o his a gumen , we include he pe - cen age o empo a y wo ke s in a uni in ou es ima ions. Un o una ely, plan C did no epo any in o ma ion on he sha e o empo a y wo ke s. In hypo hesis H4 we s a e ha he sha e o eam membe s wi h heal h impai men s should inc ease uni -le el absence o wo easons. Fi s , mos ob iously hese wo ke s can be expec ed o be mo e p one o ab- sence. Second, heal hy wo ke s migh su e addi ional wo kload. We ake his hypo he- sized ela ionship in o accoun by including he pe cen age o eam membe s ha ha e any kind o ce i ied heal h impai men in a gi en mon h. Economic in luences / incen i es (ECON) In o de o s udy po en ial economic in luences and incen i e e ec s on employee absence, we ex end he company da a se wi h en i onmen al a iables iden i ied by p e ious e- sea ch o de e mine employee absen eeism. A his poin , we bene i om he in e na ional con ex o ou s udy since we a e able o compa e di e en social secu i y sys ems and employmen p o ec ion laws as well as a ying na ional economic en i onmen s. This al- lows us o conside po en ial economic in luences along he ou main opics in oduced in H5 o H8. Fi s , as p oposed among o he s by Riphahn and Thalmaie (2001) as well as Ichino and Riphahn (2005) employmen p o ec ion may ha e a signi ican impac on em- ployee absence beha io . To accoun o a ying legisla i e se ings in Ge many, Spain and he UK we inco po a e he OECD’s Indica o s o Employmen P o ec ion (OECD 2014a). Th ee syn he ic indica o s a e he s ic ness o a coun y’s egula ions on indi id- Employee Absen eeism: De e minan s in he In e na ional Con ex 59 ual and collec i e dismissal as well as on ixed- e m and empo a y wo k agency con ac s on a scale om weak=0 o e y s ic =6. Second, we conside sugges ions s a ing ha ab- sen eeism inc eases wi h mo e gene ous sickness bene i s (e.g. F ick, Malo 2008, Johans- son, Palme 2002, 2005) in ollowing an app oach p oposed by F ick and Malo (2008). They use he so-called MISSOC epo s (Mu ual In o ma ion Sys em on Social P o ec ion) published by he Eu opean Commission ha desc ibe na ional sickness bene i s conce ning hei co e age, wai ing pe iod, maximum du a ion and eplacemen a e (Eu opean Com- mission 2014a). We con e he epo ’s ex ual desc ip ions in o a scale om weak=0 o gene ous=6 sickness bene i s in o de o ma ch he OECD’s Indica o o Employmen P o- ec ion scales. Thi d, ollowing p e ious e idence (e.g. Leigh 1985, Askildsen, B a be g, Nilsen 2005) we expec high unemploymen a es o signi ican ly in luence employee ab- sence beha io . Hence, we con ol o na ional (un)employmen in wo ways. On he one hand, we use seasonally adjus ed mon hly na ional unemploymen a es. On he o he hand, we include he sha e o pe sons employed in manu ac u ing (NACE Classi ica ion D) in pe cen o o al employees o model wo ke s ou side op ions. Da a is based on qua - e ly Eu os a s a is ics o bo h unemploymen (Eu os a 2014a) as well as employmen in o ma ion (Eu os a 2014b). Finally, we ake in o conside a ion esea ch s a ing a link be ween na ional economic si ua ion and employee absence beha io (e.g. Vi anen e al. 2005, Audas, Godda d 2001) by inco po a ing Wo ldbank in o ma ion on annual pe cen - age changes o GDP by coun y (Wo ldbank 2014). Wo ke cha ac e is ics (WORK) To accoun o he in luence o wo ke cha ac e is ics on absence we chose ou main po- en ial de e minan s o absen eeism. On he one hand, we include he classical non-wo k- ela ed demog aphic ea u es o age and gende o es hypo heses H9 o H11. Fi s , age is included as uni mean age. Al hough he measu e o age on he g oup le el seems some- how no ideal since simila age alues in g oups may be caused by o ally di e en age dis ibu ions (e.g. K is ensen e al. 2006), we belie e ha con olling o he coe icien o a ia ion (CV) o age should le el ou po en ial biases. Second, we use he sha e o males o model he gende dis ibu ion wi hin a eam. On he o he hand, we aim a add essing po en ial in luences o wo k- ela ed indi idual demog aphics on absence. We, he e o e, use uni mean enu e and he espec i e CV as p oxies o p o essional expe ience. Follow- ing his logic, a high mean enu e indica es uni s wi h wo ke s ha ha e accumula ed high i m-speci ic human capi al and, hus, should ha e be e s a egies o cope wi h wo k- Employee Absen eeism: De e minan s in he In e na ional Con ex 60 ela ed s esso s. Al hough age and enu e a e highly co ela ed (0.698), esul s do no change when lea ing enu e ou om he es ima ions. The e o e, i seems ha age and en- u e a iables a e ac ually measu ing wo di e en phenomena. This is in line wi h among o he s Ba mby, E colani and T eble (2002) who obse e a enu e e ec on absence e en when age is con olled o . Un o una ely, CVs we e calcula ed nei he o age no o en- u e om plan D. E en ually, we add ess esea ch on acu e heal h condi ion wi h in luenza and ILI being among he mos s a ed easons o s aying away om wo k (e.g. Tsai, Zhou, Kim 2014, Keech, Sco , Ryan 1998). 23 As sugges ed in hypo hesis H13, we con ol o ILI and in luenza by using in luenza eco dings om he Wo ld Heal h O ganiza ion (WHO). These s a is ics assess in luenza ac i i y om no ac i i y o widesp ead ou b eak on a na- ional le el (WHO 2014). We con e weekly ex ual e alua ions o mon hly o dinal scales o no ac i i y=0, spo adic=1, local ou b eak=2, egional ou b eak=3 and widesp ead ou - b eak=4. Wo king condi ions (COND) In H14 we add ess esea ch s a ing ha shi wo k usually inc eases absence due o highe bu dens conce ning sleeping habi s, diso de ed diu nal hy hm and complica ed amily and social li e (Slany e al. 2014, Mo ikawa e al. 2001, Fekedulegn e al. 2013). By aking a no mal se en hou one-shi sys em (only mo ning shi ) as e e ence ca ego y, we model addi ional shi s based on a wo-shi (mo ning and day shi ) o h ee-shi sys ems (mo n- ing, day and nigh shi ). Con ol a iables (CONT) In o de o u he exclude con ounding e ec s on employee absence beha io , we add u he con ols o he company da a se . Fi s , we add ess e idence on he in luence o wea he on absen eeism (e.g. Shi, Sku e ud 2015) by including in o ma ion on mon hly mean empe a u e, p ecipi a ion and sun hou s. 24 Second, we use mon h and yea dummies o con ol o po en ial seasonali y o o he iming e ec s o absence. 23 Ye , we do no ha e in o ma ion on he indi idual heal h condi ion o employees o wo k uni s. 24 Since we aim o ensu e da a compa abili y by elying on single-sou ced da a p o ided om he in e na- ional ne wo k o he Ge man Me eo ological Se ice (DWD 2014), we only ha e a limi ed choice o me- eo ological s a ions close o he plan s’ loca ions. The e o e, wea he in o ma ion was ga he ed a me e- o ological s a ions 130km away om si es a maximum. We do no assume his o be a p oblem. Employee Absen eeism: De e minan s in he In e na ional Con ex 61 3.5 Empi ical Resul s As a i s s ep, we conduc one-way ANOVAs in o de o check o signi ican mean a i- ance in absence a es be ween and/o wi hin o ganiza ional uni s. While we ind be ween- uni a iance in absence a es o be highly signi ican a all plan s (plan A (F(5, 192)=6.56, p=.000), plan B (F(7, 256)=8.60, p=.000), plan C (F(103, 3564)=9.67, p=.000) and plan D (F(41, 1419)=4.28, p=.000)), we obse e wi hin-uni absence o a y signi ican ly a only 69 ou o 160 uni s o e ime (43%). This phenomenon is in line wi h p e ious esea ch ha ound absence a iance o be g ea be ween uni s bu smalle wi hin uni s (e.g. Ha ison, Ma occhio 1998, Xie, Johns 2000). In esponse o his inding, we chose ixed e ec s eg ession models o be p e e able o e andom e ec s models since he e seems o be some g oup speci ic e ec s ha emain cons an o e ime. Fixed e ec s may be o psychological na u e bu may also be a ibu a- ble o a ying wo kplace cha ac e is ics ha migh inc ease e gonomic bu dens o wo k. We, he e o e, es ima e ixed-e ec models on all h ee absence measu es o es o po en- ial de e minan s o employee absen eeism. 25 In his p ocess, we ollow a wo-s aged es i- ma ion s a egy. On he one hand, we pool da a o wo plan s (bo h olume ca p oduc ion si es and bo h luxu y ca p oduc ion si es, espec i ely) and apply pai -le el es ima ions by using in e ac ion e ms o economic and social pee in luences. This p ocess seeks a a oiding po en ial biases a ising om he ac ha a iables may no only a y be ween plan s bu may de e mine absence beha io a each plan in di e en ways. On he o he hand, we apply plan -le el es ima ions o each loca ion independen ly o u he con ol o plan -speci ic e ec s ha migh ge los in pai -wise eg essions. Fo all es ima ions we use a ious model speci ica ions including di e en se s o explana o y a iables o con ol o he obus ness o ou indings. Subsequen o he empi ical es ing, we conduc ed a o al o 18 on-si e expe in e iews a bo h in e na ional si es (B, C) in o de o double-check ou own expe ience om na ional plan s. Expe s a e wo king in di e en o ganiza ional ields including HR expe s, wo ke ep esen a i es, shop loo manage s and shop loo s a . Thei ope a ional expe ience and p ac ical knowledge helped us o in e p e ou indings comp ehensi ely and de i e impli- ca ions ha a e easible and ealizable in o ganiza ional implemen a ion. 25 O he esea che s o en use coun da a me hods such as (ze o-in la ed) poisson o nega i e binomial e- g essions o s udy absence du a ion. Un o una ely, ou du a ion da a ep esen s a uni ’s mean and, he e- o e, does no mee he in ege alue assump ion necessa y o bo h eg ession models. Employee Absen eeism: De e minan s in he In e na ional Con ex 62 In he ollowing, we s a he de ailed discussion o ou indings by i s p esen ing es ima- ion esul s on he e ec s o social pee in luences on wo ke absence (H1 o H4). We hen go on o p esen ou indings wi h espec o economic in luences / incen i es (H5 o H8), wo ke cha ac e is ics (H9 o H13) and wo king condi ions (H14). Wi hin each pa ag aph we discuss pooled pai -le el as well as plan -le el esul s and u he dis inguish ou ind- ings along he h ee main dependen a iables absence a e, absence spell du a ion and absence equency. O e all indings sugges ha de e minan s o absence a y by plan . In o he wo ds, we only ind ew pa e ns in he da a ha a e common a all ou plan s. Main esul s o pooled in e ac ion es ima ions a e epo ed in Tables 3.3 o 3.5, main esul s o plan -le el es ima ions a e epo ed in Tables A.1 o A.3 in he appendix. Es ima ions in all ables a e numbe ed o a be e o ien a ion and e e ed o h oughou he ex . Due o ea- sons o b e i y we ocus on p esen ing he main esul s in all ables. Social pee in luences (PEER) S a ing wi h pee e ec s, we ha e analyzed ou po en ial de e minan s: u no e as a p oxy o g oup no ms, uni size as a p oxy o shi king, he sha e o empo a y wo ke s and he sha e o wo ke s wi h empo a y o pe manen heal h impai men s. Resul s will be discussed in he speci ied o de . In hypo hesis H1, we p edic absence o dec ease wi h u no e because he en o ceabili y o g oup no ms – ha we belie e o inc ease olun a y absence – may be limi ed wi h a iance in eam composi ion. Ye , we ind only weak sup- po o his a gumen as he ela ionship be ween u no e and absence spell du a ion is only signi ican a plan D (es ima ion VI-4). Howe e , since du a ion measu es a e usually used o p oxy o in olun a y absence, we e ain om in e p e ing his inding in suppo o H1. Ins ead, plan le el es ima ions o loca ion D (V-4) as well as in e ac ion e m es- ima ions o loca ions C and D (III-4, -5, -6) indica e ha high u no e inc eases absence equency. Resul s o absence a e do no each s a is ical signi icance. The g ea majo i y o local expe s con i med he empi ical indings as hey obse e absence o be usually highe in uni s wi h high employee u no e . Acco ding o hei expe ience, his beha io is a ibu able o employees eeling unse led and unce ain a he han o changing g oup no ms. Mo eo e , u no e and absen eeism a e bo h iden i ied o be s a egies wo ke s may choose in esponse o indi idual job dissa is ac ion ( heo y o exi , oice and loyal y, e.g. Hi schman 1970, Fa ell 1983). Thus, high u no e may educe he wo ke s’ need o o he esponses o job dissa is ac ion such as absence. As a consequence, absence can be assumed o be low in high u no e si ua ions. In o al, we ha e o ejec hypo hesis H1. Table 3.3: Fixed-E ec s In e ac ion Te ms on Absence Ra es (I) luxu y si es olume si es I-(1) I-(2) I-(3) I-(4) I-(5) I-(6) VARIABLES Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Mean age (in yea s) -0.949 -0.611 -0.206 0.703* -0.396 -1.512 -0.476 1.487 0.021 0.501** 1.601 0.257 (7.203) (2.612) (0.494) (0.385) (7.600) (2.699) (1.501) (1.908) (0.163) (0.223) (2.635) (2.762) Mean age² (in yea s) 0.00949 0.0170 --- --- 0.00260 0.0193 0.0059 -0.0134 --- --- -0.022 0.00293 (0.0994) (0.0287) (0.105) (0.0299) (0.0201) (0.0236) (0.035) (0.0338) Sha e o males -19.74 5.273 -17.50 7.464 -15.52 -0.192 -2.78* -0.297 -6.988*** 0.972 -6.751*** -0.364 (in %/100) (18.96) (8.150) (15.06) (7.947) (20.10) (9.218) (1.429) (4.916) (2.406) (6.67) (2.342) (6.735) Mean enu e (in yea s) 2.533 -2.223* -3.027** -0.344 -0.967 -1.472 0.0492 0.5573 0.2374 -0.5294 0.1902 1.457 (3.193) (1.223) (1.378) (0.226) (4.925) (1.015) (0.4074) (1.1011) (0.1652) (0.4025) (0.5842) (1.520) Mean enu e² (in yea s) -0.405 0.0521 --- --- -0.131 0.0426 0.0012 -0.0445 --- --- -0.0002 -0.0761 (0.260) (0.0336) (0.379) (0.0289) (0.0148) (0.0401) (0.0218) (0.0553) Uni size (in pe sons) 0.186** -0.0174 0.160*** -0.0322 0.136* -0.0264 --- -0.122 0.0464 0.0595 0.0611 0.0647 (0.0624) (0.0541) (0.0491) (0.0521) (0.0698) (0.0502) (0.148) (0.06004) (0.168) (0.058) (0.158) In luenza 0.0602 0.101 0.144 (0.248) -0.328 -0.5543 -0.816* (0=no ac i i y,…) (0.275) (0.255) (0.291) (0.454) (0.460) Employmen p o ec ion 6.857 6.423 --- 1.924 0.867 --- legisla ion (0=weak,…) (3.882) (3.743) (3.875) (8.749) Employmen in 106.2 88.02 --- 128.565 153.555 --- manu ac u ing (in %/100) (180.0) (202.2) (91.666) (1.981) Unemploymen a e -11.46 -30.00 --- 3.3119 13.334* --- (in %/100) (96.23) (96.09) (14.264) (26.688) P ospe i y le el (annual 10.56* 10.06* --- 6.1809** 3.495 --- change in GDP in %/100) (5.490) (5.673) (2.541) (4.9687) 2-Shi sys em (yes=1) 1.442 1.281 0.826 1.921 2.697 1.608 (1.083) (1.136) (1.229) (1.262) (1.699) (1.577) 3-Shi sys em (yes=1) n/a n/a n/a 1.208 1.787 1.261 (1.152) (1.303) (1.266) Cons an 3.501 -13.69 50.39 -21.86 -8.137 -16.24 (60.35) (35.07) (60.45) (25.73) (24.844) (35.19) Obse a ions 471 471 471 4,878 2,128 2,128 R-squa ed 0.189 0.182 0.157 0.042 0.07 0.068 Numbe o uni s 14 14 14 146 146 146 Robus s anda d e o s in pa en heses | *** p<0.01, ** p<0.05, * p<0.1 + all es ima ions including mon h and wea he dummies & age and enu e coe icien s o a ia ion | only main esul s p esen ed | n/a= no a ailable Employee Absen eeism: De e minan s in he In e na ional Con ex 63 Table 3.4: Fixed-E ec s In e ac ion Te ms on Absence Spell Du a ion (II) luxu y si es olume si es II-(1) II-(2) II-(3) II-(4) II-(5) II-(6) VARIABLES Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Mean age (in yea s) -0.276 -5.748 -0.274 0.0202 -0.540 -6.717 -2.24 0.950 0.0628 0.653** 2.566 -1.882 (2.570) (4.142) (0.388) (0.384) (2.407) (4.601) (2.345) (2.588) (0.2221) (0.302) (3.603) (3.091) Mean age² (in yea s) 0.000915 0.0673 --- --- 0.00411 0.0837 0.0301 -0.00550 --- --- -0.0347 0.0307 (0.0355) (0.0470) (0.0337) (0.0527) (0.0317) (0.0317) (0.0484) (0.0379) Sha e o males -21.42* 4.607 -24.3** 6.889 -22.85* 4.248 -2.721 -4.406 -5.2565* -4.135 -5.017* -5.513 (in %/100) (11.35) (6.781) (9.569) (7.457) (12.78) (6.197) (1.789) (6.571) (3.0911) (8.806) (3.027) (8.841) Mean enu e (in yea s) -5.607 -0.563 -1.594 0.102 -5.164 -1.798 -0.1032 -0.604 0.4505** -0.734 -0.0047 1.354 (4.232) (1.034) (1.612) (0.222) (4.623) (1.029) (0.4351) (1.583) (0.2127) (0.566) (0.722) (2.377) Mean enu e² (in yea s) 0.295 0.0179 --- --- 0.283 0.0462 0.013 0.000880 --- --- 0.0167 -0.0799 (0.278) (0.0285) (0.292) (0.0287) (0.0174) (0.0594) (0.0291) (0.0926) Uni size (in pe sons) 0.0475 -0.0163 0.0735* -0.0612 0.0691* 0.00821 --- -0.0781 0.2712*** 0.399* 0.299*** 0.428* (0.0362) (0.0888) (0.0399) (0.0704) (0.0373) (0.0927) (0.146) (0.0845) (0.236) (0.0833) (0.232) Sha e o empo a y -0.306 -0.112 -1.233 --- --- --- wo ke s (in %/100) (1.902) (1.704) (2.073) Employmen p o ec ion -4.271 -4.687 --- 4.748 1.586 --- legisla ion (0=weak,…) (3.690) (3.655) (5.002) (10.27) Unemploymen a e -124.6 -133.0 --- 9.535 -23.96 --- (in %/100) (105.3) (102.0) (18.03) (29.68) P ospe i y le el (annual 1.868 3.022 --- 7.583** 2.167 --- change in GDP in %/100) (4.328) (4.416) (3.558) (6.486) 2-Shi sys em (yes=1) 0.321 0.460 1.037 3.652** 3.606* 3.606* (0.765) (0.695) (0.764) (1.714) (2.144) (2.144) 3-Shi sys em (yes=1) n/a n/a n/a 2.503* 2.593* 2.593* (1.371) (1.499) (1.499) Cons an 99.98 14.80 120.3* -4.780 -12.28 -15.59 (63.79) (38.29) (58.64) (37.53) (30.78) (45.73) Obse a ions 471 471 471 4,817 2,129 2,129 R-squa ed 0.141 0.134 0.123 0.033 0.052 0.049 Numbe o uni s 14 14 14 143 143 143 Robus s anda d e o s in pa en heses | *** p<0.01, ** p<0.05, * p<0.1 + all es ima ions including mon h and wea he dummies & age and enu e coe icien s o a ia ion | only main esul s p esen ed | n/a= no a ailable Employee Absen eeism: De e minan s in he In e na ional Con ex 64 Table 3.5: Fixed-E ec s In e ac ion Te ms on Absence F equency (III) luxu y si es olume si es III-(1) III-(2) III-(3) III-(4) III-(5) III-(6) VARIABLES Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan A (GER) Plan B (UK) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Plan C (ESP) Plan D (GER) Mean age (in yea s) -0.191 0.202* 0.00238 0.0199** -0.176 0.180 -0.00762 0.0188 0.0024 -1.10e-06 0.0237 0.0461 (0.188) (0.110) (0.0118) (0.00886) (0.193) (0.109) (0.0244) (0.04497) (0.0027) (0.00428) (0.0384) (0.0590) Mean age² (in yea s) 0.00261 -0.00205 --- --- 0.00238 -0.00198 0.00011 -0.00025 --- --- -0.00028 -0.00056 (0.00267) (0.00121) (0.00273) (0.00122) (0.00032) (0.00054) (0.0005) (0.00069) Sha e o males (in %/100) 1.074 0.0559 1.206* 0.0700 1.152 -0.0761 -0.056** 0.0593 -0.087** 0.04906 -0.081** 0.0201 (0.626) (0.301) (0.598) (0.327) (0.653) (0.293) (0.0256) (0.0795) (0.0368) (0.113) (0.0362) (0.107) Mean enu e (in yea s) 0.402** -0.0723 -0.0490 -0.0112* 0.293* -0.0609 0.0044 0.0442 0.0018 -0.0002 0.0072 0.0370 (0.157) (0.0612) (0.0517) (0.00594) (0.165) (0.0577) (0.0055) (0.0371) (0.00094) (0.00781) (0.00797) (0.0419) Mean enu e² (in yea s) -0.0331** 0.00173 --- --- -0.0249** 0.00168 -0.0002 -0.00139 --- --- -0.00027 -0.00141 (0.0115) (0.00174) (0.0114) (0.00166) (0.00021) (0.00139) (0.00031) (0.00152) Uni size (in pe sons) 0.00392* -0.00297 0.00211 -0.00176 0.00263 -0.00304 --- -0.00105 0.00104 -0.00228 0.00104 -0.00233 (0.00220) (0.00176) (0.00190) (0.00138) (0.00234) (0.00175) (0.0033) (0.00094) (0.00340) (0.00094) (0.00335) Tu no e 0.000184 0.000300 0.000199 0.0011* 0.00157** 0.00179** (a i als & exi s) (0.000365) (0.000412) (0.000365) (0.00064) (0.00082) (0.000814) Sha e o empo a y 0.105 0.0847 0.157* --- -0.0114 -0.00426 wo ke s (in %/100) (0.0768) (0.0920) (0.0815) (0.0598) (0.0565) Employmen p o ec ion 0.161** 0.159** --- -0.0984 -0.0122 --- legisla ion (0=weak,…) (0.0638) (0.0680) (0.0782) (0.1597) Employmen in 6.746 4.146 --- 3.081 5.511** --- manu ac u ing (in %/100) (5.214) (4.832) (1.909) (3.0594) Unemploymen a e -0.359 -0.905 --- -0.0436* 0.6342 --- (in %/100) (1.980) (2.365) --- (0.2449) (0.641) P ospe i y le el (annual 0.245 0.179 0.000921 0.3009*** --- change in GDP in %/100) (0.156) (0.172) (0.000630) (0.1112) 3-Shi sys em (yes=1) n/a n/a n/a 0.0201 0.0290* 0.0268 (0.0155) (0.0174) (0.0173) Cons an -3.049 -1.400 -1.242 (2.060) -0.1187 -0.8429* -0.595 (1.977) (1.045) (0.4345) (0.5814) (0.611) Obse a ions 471 471 0.239 471 4,810 2,123 2,123 R-squa ed 0.272 0.250 14 0.053 0.064 0.061 Numbe o uni 14 14 143 143 143 Robus s anda d e o s in pa en heses | *** p<0.01, ** p<0.05, * p<0.1 + all es ima ions including mon h and wea he dummies & age and enu e coe icien s o a ia ion | only main esul s p esen ed | n/a= no a ailable Employee Absen eeism: De e minan s in he In e na ional Con ex 65 Employee Absen eeism: De e minan s in he In e na ional Con ex 72 ide suppo o hypo hesis H13 – he e is no link be ween absence and acu e illness meas- u ed by in luenza s a is ics de ec able in ou da a. Wo king condi ion a iables (COND) Following empi ical esul s by Slany e al. (2014), Mo ikawa e al. (2001) and Fekedulegn e al. (2013), we p edic in H14 ha wo king hou s o he han he usually applied one-shi sys ems (day) inc ease absence due o nega i e heal h and social impac s coming along wi h i egula and/o a ypical wo king hou s. In ou es ima ions, we obse e wo-shi sys- ems o inc ease absence a es in compa ison o he e e ence one-shi sys em a bo h in- e na ional plan s B and C by 2.8 (IV-2) and 2.3 (IV-3) pe cen age poin s. Fo plan A, we ind a signi ican e ec in he opposed di ec ion (IV-1). Howe e , we belie e his esul o be highly biased since a plan A only h ee uni s wo ked a one-shi sys em o only six mon hs, while ope a ing on a wo-shi sys em o he emaining obse a ion pe iod. Con- ce ning absence spell du a ion, pooled es ima ions o bo h olume ca p oduce s C and D e eal s a is ically signi ican e ec s o wo-shi and h ee-shi sys ems (II-4, -5, -6). Plan -le el es ima ions o wo-shi sys ems a plan s B (V-2) and C (V-3) and o a h ee- shi sys em a plan C (V-3) suppo his inding. Re e ing o absence equency, we ind a weak suppo o he nega i e impac o a wo-shi sys em a plan s B (VI-2) and C (VI- 3) and a h ee-shi sys em a plan s C and D (III-5). On-si e expe s acknowledge ha nigh shi s a e usually he mos exhaus ing wo king hou s o he majo i y o he wo k o ce as hey a e opposing he human sleeping pa e n. In pa icula , he ansi ion om nigh shi o he subsequen day shi is excep ionally bu dening. In e es ingly, expe s epo ha absence is lowes in nigh shi weeks since employees do no wan o o ei nigh shi p emiums which can amoun up o 30% o hei no mal pay (see S ein (2015) o empi ical suppo ). S ill, we a e able o pa ially con i m ou hypo hesis H14. Addi- ional shi s signi ican ly inc ease employee absence a some plan s. Con ol a iables (CONT) Al hough no hypo heses we e de i ed o da e and wea he con ols, ou es ima ions e eal some in e es ing insigh s. 31 A bo h olume ca p oduce s, absence a es and spell du a ion a e signi ican ly lowe du ing mos summe mon hs om Ap il o Sep embe in compa i- son o he e e ence pe iod (Janua y) – e en when in luenza ac i i y and wea he is con- olled o . Howe e , his phenomenon is mos likely a ibu able o he summe holidays 31 Resul s o con ol a iables a e no abula ed in his pape . Employee Absen eeism: De e minan s in he In e na ional Con ex 73 as absence is usually no epo ed while being on holiday. Findings o absence equency a e weak and inconsis en wi h no clea pa e n. Conce ning wea he , we ind suppo o a gumen s by Shi and Sku e ud (2015) who s a e ha employees adap hei absence be- ha io o he wea he condi ions. A he Spanish si e, we ind absence a e, du a ion and equency all inc ease signi ican ly wi h empe a u e. Fo ins ance, uni -le el absence spell du a ion inc eases by 0.47 days wi h each deg ee Celsius a he 1%-le el o signi icance. This inding is in line wi h he imp essions o on-si e expe s a his plan . Same esul s a e ound a leas o absence spell du a ion a plan D (0.2 days, 1%-le el). No o he e ec s o empe a u e, sunshine hou s and p ecipi a ion a e ound. In o he wo ds, we obse e a link be ween wea he and absence a some plan s indica ing ha employees may adap hei absence beha io o a o able wea he condi ions. This beha io can be in e p e ed as shi king. In o de o con ol o he obus ness o ou indings, we e-es ima e ou models using quan ile eg ession. 32 Al hough his app oach deli e s addi ional signi icance a some quan iles, o e all esul s emain he same. Fu he mo e, we pool ou da a o e all ou plan s. Ye , we ail o ind o e all s a is ical pa e ns. E en ually, we limi ou da a o uni s wo king in he assembly line since assembly line p oduc ion is he only di ision ha is obse ed in all ou plan s. Ye , his does no deli e u he insigh s on po en ial de e mi- nan s o wo ke absen eeism. Un o una ely, ou s udy en ails limi a ions a some ins ances. Fi s , by ocusing on agg e- ga ed da a we gain e idence on eam-le el de e minan s o absence, ye a he expense o limi ed insigh s on indi idual-le el in luences. Second, al hough using da a on absence spell du a ion and equency o p oxy o olun a y and in olun a y absence, ou s udy is limi ed in so a ha bo h measu es may con ain hei espec i e coun e pa as well (Thomson, G i i hs, Da idson 2000). Thi d, we a e no able o con ol o indi idual o g oup-le el heal h condi ions o employees o he han by applying na ion-wide WHO in- luenza ou b eak s a is ics. In doing so, we a e qui e con iden o a leas p oxy o em- ployee heal h condi ions conce ning acu e lu pandemics. Howe e , we do no ha e an app op ia e measu e o o he illnesses such as musculoskele al diso de s o s ess-induced illnesses p o oked by mono ony, epe i i eness and cycle ime. Ye , we assume physical as well as men al sicknesses o be dis ibu ed homogeneously among shop- loo eams a e 32 Resul s a e no epo ed in his pape . Employee Absen eeism: De e minan s in he In e na ional Con ex 74 con olling o age and gende . Fou h, by wo king on mon hly in e als we only p o ide e idence on de e minan s ha a e ela i ely s able o e ime (Ha ison, Ma occhino 1998). Con e sely, his means ha daily o o he sho - e m e ec s on absen eeism a e le eled ou o e he agg ega ed mon hly pe iods and, he e o e, could no be s udied in his pape . 33 Fi h, we a e no able o con ol o educa ional and cul u al in luences on absence beha io . This would ha e been pa icula ly in e es ing since discussions wi h on-si e ex- pe s ha e shown ha absence cul u es a y widely be ween cul u al sphe es. Addi ionally, we a e no able o include mo e de ailed wo kplace cha ac e is ics o he han he imple- men ed shi sys em o assu e he anonymi y o eams. E en ually, since his pape is buil on inside da a, indings migh su e om a lack o gene alizabili y o o he wo king con- di ions and o ganiza ions. 3.6 Conclusion and Implica ions Using a hi he o una ailable da a se ga he ed om inside econome ic da a o ou in e - na ional manu ac u ing plan s o a la ge Eu opean au omobile company, we a e able o p o ide e idence on de e minan s o absen eeism. In o al, we in es iga e eam-le el ab- sence o blue-colla employees o ganized in 160 wo k uni s (app ox. 2,150 wo ke s). Based on a comp ehensi e li e a u e e iew we de i e ou een hypo heses co e ing ou main a eas o po en ial de e minan s on absence: social pee in luences, economic in lu- ences / incen i es, wo ke cha ac e is ics and wo king condi ions. We ocus ou analysis on h ee uni -le el dimensions o absence: absence a e (pe cen age o scheduled ime los due o absence), absence spell du a ion (a e age numbe o days los pe absence occasion) and absence equency (numbe o absence inciden s di ided by uni size). Ou empi ical analyses ollow a wo-s age esea ch s a egy ha complemen s inside econome ic da a analyses wi h in e iews wi h on-si e expe s. In e es ingly, we obse e e ec s o a y by p oduc ion si e, howe e , wi h only e y ew consis en indings. Since a de ailed desc ip- ion o esul s by plan has been gi en h oughou chap e 3.5, we e ain om ecapi ula - ing e ec s by si e in wha ollows, bu ins ead summa ize he o e all e ec s. Conside ing social le el de e minan s, we p edic ed u no e o dec ease absence since g oup no ms a e less en o ceable when eam composi ion is changing. Ye , esul s a e 33 Al hough empi ical e idence is missing a his poin , we we e able o gain some anecdo al insigh s by means o he expe s’ discussions. Fo ins ance, expe s con i med indings by Rosenbla , Shapi a- Lishchinsky and Shi om (2010) o inc eased absence a ound school holidays and weekends (“Monday mo ning lu”). A one plan , expe s could also iden i y highe absence a es a ma ch days o he local socce eam in he a ec ed shi and, hus, suppo a gumen s by Thou sie (2004). Employee Absen eeism: De e minan s in he In e na ional Con ex 75 poin ing in he opposi e di ec ion, p esumably due o a eeling o unce ain y and he lack o an o e all eam spi i in eams wi h high u no e . In acco dance wi h ou expec a ions, we obse e absence o inc ease wi h uni size. Al hough esul s may sugges shi king, we ollow on-si e expe s in a guing ha line manage s’ a endance managemen may be mo e ca ing and amilia in smalle eams. Wi h ega d o pee e ec s, he sha e o empo a y eam ma es is posi i ely ela ed o olun a y absence o pe manen wo ke s, p esumably due o shi king. Su p isingly, he sha e o wo ke s wi h heal h impai men s ails o each s a is ical signi icance in all model speci ica ions. E idence on economic in luences and incen i es e eals he expec ed posi i e link be ween employmen p o ec ion laws and absence. In o he wo ds, wo ke s end o inc ease hei absence beha io when eeling p o- ec ed by s ic employee igh s. Mo eo e , we hypo hesized a nega i e ela ionship be- ween absence and ou side op ions exp essed by na ional (un)employmen igu es. Su p is- ingly, we obse e his ela ion o be inconclusi e. Fu he mo e, we a e able o con i m he expec ed posi i e link be ween p ospe i y le el and absence beha io . This e ec seems o be a ibu able o a ce ain eeling o job secu i y since on-si e expe s obse ed he in e se ela ionship du ing he inancial c ises o he p e ious yea s. We we e no able o p ope ly s udy he e ec s o na ional sickness bene i sys ems on absence since he company a hand g an s addi ional sick pay o compensa e o weaknesses in s a u o y sickness bene- i s. As a consequence, di e ences in na ional legisla ion a e le eled ou . Rega ding wo k- e cha ac e is ics, esul s sugges ha absen eeism inc eases wi h he sha e o emale em- ployees wi hin a uni . Acco ding o on-si e expe s, his e ec is o a la ge deg ee a ibu - able o he double bu den o wo k and amily obliga ions which is e en oday subs an ially mo e p onounced o emales han o males. Fo some si es, we we e able o con i m he p edic ed posi i e ela ionship be ween age and absence a e and spell du a ion. Thus, i seems ha mee ing wo kplace bu dens is subs an ially mo e demanding wi h age. Ye , agains ou expec a ions we obse e weak e idence ha olun a y absence, oo, inc eases wi h age. Resul s o mean o ganiza ional enu e and acu e heal h condi ions exp essed by na ional in luenza ou b eak s a is ics a e inconclusi e. Consis en wi h he hypo hesized ela ionship, we ind wo-shi as well as h ee-shi sys ems o induce signi ican ly highe absence han he s anda d one-shi sys em. Acco ding o on-si e expe s his e ec is la gely a ibu able o he non-co espondence o wo king hou s and human bio- hy hm as well as social li e. All indings a e obus on a ious speci ica ions and su i e a la ge numbe o obus ness checks. Employee Absen eeism: De e minan s in he In e na ional Con ex 76 Subsequen o he da a analyses, we conduc a o al o 18 expe in e iews a bo h in e na- ional plan s wi h shop loo s a , wo ke ep esen a i es, line supe iso s and HR manag- e s in o de o double-check ou own expe ience om he na ional plan s. These in e iews helped us o u he in e p e ou esul s and o lea n mo e abou he ins umen s o a end- ance managemen in p ac ice a o he loca ions. Based on ou empi ical indings and he p ac ical expe ience o on-si e expe s, we yield some impo an implica ions ha may help o inc ease bo h o e all heal h condi ions and a endance o blue-colla wo ke s. The esul s on uni size seem o sugges ha small uni s a e ad an ageous o e la ge uni s in e ms o employee a endance. As discussed, we p edic ed his phenomenon o a ise om be e shi king op ions in la ge eams. Howe e , we lea ned om expe s ha quali y o and amilia i y wi hin he manage -employee ela ionship is mode a ed by eam size. We, he e o e, sugges manage s o in es in de eloping us ul and since e social ies wi h hei subo dina es based on a ai and open cul u e. O cou se, building up in ensi e ela ions wi hin eams equi es ime line manage s usually do no ha e. Thus, we call o HR and plan managemen o suppo line manage s by g an ing a su icien amoun o ime o in es in socializing wi h hei eams. We belie e he long- e m ou comes in e ms o educed absen eeism will le e age any ini ial ime loss. Wi h ega d o economic in luences i seems o be in easible o a company o ake ac ions since poli ics (e.g. sickness bene i s, employmen p o ec ion) as well as o e all economic si ua ion (e.g. unemploymen , p ospe i y le el) a e usually ou o he ange o in luence o a company. Ye , i migh be a easonable s a egy o openly communica e he cu en eco- nomic si ua ion o he company, in pa icula he compe i i eness on he global ma ke s, o he employees o inc ease hei awa eness o po en ial menaces. This could mo i a e em- ployees who eel oo secu e. As men ioned abo e, bo h in e na ional plan s compensa e o weaknesses in s a u o y sick pay by o e ing addi ional sickness bene i s. Howe e , a one si e employees wi h mo e han h ee non-wo k ela ed absence occasions pe 12 mon hs a e ge ing an o icial wa ning and a e wi hheld addi ional sickness bene i s on hei nex absence inciden ( epea ed absence a ising om known long- e m sicknesses e.g. cance is no aken in o accoun ). Acco ding o on-si e expe s, his policy has p o en o be success- ul in educing absence equency. O e all, empi ical e idence on en i onmen al de e mi- nan s is poin ing in he di ec ion ha d awing in e na ional compa isons o absence igu es wi hin he company migh be biased due o na ional a iance in laws and economic si ua- ions. In o he wo ds, c i ical assessmen s o in e na ional loca ions a e no necessa ily Employee Absen eeism: De e minan s in he In e na ional Con ex 77 accu a e and ue and, as a consequence, policy decisions based on in e na ional compa i- sons may be misleading. In esponse o empi ical e idence con i ming he p edic ed posi i e link be ween age and absence, we emphasize he impo ance o ha nessing all possible heal h-p omo ing po en- ials in e gonomic wo kplace design and occupa ional heal h ca e (OHC). We acknowledge ha a all ou plan s wo king condi ions a e cons an ly imp o ing in e ms o e gonomics. Addi ionally, i seems impo an o men ion ha medical a endance o musculoskele al illnesses by OHC is e y comp ehensi e and high in quali y. The same is ue o medical check-ups o e ed o employees o ee in he company’s on-si e OHC cen e s. Howe e , we lea ned om local expe s ha wo k- ela ed and non-wo k ela ed men al s esso s a e o inc easing impo ance e en in manu ac u ing. Expe s e eal ha on-si e OHC cen e s s ill lag in de eloping a s a egy o ea s ess-induced illnesses as comp ehensi ely as musculoskele al illnesses. Mo eo e , some indings emphasize he ad an ageousness o eams ha a e mixed wi h ega d o age since younge wo ke can bene i om olde ones and ice e sa. I , he e o e, seems easonable o yield a he e oge- neous eam composi ion. Conside ing empi ical e idence sugges ing emales o su e om he job- amily double bu den mo e han men i seems impo an o suppo amilies in dealing wi h he econcili- a ion o wo k and amily li e. F om on-si e expe s we lea ned ha a di e en se o sup- po ing ac i i ies is in p ac ice a each si e. Ins umen s ha enhance indi idual lexibili y ha e p o en pa icula ly help ul in inc easing a endance o pa en s. We, he e o e, sugges o e ing pa en s high lexibili y gi en o ganiza ional cons ains. Fo ins ance, a one loca- ion pa en s can wo k educed hou s e en when wo king shi s. He e, wo pe sons on e- duced shi s sha e one job. In si ua ions whe e bo h pa en s a e employed on he shop loo , hey a e o e ed opposing shi s i necessa y o gua an ee child ca e a ound he clock. A some plan s, employees can use dependency lea e (o hi d-pa y lea e) when child en o elde s a e ill and need in ensi e ca e. Al hough hi d-pa y lea e has p o en o be e y use ul in allowing pa en s o ca e o hei sick child en wi hou ha ing o ake own illnesses, employees epo ha making use o his ins umen is o en c i icized by supe iso s and HR. G ea e unde s anding seems o be necessa y a his poin . E en ually, ou indings on shi sys ems p omp he hough ha shi -wo k should be a oided as we obse e wo-shi and h ee-shi sys ems o inc ease absen eeism in com- Employee Absen eeism: De e minan s in he In e na ional Con ex 78 pa ison o he no mal one-shi (day) sys ems. Ye , om local expe s we lea n ha i is some imes p e e able o add a second o hi d shi o be e dis ibu e wo kload ins ead o using o e ime. I shi wo k is una oidable, i seems impo an o ollow he la es scien- i ic insigh s on heal hy shi wo k design. In gene al, he indings o his wo k depic an impo an con ibu ion o (pe sonnel) eco- nomics absence esea ch. In pa icula , we bene i om excep ional and unique inside da a ha allow o comp ehensi e econome ic analyses. Howe e , in his s udy we only obse e au omobile wo ke s. As a consequence, we migh su e om selec ion bias ha en ails endogenei y p oblems – ye a common issue in inside econome ic s udies. We, he e o e, encou age o he esea che s o u he s udy absen eeism in di e en o ganiza- ional con ex s in o de o gain a comp ehensi e idea o he unc ioning o employee ab- sen eeism and o e ine ways o manage a endance. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 79 4 ON THE ROAD AGAIN: CROWDING-OUT EFFECTS OF EXTRIN- SIC MOTIVATION IN COMMERCIAL TRUCKING 4.1 In oduc ion Today, he use o a iable compensa ion schemes in he o m o exogenous inancial in- cen i es has become a common p ac ice in o ganiza ions o align in e es s o p incipals and agen s. 34 In his pape , we seek o s udy how ex insic ewa ds may in luence employ- ee e o choice decisions and, hus, wo ke p oduc i i y. We analyze pe o mance unde a ying incen i e schemes based on unique da a on comme cial uck d i e s. The ucking indus y appea s pa icula ly sui able o his pu pose since he ca ie -d i e ela ionship ep esen s he classic p incipal-agen se ing as modeled by economic heo y (e.g. Oye , Schae e 2011) wi h agen s p oducing hei ou pu a beyond he moni o ing ange o he p incipal. 35 Addi ionally, uck d i e pe o mance is en i ely achie ed on an indi idual le el and, he e o e, is no p one o any kind o eam bias. 36 To sol e his e y speci ic p incipal-agen p oblem, asse owne ship is o en conside ed a pa icula ly cheap and e i- cien moni o ing ins umen . Ye , i s applicabili y is limi ed since uck d i e s a e as- sumed o be isk a e se and limi ed in hei access o capi al (e.g. A unada, Gonzalez- Diaz, Fe nandez 2004, Nicke son, Sil e man 2003, Sheikh 2007). Recen ly, mode n GPS- based lee managemen sys ems ad ance d i e moni o ing om only obse ing he sim- ple ou pu (e.g. a i al ime and goods ca ied) o mo e comp ehensi e eal- ime su eil- lance o d i e pe o mance while on he oad (e.g. Bake , Hubba d 2004, Ba la e al. 2010, Hubba d 2000). This allows o new app oaches in incen i e design o d i e s. T ucking is, by and la ge, an inc easingly compe i i e business en i onmen o haule s as well as d i e s. Nowadays, uel cos s accoun o a ound 30% o he o al li e-cycle cos s o a hea y du y uck (e.g. Schi le 2003). In his si ua ion, indings ha only sligh ly in- c ease ope a ional e iciency may le e age p o i abili y and, he e o e, a e app ecia ed by p ac i ione s. A he same ime, he oad anspo a ion indus y is a subs an ial consume 34 Exogenous ma e ial incen i es as unde s ood in his pape o comp ise any wage con igu a ions ha a leas pa ially in ol e a iable emune a ion such as bonus paymen s, piece a es and pay o pe o - mance schemes. Al hough no exis en in he o ganiza ional se ing a hand, he e migh be non-mone a y ex insic incen i es as well (see, o ins ance, seminal pape s on ou namen s by Rosen (1986) and ca ee conce ns by Fama (1980)). 35 Ve non and Meie (2012) discuss he peculia i ies o p incipal-agen se ings in ucking in mo e de ail. 36 S ill, exogenous in luences de e mine d i ing beha io o a la ge ex en , e.g. wea he , oad condi ions and a ic densi y. Resea ch ha does no include hese con ounding impac s mos likely p o ides inco ec and allacious conclusions. Thanks o ou ex ensi e da a, we a e able o con ol o a ious ex e nal in lu- ences. All con ol a iables a e discussed in he emainde o his pape . On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 80 o ossil uels. Wi hin he EU, CO2 emissions o hea y du y ehicles – ucks, buses and coaches – ha e inc eased by 36% be ween 1990 and 2010 and now accoun o abou 25% o o e all CO2 emissions and 5% o all g eenhouse gas emissions om oad anspo (Eu- opean Commission 2014a). Thus, mo i a ing comme cial uck d i e s o adap uel e i- cien d i ing p o ides a p omising oppo uni y o con ibu e signi ican ly o he educ ion o bo h uel consump ion and CO2 emissions o economic and en i onmen al bene i s. This is whe e mone a y incen i es o d i e s come in o play. Academic heo y knows wo opposing e ec s o exogenous incen i es on mo i a ion. Classic pe sonnel economics heo y assumes exogenous incen i es o inc ease employee e o choice and p oduc i i y by aligning he in e es s o p incipals and agen s (e.g. Gib- bons, Robe s 2013, Oye , Schae e 2011). The e is b oad empi ical and expe imen al e i- dence on his issue in a ying o ganiza ional se ings (e.g. Kahn, Sil a, Ziliak 2001, La y 2002, 2009, Lazea 2000b, Shea e 2004). Ne e heless, beha io al incen i e heo y as- sumes exogenous incen i es o c owd ou in insic o social mo i es ha a e ound o be equally impo an in de e mining employee e o choice and p oduc i i y (e.g. Bénabou, Ti ole 2003, 2006, Gneezy, Meie , Rey-Biel 2011). As a consequence, o e all employee pe o mance migh be lowe when being mone a ily incen i ized. Again, he e is comp e- hensi e suppo o hese a gumen s in he li e a u e (e.g. A iely, B acha, Meie 2009, F ey, Obe holze -Gee 1997, Gneezy, Rus ichini 2000a, 2000b). In esponse o he concu en heo e ical assump ions and he con lic ing indings on he e ec s o ex insic incen i es, his s udy aims a analyzing employee e o choice deci- sions unde a ying incen i e schemes gi en he speci ic o ganiza ional se ing o he ucking indus y. We use hi he o una ailable da a on he pe o mance o 37 comme cial uck d i e s. Da a is ga he ed om he GPS-based lee managemen sys em o an in- house haule o a la ge Eu opean uck manu ac u e . The pa icula ea u es o his e- sea ch a e wo old. Fi s , in con as o mos s udies on ex insic incen i es we do no ocus on he in oduc ion o a bonus pay scheme, bu on i s aboli ion. Second, ou analyses a e based on ou pe o mance measu es eco ded by in- ehicle compu e s o he haule ’s lee managemen sys em ha collec eal- ime in o ma ion on uck posi ioning, echnical pa- ame e s and d i ing beha io . This da a allows us o con ibu e o he ongoing discussion on he bene i s and pi alls o ex insic incen i es. I classic pe sonnel economics heo y holds ue, we expec d i e s o pe o m wo se a e he incen i es a e abolished. In con- as , i beha io al incen i e heo y holds ue, we expec d i e pe o mance o inc ease On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 81 al hough d i e s a e no longe ex insically incen i ized. The empi ical e idence o his wo k e eals a signi ican inc ease in d i e pe o mance a e incen i es ha e been abol- ished, independen o he pe o mance measu e. Thus, ou indings con i m assump ions s a ed by beha io al incen i e heo y as ex insic incen i es a e shown o educe employee pe o mance. Following p e ious esea ch (e.g. F ey 1994, F ey, Obe holze -Gee 1997, Gneezy, Meie , Rey-Biel 2011), we a ibu e his phenomenon o he c owding-ou e ec o ex insic incen i es. O he impo an in insic mo i es – such as in e nal compe i ion o d i e s’ en i onmen al belie es – a e li e ally bough o by mone a y incen i es. Wi h his s udy, we con ibu e o he li e a u e in se e al ega ds. Fi s , we add o he cu - en discussion on he ad an ages and d awbacks o ex insic incen i es on wo ke p oduc- i i y by ex ending he empi ical li e a u e on employees’ e o choice decisions unde di e en incen i e schemes. Ou indings con i m he c owding-ou app oach sugges ed by beha io al incen i e heo y. Second, while nume ous s udies analyze he swi ch om ixed pay o pe o mance pay (e.g. Lazea 2000b, Shea e 2004), li le e idence is p o ided on si ua ions whe e incen i es a e comple ely abolished. One o he ew excep ions is a s udy by F eeman and Kleine (2005) who s udy he change om piece a e pay o hou ly wages. As a esul , hey obse e a signi ican p oduc i i y loss. We add o hei esea ch by o e - ing a comp ehensi e analysis o employees’ beha io al esponses o he aboli ion o ex- insic incen i es. Thi d, o ou knowledge we a e he i s o use uck d i e pe o mance da a ga he ed by mode n GPS-based lee managemen sys ems o analyze employee be- ha io . This seems su p ising since his da a consis s o b oad objec i e pe o mance in- o ma ion. We, he e o e, hope o pa e i s way in o pe sonnel economics. Finally, ou wo k con ibu es o he eme ging inside econome ics app oach in oduced in seminal pape s by Ichniowski, Shaw and P ennushi (1997) as well as Lazea (2000b). This ap- p oach applies elabo a e econome ic me hods o (panel) da a o only one o ganiza ion. Thus, i deli e s high-quali y indings on a mic o-pe spec i e, ye a he expense o a lim- i ed gene alizabili y. This pape p oceeds as ollows. The nex pa ag aph discusses he wo compe ing heo ies on he impac o ex insic incen i es on employee beha io . Sec ion h ee desc ibes he unique o ganiza ional se ing o his s udy while sec ion ou explo es he es ima ion s a - egy as well as he esul s o ou econome ic analyzes. E en ually, sec ion i e o e s an in e p e a ion and inal concluding ema ks. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 88 o mance and p oduc i i y a e he mone a y incen i es a e abolished. Howe e , i beha - io al incen i e heo y p o es ue, we expec o ind a educed pe o mance while bonus pay is in p ac ice due o c owding-ou . In o he wo ds, we should obse e pe o mance o inc ease a e he incen i es a e abolished since d i e s migh d aw hei mo i a ion om o he in insic o social mo i es when no longe inancially incen i ized. 4.3 Da a Se and Desc ip i e S a is ics In his pape we seek o analyze beha io al esponses o wo ke s o shi s om a bonus pay scheme o a non-incen i e en i onmen . The e o e, we use a hi he o una ailable panel da a se ha includes ip-wise pe o mance in o ma ion o comme cial long-haul uck d i e s o an in-house haule o a la ge Eu opean uck manu ac u e . The da a is based on in o ma ion eco ded by GPS-based in- ehicle compu e s o he haule ’s lee managemen sys em. Pe o mance da a co e s an ex ended pe iod o hi y-six mon hs (Janua y 2011 o Decembe 2013). Wi h he beginning o he obse a ion window, he haule ins alled a mone a y pe o mance p emium o ewa d good d i e pe o mance. Howe e , a e wo yea s in p ac ice, he haule decided o dises ablish he incen i e scheme again. Wi hin he o ganiza ional s uc u e o he pa en company, he haule ealizes jus -in- ime deli e y o in e media e goods om one p oduc ion plan o he uck manu ac u e o an- o he . While on he oad, d i e s a el oughly 1,250 kilome e s ac oss ou No he n Eu- opean coun ies. To comply wi h d i e s’ manda o y es ing pe iods 44 , he o al dis ance is spli up in o wo s ages wi h a ixed loca ion o he d i e s o es . A his loca ion, he haule ope a es a p emise ha p o ides sleeping, cooking and spo ing acili ies o d i e s du ing hei es . Upon hei a i al a he loca ion, d i e s pass on hei ucks o ano he d i e who con inues he jou ney on he e y same uck. In o he wo ds, d i e A a i es a he es ing loca ion, whe e d i e B has jus inished his manda o y es . D i e B akes o e he uck and immedia ely d i es on o he inal des ina ion, whe eas d i e A s ays o his es and awai s he a i al o he nex uck o con inue his jou ney. Thus, d i e s a el a ound- ip wi h ou s ages on ou di e en ucks be o e e u ning home. This p ocedu e secu es jus -in- ime supply o goods wi hou ans e ing he company’s wa e- 44 Due o Eu opean oad haulage egula ions d i e s a e no allowed o exceed 9 hou s o d i ing pe day wi h a subsequen manda o y es pe iod o a leas 11 hou s. The e a e ew excep ions enabling d i e s o b eak hese egula ions in o de o eac o cu en a ic condi ions o sho - e m wo kload peaks (Eu o- pean Commission 2014b). On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 89 house on o he s ee s. Ob iously, d i e - uck so ing in his se ing is o ally andom since d i e s a e equi ed o ake he nex uck a ailable wi hou ha ing any choice. Ou ini ial da a se includes 81 comme cial uck d i e s esul ing in n=11,439 d i e - ip obse a ions. 45 We ocus exclusi ely on d i e s on pe manen con ac s o a oid con ound- ing e ec s a ibu able o incen i es a ising om empo a y con ac cha ac e is ics as ob- se ed in o he s udies (e.g. chap e i e o his hesis, B adley, G een, Lee es 2012). All ucks a e epo ed o be deployed almos exclusi ely on long-haul assignmen s om one p oduc ion acili y o he o he . To ully exclude sho -haul assignmen s we es ic he da a o ips wi h a leng h be ween 520 kilome e s and 720 kilome e s. We selec ed his ange since he wo p oduc ion acili ies a e 570 km espec i ely 670 km away om he es ing acili y. 46 Mo eo e , we exclude hose obse a ions wi h a epo ed uel consump- ion ou side he ange o minus wo imes and plus ou imes he s anda d de ia ion a ound he mean. Finally, we only include hose d i e s ha we can obse e a leas on 80 ips du ing he obse a ion pe iod (i.e. app oxima ely ou o i e mon hs o egula d i - ing) and ha ha e expe ienced he incen i e scheme o a leas h ee mon hs. These con- s ain s educe he ini ial da ase hi y-se en uck d i e s and n=6,825 obse a ions. 47 We me ged he ip-wise pe o mance in o ma ion wi h demog aphic da a o d i e s (age, gende and enu e). I appea s om he summa y s a is ics p esen ed in Table 4.1 ha he sample is domina ed by male d i e s, a ac ha comes as no su p ise in he ucking indus- y. S ill, h ee emale d i e s a e included in he da a accoun ing o 580 obse a ions (8.5% o o al obse a ions). Repo ed d i e age co e s almos he en i e wo king li e span om 22 o 64 yea s (mean=44.08, s.d.=11.07). This a iance allows o a ho ough analysis o po en ial age e ec s. Tenu e as a p oxy o d i ing expe ience is included on a mon hly le el gi en he exac da e a d i e joined he company (mean=25.59 mon hs, s.d.=12.02). Since we exclude d i e s om he sample ha a e employed on a empo a y con ac , employmen s a us is simila o all d i e s. Mo eo e , we could neglec any cul- 45 The panel da a se is unbalanced due o d i e and ehicle u no e du ing he obse a ion pe iod. Fo ins ance, a his company ucks a e eplaced a e a o al mileage o 1m kilome e s. 46 We ex ended bo h dis ances by +/- 50km o accoun o d i e s who exceed he maximum allowed d i - ing ime be o e eaching hei des ina ion o need o ake al e na i e ou es due o cons uc ion wo k o a ic conges ion. 47 Howe e , no all models could be es ima ed wi h he ull da a se since no all ehicles a e equipped wi h he la es e sion o he ele an on-boa d communica ion uni s. Hence, incomple e in o ma ion p oduces missing alues a some ins ances. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 90 u al di e ences due o he ac ha all d i e s o igina e om he coun y domes ic o he haule . Table 4.1: Summa y S a is ics - D i e Demog aphics Va iable mean s.d. min. max. D i e Age (in yea s) 44.08 11.07 22 64 D i e Gende (male=1) .915 - 0 1 D i e Tenu e (in mon h) 25.59 12.02 1 64 Bonus (Bonus=1) .58 - 0 1 Numbe o d i e s: 37 Numbe o d i e - ip obse a ions: 6,825 The GPS-based in- ehicle compu e s p o ide he haule wi h ex ensi e eal- ime in o - ma ion o e ing conside able oppo uni ies o manage and moni o lee ac i i ies including uck posi ioning, uel consump ion, d i ing beha io , uck handling and ehicle s a us. Table 4.2 p esen s he uel consump ion in li e s pe 100 km (mean=27.89 l, s.d.=3.02 l) as well as a sho lis o he mos impo an d i ing a iables (e.g. b ake applica ions, speed- ing, a e age speed, maximum speed). Table 4.2: Summa y S a is ics - D i ing S yle Va iable mean s.d. min. max. Fuel Consump ion (in l/100 km) 27.89 3.02 21.2 39 A e age Speed (in km/h) 75.32 3.34 42.1 84.7 Idling (in % o engine unning ime) .023 .02 .002 .238 Coas ing (in % o engine unning ime) .14 .05 .007 .399 Speeding (in % o engine unning ime) .079 .11 0 .892 B ake Applica ions (# pe 100km) 12.62 7.83 .2 149.1 Ha sh B ake Applica ions (# pe 100km) .11 .25 0 10.1 Ha sh Accele a ions (# pe 100km) .07 .27 0 4.3 Maximum Vehicle Speed (in km/h) 96.23 5.55 84 114 Numbe o d i e s: 37 Numbe o d i e - ip obse a ions: 6,825 On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 91 The desc ip i e da a al eady e eal in e es ing insigh s. Fi s , idling – which is o en la- beled o be one o he mos impo an de e minan s o uel combus ion (e.g. Lu sey e al. 2004) – is no o majo ele ance in he o ganiza ional se ing a hand (mean=2.3% o en- gine unning ime). This migh be due o he ac ha d i e s spend mos o hei ime on highways and a ely ace s op-and-go a ic in ci ies. Second, d i e s a e equi ed by na- ional legisla ion o he ansi coun ies o d i e 80 km/h a max. Howe e , maximum al- ues o a e age speed (max=84.7 km/h) and maximum ehicle speed (max=114 km/h) sugges ha d i e s o e ide na ional speed limi s occasionally. This migh indica e ha a imes d i e s need o ca ch up o a ic-induced delays o secu e jus -in- ime supply. Thi d, he simple min-max compa ison o se e al d i ing pa ame e s e eals a high he e - ogenei y (e.g. b ake applica ions a y om .2 o 149.1 pe 100 km). This indica es ha d i ing s yle migh o a la ge deg ee be in luenced by exogenous e ec s such as a ic densi y o wea he condi ions. To add ess hese issues we add ex e nal in o ma ion o he da a ha enables us o con ol o exogenous condi ions. All con ols will be discussed in de ail in he nex sec ion. Based on he in o ma ion eco ded by i s in- ehicle de ices, he lee managemen so - wa e is p og ammed o compu e sco es ha assess d i e pe o mance in ela ion o a ge alues p ede ined by he haule . These sco es co e d i e s’ an icipa ion beha io , choice o gea , use o b akes and hill d i ing beha io . 48 In addi ion, he so wa e uses hese sco es as well as u he d i ing pa ame e s o assess o e all d i e pe o mance o each ip as being ei he “good”, “medioc e” o “poo ”. Exp essed by a g een-yellow- ed a ic ligh isualiza ion, he ip e alua ion is epo ed back o he haule ia he lee manage- men sys em. Fo anspa ency easons, all esul s a e accessible o all d i e s ia an online ool. Al hough uel consump ion is no di ec ly e alua ed by he so wa e, a “good” d i ing beha io is cong uen wi h an eco- iendly d i ing s yle. I appea s om Table 4.3 ha d i ing sco es on a e age display high means a ying om 63% o 81%. Ye , we ques ion he choice o gea sco e since a mean o 97% appea s o a ise om he usage o au oma ic ansmission ehicles in he majo i y o he lee . The limi ed numbe o obse - a ions on his sco e i s his pic u e. Conce ning he a ic ligh e alua ion sco es, exac ly hal o all ips a e e alua ed as being “good” (50%), while sligh ly ewe ips a e epo ed as being “poo ” (45%). Only a small numbe is a ed as being “medioc e” (5%). 48 The unde lying algo i hms ha assess d i e pe o mance and gene a e sco es a e co po a e sec e s and we e no e ealed o us. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 92 Table 4.3: Summa y S a is ics - T a ic Ligh Pe o mance E alua ion Tool Va iable mean s.d. min. max. Obse a ions* D i ing Sco es An icipa ion (Sco e in %/100) .81 .16 0 1 5,251 Choice Gea (Sco e in %/100) .97 .14 0 1 3,673 Use B akes (Sco e in %/100) .77 .19 .1 1 5,064 Hill D i e (Sco e in %/100) .63 .25 0 1 5,028 T ip E alua ion O e all (1=g een, 2=yellow, 3= ed) 1.94 .97 1 3 6,801 Good (yes=1) .50 - 0 1 - Medioc e (yes=1) .05 - 0 1 - Poo (yes=1) .45 - 0 1 - Numbe o d i e s: 37 *Va ying numbe o obse a ions since no all on-boa d compu e s epo comple e da a a all ins ances. Fo an o e iew o mean, s anda d de ia ion, as well as wi hin and be ween a iance o all dependen a iables lis ed sepa a ely o he pe iod wi h incen i es in p ac ice and he pe- iod a e incen i es ha e been abolished, see Table A.6 in he appendix. Ou s udy design has nume ous ad an ages and con ibu es o pe sonnel economics in se - e al ega ds. Fi s , he o ganiza ional se ing allows s udying he e ec s o he aboli ion o incen i es. As men ioned be o e, he haule ins alled a pe o mance bonus as o Janua y 1s 2011. Based on he e alua ion sco es epo ed by he lee managemen so wa e, d i e s we e ewa ded inancially o good pe o mances. We like o poin ou again ha low uel consump ion is only indi ec ly ewa ded. The maximum amoun a d i e could ea n on incen i es did no exceed 5% o his mon hly ix pay. Two yea s la e , he bonus sys em was abolished by he haule a Decembe 31s 2012. In e es ingly, he haule ins alled he pe o mance bonus wi hou elling he d i e s o he i s h ee mon hs. Despi e he de- layed disclosu e, we doub ha he in oduc ion o he incen i e scheme emained com- ple ely unno iced by d i e s wi hou any umo s sp eading o p ema u e in o ma ion leaks. This would bias any indings. Thus, we ocus ou main es ima ions exclusi ely on he abo- li ion o he bonus a he end o Decembe 2012. This seems jus i ied since he e is b oad e idence on p oduc i i y e ec s ollowing he ins alla ion o incen i es (e.g. Lazea 2000b, Shea e 2004), bu o ou knowledge we a e among he i s esea che s who wo k on he On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 93 aboli ion o an incen i e scheme in p ac ice, wi h F eeman and Kleine (2005) being a a e excep ion. Ou se up, he e o e, migh o e new insigh s on he unc ioning o incen i es ha ha e no been co e ed by academic esea ch so a . Second, compa isons among uck d i e s a e o en hampe ed due o inconsis en exogenous in luences coming along wi h di e en assignmen s, des ina ions and ou es. Since he ound- ip in he o ganiza ional se ing a hand is iden ical o all d i e s and cons an o e ime, we can assume each d i - e o be exposed o compa able exogenous in luences while on he oad (e.g. oad condi- ion, cons uc ions, de ou s e c.). This allows o comp ehensi e analyses o d i e pe o - mance. Thi d, d i e - uck so ing is o ally andom in he o ganiza ional se ing o his s udy since d i e s canno in luence which uck is he nex a ailable a ei he s age o hei ound- ip. Thus, we a e able o con ol o good o poo d i e - uck ma ches. Fou h, we can neglec any selec ion bias o agen s so ing in o speci ic incen i e schemes since mos d i e s we e al eady employed p io o he ins alla ion o he incen i e scheme. Fi h, de ailed da e and ime in o ma ion o each ip allows us o con ol o exogenous e ec s (e.g. wea he condi ions) and empo a y a ic densi y due o holiday a ic o dai- ly ush hou s. Bo h wea he and a ic densi y can be assumed o highly de e mine uck handling and d i e pe o mance. Finally, ehicle load ac o s need o be simila o e ime o a oid biases a ising om unequal ehicle maneu e abili y. Da a by he Eu opean Commission (2014c) sugges ha almos one qua e o all ehicle-kilome e s o ucks wi hin he EU is un emp y. 49 Howe e , he haule epo s a e age load ac o s abo e 90%, hus we can assume simila ca go weigh s and equal maneu e abili y o ucks o all ips. This allows us o neglec his issue o ou da a. 4.4 Empi ical Models and Es ima ion Resul s Ou da a o e s b oad oppo uni ies o measu e uck d i e pe o mance. In o al, ou es- ima ions a e based on ou dependen a iables ha p oxy o d i e s’ e o choice and p oduc i i y. Fi s , we use he a e age uel consump ion in li e s pe 100 kilome e s. Al - hough low uel consump ion is no di ec ly incen i ized, all a ge alues se wi h ega d o d i ing pa ame e s and pe o mance e alua ions a e based on eco- iendly d i ing beha - io and esul in low uel use. Since d i e s ecei e indi idual uel e iciency aining on a non- egula basis, we can assume hem o be well in o med on how o mee company 49 Acco ding o he Eu opean Commission (2014c) his is mainly a ibu able o s ic c oss-bo de ansi egula ions. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 94 s anda ds. 50 D i e s de ia ing om hese s anda ds, he e o e, may be assumed o ha e chosen low e o le els. Thus, we a e con inced ha uel consump ion used as dependen a iable may se e as a good p oxy o d i e s’ e o choice decisions. In ou sample, a - e age uel consump ion is 27.89 l/100 km wi h a s anda d de ia ion o 3.02 li e s (see Ta- ble 4.2). Second, we deploy he mos impo an d i ing pa ame e s such as speeding, b ake applica ions and a e age speed (see Table 4.2). Due o uel e iciency aining, d i e s know how o mee company s anda ds. Di e ing alues, he e o e, may again be an app o- p ia e p oxy o d i e e o choice. Thi d, we use he ou main d i ing sco es (an icipa- ion, use o b akes, choice o gea , hill d i e) eco ded by he in- ehicle compu e s (Table 4.3). These sco es di ec ly in luence ip e alua ion and a e highly ele an in de e mining he amoun o bonus paid. Full online excess o he lee managemen da a base should aise d i e s’ awa eness on how hey pe o med on each sco e. Hence, we assume d i e s o know hei indi idual s eng hs and weaknesses in d i ing beha io qui e accu a ely. O he wise pu , d i e s a e awa e o how hey can inc ease indi idual bonus pay by im- p o ing on hose sco es ha do no ye mee company s anda ds. In ou es ima ions, we model low d i ing sco es o p oxy o low e o choices o d i e s. Fou h, we deploy in e nal a ic ligh e alua ion esul s based on so wa e calcula ions (see Table 4.3). Again, d i e s a e o e ed ull online excess o he e alua ion epo s ha a e ips as ei- he “good”, “medioc e” o “poo ”. In o he wo ds, d i e s know abou hei indi idual pe o mance and should be awa e o any misconduc . 51 Since hey ha e bo h he abili y and he knowledge o d i e acco ding o company s anda ds, we belie e ip e alua ions o he han “good” o esul om low e o choice decisions. All ou pe o mance measu es a e insensi i e o any subjec i e a e bias since hey a e eco ded and assessed by in- ehicle compu e s au oma ically. While uel consump ion, d i ing pa ame e s and e alua ion sco es a e ca ego ical a iables, ip e alua ion is o o dinal na u e. The es ima ed models ha e he ollowing gene al o m (wi h FUELC being he uel con- sump ion, DRIVPAR ep esen ing he d i ing pa ame e s, SCORE ep esen ing he e alua- ion sco es and EVAL being he a ic ligh e alua ion epo s): 50 Acco ding o an Mie lo e al (2004) (ca ) d i e s know by in ui ion how o d i e eco- iendly. Thus, i does no eally ma e ha we ha e no in o ma ion nei he on he equency no he exac da es o indi- idual uel e iciency aining since we belie e any aining e ec o a ise om ecollec ion ins ead o new insigh s. 51 Acco ding o Huang e al. (2005) and Roe ing e al. (2003), uck d i e s app ecia e eedback om ech- nology. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 95 FUELC DRIVPAR SCORE EVAL = α + βDRIV + γVEH + δTRAF + φWEA + ωCONTR + ε (1) (2) (3) (4) whe e α is he cons an , DRIV desc ibes d i e pe o mance as exp essed by equa ion (5), VEH con ols o ehicle e ec s, TRAF desc ibes a ec o o ex e nal a ic condi ions, WEA desc ibes a ec o o wea he condi ions, CONTR is a ec o combining u he con- ols, β, γ, δ, φ and ω a e he es ima ed coe icien s and ε is he e o e m. The independen a iables a e discussed in mo e de ail below. D i e pe o mance (DRIV) is modeled as a unc ion o he ollowing o m: DRIV = ((AGE + HUMC) * EFCH) (5) wi h EFCH = In insic Incen i es + Ex insic Incen i es (6) whe e AGE measu es d i e age and HUMC p oxies d i e human capi al as d i ing expe- ience wi h he haule (mon hly enu e). Age and expe ience a e impo an pe sonal cha - ac e is ics in de e mining d i e pe o mance. On he one hand, age has been p o en o be de imen al o impo an abili ies ele an o d i ing. Fo example, olde d i e s display an ex ended esponse ime as well as lowe isual and psychomo o abili ies (e.g. Llane as e al. 1998). On he o he hand, expe ience is ound o compensa e hese age- ela ed impai - men s (e.g. B ock, Llane as, Swezey 1996, Gues , Boggess, Duke 2014). The impo ance o expe ience is e en mo e p onounced when aking acciden isk in o accoun since younge and less expe ienced d i e s ha e a highe isk o be in ol ed in acciden s (e.g. Häkkänen, Summala 2001, McCall, Ho wi z 2005, Rod iguez, Ta ga, Belze 2006). Thus, any analysis wi hou including age and expe ience in o he es ima ions would p oduce bi- ased esul s. In lack o in o ma ion on skills and quali ica ion, we in e p e age and expe i- ence as he o e all pe o mance po en ial o employees. Howe e , he pe o mance po en- ial is a any momen subjec o a d i e s’ indi idual e o choice decision (EFCH). Pe - o mance po en ial and e o choice e en ually ansla e in o ac ual d i e pe o mance. We model EFCH o be he join e ec o bo h in insic and ex insic pe o mance incen i e (see equa ion (6)). While in he o ganiza ional se ing a hand in insic incen i es migh include mo i es such as social p e e ences, al uism o eco- iendliness, ex insic incen- i es comp ise he mone a y pe o mance pay incen i es. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 96 We use u he a iables o con ol o exogenous e ec s ha migh in luence uck han- dling and d i ing beha io as modeled in equa ions (1) o (4). Fi s , we use a ehicle dummy a iable VEH o con ol o di e en uck con igu a ions. A di e en se o ech- nological equipmen migh acili a e uck handling a some ins ances o in luence uel consump ion solely o echnological easons. Second, we use a ec o o da e and ime a iables TRAF ha p oxy o a ic densi y while on he oad. Conce ning he ime o he day, we assume ush hou s in he mo ning (7 o 9 a.m.) and e ening (4 o 6 p.m.) o im- pose high a ic densi y, while d i e s being on he oad a nigh migh bene i om open highways. Dummy a iables con ol o ush hou s and ips a nigh alike ( ips om 8 p.m. o 7 a.m.). Addi ionally, we con ol o day o he week e ec s since we expec e.g. Mondays and F idays o be p one o conges ed oads due o weekend commu e s. Fu - he mo e, we include a dummy a iable co e ing a h ee-day window a ound in e na ional holidays since we belie e oads o be conges ed due o holiday a ic a hese da es. 52 Thi d, a ec o o a iables, WEA, con ols o wea he condi ions du ing each ip. This seem necessa y since p e ious esea ch ound wea he o highly a ec d i ing beha io , gene al a ic condi ions and c ash isks (e.g. Kilpeläinen, Summala 2007, No man, E iksson, Lindq is 2000). We, he e o e, add con ol a iables o empe a u e, p ecipi a- ion, snow dep h and wind speed. 53 E en ually, we apply dummy a iables o all hi y-six mon hs o he obse a ion pe iod o accoun o possible end o iming e ec s. Likewise, we use dummy a iables o all i y- wo weeks o a gi en yea o accoun o seasonali y. Taken oge he , his b oad ange o con ol a iables excludes he majo sou ces o po en- ial exogenous in luences on d i e pe o mance. Thus, we a e con iden ha any pe o - mance e ec ha we will de ec in ou es ima ions is en i ely a ibu able o d i e s’ e o choice decisions as a esponse o he incen i e scheme. In he ollowing, we discuss esul s on uel consump ion as modeled by equa ion (1) i s , hen epo ou indings on d i ing pa ame e s as modeled by equa ion (2). A e wa ds, we p esen esul s on d i ing sco es as modeled by equa ion (3) and e en ually discuss he es ima ions on ip e alua ion as mod- eled by equa ion (4). The key es ima ions a e epo ed in Tables 4.4 o 4.7, espec i ely. O e all, ou esul s suppo he c owding-ou e ec o ex insic incen i es since uck 52 Only hose holidays ha a e manda o y in all ou coun ies d i e s pass h ough du ing hei ound ips a e conside ed (e.g. New Yea ’s Day, Eas e , Ch is mas, e c.). 53 Since we do no ha e de ailed uck posi ioning da a, we ga he ed wea he in o ma ion a a me eo ologi- cal s a ion app oxima ely hal way he o al dis ance. This should le el di e ing clima e condi ions a he s a ing and a i al poin s. Da a was p o ided by he Ge man Me eo ological Se ice (DWD 2014), a Ge man s a e ins i u ion. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 97 d i e s’ e o (and as a consequence hei pe o mance and p oduc i i y) is signi ican ly lowe wi h mone a y pe o mance incen i es in p ac ice. Fi s isual e idence on di e gen le els o uel consump ion wi h and wi hou incen i es in p ac ice is p esen ed in Figu e A.4.1 in he appendix. Compa ing he ke nel densi y unc ions o bo h incen i e en i onmen s, i appea s ha uel consump ion shi s o he le a e incen i es ha e been abolished. This indica es lowe uel use wi hou incen i es in p ac ice. As a i s empi ical es o signi ican beha io al esponses, we un a wo-way ANOVA on he hi y-se en d i e s om Ap il 2011 o Decembe 2013 o examine he e ec o incen i es on uel consump ion as a p oxy o d i e pe o mance and p oduc i i- y (see Table A. 7 in he appendix). We ocus on a ime ame s a ing in Ap il 2011 since he i s h ee mon hs o he obse a ion window migh be biased due o he sec e in o- duc ion o he pe o mance pay. 54 We ound a signi ican e ec o he bonus pay on d i - e ’s uel consump ion indica ing ha signi ican ly di e en quan i ies o uel a e used de- pending on he exis ence and non-exis ence o mone a y incen i es, F(1, 6,252)=4.31, p=.0379. Since he ANOVA does no epo any signi ican di e ences in he simple com- pa ison o d i e s (p=0.5263) o he d i e -bonus in e ac ion (p=0.4610), we a e con iden ha any pe o mance e ec es ima ed in his s udy may be explained by beha io al e- sponses en i ely a ibu able o he (non-)exis ence o he incen i e scheme. Mo eo e , his inding suppo s ou a gumen ha d i e s adap hei indi idual e o choice o he incen- i e en i onmen . To ollow hese indica ions, we apply elabo a e empi ical es ima ions o ou da a. We s a wi h ixed-e ec speci ica ions since we a e pa icula ly in e es ed in wi hin-d i e beha io al esponses o he dises ablishmen o he incen i e scheme. Mo e- o e , ixed-e ec s models accoun o p esen d i e ixed-e ec s ha a e no epo ed in he da a. These e ec s can be ei he ime-in a ian (e.g. alen , p io d i ing expe ience, e c.) o quasi ime-in a ian wi hou any o only sligh modi ica ions o e ime (e.g. d i - ing s yle, en i onmen al iendliness, p o essional e hos, e c.). To accoun o hese unob- se ed e ec s ha a e seldom unco ela ed wi h explana o y a iables, ixed-e ec speci i- ca ions a e p e e able (e.g. Woold idge 2013). Since he incen i e scheme was in oduced wi hou elling he d i e s o he i s h ee mon hs we es ic ou es ima ions o a pe iod om Ap il 2011 o Decembe 2013 o a oid any bias a ising om p ema u e in o ma ion leaks. 54 The sec e ins alla ion migh no ha e been as sec e as in ended by he haule since we assume umo s o ha e ci cula ed among he d i e s p io o he in oduc ion o he new emune a ion scheme. On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 104 s a is ically signi ican in luence on d i e pe o mance in any model speci ica ion. Ye , enu e exhibi s a mino bu signi ican impac on e alua ion pa ame e s wi h an icipa ion sco es educing by .4 pe cen age poin s and use o b akes sco e educing by .5 pe cen age poin s wi h each mon h a d i e is employed a he haule . This esul opposes heo e ical conside a ions on human capi al as an impo an de e minan o employee pe o mance (e.g. Rod iguez, Ta ga, Belze 2006). We belie e dec easing a en i eness as a esul o ou ine o be one possible explana ion since uck d i ing is a e y mono onous occupa ion – especially in se ings wi h cons an ou es and on highways. Howe e , he limi ed obse - a ion window o only 33 mon hs as well as he o e all limi ed enu e (max o 64 mon hs) migh bias ou indings a his poin . Vehicle dummies only seldom each s a is ical signi - icance. Coe icien s o wea he condi ion as well as da e and ime dummies a e incon- sis en o all ou a iables. The e o e, we e ain om epo ing hese esul s in de ail. In a inal s ep, we es ima e models on he a ic ligh ip e alua ion assessed by in- ehicle compu e s ( a iable EVAL in equa ion (1)). We use ip e alua ion as a p oxy o d i e e o choice decisions (EFCH) and, hus, a e con iden o de ec po en ial pe o mance e ec s o incen i e. Following Guja a i and Po e (2009) as well as Woold idge (2013) we es ima e o de ed logi and o de ed p obi eg ession models o add ess he o dinal na- u e o ip e alua ion da a. Howe e , since p obi eg ession only cons i u es a mino modi ica ion o logi eg ession and yields ba ely di e en indings (G eene 2003), we ocus on logi eg ession esul s in his pape due o easons o b e i y. Table A.10 in he appendix displays esul s o he o de ed logi es ima ions wi h ip e alua ion being he dependen a iable. I becomes ob ious ha e alua ions a e signi ican ly wo se when pe - o mance incen i es a e in p ac ice. Pos -es ima ion p obabili ies o ip e alua ion wi h and wi hou incen i es in p ac ices a e gi en in Table 4.7. Table 4.7: Es ima ed P obabili ies o T ip E alua ion in bo h Incen i e En i onmen s Va iables Incen i es in P ac ice Incen i es Abolished Change 95% CI o Change T ip E alua ion Good .4034 .6894 .2860 -.3421 -.2298 Medioc e .0619 .0513 -.0106 .0072 .0139 Poo .5347 .2594 -.2754 .2219 .3289 On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 105 Wi h incen i es in p ac ice he p obabili y o a ip being assessed as “good” by in- ehicle compu e s is 40.34% compa ed o 68.94% a e he aboli ion o incen i es. This is a di e - ence o 28.6 pe cen age poin s o 70.89%, espec i ely. A he same ime, he p obabili y o a ip being e alua ed as “poo ” d opped by 27.54 pe cen age poin s o 51.5% om 53.47% o 25.94%. P obabili ies o “medioc e” ip e alua ions a y only sligh ly by 1.06 pe cen age poin s o 17.2% a e he aboli ion o he incen i es. In o he wo ds, wi h pe - o mance incen i es in p ac ice he p obabili y o a good uck d i e pe o mance is ma kedly lowe compa ed o a si ua ion wi hou any ex insic incen i es. This indica es d i e s o choose lowe le els o e o when mone a ily incen i ized. Simila o he o he esul s o his pape , his inding suppo s he c owding-ou assump ion o ex insic incen- i es back i ing on he e y beha io hey in end o p e en . All es ima ions ep esen ed in Tables 4.5 o 4.7 a e based on he ull model speci ica ion. To con ol o biases a ising om con ol a iables we e-es ima ed all models unde di - e en speci ica ions. Resul s p o ed o be obus o all es ima ions wi h li le a ia ion in coe icien s. The same holds ue when excluding he choice o gea sco e om he SUR models. Findings o he o he dependen a iables emain obus . We u he check o he obus ness o he esul s in wo ways. Fi s , we e-es ima e ou models excluding he h ee women om he da a se . Second, we include all empo a y d i e s in he es ima ions. Al - hough we expec ed an incen i e e ec de i ing om he con ac cha ac e is ics (chap e i e o his hesis, B adley, G een, Lee es 2012), he baseline esul s a y only ma ginally in coe icien s. Thus, we a e e y con iden ha ou indings a e obus . S ill, ou s udy en ails some limi a ions as a consequence o i s o ganiza ional se ing. Fi s , we do no ha e de ailed in o ma ion nei he on he algo i hm ha de e mines pe o - mance e alua ions no on he unde lying mechanisms ha de e mine he o al amoun o incen i es. I would ha e been in e es ing o wo k on he incen i e scheme in mo e de ail, e.g. by obse ing only hose d i e s a he e ge o a highe incen i e le el. Second, he incen i es we e in oduced o he i s h ee mon h wi hou in o ming he d i e s. S ill, he e is he possibili y ha he in o ma ion migh be sp ead p ema u e. Since we expec he esul s o his pe iod o be biased due o in o ma ion leaks, we exclude he espec i e h ee mon hs window om ou baseline es ima ions. Ye , checking o he in oduc ion o incen i es a e h ee mon hs suppo s ou o e all indings (see Table A.8 in he appendix). Thi d, he bonus paymen s incen i ize low uel consump ion only indi ec ly by ewa ding an eco- iendly d i ing s yle. The e o e, one migh ques ion he alidi y o ou indings on On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 106 he link be ween incen i e scheme and uel consump ion. Howe e , all a ge alues o he d i ing pa ame e s and e alua ion sco es se by he company aim a eaching eco- iendly d i ing beha io ha yields a low uel-consump ion. Mo eo e , ou indings on uel con- sump ion a e en i ely suppo ed by es ima ions o aw pe o mance indica o s such as d i ing pa ame e s and e alua ion sco es. Thus, we a e con iden o add ess doub s on he indi ec ela ionship o incen i es and uel consump ion. Finally, ou da a se includes only h ee women. Al hough he e is suppo o a gende e ec in sel -selec ion and pe o - mance unde incen i es (e.g. C oson, Gneezy 2009, Dohmen, Falk 2011), we a oid wo k- ing on his issue o accoun o he unequal gende dis ibu ion among d i e s. Ye , he low sha e o women is indus y speci ic and, he e o e, should no bias o impac ou o e all indings. 4.5 Conclusion Wi h his s udy we o e new insigh s o he ongoing deba e on he bene i s and d awbacks o ex insic incen i es. While classic economic heo y assumes employees o eac o ex- insic (mone a y) incen i es by inc easing pe o mance and p oduc i i y, beha io al in- cen i e heo y sugges s ex insic ewa ds o c owd ou o he impo an in insic and social incen i es ha mo i a e people. The e is b oad expe imen al and empi ical suppo o bo h heo e ical assump ions. To con ibu e o his discussion, we use a hi he o una ailable da a se on he pe o mance o comme cial uck d i e s collec ed om he in e nal lee managemen sys em o he in- house haule o a la ge Eu opean uck manu ac u e . T uck d i e s ecei e bonus pay- men s o good pe o mance wi hin he i s wen y- ou mon hs o he hi y-six mon hs obse a ion pe iod. Subsequen ly, managemen decided o abolish he incen i es. We em- pi ically measu e he e ec o he aboli ion o he incen i e scheme on d i e e o choice decisions by analyzing ou sepa a e se s o dependen a iables o d i e pe o mance: uel consump ion, d i ing pa ame e s, d i ing sco es and ip e alua ion. Consis en wi h beha io al incen i e heo y, we ind comp ehensi e suppo o c owding-ou e ec s o he bonus pay scheme applied by he haule . In o he wo ds, we obse e d i e s o display signi ican ly lowe e o and pe o mance when ex insically incen i ized compa ed o he non-incen i e en i onmen . These indings a e in line wi h p e ious esea ch (e.g. Deci 1971, Gneezy and Rus ichini 2000a). Mo e p ecisely, uel consump ion is signi ican ly highe when incen i es a e in p ac ice, adding up o addi ional 530 li e s used pe d i e On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 107 and yea which ansla es o ex a uel cos s o €800 pe d i e and yea . We in e p e his amoun as hidden cos s o incen i es. Combined wi h he ne bonus paymen s, he hidden cos s cons i u e a se ious inancial bu den o he company’s baseline esul s – especially in a business as compe i i e as he ucking indus y. When na owing he obse a ion pe iod a ound he lagged in o ma ion on he in oduc ion and he inal aboli ion o he bonus pay, we obse e s ong ini ial e ec s o bo h e en s. Fu he analyses o d i ing pa ame e s sugges ha d i e s pu less e o in adap ing an eco- iendly d i ing s yle. This esul s in signi ican ly lowe alues on d i ing sco es: nine pe cen age poin s wo se on hill d i ing, hi een pe cen age poin s wo se on an icipa ion beha io and e en six een pe cen age poin s wo se on use o b ake. Conce ning ip e alua ion da a, we obse e a mo e han 70% highe p obabili y o a ip being a ed as “good” subsequen o he aboli ion o incen- i es. In sho , ou esul s indica e ha d i e s espond o incen i es no he way in ended by he haule . Ins ead, i is he aboli ion o he incen i e sys em ha signi ican ly inc eases d i e pe o mance. These indings emain obus on all ou pe o mance measu es unde a ious speci ica ions and obus ness checks. The esul s clea ly suppo c owding-ou e ec s o ex insic ewa ds as s a ed by beha io al incen i e heo y. In his sense, he em- pi ical e idence is in clea con as o classical economic a gumen s on he bene i s o mone a y incen i es (e.g. Gibbons 1998, Oye , Schae e 2011). We belie e ou indings o a ise om mone a y incen i es c owding ou o he impo an in insic and social mo i es ha seem o play a majo ole in uck d i e mo i a ion. In o he wo ds, d i e s a e o some easons in insically mo i a ed o choose a ce ain le el o e o . Though, an ex insic incen i e – such as a inancial ewa d – seems o diminish o e en dispe se hei in insic mo i a ion. As a consequence, o e all ne e o choice and pe o mance may be lowe han wi hou ecei ing inancial ewa ds. The e a e some in- insic and social mo i es ha migh se e as in e p e a ion a his poin . Fi s , esea ch iden i ied indi idual en i onmen al awa eness and conce ns o be essen ial o adap ing p o-en i onmen al beha io (e.g. Eagly, Kulesa 1997, F ansson, Gä ling 1999). The e o e, we assume d i e s’ en i onmen al belie s o be a majo in insic mo i e o adap eco- iendly d i ing. 57 In line wi h his a gumen , ex insic incen i es a e obse ed o c owd ou p o-en i onmen al beha io (e.g. Thøge sen 1994). Second, p io esea ch has iden i- ied social image, s a us and epu a ion as powe ul in insic incen i es o people’s mo i a- 57 Fo an o e iew on p o-en i onmen al beha io o uck d i e s see Schwei ze , B od ick and Spi ey (2008). On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 108 ion (e.g. A iely, B acha, Meie 2009, Ellingsen, Johannesson 2008). Hence, we assume d i e s o ha e some social conce ns on ha ing an eco- iendly image and epu a ion, in pa icula as hey a e wo king in a high-pollu ing indus y. This migh be e en s eng h- ened as he coun y domes ic o he d i e s is a below OECD a e age on CO2 emissions (OECD 2014b). Thi d, pee compe i ion among d i e s may ep esen an in insic incen- i e, oo. Acco ding o company s a , he i s hing d i e s do when e u ning om a ound- ip is o check hei a ic ligh e alua ion ela i e o pee -g oup pe o mance. Theo y and empi ical e idence sugges ha pee -g oup compe i ion unc ions as social incen i e. Fo ins ance, e o choice is highe when ela i e pe o mance ankings a e e- po ed (e.g. Cha ness, Mascle , Ville al 2013). We belie e ha mone a y incen i es – a leas pa ially – bough o he game-like cha ac e o pee compe i ion. The same holds ue o classic pee e ec s ha mo i a e d i e s o good pe o mance ia social p essu e (e.g. Falk, Ichino 2006, Mas, Mo e i 2009). E en ually, in oducing incen i es in he i s place migh be in e p e ed by d i e s as an ac o dis us . Acco ding o Sliwka (2007), eeling dis us migh u n a pe son in o sel ish beha io . This e ec could e en be in ensi- ied by he sec e in oduc ion o he incen i e scheme ha mos likely was seen as a lack o us by d i e s. Ac ing sel ish migh educe uck d i e s’ conce ns abou bo h he en i- onmen and he compe i i eness o hei employe . Ins ead, hey migh inc ease hei own u ili y by pu ing less e o in economic d i ing beha io . Fu he mo e, ou indings sup- po p io e idence on people esponding s ongly o bo h well-s uc u ed and badly de- signed incen i es (e.g. Robe s 2010). In pa icula , incen i es ha ewa d pe o mance wi h an amoun oo small may coun e ac he in ended beha io (e.g. Gneezy, Rus ichini 2000a, 2000b). By o e ing less han 5% o he mon hly ne income, he haule migh ha e badly designed he incen i e by choosing an amoun below he incen i izing h eshold. This especially holds ue gi en ha d i e s a e wo king in a high-paying company in a coun y wi h a high wage le el and, he e o e, may anecdo ally be anked among he op- ea ning uck d i e s in Eu ope. Howe e , ansla ing ou indings in o clea implica ions o p ac i ione s p o es o be di i- cul . I seems ha p incipals need o abandon he idea ha ex insic incen i es se e as a panacea o mo i a ional issues. Ins ead, mone a y incen i es do no wo k as in ended pe se and employee pe o mance and p oduc i i y depend on o he in insic and social mo- i es, oo. In oducing ex insic incen i es wi hou se iously e lec ing on hese issues migh p o e de imen al o o e all p oduc i i y in he end. Thus, we highly ecommend On he Road Again: C owding-Ou E ec s o Ex insic Mo i a ion in Comme cial T ucking 109 conside ing he in e nal and ex e nal i o incen i e schemes and gene al HRM ins u- men s. A his poin , we ollow Ichniowski, Shaw and P ennushi (1997) who a gue ha incen i e plans wo k bes when in oduced wi h complemen a y HRM p ac ices. The e- o e, we like o encou age p ac i ione s o ca e ully moni o incen i e sys ems and o ake app op ia e ac ions i incen i es appea o ha e coun e -p oduc i e e ec s. Ne e heless, i seems o be good news ha mis akes once made in incen i e in oduc ion o design seem e e sible since employees quickly espond o new se ings. E en ually, he e y pa icula o ganiza ional se up and he inside econome ic app oach o his pape migh aise doub s on he gene alizabili y o ou indings. Howe e , gi en ha ou esul s suppo heo e ical assump ions s a ed by beha io al incen i e heo y and a e in line wi h p e ious esea ch, we a e con iden ha ou e idence is applicable o simi- la si ua ions a leas o some ex en . Ye , we a e no able o p o ide clea e idence on he gene alizabili y o ou indings o o he o ganiza ional se ings. We he e o e encou age esea che s o u he con ibu e o he ongoing deba e on bo h c owding-ou and p ice e ec o ex insic incen i es. S udies examining di e en o ganiza ional se ings apa om inside econome ics migh widen gene al insigh s on incen i e e ec s on employees’ e o choice decisions. In esponse o he so a inconsis en indings on ex insic incen- i e e ec s on employee e o , esea ch on mode a ing and media ing e ec s o incen i es and he i be ween in insic and ex insic incen i es seems o be highly app ecia ed. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 110 5 BEHAVIORAL CONSEQUENCES OF THE TRANSITION FROM TEMPORARY TO PERMANENT EMPLOYMENT 5.1 In oduc ion Tempo a y employmen con ac s ha e become a common p ac ice in mode n o ganiza- ions o ci cum en labo ma ke egula ions conside ed as es ic i e and o lexibly eac o changing s a equi emen s ha a e due o changes in p oduc demand (e.g. Houseman 2001). Mo eo e , empo a y employmen o e s i ms oppo uni ies o educe labo cos s as well as adminis a i e complexi y (De Cuype e al. 2008). Acco ding o he mos ecen Eu os a Labo Fo ce Su ey, abou 14% o all employmen con ac s in he EU-28 in 2013 we e empo a y in na u e 58 (Eu os a 2014c). Among hose on empo a y con ac s, young- e wo ke s a e hea ily o e ep esen ed (OECD 2014c). In his pape we de ine empo a y con ac s o be di e en om s anda d con ac s wi h espec o he ollowing h ee dimen- sions: limi ed du a ion, a ious s a u o y d awbacks (no o li le employmen p o ec ion, no minimum wage e c.) and he di e en o ganiza ional se ing ( empo a y con ac s a e o en wi h empo a y-wo k agencies (e.g. De Cuype e al. 2008)). 59 Tempo a y employmen is no only used o a oid labo sho ages and educe labo cos s bu is o en conside ed a “s epping s one” (Boo h, F ancesconi, F ank 2002) o a “po o en y” in o pe manen employmen (Be on, De incien i, Pacelli 2011, Buddlemeye , Wooden 2011). Recen esea ch emphasizes he na u e o empo a y employmen as a sc eening de ice allowing employe s o es he quali y o a pa icula job ma ch wi hou he isk o long- e m con ac ual obliga ions (e.g. Güell, Pe ongolo 2001, Gaglia ducci 2005). Mo eo e , a g owing body o li e a u e demons a es ha con ac cha ac e is ics signi ican ly a ec employees’ beha io (e.g. Ichino, Riphahn 2005, Guadalupe 2003) in he sense ha e.g. empo a y and pe manen wo ke s a e ound o di e signi ican ly in hei choice o e o le els (e.g. Engelland , Riphahn 2005). Al hough pe manen ly em- ployed wo ke s a e supposed o pe o m be e han empo a y wo ke s because hey e- cei e mo e ( i m-speci ic) aining and epo highe wo k sa is ac ion, empi ical s udies ound he opposi e o be ue by demons a ing ha empo a y wo ke s some imes ou pe - 58 In he coun y whe e he headqua e s o he hauling company is loca ed, he pe cen age o empo a y employees is sligh ly abo e EU-28 a e age (OECD 2014). 59 In his s udy, he e m “ empo a y employmen ” e e s o all kinds o con ingen , ixed- e m, casual o non-pe manen employmen ela ions wi h an employe cha ac e ized by an ex an e de ined limi ed con- ac du a ion. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 111 o m pe manen wo ke s (e.g. Li anos, Zangelidis 2013, Amuedo-Do an es 2002). We belie e he easons o be wo old. On he one hand, employe s e y o en use empo a y employmen as a sc eening de ice. On he o he hand, employees conside mo i a ion and e o as s ong signals o an employe ha a e likely o inc ease hei p obabili y o being p omo ed o a pe manen con ac . Bo h e ec s incen i ize empo a y wo ke s o ou pe - o m pe manen wo ke s in e ms o e o le els. Howe e , his holds ue only as long as empo a y wo k is associa ed wi h in e io job cha ac e is ics compa ed o pe manen em- ploymen . So a , empo a y posi ions ha e been ound o be associa ed wi h less a o able wo king condi ions (e.g. Paoli, Me llié 2001), lowe wages (e.g. Me ens, Gash, McGinni- y 2007), less aining (e.g. A ulampalam, B yan, Boo h 2004) and lowe le els o job sa - is ac ion (e.g. Boyce e al. 2007). Mo eo e , he unce ain y abou u u e job p ospec s can be a men al bu den o empo a y wo ke s esul ing in heal h p oblems (S e ke, Hellg en, Näswall 2002). Finally, empo a y wo ke s do no bene i om manda o y job p o ec ion legisla ion o he same ex en as pe manen wo ke s. Gi en hese disad an ages, empo a y wo ke s should ha e s ong incen i es o demons a e mo i a ion and e o o “quali y” o a p omo ion om a empo a y o a pe manen con ac . These conside a ions ine i ably lead o he ques ion o whe he and o wha ex en o me - ly empo a y wo ke s adjus hei e o le el a e ha ing signed a pe manen con ac . One migh expec signi ican beha io al esponses om employees who ha e been o e ed mo e a o able wo king condi ions, including dismissal p o ec ion legisla ion. Using an unbalanced panel o uck d i e s om he GPS-based lee managemen sys em o an in- house haule o a la ge Eu opean uck manu ac u e we a e among he i s o add ess his issue wi h indi idual p oduc i i y in o ma ion om employees whose con ac s a us was changed om empo a y o pe manen . Ou esul s con i m a signi ican educ ion in comme cial uck d i e s’ e o le els esul ing in highe uel consump ion and a poo e o e all d i ing pe o mance a e being p omo ed om empo a y o pe manen employ- men . Ou s udy con ibu es o he li e a u e in se e al ega ds. Fi s , we add o he g owing e- sea ch on he in luence o employmen con ac cha ac e is ics on employee beha io . In pa icula , we ex end he li e a u e on wo ke s’ e o choices when con ac cha ac e is ics become mo e employee- iendly (e.g. Ichino, Riphahn 2005). Second, we a e among he i s o use da a om a mode n GPS-based lee managemen sys em o analyze d i e be- ha io . This usually ex ensi e sou ce o da a has no ye made i s way in o he pe sonnel Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 112 economics li e a u e. Thi d, we ollow he inside econome ics adi ion, which has en- joyed inc easing popula i y among pe sonnel economis s since he publica ion o he semi- nal pape s by Ichniowski, Shaw and P ennushi (1997) on HRM p ac ices in U.S. s eel plan s and Lazea (2000b) on mone a y incen i es o windshield ins alle s. The inside econome ics app oach applies sophis ica ed econome ic me hods o a (panel) da a se ga he ed in one o a ew companies. The esul ing nano-pe spec i e allows de ailed in- sigh s in o he beha io o indi iduals upon changes in incen i es, wo king condi ions, e c. The emainde o his pape is s uc u ed as ollows. The nex sec ion p o ides bo h a e- iew o he li e a u e on he incen i e e ec s o empo a y employmen as well as on em- ployees’ beha io al esponses o changes in con ac cha ac e is ics. The speci ic o ganiza- ional se ing and he da a se a e desc ibed in de ail in sec ion h ee. Sec ion ou p esen s he es ima ion s a egy as well as he esul s while sec ion i e concludes. 5.2 Li e a u e Re iew and Theo e ical F amewo k I is a s ylized ac in he pe sonnel economics li e a u e ha empo a y wo ke s ecei e less aining by employe s due o he usually a he sho du a ion o he employmen ela- ion (e.g. A ulampalam, B yan, Boo h 2004, Hoque, Ki kpa ick 2003, Fo ie , Sels 2003, A ulampalam, Boo h 1998). As a consequence, empo a y wo ke s accumula e less i m- speci ic knowledge han pe manen ly employed wo ke s. Since i m-speci ic skills a e c i ical o pe o mance (e.g. Ha ch, Dye 2004, Hi e al. 2001) pe manen ly employed wo ke s a e expec ed o ou pe o m empo a y wo ke s. Howe e , he e is also e idence showing ha despi e hei lowe le els o aining empo- a y wo ke s o en ou pe o m employees wi h a pe manen con ac . The pe o mance di e en ial is likely due o empo a y wo ke s’ highe le els o mo i a ion and wo k e o . Using he Swiss Labo Fo ce Su ey Engelland and Riphahn (2005), o example, ind ha he p obabili y o wo king unpaid o e ime is 60% highe o empo a y compa ed o pe manen wo ke s. Mo eo e , in a la ge da a se om Aus alia B adley, G een and Lee es (2007) ind e idence o signi ican ly lowe le els o absen eeism ( hei p e e ed measu e o employee e o ) among empo a y wo ke s. This la e esul is con i med by Li anos and Zangelidis (2013) o he EU, A ai and Thou sie (2005) o Sweden and Amuedo-Do an es (2002) o Spain. Mo eo e , using da a om a Spanish su ey Guada- lupe (2003) inds ha o he hings equal empo a y wo ke s ha e a i e pe cen age poin s Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 113 highe acciden p obabili y which may be due o lowe in es men s in empo a y wo ke s’ human capi al as well as hei p onounced incen i es o demons a e high le els o mo i a- ion and e o in he sense o aking highe isks. The incen i e e ec s o empo a y con ac s ha a e assumed o esul in highe mo i a ion and e o le els o empo a y employees a e due o hei speci ic sc eening and signaling p ope ies (Spence 1973). Fi s , empo a y con ac s a e o en used as a sc eening de ice o es newly hi ed wo ke s be o e hey a e being o e ed a pe manen con ac (e.g. G een, Lee es 2004). Second, empo a y wo k may be conside ed a “s epping s one” in o pe ma- nen employmen (Boo h e al. 2002). Consis en wi h he la e a gumen Güell and Pe ongolo (2001) ind ha empo a y wo ke s o en sign a pe manen con ac be o e hei ini ial empo a y con ac had expi ed. This sugges s ha i ms use empo a y employmen as a sc eening oppo uni y o disco e wo ke s’ “ ue” quali ies and o e pe manen con- ac s as soon as he equi ed abili ies ha e been iden i ied (see also Buddlemeye and Wooden (2011) wi h compa able e idence o Aus alia and Gaglia ducci (2005) o I aly). Summa izing, hese s udies sugges ha empo a y con ac s a e used as a sc eening in- s umen ha enables employe s o educe in o ma ion asymme ies abou wo ke s’ mo i- a ions and e o choices wi hou he isk o long- e m con ac ual obliga ions. Summa izing, he cha ac e is ics o empo a y con ac s should p o ide s ong incen i es o empo a y wo ke s o choose high e o le els o p oduce hose signals ha maximize he p obabili y o being hi ed on a pe manen con ac . This holds ue, howe e , only as long as pe manen employmen p o es o be ad an ageous o e empo a y employmen . Indeed, empo a y employmen is usually associa ed wi h poo e wo king condi ions and lowe pay le els o equal pe o mance. On he one hand, empo a y wo ke s ha e been ound o bene i less om in es men s in heal h ca e o e gonomics leading o highe le - els o job dissa is ac ion, a igue and muscula pain (Bena ides, Benach 1999, Bena ides e al. 2000). On he o he hand, hey a e exposed mo e o en o epe i i e asks and mo e- men s as well as hea y loads (Paoli, Me llié 2001), enjoy lowe le els o wo k au onomy and epo hemsel es he leas well-in o med abou hei wo k en i onmen (A onsson 1999). Fu he mo e, empo a y wo ke s o en conside hemsel es subjec o “s a us s ig- ma iza ion” leading o lowe le els o well-being, job sa is ac ion, and pe o mance (Boyce e al. 2007). Gene ally speaking, empo a y jobs a e ound o be o an o e all lowe job quali y (G een, Kle , Lee es 2010). The wage penal y o empo a y wo ke s is well documen ed in he li e a u e: Me ens, Gash, McGinni y (2007) o example ind ha he Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 120 I is impo an o bea in mind ha he adi ionally used measu e “ uel consump ion pe 100 km” is no included in he au oma ic e alua ion o he indi idual ips. Howe e , d i - ing beha io ha leads o a “good” e alua ion is cha ac e ized by an eco- iendly d i ing s yle (no ha sh accele a ions, no speeding, e c.). I appea s om Figu e A.1 in he Appen- dix, ha uel consump ion and ip e alua ion a e unco ela ed: A e age uel use o ips e alua ed as “good” is 28.1 li e s pe 100 km, o ips e alua ed as “medioc e” i is 28.2 li e s and o ips e alua ed as “poo ” he espec i e alue is 27.9 li e s (F=0,49; no sig- ni ican ). Mo eo e , i appea s om Table 5.2 ha sligh ly mo e han hal o all ips (54%) a e assessed as “good” and 43% as “poo ”. Only 3% o all ips a e e alua ed as “medioc e”. Thus, apa om he s anda d pe o mance measu e “ uel use pe 100 km” we use in ou es ima ions a second pe o mance measu e (“quali y o ip as assessed by he company’s lee managemen sys em”) ha is unco ela ed wi h he i s one. Ou s udy design has a numbe o ad an ages a oiding majo weaknesses o p e ious e- sea ch in he economics o ucking. Fi s , d i e s can usually no be compa ed as hey wo k unde highly a iable condi ions, expe iencing di e en exogenous e ec s o di e - en ou es and des ina ions (e.g. oad condi ions, cons uc ions, de ou s, e c.). Due o he o ganiza ional se ing o ou s udy we a e able o assume equal oad and en i onmen al condi ions since all d i e s use he same p ede ined ou e. Second, based on de ailed in- o ma ion on da e and ime o each ip, we a e able o con ol o wea he as well as o da e- (e.g. holiday) o ime-o - he-day ela ed (e.g. ush hou s) peaks in a ic densi y since “ iming” is likely o in luence d i ing pe o mance. Thi d, d i e s a e obse ed on di e en ucks wi hou ha ing any in luence on uck selec ion. Thus, we can accoun o ehicle speci ic e ec s and elimina e any impac o d i e - uck ma ches. Fou h, almos one qua e o all ehicle-kilome e s o ucks in Eu opean c oss-bo de a ic occu s on emp y ucks (Eu opean Commission 2014c). Any s udy ha is unable o con ol o load ac o s is hus likely o su e om an omi ed a iable bias as ull and emp y ucks di e in maneu e abili y and uel consump ion. Since he haule whose da a we use he e epo s a e age load ac o s o mo e han 90% on i s ou as well as i s e u n jou neys, his issue can be neglec ed. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 121 5.4 Empi ical Models and Es ima ion Resul s Recall ha ou empi ical analyses a e based on wo di e en , ye somewha ela ed pe - o mance measu es ha can be conside ed p oxies o choice o e o . Fi s , we use in ou es ima ions a e age uel consump ion pe 100 km as he dependen a iable. I appea s om Table 5.2 ha mean uel consump ion is 28.0 li e s/100km wi h a s anda d de ia ion o 2.9 li e s. Since all d i e s ecei e indi idual uel e iciency aining in i egula in e - als, we conside hem well-in o med on how d i ing beha io and s yle a ec s uel con- sump ion. 65 Thus, we sugges any di e ences in uel consump ion o de i e om d i e s’ indi idual e o choices. Second, we use he epo s o he in e nal ip e alua ion ool assessing each ip as ei he “good”, “medioc e” o “poo ” (see Table 5.2). Since d i e s a e o e ed ull online access o all e alua ion epo s hey a e awa e no only o hei pe - o mance ela i e o hei pee s, bu also o any kind o “misconduc ” in d i ing beha io . In e es ingly, publicly a ailable pe o mance e alua ions ha e ecen ly been ound o elim- ina e social loa ing and inc ease pe o mance a leas in g oup wo k se ings (Loun , Wilk 2014). Fu he esea ch indica es eedback o be highly app ecia ed by uck d i e s wi h eedback om ad anced echnology being less desi ed han eedback om supe iso s (Huang e al. 2005). Thus, d i e s ecei ing pe o mance eedback and s ill no pe o ming in acco dance wi h he employe ’s expec a ions a e assumed o d i e wi h educed e o . While uel consump ion is a ca ego ical a iable, he ip e alua ion a iable is o dinal. Bo h ou come a iables a e compu ed by he in- ehicle compu e s and a e no p one o any subjec i e a e bias o manipula ion by he d i e s. Ou models ha e he ollowing gene al o m (wi h FUELC deno ing uel consump ion pe 100km and EVAL being he composi e e alua ion measu e): FUELC = α + βDRIV + γVEH + δTRAF + φWEA + ωCONTR + ε (1) EVAL = α + βDRIV + γVEH + δTRAF + φWEA + ωCONTR + ε (2) whe e α is he cons an , DRIV ep esen s he d i e pe o mance (see equa ion (3) o de- ails), VEH is a se o ehicle dummies, TRAF a ec o o a ic condi ions, WEA a ec o o wea he condi ions and CONTR a ec o o u he con ols, β, γ, δ, φ and ω a e he es- 65 Al hough we ha e no in o ma ion on he equency o he exac da es o indi idual uel e iciency ain- ing we assume his no o be a p oblem since ecen esea ch has con incingly demons a ed ha (ca ) d i e s seem o know by in ui ion how o d i e eco- iendly ( an Mie lo e al. 2004). The e o e, we as- sume ha uel e iciency aining does no undamen ally change d i e beha io , i.e. uel consump ion goes down a e aining bu e u ns o i s p e- aining le el a he quickly. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 122 ima ed coe icien s and ε is he e o e m. All a iables a e discussed in mo e de ail be- low. D i e pe o mance (DRIV) is modeled as ollows: DRIV = ((AGE + HUMC) * EFCH) (3) whe e AGE is d i e age and HUMC ep esen s a d i e s’ human capi al in he o m o pas d i ing expe ience wi h he cu en employe ( enu e in h ee-mon h in e als). 66 P e- ious esea ch p o ides e idence o he nega i e e ec o age on a ious abili ies ele an o d i ing, e.g. esponse ime, isual and psychomo o abili ies (e.g. Llane as e al. 1998). O he indings, howe e , sugges ha olde d i e s can compensa e o age- ela ed im- pai men s by expe ience (e.g. Gues , Boggess, Duke 2014). Addi ionally, young and less expe ienced d i e s a e ound o ha e a highe p obabili y o being in ol ed in acciden s (e.g. Häkkänen, Summala 2001), which u he unde lines he impo ance o expe ience in ucking. Hence, we include age as well as d i ing expe ience in ou model o accoun o hese e ec s. Bo h a iables aken oge he ep esen he pe o mance po en ial o an em- ployee, which is hen subjec o d i e s’ cu en e o choice (EFCH). Depending on he choice o e o , pe o mance po en ial e en ually ansla es in o ac ual d i e pe o mance. In ou es ima ions EFCH is ep esen ed by a dummy a iable aking a alue o 1 o a d i - e is on a empo a y con ac and 0 i he is on a pe manen con ac . In addi ion, we con ol in ou es ima ions o exogenous e ec s ha may in luence uck d i e pe o mance. VEH is a se ies o ehicle dummies accoun ing o di e ences in echnological con igu a ions o ucks ha migh acili a e d i ing and/o dec ease uel consump ion. TRAF is a ec o o a iables con olling o a ic densi y on he ips in wo di e en ways. On he one hand, we include a iables o con ol o he ime o he day when he s ee s a e p esumably less conges ed (e.g. du ing nigh ime, be ween 8 p.m. and 7 a.m. he nex day) o mo e conges ed (e.g. du ing ush hou s, be ween 7 a.m. and 9 a.m. in he mo ning as well as be ween 4 p.m. and 6 p.m. in he a e noon) when a ic densi y is pa icula ly high due o commu e a ic. On he o he hand, we include he day o he week as we belie e some days o be sys ema ically di e en (e.g. Mondays o F i- days) om o he s. Fu he mo e, we con ol o a h ee-day window a ound in e na ional 66 We a e well awa e ha enu e on a mon hly basis would be p e e able o he i s ew mon hs; howe e , we chose h ee-mon h in e als because du ing ou obse a ion pe iod o 36 mon hs any e ec o mon h- ly enu e is likely o disappea quickly as ime elapses. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 123 holidays du ing which we suppose a ic densi y o be highe due o holiday a ic (public holidays in all coun ies he d i e s pass on hei ips, e.g. New Yea ’s Day, Eas e , Ch is mas, e c.). WEA is a ec o o wea he dummies including wind speed, p ecipi a ion, snow dep h, and empe a u e 67 because p e ious esea ch has ound d i ing beha io , a - ic condi ions as well as c ash isks o be associa ed wi h wea he condi ions (e.g. Kil- peläinen, Summala 2007). Mo eo e , we include dummy a iables o all hi y-six mon hs o accoun o possible end o iming e ec s. Finally, o con ol o season e ec s we include dummies o all i y- wo weeks o a pa icula yea . Gi en he wide ange o con- ol a iables, we a e con iden o ha e excluded mos exogenous in luences on d i e pe - o mance and a e hus able o s udy he “clean” e ec o d i e s’ beha io al esponses o a change in hei con ac s a us. We s a he discussion o ou indings by looking i s a he impac o con ac s a us on uel consump ion (equa ion (1) abo e) in he sho - e m, ollowed by a de ailed analysis o i s impac in he long- un. We hen go on o p esen ou indings wi h espec o he impac o con ac s a us on ip e alua ion (equa ion (2) abo e). The main esul s a e epo ed in Tables 5.3, 5.4, and 5.5 sugges ing ha comme cial uck d i e s signi ican ly change hei beha io ollowing changes in hei con ac s a us, i.e. educe hei e o le els a e ha - ing been p omo ed om a empo a y o a pe manen con ac . The baseline esul s o ou ixed-e ec s es ima es wi h uel consump ion as he dependen a iable a e displayed in Tables 5.3 and 5.4. We use a ious speci ica ions including di e en se s o explana o y a iables o con ol o he obus ness o ou indings. To iden i y he impac o a change in con ac s a us on uel consump ion in he sho - un, we na ow he obse a ion pe iod o a six-mon h window and compa e o each d i e his pe o mance in he las h ee mon hs unde he empo a y con ac wi h ha du ing he i s h ee mon hs unde he pe manen con ac . I appea s om Table 5.3 ha he immedia e consequence o p omo ing a d i e om a empo a y o a pe manen con ac is a s a is i- cally signi ican and economically ele an inc ease in uel consump ion o mo e han 10% (3.04 l/100km). While his inding suppo s ou assump ion o a s ong ini ial beha io al esponse in he i s ew mon hs a e ha ing been p omo ed o a pe manen con ac , i is 67 Wea he in o ma ion was collec ed om a me eo ological s a ion loca ed app oxima ely in he middle o he ound- ip ou e. This admi edly c ude wea he p oxy helps o o e come he p oblem ha we canno con ol o uck posi ion in de ail. Thus, we assume ha wea he in o ma ion om hal he dis ance ade- qua ely e lec s he sligh ly di e ing clima e zones o he ips’ s a ing and a i al poin s. Da a was p o- ided by he Ge man Me eo ological Se ice (DWD, 2014). Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 124 no ye clea whe he his ini ial eac ion emains cons an o e ime o whe he d i e s change hei beha io once mo e a e some ime has elapsed. Table 5.3: Impac o Con ac S a us on Sho -Te m Fuel Consump ion Va iables (1) (2) (3) Tempo a y Con ac (yes=1) -1.887** -1.887** -3.035** (.628) (.628) (.903) Age (in yea s) .480 1.079 (4.215) (4.188) Tenu e (in h ee-mon h in e als) -1.445 (.872) Vehicle Dummies YES YES YES Da e Dummies YES YES YES Time Dummies YES YES YES Wea he Dummies YES YES YES D i e Fixed-E ec s YES YES YES Cons an 37.31 25.46 -1.461 (72.72) (170.0) (171.5) Numbe o obse a ions 308 308 308 Numbe o d i e s 8 8 8 R-squa ed .343 .343 .350 Robus s anda d e o s in pa en heses *** p<.01, ** p<.05, * p<.10 Table 5.4 displays he esul s o ou second es ima ion, iden i ying he long- un beha io al esponses o comme cial uck d i e s o changes in con ac s a us. Acco ding o he poin es ima es, d i e s on a e age educe hei indi idual e o le el upon a change in con ac (e.g. om empo a y o pe manen ). The coe icien o con ac s a us in ou p e e ed speci ica ion (model 3) is -.475 (s anda d e o = .150). This implies ha , o he hings equal, a d i e uses .475 l/100km less uel when employed on a empo a y con ac . On a e age, d i e s co e an annual dis ance o 125,765 km and use 35,185 li e s o uel. The es ima ed coe icien ansla es in o an inc eased uel use o 600 li e s pe yea a e a change in con ac s a us. Assuming a uel p ice o €1.50 pe li e , his amoun s o addi- ional uel cos s o a ound €900 pe d i e and yea o e e ybody who is p omo ed om a empo a y o a pe manen job. Thus, hese cos s can be in e p e ed as “hidden cos s o con- ac con e sion” ( om empo a y o pe manen ). Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 125 Table 5.4: Impac o Con ac S a us on Long-Te m Fuel Consump ion 68 Va iables FE RE (1) (2) (3) (4) (5) Tempo a y Con ac -.397* -.415** -.475** -.487** -.487*** (yes=1) (.180) (.161) (.142) (.150) (.151) Tenu e (in h ee-mon h -.0740 -.0026 in e als) (.208) (.209) Age (in yea s) -.604 -.0740 (1.550) (.027) Da e Dummies YES YES YES YES YES Vehicle Dummies YES YES YES YES Time Dummies YES YES YES Wea he Dummies YES YES YES D i e Fixed-E ec s YES YES YES YES Cons an 25.51*** 25.65*** .118 39.11 0 (1.085) (1.126) (25.37) (64.72) (0) Numbe o obse a ions 1,299 1,299 1,299 1,299 1,299 Numbe o d i e s 8 8 8 8 8 R-squa ed .090 .116 .128 .128 Robus s anda d e o s in pa en heses *** p<.01, ** p<.05, * p<.1 To u he con ol o he in luence o age and enu e we es ima ed a andom-e ec s mod- el, he esul s o which a e displayed in column 5 o Table 5.4. Con a y o p e ious e- sea ch (e.g. Llane as e al. 1998), we ind d i e pe o mance o be una ec ed by age and/o expe ience. This is pe haps su p ising as he d i e s in ou da a se a e e y he e o- geneous in e ms o age (19 o 61 yea s) and enu e (up o 38 mon hs). The ac ha a d i - e ’s human capi al (in e ms o expe ience) seems o be i ele an o his s yle o d i ing ende s he impac o mo i a ion and choice o e o (EFCH) e en mo e impo an . Pe - haps su p isingly, none o he hi y-eigh ehicle dummies comes close o s a is ical sig- ni icance (coe icien s no displayed in able o sa e space). This inding migh sugges ha ehicle echnology is no as impo an as d i e beha io in de e mining uel con- sump ion. Ye , we belie e his inding o be due o a mo e s aigh o wa d explana ion. The haule so s ou ehicles om i s lee a e hey ha e been used o abou one million kilome e s. Since his h eshold le el is usually eached a e abou wo o h ee yea s, he haule ’s lee consis s o almos (b and) new ucks and all d i e s a e equipped wi h s a e- o - he-a echnology. The e o e, i seems easonable ha ou dummies ep esen ing di e - ences in ehicle echnology ail o each s a is ical signi icance. In e es ingly and in con- as o p e ious esea ch (Kilpeläinen, Summala 2007), he only s a is ically signi ican 68 Age and enu e e ec s we e es ima ed indi idually bu ail o each s a is ical signi icance. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 126 wea he a iable in e ms o uel consump ion is wind speed (+.11 l). This is again plausi- ble as ucks a e p one o windage due o hei size and un a o able ae odynamic ea u es. None o he emaining wea he a iables, e.g. empe a u e, snow all and p ecipi a ion came close o s a is ical signi icance. In pa icula , we expec ed high empe a u es o ha e a no iceable in luence on uel consump ion since ai condi ioning sys ems a e widely known o inc ease uel use. Wi h espec o he ime dummies (mon hs and weeks), no clea pa e n eme ges. On Tuesdays uel consump ion le els a e highe (plus .77 li e s pe 100 km) highe han on he e e ence day (Mondays). No o he day o he week e ec can be iden i ied. 69 The esul s o ou ixed-e ec es ima ions s ill hold a e including a dummy a iable ep- esen ing p esence o a bonus egime du ing he i s wo yea s o he obse a ion pe iod. 70 The bonus egime migh be an addi ional ac o in luencing d i e beha io ia choice o e o (EFCH). In pa icula , he bonus egime migh impac he mo i a ional e ec s ha we a ibu e exclusi ely o beha io al esponses ollowing he change in con ac s a us. Howe e , we ail o ind any signi ican e ec o he bonus egime (see Table A.11 in he appendix wi h es ima ions including he bonus egime). Thus, we a e con iden ha he beha io al esponses iden i ied a e en i ely a ibu able o changes in mo i a ion and e o choice igge ed by a change in con ac s a us. Since we a e pa icula ly in e es ed in he iming as well as he pe sis ence o he indi idu- al d i e s’ beha io al esponse, we included in ou es ima ions a end a iable coun ing he numbe o mon hs a e con ac con e sion o each d i e . The espec i e coe icien indica es a s a is ically signi ican nega i e e ec (-.140), sugges ing an inc ease o uel consump ion a e he change in con ac s a us (see Table A.12 in he appendix). The coe - icien o he squa ed ime end, howe e , is posi i e and signi ican (.004) indica ing a u- shaped pa e n (wi h he u ning poin a e 18 mon hs). This, in u n, sugges s ha uck d i e s in he long un e u n o hei ini ial pe o mance le els. Appa en ly, u ili y- maximizing uck d i e s seem o no longe eel he need o displaying high le els o e o 69 The coe icien s o ips a nigh (-.008 l) and du ing e ening ush hou s (+.26 l) ha e he expec ed signs bu ail o each s a is ical signi icance. Su p isingly, he coe icien o mo ning ush hou s (-.11 l) is nega i ely signed, bu also ails o each s a is ical signi icance. 70 The bonus egime was in p ac ice om he beginning o he obse a ion pe iod in Janua y 2011 un il Decembe 2012 when he haule decided o abolish i . Al hough a low uel consump ion was no di ec ly ewa ded by he incen i e sys em, d i e s we e inancially incen i ized o pe o m acco ding o he measu es o he on-boa d e alua ion ool. Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 127 a e ha ing been signed o a pe manen con ac . Indeed, indi iduals educe hei e o le els subs an ially and pe o mance de e io a es conside ably once a ansi ion in con ac s a us has occu ed. Howe e , wi h espec o uel consump ion his change in beha io seems o ge weake o e ime. We now p esen he indings o ou second es ima ion using he au oma ically gene a ed ip e alua ion as a measu e o d i ing pe o mance and p oxy o d i e choice o e o . To accoun o he o dinal na u e o he da a, we es ima e a se ies o o de ed logi models as sugges ed by Guja a i and Po e (2009) as well as Woold idge (2013). I appea s om Table A.12 in he appendix ha he e a e signi ican beha io al esponses in e ms o high- e le els o e o depending on con ac s a us ( empo a y s. pe manen ). Table 5.5 p e- sen s he pos -es ima ion p obabili ies o indi idual ip e alua ions. Table 5.5: Es ima ed P obabili ies o T ip E alua ion Va iables Tempo a y Pe manen Change 95% CI o Change T ip E alua ion Good .7471 .4140 -.3331 -.4837 -.1825 Medioc e .0404 .0559 .0155 -.0045 .0355 Poo .2124 .5300 .3176 .1682 .4671 Acco ding o ou es ima ions he p obabili y o a ip being e alua ed as “good” is 74.7% when a d i e is on a empo a y con ac and 41.4% when a d i e is pe manen ly em- ployed. Thus, he p obabili y o a pe o mance ha is in acco dance wi h he employe ’s expec a ions is 33.3 pe cen age poin s (o 80.4%) highe in he case o he o me wo ke s. An in e se pa e n appea s o ips e alua ed as “poo ”. He e he p obabili y o empo- a y wo ke s is 21.2% compa ed o 53.0% o d i e s on pe manen con ac s. This yields a di e ence o 31.8 pe cen age poin s o 150%, espec i ely. This esul con i ms ou main inding ha d i e s on empo a y con ac s ha e s ong incen i es o signal mo i a ion and high le els o e o . A e ha ing signed a pe manen con ac , d i e s quickly educe hei e o le els leading o a lowe o e all d i ing pe o mance e alua ion. Howe e , a close look a he da a e eals ha when compa ing bo h uel consump ion and ip e alua ion be o e and a e p omo ion o a pe manen con ac , only hal o he d i e s pe o m wo se a e hey ha e been p omo ed. Thus, while some d i e s show (nega i e) beha io al eac- Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 128 ions o a change in con ac s a us, o he s do no ( his inding is in line wi h e.g. Nagin e al. 2002 as well as Riphahn and Thalmeie 2001). We admi ha due o he o ganiza ional se ing o ou s udy we ha e o be awa e o some limi a ions ha p eclude gene aliza ion o ou indings. Fi s , he numbe o d i e s (n=8) is qui e small. Howe e , he da a se consis s o a leas 100 obse a ions pe d i e esul - ing in a o al numbe o obse a ions o 1,299. We a e, he e o e, con iden ha ou esul s a e obus and eliable. Second, due o he haule ’s da a p i acy es ic ions o d i e s’ pe sonnel eco ds, we a e unable o include in ou es ima ions measu es o d i e quali ica- ion and o me expe ience wi h o he employe s. Ye , since we con ol o age and enu e wi h he cu en employe , we belie e o ha e included su icien p oxies o quali ica ion and pas d i ing expe ience. Mo eo e , he coe icien s o bo h, age and enu e ailed o each s a is ical signi icance indica ing a negligible impo ance o quali ica ion and ( o - me ) d i ing expe ience. 5.5 Conclusion Using a hi he o una ailable da a se compiled om he lee managemen sys em o an in- house haule o a la ge Eu opean uck manu ac u e , we p o ide obus e idence on em- ployees’ beha io al esponses o changes in con ac s a us (i.e. ollowing he ansi ion om a empo a y o a pe manen con ac ). We analyze comme cial uck d i e s’ e o choices be o e and a e hey ha e been signed o a pe manen con ac using wo pe o - mance a iables – uel consump ion pe 100 km and an au oma ically gene a ed ip e alu- a ion measu e. In line wi h p e ious esea ch (B adley e al. 2012), ou indings sugges ha comme cial uck d i e s choose hei e o le els depending on he na u e o hei con ac s, i.e. hey exe mo e e o (in he sense o using less uel and be e ip e alua ions) when on em- po a y con ac s. A e ha ing been p omo ed d i e s adjus hei e o le els, i.e. use mo e uel and ecei e mo e “bad” e alua ions. These indings a e obus on a ious speci- ica ions and su i e a numbe o obus ness checks. D i e s on empo a y con ac s (o mo e gene al: wo ke s) – being awa e o o jus assuming sc eening ac i i ies o be used by he haule (o mo e gene al: i ms) – choose high le els o e o o signal mo i a ion and dedica ion. Howe e , as soon as d i e s (wo ke s) a e signed o pe manen con ac s, hey educe e o back o “no mal” le els. When employed unde a pe manen con ac , uel consump ion o d i e s is 600 li e s highe pe yea ( he addi ional cos s pe d i e a e Beha io al Consequences o he T ansi ion om Tempo a y o Pe manen Employmen 129 a ound 900 € pe yea ) han when employed unde a empo a y con ac . We in e p e his as he hidden cos s o changes in con ac s a us ( om empo a y o pe manen ). This nega i e (expensi e) beha io al esponse is pa icula ly s ong in he i s mon hs a e con ac con e sion bu hen declines o each he ini ial le el again a e abou 18 mon hs. Mo eo e , o d i e s on empo a y con ac s he p obabili y o a ip being e alua ed by he ucks’ compu e sys em as “good” is 55% highe han o d i e s on pe manen con- ac s. Ye , we ind only hal o he d i e s o display hese beha io al changes when being p omo ed om a empo a y o a pe manen con ac . The o he hal does no espond o he new con ac s a us bu ins ead pe o m a hei ini ial e o le els. This inding is in line wi h Nagin e al. (2002) who also ind ha only a ac ion o employees beha e as “ a ion- al chea e s” when gi en he oppo uni y o do so while many o he s esis ha emp a ion. Whe e do he incen i e e ec s o empo a y con ac s come om? Pe manen con ac s ha e been ound o be associa ed wi h be e wo king condi ions, highe pay, mo e job secu i y and highe le els o employmen p o ec ion. Mos empo a y wo ke s wan o ec- ommend hemsel es o pe manen con ac s o enjoy inc eased le els o job secu i y and dismissal p o ec ion legisla ion. This is pa icula ly ue in ou case as he coun y whe e he headqua e s o he haule is loca ed has mos ecen ly been anked below a e age on OECD’s “Indica o s o Employmen P o ec ion” o wo ke s on empo a y con ac s and abo e a e age o wo ke s on pe manen con ac s (OECD 2013). Mo eo e , we a e no awa e o any di e ences in wo king condi ions o empo a y and pe manen ly employed d i e s a his pa icula i m. Wha a e he p ac ical implica ions ha can be de i ed om ou esul s? F om an o ganiza- ional poin o iew, he esul s seem o sugges keeping employees on empo a y con ac s as long as possible o maximize he e u ns om highe e o le els. Ye , labo laws ule ou ecu ing ex ension o empo a y con ac s. I may, he e o e, appea an e en mo e p omising s a egy o employ empo a y wo ke s only. Howe e , in he absence o p omo- ion oppo uni ies ( om empo a y o pe manen con ac s) empo a y wo ke s lack he necessa y incen i es o choose high le els o e o . Dolado and S ucchi (2008) ind ha i ms wi h a high con ac con e sion a e (by p omo ing wo ke s om empo a y o pe - manen con ac s) display highe le els o labo p oduc i i y. Thus, i seems a easonable s a egy o openly communica e he sc eening na u e o empo a y con ac s since hen empo a y wo ke s will be mo i a ed o choose high le els o e o . Mo eo e , o e ing p ope ly designed mone a y incen i es du ing he weeks and mon hs a e con ac con e - Summa y and Fu u e Ou look 136 au ho ’s knowledge – has no ye been used in pe sonnel economics esea ch. Since a ail- abili y and quali y o his objec i e eal- ime da a is o e whelming, i s s udy may o e b oad ad an ages in analyzing employee beha io . This hesis, he e o e, makes a ele an claim o include his inno a i e and p omising ype o da a in o pe sonnel economics. Sec- ond, he use o objec i e pe o mance measu es as indica o s o wo ke p oduc i i y de- pic s a dis inc ad an age o e mos exis ing esea ch ha is o en based me ely on subjec- i e (sel -) a ed pe o mance. This is pa icula ly ue o absen eeism esea ch since eg- is e absence da a is o en una ailable o esea che s. As a consequence, mos exis ing indings on absence a e based on sel - epo ed absence igu es ha a e known o be biased due o unde es ima ion wi h ac ual absence being wice as high (Johns 1994). Hence, he objec i e absence da a used in chap e h ee depic s a clea ad an age. Simila ly, analyses in chap e s ou and i e bene i om compu e ized pe o mance e alua ions ha a e col- lec ed au oma ically and objec i ely a e pe o mance based on p ede ined algo i hms. In gene al, he use o high quali y objec i e pe o mance measu es cons i u es a majo ad- an age o he disse a ion a hand. I add esses weaknesses which can be ound in pa s o he exis ing p oduc i i y esea ch ha o en ely on subjec i e pe o mance measu es. Based on objec i e pe o mance da a, he p esen ed indings complemen exis ing e i- dence and, hus, con ibu e o he p og ess o he discussion in pe sonnel economics. Thi d, despi e he ac ha incen i es a e o en amed o be he co e o pe sonnel economics (e.g. Lazea 2000b) and as such a e p obably one o he mos in ensi ely s udied ields in HRM esea ch, e y li le is known abou esponses o employees o he aboli ion o incen i es. To he bes o he au ho ’s knowledge, he aboli ion o incen i es has no been s udied ha o en – F eeman and Kleine (2005) ep esen ing a a e excep ion. By s udying employees who expe ience he aboli ion o an exis ing incen i e scheme, his wo k may con ibu e o a be e unde s anding o he gene al unc ioning o incen i es. E en ually, his hesis con- ibu es o absence esea ch in h ee ways. Fi s , as men ioned abo e he use o egis e da a in absence esea ch depic s a clea ad an age o e mos o he exis ing esea ch ha lacks access o objec i e absence igu es om inside a company. Second, he absence da a used in his s udy is ad an ageous in ha i was compiled in di e en in e na ional plan s o he same company. Hence, he da a is cohe en ly eco ded using iden ical co po a e s anda ds a all plan s. This allows o in e na ional compa isons wi hou acing he com- mon d awbacks o a ying epo ing me hods usually associa ed wi h in e na ional absence da a (Eu opean Founda ion o he Imp o emen o Li ing and Wo king Condi ions 2010). Thi d, absence he e is p ima ily e alua ed a he g oup le el. Despi e he ac ha absence Summa y and Fu u e Ou look 137 is a social concep and absence beha io is o a la ge ex en in luenced by wo ke ’s pee s, pe sonnel economic knowledge on absen eeism is o a g ea pa based on indi idual-le el da a (e.g. Ren sch, S eel 2003). Chap e h ee add esses his gap in absence esea ch. In o al, he indings o his hesis con ibu e o he ongoing discussion o he social and eco- nomic de e minan s in pe sonnel economics by o e ing impo an insigh s on eamwo k and incen i es based on unique da a. Despi e i s aluable con ibu ion o pe sonnel economics he wo k a hand aces some limi- a ions o igina ing om bo h econome ic me hods and da a. In gene al, me a-analyses ace wo pa icula challenges (e.g. Egge , Smi h, S e ne 2001): he publica ion bias as well as he ga bage-in-ga bage-ou issue. Fi s , s udies wi h posi i e and signi ican ind- ings a e mo e likely o be published in pee - e iewed jou nals. As a esul , me a-analyses ha ocus solely on published a icles only co e signi ican indings and lea e ou con o- e sial o inconclusi e esul s, leading o he eme gence o a publica ion bias. Second, he quali y o any me a-analysis is dependen on he quali y o he s udies e iewed. To ac- coun o his pi all – e e ed o as ga bage-in-ga bage-ou e ec – only s udies published in high-quali y pee - e iewed jou nals a e included in he me a-analysis p esen ed in chap- e wo. Albei he ocus on pee - e iewed a icles migh p o oke a publica ion bias, his p ocedu e seems o be pa icula ly c ucial o iden i y he mos impo an con ibu ions o he ex ensi e li e a u e on eam di e si y. Mo eo e , any publica ion bias is mi iga ed by he ac ha he indings included in he me a-analysis in chap e wo show posi i e as well as nega i e co ela ions. A common issue o inside s udies is cen e ed on i s single- i m da a sou ces. Al hough his allows o high-quali y esea ch a he mic o-pe spec i e, he gene alizabili y o esul s is limi ed (Ichniowski, Shaw 2013). Mo eo e , his esea ch is likely o su e om selec- ion bias and, he e o e, i also su e s om endogenei y in he choice o wo ke s and man- age s as i ocuses exclusi ely on a speci ic indus y. Resul s based solely on da a and in- o ma ion o igina ing om he au omo i e indus y may no be p esumed o be uni e sally alid and, he e o e, should no se e as a bluep in o o he indus ies wi hou u he in es iga ion. A dis inc i e limi a ion o all non-expe imen al da a – and hus o inside econome ics as well – is ha he choice o a ea men by means o a pa icula manage- men p ac ice is no andom bu ins ead is he esul o a maximiza ion decision aken by he company. Summa y and Fu u e Ou look 138 In conclusion, he e idence and implica ions discussed in his hesis p o ide impo an con ibu ions o he s ands o inside econome ics and pe sonnel economics, in pa icula o he ields o incen i e design and wo k o ganiza ion in eams. Wi h ega d o he me h- ods applied and opics s udied, he wo k a hand concludes by poin ing ou p omising sug- ges ions o u u e esea ch. Fi s , al hough ha ing a long adi ion in o he scien i ic ields such as medicine, he me a-analy ic app oach is only slowly gaining c edence in economic esea ch. Howe e , esul s o his hesis demons a e ha me a-analyses a e a use ul means o o e come limi a ions o small sample sizes and gain an o e iew on opics wi h ex en- si e, bu so a inconclusi e empi ical e idence. In pa icula , opics ha can be assessed as being “o e - esea ched” may bene i mo e om me a-analyzed conclusions han om addi ional empi ical indings. Second, mode n imes a e changing he ci cums ances o i ms and employees alike as complexi y o coope a ion and p ocesses inc eases. The e- o e, economic models need o be adap ed and e ined by inco po a ing in o ma ion and da a om wi hin companies. On his accoun , inside s udies a e a key o imp o ing he quali y o bo h heo e ical models and implica ions o p ac i ione s de i ed om esea ch. Thus, u he con ibu ions o inside econome ics a e highly app ecia ed. Thi d, in o de o gain a comp ehensi e unde s anding o an economic phenomenon, i is c ucial o ana- lyze i in all i s ace s. Fo ins ance, in esponse o he well-s udied ield o incen i e in o- duc ion, his wo k analyses he beha io al consequences o i s aboli ion. As a conse- quence, he p esen ed indings may con ibu e o an o e all be e unde s anding o incen- i es. The e o e, mo e s udies ha depa om amilia pa hs and shed ligh on hi he o uns udied aspec s o well-known subjec s a e needed. Fou h, as men ioned ea lie , he use o da a om GPS-based lee managemen sys ems has p o en o con ibu e signi ican indings o pe sonnel economics. Thus, schola s should keep hei eyes open o new and inno a i e da a sou ces ha so a ha e no been aken in o accoun o scien i ic esea ch. The digi al age can be assumed o o e u he “ easu e ches s” in e ms o compu e ized da a. E en ually, his disse a ion emphasizes he ad an ageousness o he coope a ion be ween academia and businesses in he con ex o inside econome ics. The au ho no only acknowledges he g ea oppo uni y o wo k wi h unique and o he wise una ailable da a om inside a company bu also highly app ecia es he chance o discuss indings wi h in e nal expe s in o de o e ine in e p e a ions and de i e mo e comp ehensi e implica- ions. The au ho o his hesis, he e o e, in i es esea che s as well as p ac i ione s o join hei o ces in he sea ch o aluable insigh s ha bene i bo h scien i ic p og ess and companies’ bo om-lines. Appendix IX APPENDIX Table A.1: He e ogenei y Measu es Lis ed by F equency o Use Di e si y measu e Blau CV Teachman s.d. O he * Age 4 14 2 5 3 Gende 14 1 5 1 8 Cul u e 13 1 3 - 2 Func ion 14 - 5 - 3 Tenu e 2 16 - 4 1 Educa ion Backg ound 7 - - 3 3 Le el 4 3 2 - 1 * O he he e ogenei y measu es include He indal index, Gini index, mean, pe cen age sha e and au ho s’ own modi ica ions. Figu e A.1 : Fo es Plo – Age Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix X Figu e A.2: Fo es Plo – Gende Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix XI Figu e A.3: Fo es Plo – Cul u e Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix XII Figu e A.4: Fo es Plo – Tenu e Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix XIII Figu e A.5: Fo es Plo – Func ional Backg ound Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix XIV Figu e A.6: Fo es Plo – Educa ional Backg ound Di e si y and O e all Pe o mance Sou ce: Own calcula ions. Appendix XV Figu e A.7: Fo es Plo – Educa ion Le el Di e si y and O e all Pe o mance Sou ce: Own calcula ions.