June 2015 The relationship between economic growth and employment

Transcription

1 The relaionship beween economic growh and employmen K. Burggraeve G. de Walque H. Zimmer Inroducion The quesion of he relaionship beween economic growh and employmen is highly relevan in he recen economic conex dominaed by he grea recession, in which Belgium sood ou from some of is parners hanks o a cerain degree of labour marke resilience. This aricle akes sock of how ha relaionship has changed over ime, hrough various recession episodes, and of he underlying rends in gross domesic produc (GDP), he volume of labour, and produciviy. The aricle begins by describing he respecive movemens, in Belgium, in he variables ha explain he paern of GDP, boh hrough he various economic cycles and in a long erm perspecive. The firs secion ses ou he accouning relaionships beween aciviy and employmen, while he second secion analyses he deailed breakdown of GDP. Nex, he hird secion explains he long erm rends in each elemen of ha breakdown, making i possible o address issues such as he job inensiy of growh. The fourh secion examines he developmen of he various branches of aciviy and heir conribuion o he changes menioned above. Afer ha, he fifh secion of he aricle examines more paricularly he quanificaion of he relaionship beween growh and employmen hroughou he business cycle and during various successive cycles. Here, he analysis only considers he cyclical componens of he variables concerned. A ha level, i endeavours o measure he sensiiviy of he cyclical componen of employmen o he oupu gap in Belgium, o compare i wih he sensiiviy observed in oher developed economies, and o verify wheher i is sable over ime or wheher here is any asymmery beween expansion and recession episodes. Finally, in he sixh secion, his empirical sudy along he business cycle is exended o a secoral breakdown and he oal volume of labour raher han employmen in erms of persons. The conclusion aemps o place he salien findings of he sudy in perspecive while idenifying a number of ideas worh exploring. 1. relaionship beween aciviy and employmen 1.1 Aaccouning relaionships GDP represens all goods and services produced and supplied in he economy during a given period. Among oher hings, ha oupu (Y) has o mee he needs of he oal populaion (P), only some of whom, namely he labour force (A), ake par in he producion process. In an economy wih under- employmen, ha labour supply is larger han he needs of he producion sysem so ha he labour force is sub divided ino workers (E) and unemployed persons (U). une 215 The relaionship beween economic growh and employmen 31

2 In accouning erms, GDP is broken down as follows : of which TH = E H = oal hours worked in he economy H = average hours worked per worker Y / TH = hourly labour produciviy Simplifying by employmen E, i emerges ha GDP (Y) depends on wo facors : he volume of labour used (TH), which is equal o employmen in persons (E) imes he average hours worked (H), i.e. a quaniaive aspec, on he one hand, and apparen hourly labour produciviy (Y / TH), or a qualiaive aspec, on he oher. All oher hings being equal, and leaving aside capial and echnical progress, if produciviy oupaces economic growh, he volume of labour diminishes. The relaionship beween employmen and aciviy is cenral o Okun s law. In principle, hese wo variables have o move in he same direcion. In periods of expansion, he producion sysem needs more workers o saisfy demand, so ha employmen rises and unemploymen falls. Bu alhough one addiional worker poenially reduces he numbers unemployed by one, he unemploymen rae (U / A) will no show a decline proporionae o he growh in employmen, he main reason being he Char 1 Y/TH DIAGRAM OF THE ACCOUNTING RELATIONSHIPS BETWEEN ACTIVITY AND EMPLOYMENT DEMAND FOR LABOUR Hourly labour produciviy Aciviy Y * Y = Y. TH. E (1) TH E Volume of labour TH Okun s law LABOUR SUPPLY Populaion of working age T Labour force A specific dynamics of he labour force. Those dynamics are deermined by he demography of he oal populaion and by he labour marke paricipaion rae, which is iself influenced in paricular by he prevalence of working women, insiuional facors such as compulsory educaion, reiremen age and rules regarding unemploymen exclusion, and by he business cycle specific o he labour force. Char 1 offers a general view of he relaionships oulined here. 1.2 Hisorical paern of aciviy and employmen Char 2 compares economic growh wih employmen growh over a long period. I clearly reveals a posiive correlaion beween hem. However, he peaks and roughs in he economic cycle do no coincide wih hose corresponding o he growh of employmen. In general, i akes some ime before flucuaions in demand affec he growh of employmen. This response ime has no been sable over he pas fify years and depends on such facors as he deph of he recession, is origin, is expeced duraion, and recourse o flexibiliy insrumens on he par of employers. The adjusmen of producion capaciy in line wih he changing oulook for aciviy is an expensive process and one ha akes ime. Before dismissing workers (or recruiing addiional saff), firms respond firs o a decline (or increase) in aciviy by exploiing heir inensive producion margin (average working ime and/or hourly produciviy). By using various organisaional mehods such as adjusing overime, swiching o par ime working, or emporary lay offs, firms can align heir use of labour more closely wih he needs of producion (inernal quaniaive adjusmen). In a scenario in which adverse economic condiions persis, if hese margins have been used up and he financial resilience of firms is no longer assured, redundancies are unavoidable (exernal quaniaive adjusmen). In addiion, in view of he procedures o be respeced, job losses under collecive redundancy programmes ake some ime o have an effec ; in pracice, several monhs may elapse beween he announcemen of collecive redundancies and he acual job losses ; ha conribues o he ime lag. Average working ime H * Employmen in persons E Y = Y/TH * TH/E * E Unemploymen U In he opposie scenario of an economic upurn, i is only once he flexibiliy levers available o firms have been used and he growh of demand is confirmed ha firms ake on addiional labour ; ha recruimen procedure also akes some ime. 32 The relaionship beween economic growh and employmen NBB Economic Review

3 Char 2 HISTORICAL PATTERN OF EMPLOYMENT AND ECONOMIC GROWTH (daa adjused for seasonal and calendar effecs, percenage changes compared o he corresponding quarer of he previous year) Domesic employmen GDP in volume Sources : NAI, OECD. 2. Breakdown of GDP growh The change in GDP can be aribued o hree facors, namely changes in employmen, average working ime (hours worked per worker) and hourly produciviy (GDP per hour worked, or apparen produciviy). The relaive conribuion of each of hese facors is no sable over ime. The delayed response by employmen and he duraion of ha response can be illusraed by fixing he level of he workforce a he ime of he pre- recession GDP peak and observing he momen when employmen sars o fall and he duraion of ha fall. Char 3 demonsraes his for he five recession episodes in Belgium beween 197 and 214 : a he ime of he firs oil shock in he 197s, hree quarers elapsed before employmen responded o he decline in aciviy. Once he ne job losses se in, hey were significan for a year and persised a a slower rae for some ime; during he episode which followed he second oil shock, employmen responded immediaely and srongly. ob losses persised over a period of more han hree years ; he nex recession which began in 1992 generaed much more modes job losses, wih an almos immediae bu fairly slow conracion and a recovery in he enh quarer (i.e. afer wo years) ; a he ime of he 21 episode, he decline in real aciviy was relaively small. I ook hree quarers for ne job losses o appear, and he losses were modes in comparison wih oher economic crises ; finally, in he grea recession, employmen followed he downward rend in aciviy afer a lag of hree quarers. Quarer- on- quarer employmen growh herefore became negaive a he beginning of 29. Compared o he pre- recession peak in aciviy, ne job losses were on a much smaller scale han in previous episodes, paricularly in regard o he decline in real aciviy. Owing o he sovereign deb crisis in he euro area, employmen began falling again in Belgium beween 212 and 213 (beyond he 14 quarer horizon covered by char 3). The flexibiliy available o firms explains why he ime aken for employmen o respond o changes in GDP may be shorer in erms of hours worked han in he number of persons employed. A he ime of he firs oil shock, he decline in hours worked perpeuaed a fall ha had begun long before 1974, and refleced a downward rend in average working ime. During he second oil shock, average hours hus declined in parallel wih employmen. However, hey picked up slighly sooner han employmen. ln he early 199s, he hours worked fell more seeply han employmen bu he siuaion recovered une 215 The relaionship beween economic growh and employmen 33

4 Char 3 RESPONSE BY DOMESTIC EMPLOYMENT IN PERSONS AND AVERAGE HOURS PER WORKER DURING ECONOMIC RECESSION EPISODES (daa adjused for seasonal and calendar effecs ; indices, respecively levels of employmen and hours corresponding o he pre-recession peak in GDP in volume, = 1) 13 EMPLOYMENT IN PERSONS HOURS WORKED PER WORKER Quarers Q Q2 198 Q Q1 21 Q2 28 Quarers Sources : NAI, OECD, NBB. afer seven quarers. The prioriy use of he inensive margin (average working ime) was paricularly eviden a he ime of he recession in he early 2s and in However, he average hours declined relaively slowly in he early 2s, while he hours worked per worker were adjused more rapidly a he sar of he grea recession. A ha ime, he decline in he pace of work was spread over four quarers. The use of emporary lay offs and he crisis measures adoped in 29 are regularly cied o explain he relaive sabiliy of employmen in Belgium despie he seriousness of he crisis. People laid off emporarily remain on he firm s saff regiser even if hey do no work on he days in quesion. While he level of emporary lay offs was hisorically high and, saring from a low base, heir number increased dramaically in 29, he addiional measures aken a he ime (referred o as crisis measures) had only raher limied success. A he heigh of he recession, more han 2 manual workers were recorded as emporarily laid off, whereas he similar sysem developed for clerical workers (known as saff suspension owing o lack of work for firms in difficuly, a sysem which has since been reained) affeced no more han 8 people in 21. Moreover, he crisis ime credi scheme whereby an employer recognised as in difficuly could offer individual full ime workers a ime-credi in he form of a 5 % or 2 % reducion in heir work, concerned fewer han 3 people. In fac, recourse by Belgian firms o emporary lay offs for manual workers is largely srucural ; beween 1992 (1) and 214, here was only one quarer in which he numbers concerned came o less han 1 (on average), namely he hird quarer of 2. This is a long sanding insrumen in Belgium. Righ from he sar, he Naional Placemen and Unemploymen Office esablished in 1935 (as he forerunner o he Naional Employmen Office) made provision for a form of emporary lay offs for manual workers, alhough here was no legal framework a ha sage (2). The gap beween he change in he oal volume of labour and he change in real aciviy corresponds o he mechanical adjusmen of apparen hourly produciviy. Char 5 below presens he growh raes of he inensive and exensive margins over he period , in (1) Sar of he series expressed in number of paymen recipiens (physical unis). (2) In 1954, he various opions for suspension of an employmen conrac (such as he lack of work owing o economic reasons, bad weaher or force majeure) were pu on a legal basis wihou any resricions on access per secor of aciviy. However, here is a special scheme for he consrucion secor. As ime wen by, new rules on emporary lay offs came ino force (compulsory noificaion o he NEO, monioring procedures, level of replacemen benefis, ec.), some being applicable o all secors while ohers applied only o he consrucion secor. 34 The relaionship beween economic growh and employmen NBB Economic Review

5 Char TEMPORARY LAY OFFS (manual workers, physical unis, aking all reasons ogeher (1), seasonally adjused quarerly averages) order o sudy heir behaviour hroughou he economic cycle. Hisorically, hourly produciviy has aced as a shock absorber, boh during cyclical downurn phases when produciviy growh has weakened, and during recovery phases when growh is sronger, and firms give prioriy o resoring heir margins. However, alhough he growh of hourly produciviy is pro cyclical, i has almos always remained posiive unil he oubreak of he grea recession. Conversely, as expeced in view of he subsequen relaively limied fall in individual hours, i is hus hourly produciviy ha has been hardes hi by he decline in aciviy Char BREAKDOWN OF GDP GROWTH (daa adjused for seasonal and calendar effecs, percenage changes compared o he corresponding quarer of he previous year) To proec jobs, firms may accep a decline in hourly produciviy, while employees may agree o a cu in he hours worked, and hence in heir income. For employers, he fear ha in a siuaion of mismaches on he labour marke and populaion ageing hey migh be shor of skilled saff once he economy picks up has herefore ouweighed he fear of a emporary loss of profiabiliy due o a reducion in hourly labour produciviy. Sources : NEO, NBB. (1) Economic reasons represen around 7 % of all reasons (on average over he period ). A he ime of he grea recession, labour hoarding iniially limied he job losses in a conex of sound corporae fundamenals, bu he furher decline in GDP in 212 caused Hourly produciviy Hours worked per person Employmen in persons GDP in volume Sources : NAI, OECD, NBB. une 215 The relaionship beween economic growh and employmen 35

6 he number of workers o fall more seeply in 213 han in 29. The main reason was ha he labour reenion sysems did no ac as a buffer o he same exen as in 28 and 29. The lengh of he crisis and he hesian exi from i eroded he financial capaciy of some firms, so ha workforce adjusmens were ineviable. In addiion, he condiions governing recourse o emporary lay offs for economic reasons were ighened as an accounabiliy conribuion was inroduced. Thus, he number of people emporarily laid off declined in 213 o approach is long erm average. 3. produciviy and job inensiy of growh : long erm analysis Produciviy flucuaes in line wih he business cycle around a medium- o long erm rend. I is essenial o disinguish beween rend gains in produciviy and heir cyclical variaions. By aking average growh raes per decade, able 1 endeavours o capure he change in he rend componen. Expressed as average annual growh raes, hourly produciviy gains dropped from 4.2 % in he 197s o.9 % in he 2s, before collapsing o.1 % over he period This rend in produciviy gains mirrors ha in GDP. In he 197s, he annual growh of GDP averaged 3.4 %, hen dropped o around 2 % in he 198s and 199s. This growh rae hen slowed in he 2s and 21s, averaging no more han 1 % in recen years. The oher componen of GDP, namely he volume of labour, recorded negaive growh in he 197s and 198s, hen edged upwards o 1 % in he recen period. The difference beween he average annual growh raes for hourly produciviy and produciviy per worker reflecs he change in he average working ime. Tha has diminished over he years owing o various facors, such as a decline in he number of conrac hours, he increase in he rae of par ime working, developmen of he use of ime-credi, bu also he change in he srucure of employmen wihin he economy (see below). If he job inensiy of growh is defined as he raio beween he expansion of employmen and he change in aciviy, ha indicaes he reciprocal of he growh of produciviy per worker. In fac, he job inensiy of growh is clearly pursuing an upward rend. In parallel wih he movemen in annual GDP growh, beween he 197s and he 2s, he average annual growh of employmen increased from.2 o.9 % (.6 % in ). Thus, wihou any breakdown beween he rend and he cycle a his sage, aciviy growh of 1 % in he 197s did no creae any jobs in he economy, whereas i creaed.3 % exra jobs in he 199s and.6 % in he mos recen period. However, in order o sudy he sensiiviy of employmen o growh (elasiciy), i is necessary o focus solely on he cyclical componen of he series, an exercise ha will be conduced in secion 5. The job inensiy of growh also varies according o he naure of he aciviy. I is relaively high in he service branches whereas i is lower, or even negaive, in indusry, given he seady rise in produciviy in ha branch (see below). A he level of he economy as a whole, job inensiy depends on he srucure of he aciviy, and is paern may be influenced by he gradual shif owards a service economy. In a conex of coninuing compuerisaion of occupaions, his rise in job inensiy may appear conrary o some predicions concerning he possibiliy of many human jobs being aken over by robos. In he economic debae TABLE 1 AVERAGE ANNUAL GROWTH RATES OF THE COMPONENTS OF ACTIVITY (annual averages, in %) (1) GDP Employmen Volume of labour Produciviy per person Hourly produciviy Sources : NAI, OECD. (1) The resuls for his shorer period are influenced by he effecs of he grea recession and he sovereign deb crisis. 36 The relaionship beween economic growh and employmen NBB Economic Review

7 on he subjec, supporers of he maximalis approach go quie far, since hey consider ha he auomaion of jobs will no longer apply only o rouine asks bu will also increasingly affec occupaions involving cogniive and non rouine asks. The sudy by Frey and Osborne (213) applies a probabiliy of auomaion o hundreds of jobs in he Unied Saes. Occupaions feauring a high degree of creaiviy, social skills, percepive qualiies and manipulaion are less a risk. The same exercise was conduced by he Bruegel Insiue (Bowles, 214) for European counries. I shows ha, for Belgium, 5 % of occupaions are a risk. However, he findings of his ype of research need o be viewed wih cauion since hey are surrounded by an obvious degree of uncerainy and he analysis does no specify he ime scale of he poenial changes : he definiion of curren occupaions could change in he meanime. Indeed, he disappearance of some occupaions and he emergence of new ones (which could acually resul from hese echnological changes) is nohing new and is cenral o Schumpeer s ideas. 4. Influence of he changes in srucure of aciviy 4.1 rend in employmen in he branches of aciviy (1) indusry : funcions previously performed in house by firms in he branch have been relocaed or ousourced o service companies in order o secure greaer flexibiliy and beer cos conrol. In 214, indusrial jobs averaged 12 % of oal employmen, as agains 19 % in 1995, while marke services accouned for he larges share wih 46 % (compared o 41 % almos 2 years earlier) and non marke services 35 %, or more han a hird (3). ob losses during he grea recession were herefore limied no only by he use of radiional flexibiliy insrumens bu also by he resilience of some branches o flucuaions in aciviy, paricularly non marke services, which are generally largely subsidised by he governmen. The proporion of non marke secor jobs is relaively high in Belgium. Since 1995, he expansion of employmen in Belgium has come mainly from ne creaion of jobs financed enirely or largely by he governmen. This concerns exra saff in public auhoriies and educaion, bu also and primarily workers employed in human healh and social work and in he service voucher sysem among privae employers. Alogeher, i is esimaed ha seven ou of en jobs creaed beween 1995 and 214 are largely financed by he governmen. Employmen does no reac o cyclical flucuaions in he same way in all branches of aciviy. The marke branches, or hose sensiive o he business cycle, comprise agriculure, consrucion, indusry and marke services (2). Non marke services include general governmen and educaion, healh, social work and oher non marke service aciviies. Char 6 6 GROWTH OF ACTIVITY AND EMPLOYMENT BY BRANCH OF ACTIVITY (daa adjused for seasonal and calendar effecs, change compared o he corresponding quarer of he previous year) 8 Non marke services recorded seady growh of employmen up o he end of 211, i.e. including a he heigh of he grea recession. Since 212 he growh rae has slowed in a conex of fiscal consolidaion. In conras, changes in employmen in marke services closely reflec he changes in aciviy, and he number of persons in work here declined in 29. Finally, job losses have persised in indusry since 22, and ha rend was accenuaed during he grea recession. This srucural decline parly reflecs he reorganisaion of he producion process in (1) In he curren NAI series (ESA 21), he breakdown of employmen by branch of aciviy begins in (2) Trade ; repair of moor vehicles and moor cycles ; ranspor and sorage ; hoels and resaurans ; informaion and communicaion ; financial and insurance aciviies ; real esae aciviies ; specialis, scienific and echnical aciviies, and adminisraive and suppor service aciviies. (3) Agriculure and consrucion were no included Source : NAI GDP in volume (in %, lef-hand scale) Indusry Non-marke services Marke services (employmen in housands of persons, righ-hand scale) une 215 The relaionship beween economic growh and employmen 37

8 The shif o he service economy is one of he facors behind he fundamenal rend in oal produciviy and average working ime. 4.2 Working ime in he branches of aciviy The average working ime is generally lower in services han in indusry. A he end of 214, employees worked an average of 388 hours per quarer in indusry, compared o 361 in marke services and barely 341 in non marke services. In addiion, he non marke secor like indusry is seeing a downward rend in average hours per person. The low poin in indusry in 29 was due o he srong adjusmen of employees working hours in response o he crisis. The number of hours worked has sill no reurned o is previous level. These marked differences in level are due o he larger proporion of par ime workers in services, especially in he non marke secor. In 213, according o he resuls of he labour force survey (LFS), over a hird of workers in non marke services worked par ime. In marke services, around a quarer of employees work reduced hours, whereas in indusry only one in en workers is no employed full ime. The change in he srucure of employmen, wih an increased share of branches wih a high rae of par ime work, has herefore depressed he oal average working ime in he economy. The over- represenaion of women in cerain branches of aciviy is a facor here, as wo hirds of jobs in non marke services are filled by women ; more han 4 % of women in he workforce work par ime. Tha raio has doubled since he early 198s. In he case of male workers, only around one in en works reduced hours, bu ha figure has increased five fold since he early 198s. In all, he rae of par ime working has risen from 8 % o almos 25 % over he same period. Facors behind his rend are he rise in he labour marke paricipaion rae of women, he exension of working life among older workers, and he greaer involvemen of men in family responsibiliies. 4.3 produciviy in he branches of aciviy Changes in he relaive weigh of he branches of aciviy in he economy are also par of he reason for he rend in average labour produciviy. In pracice, i is no feasible o obain a measure of he oupu of all economic aciviies since some are non marke aciviies (1). In heir case, value added is esimaed in he naional accouns as he sum of he coss (2). The measure of produciviy is hen biased since any change in he wage bill is passed on in full o he change in value added. For ha reason, char 8 (1) The selling price does no cover 5 % of he producion coss. (2) Inermediae consumpion, compensaion of employees, oher axes on producion ne of subsidies and fixed capial consumpion. Char 7 AVERAGE HOURS WORKED PER EMPLOYEE, BY BRANCH OF ACTIVITY (level per quarer, daa adjused for seasonal and calendar effecs) Char 8 AVERAGE ANNUAL GROWTH OF HOURLY PRODUCTIVITY BY BRANCH OF ACTIVITY (1) (annual averages, in %) Indusry Marke services Non-marke services Toal Indusry Marke services Toal Source : NAI. Sources : NAI, NBB. (1) Based on deailed annual accouns, for which he mos recen year available is 213. Esimaed volume of hours for self employed workers before The relaionship beween economic growh and employmen NBB Economic Review

9 only compares he rend in produciviy for indusry and marke services. The level and growh of produciviy have always been lower in services han in indusry. In indusry, he average annual growh of hourly produciviy dipped from 4.5 % in o 1.8 % over he recen period encompassing he grea recession. In marke services, he average annual growh of produciviy surged beween 2 and 24, noably as a resul of increased capial inensiy, and paricularly ha relaing o invesmen in ICT (Federal Planning Bureau, 27), hough wihou surpassing he growh rae in indusry. In elecommunicaions, he average annual growh rae beween 2 and 24 exceeded 1 %, while i was above 5 % in rade aciviies over he same period. In financial and insurance aciviies, ha growh exceeded 5 % in he laer half of he 199s. Afer ha, he average annual growh rae in marke services declined, and even became negaive beween 28 and 213. Thus, he iniially promising developmens could no preven he downward rend in he growh of produciviy gains for he economy as a whole. 5. Esimae of he relaionship beween aciviy and employmen As explained in secion 3 above and as illusraed in able 1, he growh raes of GDP, employmen, and oal hours worked follow non linear, divergen long erm rends. This implies ha, in order o sudy and above all, measure he relaionships beween hese variables hroughou he business cycle, i is essenial o disinguish beween he cycle and he rend. Tha breakdown may operae, for example, for he variables cenral o char 1, consiuing Okun s law (1962).Okun s law represens he empirical regulariy ha Okun observed in he relaionship beween unemploymen and real GDP : U U* = α(y Y * uy ) +ε, α < where U represens he unemploymen rae, Y he neperian logarihm of real GDP, and U * and Y * he rend value of hose variables. The difference beween he observed value of a variable and is rend value is ha variable s cyclical componen, or he gap relaing o ha variable. The coefficien a herefore describes he sensiiviy of he cyclical componen of he unemploymen rae o a 1 % change in he cyclical componen of GDP. Equaion (2) above is commonly used o esimae he percenage increase (namely 1/a) in real aciviy above is long erm rend necessary o generae a 1 percenage poin fall in unemploymen compared o he long erm equilibrium. (2) Tha relaionship can be deemed o originae from wo oher empirical relaionships found, namely he posiive correlaion beween he cyclical componen of employmen and ha of GDP, and he negaive correlaion beween he cyclical componen of employmen and ha of he unemploymen rae : * * ey E E = β(y Y ) +ε, β > * * ue U U = γ(e E ) + ε, γ < wih (1) : γ = a / β and ɛ uy =ɛ ue +γɛ ey where E represens he nepe rian logarihm of employmen expressed in persons. There are various ways of making his disincion beween rend and cycle. There is a degree of consensus in favour of he mehod popularised by Hodrick and Presco (2). One advanage of ha mehod is ha i can explicily ake accoun of medium / long erm changes in he rend, unlike a growh rae analysis which implicily presupposes a consan linear rend. This breakdown is illusraed in char 9 for GDP and for employmen. The difference beween he rend and he gross series gives he cyclical componen expressed as a percenage deviaion from he rend. By consrucion, i is saionary, oscillaing around. The join observaion of he cyclical componens of employmen and real GDP is highly informaive : employmen is evidenly a delayed pro cyclical variable, i.e. i lags slighly behind GDP. Tha characerisic is more obvious han in char 2, expressed as an annual growh rae. I is also eviden ha is movemens hroughou he business cycle are equivalen in ampliude o around wo hirds of he movemen in GDP ; he closeness of he relaionship beween employmen and GDP may change from one cycle o he nex. Le us ake he example of he recen double dip recession in and A he ime of he financial crisis, he cyclical componen of employmen produced a delayed and exremely moderae reacion, boh in he growh phase and in he conracion phase occurring around 28. In conras, a he ime of he afershock sovereign deb crisis, employmen reaced a he same ime as economic aciviy and in exacly he same proporion. (1) If he erm on he righ of equaion (4) is replaced by equaion (3), ha gives an expression equivalen o equaion (2). The resricions menioned in he ex require he esimaed sysem o be consisen wih he breakdown envisaged. (2) The mehod iself was devised by he mahemaician Edmund Whiaker in I uses a coefficien of penaly λ for he firs difference from he rend. Tha coefficien is usually se a 16 for quarerly variables, which is wha applies here. (3) (4) une 215 The relaionship beween economic growh and employmen 39

10 Char 9 TREND AND ECONOMIC CYCLE IN THE CASE OF REAL GDP AND EMPLOYMENT (neperian logarihm in he upper chars and percenage deviaion from he rend in he lower char. The grey areas correspond o periods of recession) Real GDP (ln) Employmen (ln) Trend Trend 5 % 4 % 3 % 2 % 1 % % 1 % 2 % 3 % 4% 5 % 4 % 3 % 2 % 1 % % 1 % 2 % 3 % 4% Cyclical componen of GDP Cyclical componen of employmen Sources : OECD, own calculaions. 5.1 Esimaes for differen economies I mus be clearly undersood ha here can be no quesion of verifying expressions (2) o (4) from one quarer o he nex. For he purposes of he economeric esimaion of he coefficiens of hese equaions, namely a, β and γ, i is herefore preferable o use a dynamic specificaion. By inroducing dynamics, i is possible o ake accoun of he fac ha he employmen cycle lags slighly behind he cycle of real economic aciviy, as observed above. U U * = α + α 1 (Y Y * ) + α (Y Y * ) + α (Y Y * ) uy + ε ; α = α 1 + α 2 + α 3 (5a) E E* = β + β 1 (Y Y * ) + β (Y Y * ) + β (Y Y * ) ey + ε ; β = β 1 + β 2 + β 3 (5b) U U* = γ + γ 1 (E E * ) + γ (E E * ) + γ (E Y * ) ue + ε ; γ = γ 1 + γ 2 + γ 3 (5c) wih γ 3 = [ α γ 1 γ 2 ] β The following ables only presen he sum of he coefficiens associaed wih he various lags of he explanaory variable, i.e. only he coefficiens a, β and γ (1), for ease of inerpreaion (2). In order o place he value of he elasiciies a, β and γ obained for Belgium in relaion o hose of oher developed economies, he sysem of equaions (5) was also esimaed for some euro area counries (Germany, Spain, Finland, France, Ireland, Ialy and he Neherlands), he euro area as a whole, Denmark and he Unied Kingdom, and he Unied Saes. As far as possible, he esimae was based on long quarerly series from he firs quarer of 196 o he second quarer of 214. However, for some counries he daa are no available for he whole period, so ha he esimaed coefficiens are no always enirely comparable. Tha is he case, in paricular, for Ireland (1) The saisical significance ess also concern he sum of all he parial slope coefficiens for each relaionship examined. (2) By esimaing he hree equaions in a single sysem simulaneously i is possible o impose he heoreical resricion Ɣ = α /β on he coefficiens, whereas ha is no necessarily he case if he equaions are esimaed independenly. 4 The relaionship beween economic growh and employmen NBB Economic Review

11 TABLE 2 ESTIMATION OF THE ELASTICITY OF UNEMPLOYMENT TO GDP, EMPLOYMENT TO GDP AND UNEMPLOYMENT TO EMPLOYMENT FOR TWELVE DEVELOPED ECONOMIES Counry Period Elasiciy of unemploymen o GDP Elasiciy of employmen o GDP Elasiciy of unemploymen o employmen a R² b R² g R² Germany... 91Q3 14Q2.273** (.23).55.34** (.24) ** (.65).785 Belgium... 6Q3 14Q2.41** (.2) ** (.23) ** (.37).692 Denmark... 66Q3 14Q2.348** (.19) ** (.24) ** (.32).686 Spain... 76Q3 14Q2.93** (.44) ** (.52) ** (.33).831 Finland... 75Q3 14Q2.418** (.17) ** (.2) ** (.25).884 France... 66Q3 14Q2.281** (.13) ** (.16).71.68** (.28).74 Ireland... 9Q3 14Q2.387** (.31) ** (.43) ** (.35).848 Ialy... 6Q3 14Q2.144** (.14) ** (.21) ** (.4).456 Neherlands... 6Q3 14Q2.354** (.18) ** (.22).5.756** (.35).829 Unied Kingdom... 6Q3 14Q2.32** (.15) ** (.19) ** (.28).83 Euro area... 7Q3 14Q2.348** (.13) ** (.16) ** (.21).846 Unied Saes... 6Q3 14Q2.467** (.14) ** (.17) ** (.19).99 Sources : OECD, own calculaions. ** indicaes a deviaion significanly differen from zero a he 5 % hreshold, * indicaes a deviaion significanly differen from zero a he 1 % hreshold, ( ) sandard deviaions and Germany and, o a lesser exen, for Denmark, he euro area, Spain and Finland. The esimaion periods are repored sysemaically in able 2. As expeced, able 2 shows ha he sronges relaionship is beween unemploymen and employmen, as he former merely mirrors he laer via he disoring filer of he job supply, so ha he correlaion is no perfec. On his subjec, i is ineresing ha he elasiciy γ (unemploymenemploymen) is paricularly high in Belgium (1). This is probably due o he sysem of unemploymen benefis wih no ime limi (a leas unil recenly) ; as a resul of ha sysem, wihin he populaion of working age, he numbers joining or leaving he labour force are relaively few, which implies a relaively non cyclical job supply. The column showing he elasiciy of employmen o GDP indicaes ha employmen is more sensiive o flucuaions in economic aciviy in he Unied Saes (.82) han in he euro area (.57). In he core European counries, namely Belgium, he Neherlands, France and he Unied Kingdom, ha elasiciy is very uniform, hovering around.5. In principle, a counry wih a flexible (rigid) jobs marke ypically has a high (low) elasiciy of employmen o GDP. For mos counries in he sample, he ranking is no oo surprising, wih he noable excepion of Spain which emerges as he champion in erms of job marke flexibiliy (2), and Germany a he oher end of he specrum. Unil he recen reforms, he Spanish employmen marke (1) The figure for Belgium is.8, only exceeded by Germany a.9. However, he figures calculaed for Germany are no enirely comparable owing o a much shorer esimaion period. (2) The fac ha he esimaed elasiciy is significanly greaer han one for his counry is roubling, and indicaes ha he labour marke is o say he leas aypical. une 215 The relaionship beween economic growh and employmen 41

12 exhibied a marked division beween highly proeced jobs, on he one hand, and fixed erm or agency jobs offering employers massive flexibiliy. Presumably, he laer cushion shocks in eiher direcion and are responsible for he grea sensiiviy of employmen o he economic cycle. The very low sensiiviy of German employmen o economic aciviy may seem surprising, bu i is necessary o bear in mind ha he sample is much shorer han for he oher counries since i begins in 199, i.e. jus as he counry embarked on he process of reunificaion. Apar from Spain, i is he Unied Saes and Ireland ha clearly have he mos dynamic labour marke (since he 199s). They are followed by he Scandinavian counries and, finally, he old European counries, wih Ialy railing behind wih a job marke paricularly insensiive o he economic cycle. 5.2 Robusness of he esimae of he relaionship beween employmen and GDP Before going any farher in his inernaional comparison, i is appropriae o quesion he robusness of hese esimaes. The robusness of hese esimaed elasiciies can firs be assessed wih he aid of a binary variable D 85, which akes he value before he firs quarer of 1985 and he value 1 from ha dae onwards. I hus allows he period observed o be divided ino wo more or less equal halves. Tha dae also corresponds o a spae of labour marke liberalisaion measures in various economies. The employmen GDP relaionship presened in he above sysem of equaions (5b) is re esimaed by allowing he parial slope coefficiens o change in he second half of he esimaion period : E E * = β + β 1 (Y Y * ) + β (Y Y * ) + β (Y Y * ) D 85 [ β 4 (Y 1 Y * 1 ) + β 5 (Y 1 Y * 1 ) β 6 (Y 2 Y 2 )] + εey wih β = β 1 + β 2 + β 3 and β D85 = β 4 + β 5 + β 6 The coefficien β indicaes he elasiciy of employmen o GDP during he firs half of he period examined, while he sum of β + β D85 gives ha same elasiciy during he second half of he period (1). The resuls obained for he en economies wih a sufficienly large sample are se ou in char 1. There are wo lozenges corresponding o each counry, represening β and β + β D85 respecively, while he red lines above and below correspond o he confidence inerval of 95 % around hese esimaes. Overlapping confidence inervals should be inerpreed as meaning ha here is no saisically significan srucural change a he 5 % hreshold (1) I could be said ha by considering only he sum of he parial slope coefficiens, we are only esing an overall srucural change, bu i is perfecly possible ha, alhough his overall coefficien did no change afer 1985Q1, he dynamics were affeced, i.e. he relaive weigh of GDP and GDP afer a lag of one or wo quarers. This sudy does no address ha ype of srucural change. (6) Char 1 TEST FOR A STRUCTURAL CHANGE IN THE EMPLOYMENT GDP RELATIONSHIP AROUND 1985Q1 (he red lines correspond o confidence inervals of 95 % on eiher side of he esimaed elasiciies, represened by he lozenges) Q4-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 1975Q3-1984Q4 1985Q1-214Q2 197Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 196Q3-1984Q4 1985Q1-214Q2 Spain Unied Saes Finland Euro area Unied Kingdom France Neherlands Denmark Belgium Ialy Sources : OECD, own calculaions. 42 The relaionship beween economic growh and employmen NBB Economic Review

13 Char 11 TEST ON THE EFFECT OF RECESSIONS ON THE ELASTICITY OF EMPLOYMENT TO GDP (he red lines correspond o confidence inervals of 95 % on eiher side of he esimaed elasiciies, represened by he lozenges) Spain Ireland Unied Saes Finland Denmark Euro area Belgium Unied Kingdom Neherlands France Ialy Germany Sources : OECD, own calculaions. in he employmen GDP relaionship. The counries are ranked in descending order of employmen GDP elasiciy esimaed for he second half of he period observed. In general, he esimaed elasiciy of employmen o GDP is slighly greaer for he second half of he sample. However, ha difference is no significan for six ou of en counries, he excepions being Spain, he Neherlands, France and he Unied Kingdom. The specific case of Spain is readily explained by Francoism, a degree of labour marke liberalisaion and he pursui of European inegraion. The case of France has already been highlighed by Blanchard and Cohen (24) ; i corresponds o he easing of consrains on he labour marke and on job securiy. The Neherlands also inroduced fundamenal labour marke reforms following he crisis of he 198s ; he reforms were apparenly more drasic han hose in he Unied Kingdom during he Thacher era (1). The above observaions are unaffeced if accoun is aken of a possible change in he relaionship beween employmen and GDP in he mid 198s. Afer 1985, he 95 % confidence inervals show ha here is no saisical difference in his elasiciy beween Belgium, Denmark, France, Finland, he Unied Kingdom and he euro area as a whole. Finally, as regards he elasiciy of employmen o (1) If longer series had been available for Ireland, a marked srucural change would also have been idenified for ha counry, which underook fundamenal srucural reforms from he lae 198s and especially in he early 199s. GDP, Belgium is paricularly similar o Denmark, anoher small, very open economy which was also hard hi by he second oil shock. Tha comparison may appear flaering, since Denmark is he counry wih flexible securiy, bu i is also fallacious since i is valid only for he relaionship beween employmen and GDP. If he level of he unemploymen rae and he average duraion of unemploymen are considered, hose wo parameers are much lower in Denmark. A second robusness es can be conduced a he level of he symmery of he relaionship beween employmen and growh according o wheher or no he economy is in recession. For ha purpose, we consider five economic crisis episodes : 1973Q3 1976Q1 : firs oil crisis ; 198Q1 1983Q3 : second oil crisis ; 199Q2 1994Q3 : banking crisis in Finland and Sweden ( ) and EMS crisis ( ) ; 21Q1 23Q2 : echnology socks crisis and repercussions of he New York erroris aacks on he air ranspor secor and ohers ; 28Q1 212Q4 : financial crisis and sovereign deb crisis. une 215 The relaionship beween economic growh and employmen 43

14 Char 12 SCATTER DIAGRAMS FOR THE CYCLICAL COMPONENTS OF EMPLOYMENT AND REAL GDP, SHOWING OBSERVATIONS RELATING TO THE LATEST RECESSION.3 BELGIUM 196Q1-214Q2.3 EURO AREA 197Q1-214Q2.3 GERMANY 1992Q1-214Q2 27Q4 27Q4 27Q REAL GDP (CYCLICAL COMPONENT) FRANCE 196Q1-214Q2 27Q NETHERLANDS 196Q1-214Q2 27Q4.5 UNITED STATES 196Q1-214Q2.3 27Q DENMARK 1966Q1-214Q2 FINLAND 1975Q1-214Q2 IRELAND 199Q1-214Q Q4 27Q4 27Q ITALY 196Q1-214Q2 SPAIN 196Q1-214Q2 UNITED KINGDOM 196Q1-214Q Q4.2 27Q4.1 27Q Observaions Observaions from 27Q4 o 212Q2 EMPLOYMENT (CYCLICAL COMPONENT) Sources : OECD, own calculaions. 44 The relaionship beween economic growh and employmen NBB Economic Review

15 These periods are somewha arbirary in ha he recession episodes may vary from one economy o anoher. However, hey were chosen o be long enough so ha each sub period offered an adequae degree of freedom and so ha each one covered he inerval beween he peak and he rough for every counry. The es follows exacly he same procedure as he previous one, The only difference is ha, in equaion (6), he binary variable D 85 is replaced by a binary variable D recession, which akes he value 1 in he quarers included in he above lis of recessions and he value elsewhere. The resul is shown in char 11. A quick glance is enough o confirm ha he relaionship beween employmen and growh is no generally affeced by crisis episodes. I is no surprising ha he business cycle has no influence here, since Okun s law and is corollaries express a long erm relaionship. However, each recession phase is differen, being caused by differen ypes of shock, and i is conceivable ha he relaionship beween employmen and GDP may be affeced emporarily from one crisis or one counry o anoher. In paricular, a he ime of he recen grea recession various counries, including Belgium, made much of heir policies aimed a preserving jobs. Did hey acually work? Char 3 above has already given an illusraion for Belgium, while char 12 ries o exend he analysis o all welve economies under review. For each economy considered, he char presens a scaer diagram illusraing he saisical relaionship beween he cyclical componen of real GDP, on he x axis, and he cyclical componen of employmen, on he y axis. For each counry, a scaer plo emerges wih a posiive slope, synhesised by he ordinary leas squares regression line passing hrough i, wih a slope equal o he elasiciy calculaed in able 2 above. These blue scaer plos look fairly homogenous, bu ha impression disappears if he plos corresponding o he las recession are highlighed. For all he counries included here, hose observaions ake he form of a spiral, i.e. : a firs, he employmen GDP relaionship appears o be grealy aenuaed, wih GDP falling (movemen owards he lef) and employmen no responding (he movemen is almos horizonal). Tha movemen owards he lef edge of he scaer plo is iniiaed from an area close o he regression line ; (1) No forgeing he srucural change idenified in 1985 for he four counries concerned. (2) For compleeness, i mus also be poined ou ha, for he economies which saw a significan srucural change in he relaionship beween employmen and GDP in he mid 198s (see char 1), an addiional binary variable is included in he regression o ake accoun of ha facor. afer several quarers of weak or zero reacion, employmen declines, ofen a a ime when he fall in GDP has ceased, riggering a verical downward movemen ; once i sops falling, employmen sabilises while GDP edges very gradually back up owards is cenral rend, causing a reurn o he righ, owards he cenre of he scaer plo ; his o and fro movemen is ypical of he firs crisis period. The afershock crisis in he form of he sovereign deb crisis has a very differen profile. This ime, employmen conracs a he same ime as GDP, along a slope idenical wih he slope of he regression line, or even more seeply. Does his visual observaion correspond o a saisically significan change in he slope? Are hese phenomena also apparen in oher recession episodes? To answer hose quesions, we need o conduc a new es on srucural change, his ime, separaely for each recession (1). For he welve economies considered, he elasiciy of employmen o GDP is re esimaed wih a binary variable for each recession episode described above, namely he variables D 1, D 2, D 3, D 4 and D 5. Tha gives he following expression (2) : E E* = β + β 1 (Y Y * ) + β (Y Y * ) + β (Y Y * ) Ʃ i=1 [ β 1Di D i (Y Y )* + β 2Di D i (Y 1 Y * 1 ) + β 3Di D i (Y 2 Y * 2 )] + ε ey wih : β = β 1 + β 2 + β 3 and β Di = β 1Di + β 2Di + β 3Di (i = 1, 2, 3, 4, 5) D i : binary variable aking he value 1 in he quarers corresponding o he recession i (i = 1, 2, 3, 4, 5), and he value elsewhere. The esimaed coefficien β gives he sensiiviy of employmen o economic aciviy ouside recession periods, while he coefficiens β Di esimae he exen o which crisis i affecs ha elasiciy. If hey deviae significanly from, ha implies ha his recession did acually change he relaionship beween employmen and GDP. Table 3 offers more informaion concerning char 11. For he welve economies considered, i conains he resuls of he esimaed deviaions of he employmen GDP elasiciy compared o non recession periods. The figures marked wih one aserisk correspond o coefficiens significanly differen from a he 1 % hreshold, while hose marked wih wo aserisks are significanly differen a he 5 % hreshold. The counries were ranked from he one wih he smalles number of recession episodes associaed wih a significan change in elasiciy o he one wih he larges number. The following poins emerge from observaion of his able : (7) une 215 The relaionship beween economic growh and employmen 45

16 char 11 shows ha, aking all crisis episodes ogeher, here is no saisically significan impac on he elasiciy of employmen o GDP in any counry. Conversely, a horizonal reading of able 3 indicaes ha if he recessions are considered individually, some of hem seem o be associaed wih an employmen GDP elasiciy significanly differen from he figure esimaed for non crisis periods as a whole. Depending on he recession and he counry, he elasiciy is someimes increased and someimes reduced ; Belgium and Germany boh recorded a considerable increase in he elasiciy of employmen o GDP a he ime of he 21 recession and a (smaller) reducion in ha elasiciy during he las recession ; France seems o be he counry wih he leas consan relaionship beween employmen and GDP ; a verical reading of he able reveals ha, for any given recession episode, economies may see a considerable change in he elasiciy of employmen o GDP, bu he sign is never he same for all counries, excep in he mos recen crisis. In fac, here is apparenly a consensus concerning he grea recession in ha he elasiciy declines significanly (a he 1 % hreshold) for seven ou of welve economies, he oher five producing no significan change. Tha confirms he visual impression lef by he spirals in char 12 ; he economies in which here was no significan reducion in he relaionship beween employmen and GDP during he las crisis are eiher he ones where ha relaionship is exremely robus, such as he Unied Saes and Denmark, or he peripheral euro area counries which were more seriously affeced by he financial crisis and he sovereign deb crisis. As already saed, his decline in he elasiciy of employmen o GDP during he las recession mus cerainly no be inerpreed as a permanen srucural change since, during he period from he firs quarer of 28 o he las quarer of 212, he employmen GDP relaionship becomes very fla a firs, revering o is original form or TABLE 3 DEVIATIONS IN THE ELASTICITY OF EMPLOYMENT TO GDP FOR EACH RECESSION EPISODE COMPARED TO NON RECESSION PERIODS Counry Period episodes 73Q3 76Q1 8Q1 83Q3 9Q2 94Q3 1Q1 3Q3 8Q1 12Q4 R² Unied Saes... 6Q3 14Q2.12 (.8). (.7).155 (.16).8 (.154).153 (.18).855 Denmark... 66Q3 14Q2.56 (.199).52 (.153).12 (.153).14 (.247).23 (.96).675 Ialy... 6Q3 14Q2.148 ( (.17).526** (.138).12 (.222).19 (.11).45 Spain... 76Q3 14Q2.284** (.134).118 (.364).6 (.12).36 (.277).67 (.12).858 Ireland... 9Q3 14Q2 n. n..44** (.225).174 (.327).83 (.154).75 Germany... 91Q3 14Q2 n. n..117 (.155).45** (.145).173** (.73).573 Euro area... 7Q3 14Q2.125 (.92).57 (.11).271** (.71).43 (.119).198** (.61).847 Belgium... 6Q3 14Q2.91 (.11).192* (.116).71 (.11).423** (.177).153* (.92).66 Neherlands... 6Q3 14Q2.39 (.123).359** (.95).163 (.124).3 (.146).245** (.67).624 Finland... 75Q3 14Q2.248 (.197).212 (.169).232** (.66).398** (.191).24** (.61).867 Unied Kingdom Q2.188* (.112).135 (.1).17 (.14).394* (.27).249** (.9).73 France Q2.123** (.57).218** (.121).151* (.69).47 (.12).24** (.58).81 Sources : OECD, own calculaions. ** indicaes a deviaion significanly differen from zero a he 5 % hreshold, * indicaes a deviaion significanly differen from zero a he 1 % hreshold, ( ) sandard deviaions 46 The relaionship beween economic growh and employmen NBB Economic Review