Different Age Groups

Transcription

1 OSIPP Discussion Paper : DP-001-E-003 The Impac of Secoral Shifs on he Unemploymen Rae of Differen Age Groups 30 November 001 Kei Sakaa Researcher, Insiue for Social and Economic Research, Osaka Universiy C.R. McKenzie Professor, Osaka School of Inernaional Public Policy, Osaka Universiy Keywords Unemploymen; Secoral Shifs Hypohesis; Age cohor effecs; Time Series Analysis; Generaed regressors. Summary This paper examine Lilien s secoral shifs hypohesis for Japan for differen age cohors. Previous sudies of he secoral shifs hypohesis for Japan have for he mos par concenraed on he relaionship beween aggregae unemploymen and secoral shifs, and are ypically no supporive of he hypohesis. However, recen increases in he unemploymen raes of he young and he aged sugges a need o reexamine he hypohesis for differen age groups. I is found ha secoral shifs have a shor-erm posiive effec on he unemploymen of aged male workers and he effecs increase in imes of recession.

2 The Impac of Secoral Shifs on he Unemploymen Rae of Differen Age Groups Kei Sakaa a and C.R. McKenzie b a Insiue for Social and Economic Research, Osaka Universiy, Japan (ksakaa@osipp.osaka-u.ac.jp) b Osaka School of Inernaional Public Policy, Osaka Universiy, Japan (mckenzie@osipp.osaka-u.ac.jp) Absrac: This paper examine Lilien s secoral shifs hypohesis for Japan for differen age cohors. Previous sudies of he secoral shifs hypohesis for Japan have for he mos par concenraed on he relaionship beween aggregae unemploymen and secoral shifs, and are ypically no supporive of he hypohesis. However, recen increases in he unemploymen raes of he young and he aged sugges a need o reexamine he hypohesis for differen age groups. I is found ha secoral shifs have a shor-erm posiive effec on he unemploymen of aged male workers and he effecs increase in imes of recession. Keywords: Unemploymen; Secoral Shifs Hypohesis; Age cohor effecs; Time Series Analysis; Generaed regressors. 1. INTRODUCTION Lilien s [198] secoral shifs hypohesis assers ha secoral shifs of demand affec unemploymen because labour reallocaion across secors is a ime consuming process. Using annual poswar daa for he Unied Saes, Lilien [198] esimaed a reduced form unemploymen equaion ha included curren and lagged values of a cross-secoral dispersion measure (Lilien s sigma) and a moneary disurbance measure. Lilien found a saisically significan posiive relaionship beween he dispersion measure and he unemploymen rae. Abraham and Kaz [1984, 1986] argue ha aggregae demand disurbances can induce counercyclical movemens in Lilien s dispersion measure. Abraham and Kaz [1984] find ha when Lilien s sigma is purged of he influence of aggregae demand shocks, his purged proxy variable can explain only a small fracion of unemploymen flucuaions in he Unied Saes. Davis [1987] suggess ha Lilien fails o ake ino accoun he possibiliy of sage-of-business-cycle effecs on he secoral shifs hypohesis. If he value

3 of foregone producion associaed wih unemploymen is procyclical, hen here will be incenives for unemploymen spells o be shorer during expansions and longer during recessions. Therefore, a given amoun of labour reallocaion may lead o underesimaing unemploymen when aggregae macroeconomic condiions are improving and vice versa. Taking hese suggesions ino consideraion, Mills e al. [1995] provide evidence o suppor he secoral shifs hypohesis for he Unied Saes. In conras, Brunello s [1991] sudy for he aggregae unemploymen rae in Japan does no suppor he secoral shifs hypohesis. However, Sakaa [001] finds ha secoral shifs affec male unemploymen, bu no female unemploymen. He speculaes ha gender differences in he accumulaion of human capial may play a crucial par in explaining hese differences. Sakaa s [001] findings sugges ha i is imporan o examine Lilien s hypohesis for differen cohors, and ha he findings for he aggregae unemploymen rae can be decepive. If as Sakaa [001] speculaes he effecs of secoral shifs on unemploymen depend on he accumulaion of indusry-specific human capial by workers, hen secoral shocks may have a large impac on older workers who have accumulaed large amouns of human capial. In conras, secoral shocks may have a small (or no) impac on young workers who have ye o accumulae very much human capial. Sakaa [001] indicaes ha here are gender differences in he effecs of secoral shocks on unemploymen, and herefore, he analysis will focus on male unemploymen. Currenly, he unemploymen raes of workers in he age groups 15-4 and are paricularly surging. In he conex of he aging populaion and casualisaion of he youh workforce, i is imporan o scruinize he effecs of secoral shifs on hese wo groups. In his paper, incorporaing criicisms made of Lilien s work by previous sudies, he secoral shifs hypohesis is esed for males aged 15-4 and using Lilien s dispersion index and a purged index, and ess for sage-of-business-cycle effecs are conduced. The plan of his paper is as follows. In secion, he daa used are described. In order o obain esimaes of he expeced and unexpeced componens of money growh, a money growh equaion is specified and esimaed in secion 3. Tess of he secoral shifs hypohesis for he aggregae unemploymen rae and he male unemploymen rae of wo age groups are conduced in secion 4. Secion 5 explores he possibiliy ha he sage of he business cycle has an effec on labour reallocaion. Secion 6 provides some concluding remarks.. DATA Employmen daa o compue he measures of secoral shifs are aken from The Monhly Repor on Labour Force Survey (Roudouryoku Chousa Houkoku) (LFS). LFS provides a hireen-indusry division of monhly nonagriculural employmen. The hireen secors are: Fisheries; Mining; Consrucion; Manufacuring: Texile mill producs; Manufacuring: Chemical and relaed producs; Manufacuring: Meal and machinery;

4 Manufacuring: Oher; Elecriciy, gas, hea and waer supply; Transpor and Communicaion; Wholesale Trade; Reail rade, resaurans and pubs; Finance, insurance and real esae; and Services. This caegorizaion of indusries began in January As a resul, he sample is from he firs quarer of 1973 o he fourh quarer of Daa on he aggregae unemploymen rae and he unemploymen raes for males aged 15-4 and males aged are also obained from LFS. The shor-erm and long-erm ineres raes, he hree monh Gensaki rae and he 10 year governmen bond yield, are aken from he NEEDS daabase. Oher macrodaa are aken from he OECD Main Economic Indicaors (MEI) in he OECD Saisical Compendium. Money is measured as M+CD. The real values of money and GDP are compued by deflaing by he consumer price index (CPI). Energy prices are measured using he Fuel and Elecriciy Price Index. The inflaion rae is compued using he CPI. All he macroeconomic daa (excep he ineres raes) are seasonally adjused. M+CD and real GDP, respecively, and R and i be he long-erm ineres rae and he inflaion rae. In order o consruc he expeced and unexpeced componens of money growh, no conemporaneous variables appear in he money growh equaion. Uni roo ess sugges ha all variables in he money growh equaion are non-saionary I(1), and a Johansen es suggess ha here is one coinegraing relaionship beween m, y, R, and i. As a resul, he money growh equaion is esimaed as an error correcion model wih four lags. The residual from his money growh regression, DMR, is hen used as a measure of unanicipaed money growh, and anicipaed money growh is defined as DME - DMR, where denoes he firs difference of a variable. DME and DMR are used as explanaory variables in an unemploymen equaion o es he secoral shifs hypohesis. 4. TESTING THE SECTORAL SHIFTS HYPOTHESIS m = 3. MONEY GROWTH EQUATION In order o es he implicaions of he secoral shifs hypohesis, a well-specified unemploymen equaion is required. Expeced and unexpeced money demand shocks compued as he fied values and residuals from a Barro [1977] ype money growh equaions are included in he unemploymen equaion as explanaory variables. The money growh equaion is esimaed in a simple form. Le m and y be he logarihms of real In his secion, he proxy variables for secoral shocks are inroduced. To check he robusness of he resuls obained, wo differen measures, Lilien s [198] measure and Mills e al. s [1995] measure which akes accoun of Abraham and Kaz s [1984] criicism, are used. Lilien s measure has he following form: N ( ei / E ) ( log( ei ) log( E i 1 σ = )), (1) = where e i is he employmen level in secor i, i=1, N, 13, a ime, E = e i is aggregae i= 1 employmen a ime, and N=13.

5 Mills e al. [1995] purge he index of moneary shocks, regarding hese as he main driving force on he demand side. Each secoral growh rae, log( e ) log( E ), was regressed on he curren i value and four lags of expeced and unexpeced money growh, DME and DMR, as follows: e ) E ) = 4 4 a log( i j - j j - j j=0 log( - j=0 () Equaion () was esimaed for each i by ordinary leas squares, and he residuals,. + DMR + DME + ~, were combined o obain he purged dispersion index defined as: N i=1 i ~ i s = ( E ). (3) / e Augmened Dickey-Fuller (ADF) ess applied o σ and s rejec he null hypohesis of a uni roo suggesing ha hey are boh saionary. In conras, he unemploymen raes of all hree groups namely, all workers, U, males aged 15-4, U15, and males aged 55-64, U55, are found o be nonsaionary. As he dispersion measures are saionary, hey canno be coinegraed wih he non-saionary U, U15, and U55. Thus, here can be no long-erm relaionship beween he unemploymen variables and he dispersion measures. Following Wallis [1987], U, U15, and U55 are he logisic ransformaion of he unemploymen rae of each group. The unemploymen equaion for each group includes he shor-erm ineres rae, r, he logarihm of he fuel and elecriciy price index, FEPI, and he raio of impors o GDP, IMP. The inclusion of he shor-erm ineres rae aims o capure he impacs on unemploymen of changes in i he working-capial coss of firms or he ineremporal subsiuion of leisure. Impors can be assumed o reflec he effec of shocks from he global economic environmen on domesic unemploymen. The fuel and elecriciy price index is used o conrol for he impac of oil shocks in he sample period. As wih he specificaion of he money growh equaion, a vecor auoregression (VAR) is firs esimaed o deermine he order of VAR, and hen, Johansen ess are conduced o deermine he number of coinegraing relaionships. In choosing he order of VAR and in esing coinegraion, he conemporaneous values and up o four lags of DME, DMR and σ are included as exogenous variables. When he variable s and he business cycle variables are added, i is assumed ha he unemploymen equaions used for esing he secoral shifs hypohesis have he same srucure as he model specified when σ is used. The es resuls for σ and s are shown in Tables 1 and, respecively. To save he space, he repored resuls are limied o esimaes direcly relaed o he dispersion measures. Furhermore, as generaed regressors, DME and DMR, are included in he esimaed model, -saisics are compued using Newey-Wes sandard errors (Parzen weighs). For aggregae unemploymen, he resuls in Tables 1 and are ambiguous in ha here are boh posiive and negaive significan coefficiens. As shown in Sakaa [001], his may be due o reaing he labour force as a homogeneous group.

6 Table 1. The impac of σ on Unemploymen. σ T=100 (Sample Period: 1975Q1-1999Q4) Absolue -values are presened in parenheses. * and ** indicae he variable is significan a he 5% and 1% significance levels, respecively. s U Table. The impac of U U15 T ** (3.4) (0.45) (0.67) (1.0) U15 s on Unemploymen. U55 T -0.94** (.56) (0.7) (0.95) (0.36) ** (1.13) (3.33) (0.93) R T=95 (Sample period: 1976Q-1999Q4) U * ** (.43) (1.) (3.46) * (.37) (1.8) R The resuls for he males aged 15-4 show ha none of he curren or lagged values of eiher dispersion variable, σ or s, are individually significan. Wald ess confirm heir join insignificance (p-value=0.08 for σ and p- value=0.6 for s ). Alhough he Wald es for σ rejecs he null hypohesis a he 10 % significance level, none of dispersion measures are significan when he curren and lagged values of σ are added individually. On he oher hand, for males aged he impac of he dispersion index is posiive and significan for boh σ and s. The resuls are in line wih he hypohesis ha he groups ha have relaively more indusry- and firm- specific human capial are more affeced by secoral shocks. 5. STAGE-OF-BUSINESS CYCLE EFFECTS The sage-of-business-cycle effecs are modeled here using he indicaor of he sage of he business cycle (Keiki Kijyun Hiduke) announced by he Economic Planning Agency. A dummy variable, B, is consruced o ake he value uniy when he announced business cycle is in a downurn, and zero oherwise. Ineracion variables of he form B σ and is four lags are hen also added o he baseline unemploymen equaion of each group wih significan dispersion variables. In addiion, in order o disinguish he effecs of he recession from he sage-of-business-cycle effecs, a recession dummy variable is also included. Table 3 summarises he resuls of esing for sageof-business-cycle effecs using σ. Significan sage-of-business-cycle effecs are presen for aggregae unemploymen and for he unemploymen of males aged 55-64, indicaing ha he impacs of secoral shocks on unemploymen increases in imes of recession for hese wo groups. In conras, he resuls for males aged 15-4 show no significan ineracion erms. A Wald es also acceps he null hypohesis ha he curren and four lags of he sage of business cycle variables are joinly insignifican for his group (p-value=0.6).

7 Resuls consisen wih hose in Table 3 are also obained when he sage-of-business-cycle effecs are invesigaed using he purged index s. Table 3. Sage-of-Business-Cycle Effecs. B σ U T=100 (Sample period: 1975Q1-1999Q4) 6. CONCLUSION U15 U55 T (0.8) (0.) (0.1) ** (0.75) (3.1) * (3.47) (0.6) R In conras o he previous sudies for he Unied Saes, earlier sudies for Japan do no suppor he secoral shifs hypohesis. However, Sakaa [001] provides evidence ha in Japan, he findings for aggregae unemploymen can be misleading because secoral shocks have affecs ha differ according o workers levels of indusry- and firmspecific human capial accumulaion. Following Sakaa [001], his paper has examined he unemploymen of males in wo age groups, 15-4 and 55-64, whose accumulaion of human capial should differ grealy. Alhough secoral shifs do no have any long-erm impacs on unemploymen, secoral shifs have posiive effecs on he unemploymen of old males in he shor-erm. In addiion, he aggregae unemploymen rae and he unemploymen rae of males aged are srongly influenced by he sage of he business cycle. These findings sugges ha policy judgmens based on an analysis of aggregae unemploymen may be blind o he acual impacs of secoral shifs on differen cohors. 7. REFERENCES Abraham, K.G., and Kaz, L.F., Cyclical unemploymen: secoral shifs or aggregae disurbances?, NBER Working Paper No. 1410, Abraham, K.G., and Kaz, L.F., Cyclical unemploymen: secoral shifs or aggregae disurbances?, Journal of Poliical Economy, 94, 507-5, Barro, R.J., Unanicipaed money growh and unemploymen in he Unied Saes, American Economic Review, 67, , Brunello, G., Mismach in Japan in Mismach and Labour Mobiliy (F. Padoa-Schioppa, Ed.), pp Cambridge Univ Press, London/New York, Davis, S.J., Flucuaions in he pace of labor reallocaion, Carnegie-Rocheser Conference Series on Public Policy, 7, , Lilien, D.M., Secoral shifs and cyclical unemploymen, Journal of Poliical Economy, 90, , 198. Mills, T.C., Pelloni, G. and Zervoyianni, A., Unemploymen flucuaions in he Unied Saes: furher ess of he secoral-shifs hypohesis, Review of Economics and Saisics, 77, , Sakaa, K. Secoral shifs and cyclical unemploymen in Japan, Journal of he Japanese and Inernaional Economies, forhcoming, 001. Wallis, K.F., Time series analysis of bounded economic variables, Journal of Time Series Analysis, 8, , 1987.

8