MONI Juha Tarkka - Matti Vir n POPULATI
POPULATION SHOCKS AND EMPLOYMENT: TESTING THE LABOR MARKET EQUILIBRIUM HYPOTHESIS* Juha Tarkka and Matti Viren Forthcoming in Economics Letters * We are indebted to Jouni Kokko for research assistance. Financial support from the Yrjö Jahnsson Foundation is gratefully acknowledged.
ABSTRACT This note tests the labor market equilibrium hypothesis by examining the speed of adjustment of employment with respect to population shocks. The analyses in the time domain with annual cross-country data from 15 OECD countries - and the accompanying preliminary analyses in the frequency domain - are clearly at variance with the equilibrium hypothesis.
1. INTRODUCTION The question of whether the labor market equilibrium hypothesis or the labor market disequilibrium hypothesis properly explains movements in employment in western economies is still largely unsettled. One obvious reason for this state of affairs is the fact that many common operationalizations of these hypotheses are non-nested and thus very powerful tests for them cannot be arranged. Thus, one is typically forced to examine the time series properties of real wages using for instance ARIMA models or partial adjustment based wage equations (see e.g. Altonji and Ashenfelter (1980)). Even though focusing on the real wage variable certainty tells us something about the determination of employment, it may be asked whether it is possible to construct a simple test for the labor market equilibrium hypothesis which is not conditional on the way the real wage variable is modelled. We believe such a test can be arranged by examining how quickly population shocks affect employment. The idea is simply that if the labor market eguilibrium hypothesis is true and if there are changes in the working age population, these changes should instantaneously affect employment (on account of adjusting wages). If, in turn, the labor market disequilibrium hypothesis is true, the opposite holds and thus population shocks are transmitted to employment only gradually over time owing to wage rigidity.1 The advantage of focusing on population shocks is that this is one of the few variables which can properly be thought to be exogeneous in the short run. 2
2 2. TEST RESULTS Our test procedure boi1s down to estimating equation (1): where E denotes (civi1ian) emp1oyment, POP working age (civi1ian) popu1ation and u the error term. (1) is estimated in 1inear and 10g 1inear form (and it turns out that there is practica11y no difference in resu1ts; thus, on1y the resu1ts with the 1inear form are disp1ayed here). (1) can be thought to correspond to a reduced form for 1abor demand and supp1y equations cum a partia1 adjustment process with respect to emp10yment equi1ibrium. The supp1y equation is of the form E = aq + a1w + a2pop + v and the demand equation E = ao + e1w + Z, where w denotes the rea1 wage and v and z the error terms, which a1so take into account other possib1e determinants of emp1oyment. 3 Equation (1) is estimated from annua1 time series data from 15 OECD countries, the samp1e period being 1956-1984 (one observation is, however, 10st because of 1agging E).4 Estimation resu1ts are reported in Tab1e 1. The resu1ts are reported both for individua1 countries and geographica1 entities (inc1uding North America, Europe, the Nordic countries and the European Community (EEC». Severa1 features of these resu1ts merit attention. 5 First, the equation fits the data rather we11 and the parameters are typica11y precise1y estimated and of correct sign and magnitude. There are some prob1ems with autocorre1ation and heteroscedasticity, even though the 10ss of efficiency does not seem to be of a crucia1 nature. The prob1em of parameter stabi1ity is, however, more acute with some samp1e countries. Unfortunate1y, we are not ab1e to go into the detai1s of this prob1em here.5 Second, the hypothesis of instantaneous adjustment of emp10yment with respect to changes in popu1ation is c1ear1y rejected with a11
3 countries. Rather, the coefficient estimates of the lagged dependent variable suggest that the average adjustment period corresponds roughly to a typical cycle period. Third, there seems to be a difference between North America, on the one hand, and Europe (as well as Japan), on the other hand, in the sense that the speed of adjustment is much slower in the European countries (and in Japan). This finding is in agreement with the common wisdom that Europe - in contrast to North America - is characterized by real wage rigidity (cf. e.g. Newe11 and Symons (1985)). The results for all European countries are qua1itative1y very similar; there is, however, a slight difference between the Nordic countries (which have more wage rigidity) and the European Community. In addition to the analyses in the time domain we also carried out some pre1iminary analyses in the frequency domain (the number of observations does not rea11y allow for a proper spectra1 ana1ysis). In accordance with the results in the time domain it turns out that the coherency between (log differences of) employment and popu1ation increases with lower frequencies (and 10nger cyc1es). In fact, the maxima1 coherency is typically - and particularly in the case of Canada and the United States - at the frequency zero, or at 1east at a frequency which corresponds to a period which is equal or 10nger than a typica1 cyc1e period. These facts are i11ustrated in figure 1.6
4 TABLE 1. OLS Estimation Results of Equation (1) Country Constant E_l POP R 2 LMl CHOW Austria -.675.978.159.892 1.524 2.286 (1.60) (14.34) (2.38) (1.68) (19.21) (2.30) Belgium.447.983 -.062.919.611.998 {2.21l (13.58) (1.63) (2.21) (14.16 ) (1.93) Canada 1.053.431.438.995 2.073.042 (2.87) (2.55) (3.42) (2.0l) (1.42) (1.99) Denmark.123.471.347.926 2.227.511 (O.99) (2.75) (2.65) (O.96) (2.18) (2.24) Finland.113.701.174.903 1.666 2.642 (O.84) (4.70) (1.95) (O.88) (6.69) (2.76) France.771.957.005.979 2.541 6.499 (0.95) (10.26) (0.14) (O.94) (9.03) (O.ll ) F.R.G. 10.588.754 -.109.769 2.710.193 (3.19) (7.60) (2.77) (4.29) (8.90) (4.29) Ireland.315.583.074.838 3.504 1.089 (3.17) (4.90) (3.17) (3.27) (5.02) (2.84)!taly.690.887.044.820.432 4.330 (0.39) (9.70) (1.92) (0.50) (11.68) (2.56) Japan 2.555.827.094.994.305 2.326 (2.01 ) (6.71 ) (1.31 ) (2.44) (7.90) (1.52 ) Netherl ands.626.707.087.952.861 1.588 (2.10) (5.14) (2.00) (2.24) (4.05) (1.35) Norway -.396.877.252.984.952 4.392 (2.52) D2.52) (2.40) (2.68) (11.46) (2.27) Sweden -.341.894.150.960 1.396 5.072 (O.88) (lo.16) (1.19) (O.78 ) (14.24) (1.29) U.K. 6.978.835 -.084.675 2.255 1. 739 (1.94) (7.20) (1.13) (1.98) (7.65) (1.09) U.S.A 8.728.515.386.994 1.158.703 (3.40) (3.57) (3.58) (4.62) (4.77) (4.93) No rth America 10.342.503.394.995 1.115.613 (3.47) (3.53) (3.70) (4.59) (4.26) (4.62) Europe 2.554.827.093.994.305 2.326 (2.01) (6.71 ) (1.31 ) (2.44) (7.90) (1.52) Nordi c -.533.881.128.982.114 1. 713 countries (1.00) (8.44) (1.22) (1.15) (8.96) (1.41) EEC 10.994.790.006.814 2.103.518 (1.99) (5.98) (0.23) (2.46) (7.53) (0.28) ------ The numbers in parentheses immediately below the coefficient estimates are standard t-ratios; below them are White's heteroscedasticity adjusted t-ratios. LM1 denotes Godfrey's LM autocorrelation statistic and CHOW the Chow stability test statistic for the period 1971. The distribution of LM1 under the null hypothesis of no autocorrelation is N(O,l)' and the distribution of CHOW is F{3,22). The corresponding critical values at the 5 per cent level of significance are 1.645 and 3.05, respectively. North America comprises Canada and the United States, and Europe the European Community (EEC), the Nordic Countries and Austria. The EEC, in turn, comprises Belgium, France, the F.R.G., Ireland, Italy, the Netherlands and the U.K., and the Nordic countries Oenmark, Finland, Norway and Sweden.
Figure 1. 5 UNWEIGHTED COUNTRY AVERAGES OF SQUARED COHERENCIES 1.0 0.8 0.6 0.4 - - - - 1- - 0.2 e-"" - - -... ~ 1- - 0.0 0.1 0.2 0.3 0.4 0.5
6 FOOTNOTES 1 Changes in education, social security and nonmarket production of goods may of course considerably distort the relationship between working age population and employment in the long runo 2 Needless to say, (net) migration may invalidate this assumption with some countries, as, in fact, turns out to be the case later in this study. One way of eliminating the effects of migration is to examine geographical entities which are characterized by more or less perfect mobility of labor, such as the EEC and the Nordic countries. In this connection we may point out that, in fact, very few studies make use of the population variable (and obviously even fewer try to make use of exogeneity of this variable). See e.g. Bean, Layard and Nickell (1986). 3 Thus, for instance, this study does not take into account taxes and aggregate demand variables. In the case of taxes we could not solve the data problems while in the case of demand variables we could not find an acceptable and simple solution which would not create obvious problems with exogeneity (for instance in terms of aggregate supply effects and possible fiscal and monetary policy reaction functions). Nevertheless, we did experiment with the time trend, public consumption and exports as additional explanatory variables in (1) without, however, any noticeable change in results (a set of results is available upon request from the authors). Notice that the partial adjustment mechanism is specified here so that the real wage adjustment does not affect the adjustment of employment. Obviously, also this assumption could be generalized (see e.g. Chow (1983), p. 246). 4 The data are derived mainly from the OECD Labor Force Statistics. A complete description of the data and a data printout is available upon request from the authors. 5 In addition, there are also some anomalous results in terms of the coefficient estimate of POP. Thus, in the cases of Belgium, the Federal Republic of Germany and the United Kingdom this coefficient is negative. As far as Germany is concerned, we think that the problem is due to very large net immigration (in her case the ratio of net immigration to working age population is the largest among the sample countries). In the case of the United Kingdom there is clear break in the regression relationship for 1979. Thus, if (1) is estimated for the pre-thatcher period 1957-1978 only, the coefficient of POP is clearly positive. 6 Computations make use of the Parzen window, the lag length being 10.
7 REFERENCES ALTONJI, J. and o. ASHENFELTER, Wage Movementsand the Labor Market Equilibrium Hypothesis, Economica, August 1980, 47, 217-245. BEAN, C.R., LAYARD, P.R.G. and S.J. NICKELL, The Rise in Unemployment: A Multi-country Study, Economica, Supplement, 1986, 53, S1-S22. CHOW, G.C., Econometrics, McGraw-Hill, Tokyo, 1983. NEWELL, A. and J.S.V. SYMONS, Wages and Employment in the O.E.C.D. Countries, University College London, Department of Political Economy, Discussion Paper 85-24, London 1985.
2 Viren, Matti The Relationship Between Interest Rates and Inflation During the Prewar Period: Some Empirical Evidence Heinäkuu 1985, 9 s. TU 11/85 Viren, Matti Some Time Series Evidence on the Behavior of Interest Rates and Inflation in 6 OECD Countries Heinäkuu 1985, 12 s. TU 12/85 Willman, Alpo The Collapse of the Fixed Exchange Rate Regime, Wage Formation and Imperfect Capital Mobility Syyskuu 1985, 23 s. TU 13/85 Tarkka, Juha Suomalaiset pankkiluottomarkkinat ja uusklassinen rahateoria Joulukuu 1985, 156 s. TU 14/85 Miikkulainen, Pirkko Palvelujen ulkomaankauppa Suomessa vuosina 1970-1983 Joulukuu 1985, 80 s. TU 15/85 Koskenkylä, Heikki ja Kohi, Pertti On the Effects and Problems of Investment Incentive Policies from the Short- and Long-Term Viewpoints: Experiences from Finland Joulukuu 1985, 41 s. TU 16/85 1986 Aaltonen, Ari Empiirinen analyysi budjettialijäämien käyttäytymisestä kahdessatoista OECD-maassa vuosina 1953-1982 Helmikuu 1986,' 93 s. TU 1/86 Koskela, Erkki and Viren, Matti Social Security and Household Saving in an International Cross-Section: Some Further Evidence Helmikuu 1986, 13 s. TU 2/86 Stenius, Marianne and Viren, Matti Some Further Results on Rosen and Quandtls Labor Market Model: Queries and Disagreements: A Reply Helmikuu 1986, 7 s. TU 3/86
3 Koskela, Erkki and Viren, Matti Endogenous Poliey, Structural Shift, and Demand for Money Helmikuu 1986, 13 s. TU 4/86 Pikkarainen, Pentti ja Viren, Matti Mitä erilaiset kokonaistuotannon trendipoikkeamat kuvaavat? Maaliskuu 1986, 18 s. TU 5/86 Koskenkylä, Heikki Raha- ja finanssipolitiikan vaikutuksesta yritysten investointeihin Maaliskuu 1986, 25 s. TU 6/86 Willman, Alpo Imperfect Capital Mobility, the Supply of Money and Balance-of-Payments Crises Toukokuu 1986, 1~ s. TU 7/86 Pikkarainen, Pentti ja Viren, Matti New Evidence on Long Swings Toukokuu 1986, 10 s. TU 8/86 Elomaa, Arto Suomen työvoimamarkkinoiden epätasapainomalli Toukokuu 1986, 69 s. TU 9/86 Tarkka, Juha Rahoitus- ja valuuttamarkkinoiden säännöstelyn purkamisesta käydyssä keskustelussa esitettyjä näkökohtia Kesäkuu 1986, 54 s. TU 10/86 Lempinen, Urho ja Lilja, Reija Commercial Banks and Money Creation in Finland Heinäkuu 1986, 27 s. TU 11/86 Pikkarainen, Pentti ja Viren, Matti Causality Between Money, Output and Prices: Some Cross-Country Evidence from the Period 1876-1984 Heinäkuu 1986, 11 s. TU 12/86 Viren, Matti Some Time Series Evidence on the Behavior of Interest Rates and Inflation in 6 OECD Countries Elokuu 1986, 21 s. TU 13/86
4 Viren, Matti Estimating the Output Effects of Energy Price and Real Rate Shocks: A Cross-Country Study Elokuu 1986, 17 s. TU 14/86 Interest Koskela, Erkki ja Viren, Matti Modelling the Behaviour of Money Stock Under Various Monetary Targeting Procedures Elokuu 1986, 19 s., TU 15/86 Vi ren, Matti Interpreting Recent Total Factor Productivity Growth Estimates for Eastern Europe Elokuu 1986, 31 s. TU 16/86 Taimio, Hilkka Valtion menojen ja valtion julkisten palvelujen kasvu Suomessa 1900-luvun alkupuoliskolla Elokuu 1986, 115 s. TU 17/86 Willman, Alpo Balance-of-Payments Crises with Uncertain Central Bank Reactions Elokuu 1986, 10 s. TU 18/86 Tarkka, Juha ja Viren, Matti Population Shocks and Employment: Testing the Labor Market Equilibrium Hypothesis Lokakuu 1986, 7 s. TU 19/86