Faculteit der Economische Wetenschappen en Econometrie
ET
05348
1993
Serie research memoranda
057
The Influence of Labour market dynamics on Wage Formation
L Broersma F.A.C, den Butter
Research Memorandum 1993-57
November 1993
applied labour economics research team
vrije Universiteit
amsterdam
THE INFLUENCE OF LABOUR MARKET DYNAMICS ON WAGE FORMATION by Lourens BROERSMA and Frank A.G. DEN BUTTER Applied Labour Economics Research Team Department of Economics Vrije Universiteit De Boelelaan 1105 1081 HV Amsterdam
ABSTRACT This paper
surveys
empirical Phillips
curve
and wage
models for The Netherlands. It provides a simple model
with
which
an
appropriate
specification
curve
theoretical
for
Phillips
curve and wage curve models can be derived, which reckons with labour market tightness and fits into a flow model of the labour market. We find that in particular the short term unemployment-vacancy ratio and flows from unemployment to employment affect the wage formation process.
AUTHORS' NOTE Correspondence
to the first
author,
Free University,
Department
of
Eonomics,
De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands; phone: (+31) (0)205484633, fax: (+31) (0)20-6462645, e-mail:
[email protected]Ü.NL.
- 1 -
1. INTRODUCTION The
description
of
the wage formation
process is crucial to
models
of
labour market. In most traditional macroeconomic policy models, wage tion is explicitly modelled in a wage
equation,
which includes the
the
formainfluence
of some measure of unemployment on wages. The empirical relationship between wages
and unemployment
theoretical foundation.
discovered by Phillips (1958)
nowadays
has
a
Phelps (1968) has shown how the Phillips curve
firm effect
can be derived from the behaviour of the firm, whereas in a newer tradition trade union behaviour has been shown to imply the so called wage curve effect (Oswald, neither
1982; of
equation
Blanchflower
these
and
theoretical
prescribes
that
Oswald
derivations
(some
1989; of
the
transformation
Graafland
1992).
specification of)
the
of
However, the
wage
unemployment
rate
should act as a measure for labour market tightness. The theory allows for a much wider have
some
set
of
measure
indicators. Traditional models of of
unemployment
as
the labour market
explanatory
variable
in
the
just wage
equation, because it is the only indicator of labour market tightness that is endogenous in the model. • However, a new approach to modelling the labour market considers labour market dynamics by endogenising labour market flows (Blanchard and Diamond 1992). Most theoretical models concentrate
on
equilibrium
of
this
flow
unemployment
approach and
to
the labour
market
have
implicit
therefore
an
description of wage formation. Yet, for an empirical model of labour market flows, an explicit modelling of wage formation by means of a wage equation is more appropriate when the labour market is not considered to be in equilibrium in each period. This is the case for the Netherlands in the last decades. This equation
paper using
investigates alternative
empirical measures
specifications
of
labour
for
market
the
Dutch
tightness,
which
wage are
endogenous in a flow model of the labour market. The next section shortly reconsiders curve.
the
Section
theoretical 3
surveys
foundation the
of
empirical
the wage
Phillips
curve
equation
for
and the
the
wage
Netherlands,
whereas we present our own estimates of Phillips curve equations and wage curve
equations
for
the
Netherlands
in
section
4.
The
latter
section
also
compares the size of the effect of labour market tightness on wage formation according to alternative indicators. Finally, section 5 concludes.
- 2 -
2. LABOUR MARKET DYNAMICS AND WAGE FORMATION This section gives the theoretical background of equation,
which has
variable.
The
an indicator
Phillips
curve
of
labour
specification
the specification
market
is
based
tightness on
the
of
as
a wage
explanatory
theoretical
model
of Phelps (1968), where the firm is assumed to set the wages. The specification of the wage curve is described as outcome of a wage bargaining process. Cf. Graafland (1992). 2.1. The Phillips curve We slightly in
order
market
adapt to
Phelps' (1968)
obtain
dynamics
an
theoretical
appropriate
foundation
specification
of
for
the
Phillips
representing
and expectational behaviour. Starting point
is a
tive firm, which has a number of unfilled jobs. By offering
curve labour
representa-
a higher wage,
compared to wages paid elsewhere, the firm will try to fill these vacancies. It is assumed that the supply of total
unemployment.
employees
of
the
applicants for the vacancies will rise with
When unemployment firm
drops;
it
is high,
rises
with
the
the
quit
macro
rate
vacancy
of
current
rate.
The
desired wage differential of the firm is then (Wi-w)/w where flur 0 , /i OT -.>0, tUj is the wage rate the t'-th firm wishes to
pay, w is the average wage rate of all firms, ur is the macro unemployment rate, vr is the macro vacancy rate, vr,- is the number of unfilled jobs of the t-th firm. Phelps assumes that wage
differential
will
all firms behave much alike and hence the be
a
function
of
the
unemployment
rate
average and
the
aggregate vacancy rate. (w -w)/w where f2ur ° a^d
w
(2) ^
tae
average wage rate the firms are willing to
pay.
- 3 -
Next it is postulated that w = X[(w -w)/w], where
07r and u and ir are the utility
functions of the employee and employer, respectively, which depend on wages. The
workers'
utility the
utility
is
is negatively
threat
point
a
positive
related
of
the
to
function
the
worker
of
the
wages. The and
wages,
while
utility levels
employer,
u
respectively,
employers' and
and
n
are
represent
their utility obtained during a breakdown in the hargaining process, a
is a
parameter representing the relative bargaining strength of the worker. Labour market dynamics enter into this model through the assumption that bargaining vacancy
power
rate,
is
and
related
to
external
internal
pressures,
labour
like
market
(short-term)
pressures,
like
unemployment.
the
These
two can be joined in the ÜV ratio. Also the two approximations of flows on the labour
market
introduced
above,
are
capable
of
representing
such pres-
sures. The
crucial
difference
of
the
wage
curve resulting
from
the
bargaining
theory in comparison with the Phillips curve is that here the labour market variables
affect
the
wage
level,
not
the
wage
growth.
There
is
a
relation between the wage level and the variables representing labour
stable market
dynamics. The location of the wage curve will depend on the other exogenous variables
in the wage negotiation process. We assume those variables to
the
same
as
i.e.,
consumer
the prices
exogenous
variables
(p c ), producer
of
prices
the (j>y),
Phillips labour
curve
be
specification,
productivity
(h)
and
taxes and social premiums (tp). Hence, in log-linear form our wage curve is
logwt = (A +P1logpct + Pj-ogpyt + /33l°g>»t + P£°&Pt + Ps*t + £m where Xt
represents the variables of labour market
term) unemployment,
vacancies, UV ratio,
dynamics, be it
or the flows from
(9) (short-
employment
to
unemployment and vice versa and e l t is a stationary error process. The latter assumption implies that sense
that
they
contain
that
case
equation
(9)
when all variables a
unit
turns
root, into
in
these an
(9) series
are nonstationary, may
error-correction
be
cointegrated.
specification.
assuming a flexibly specified autoregressive distributed lag model, we find
- 7 -
in
the In
Again
t logPct logPyt log^t logtp t wrt
sturt vrt uvt stuvt fuet Teut
mmt
-3.123, -3.042 , -2.710 -3.820, -2.360, -1.462, -1.763, -1.978, -2.317, -2.511, -2.693, -2.720 -3.033
(k=4 (k=2 (k=l (k=3 (k=3 (k=9 (k=9 (k=7 (k=5 (k=5 (k=5 (k=4 (k=4
Phillips (1987)
Phülips-Perron
0.272, 0.217, 0.181, 0.056, -0.297, 0.384, 0.092, -2.622]
-1.991, -2.215, -0.899, 0.368,
-3.989
(1988)
-2.607, -1.801, -2.112, -2.370, -3.174, -1.845, -3.538, -6.955 , -4.009 , -4.375 , -8.388 , -11.34, -12.66
unit root hypothesis not rejected at 5 percent significance. ADF stands for augmented Dickey-Fuller test. Cf. Dickey and Fuller (1981)
- 19 -
TABLE V. Estimation and test results of wage curve equations for the Netherlands, with different measures of labour market tension. Dependent variable: Alogwt. constant
-.318 (-2.037)
-.277 (-1.679)
-.198 (-1.246)
-.176 (-1.005)
-.109 (-0.705)
-.189 (-1.184)
Alogw^
-.357 (-4.008)
-.382 (-3.803)
-.378 (-3.730)
-.375 (-3.662)
-.397 (-4.088)
-.401 (-4.201)
^log^ct-i
.986 (4.677)
.633 (3.584)
.673 (3.784)
.547 (3.050)
.627 (3.674)
.586 (3.480)
^logpyt.!
-.239 (-2.336)
.154 (3.326)
.176 (3.703)
.907 (2.865)
.942 (3.031)
Aloght_3 Alogtpt_3 Aurt_2
1.05 (1.764) .202 (5.169)
.162 (3.262)
.145 (2.997)
.103 (1.856)
-1.44 (-4.361)
Ammt
5.45 (1.451)
Afuet-2 * \ogrwht_!
-.071 (-2.079)
logPyt-l w,_i
-.088 (-2.521)
-.073 (-2.122)
-.065 (-1.700)
-.055 (-1.640)
-.070 (-2.067)
-.027 (-3.608)
-.029 (-3.873)
-.026 (-3.257)
-.030 (-4.252)
-.028 (-3.937)
-.210 (-5.296)
s«uv t _i(*1000)
R2 S.E. T
-.339 (-1.827)
-.493 (-2.794)
UÜ^IOOO)
-.480 (-2.868) -.623 (-2.359)
-.679 -2.465) .750 .0094 85
.698 .0103 85
.692 .0104 85
.717 .0101 84
.723 .0099 85
.732 .0097 85
*U*(1,72) F^e(5,68)
.065 .486
.883 .454
.796 .426
.317 1.03
1.53 .677
1.29 .683
X^(2)
.258 .133 1.31
.349 .117 .105
.542 .143 .126
.032 .032 .322
1.56 .313 .074
1.32 .743 .170
FARCH^M) FARCH&W)
logrwht = logtüt - logp c t - loght.
- 20 -
TABLE VI. Eff ects of a permanent increase in unemployment on wage growth f or a number of quarters (q), in percentage points. Effect after
lq.
2q.
3q.
4q.
5q.
6q.
7q.
8q.
12 q.
16 q.
20 q.
Phillips curve: column (2) (5) (6)
.0 .0 .0
.0 -.12 -.23 -.34 -.46 -.33 -.21 -.03 -.05 -.07 -.11 -.09 -.06 -.03 -.10 -.11 -.12 -.12 -.03 -.03 -.02
.03 .02 .0 .0 -.02 - .02
.02 .0 -.02
.0 .0 .0
-.08 -.72 -.58 -.68 -.67 -.06 -.28 -.02 -.26 .02 -.12 -.04 .18 -.09 -.02 -.26 .04 -.10 -.04 .17 -.13
-.24 -.01 -.04
-.25 -.01 -.05
wage curve: column (1) (5) (6)
-.24 -.01 -.04
TABLE VIL Eff ects of a permanent decrease in vacancies on wage growth f or a number of quarters (q), in percentage points. Effect after
lq.
2q.
3q.
4q.
5q.
6q.
7q.
8q.
12q.
16q.
20q.
Phillips curve: column (2) (5) (6)
.0 .0 .0
.0 .0 .0 .0 .0 .0 .0 .0 -.07 -.11 -.23 -1.1 -1.3 -1.0 -.08 -.32 .0 .04 .04 .04 .04 .0 .0 .0
.0 .0 .0
.0 .0 .0 .0 .0 .0 .0 -.05 .04 -.20 -.77 -.79 -.64 -.43 -.09 -.01 -.30 -1.2 -1.2 -1.0 -.70
.0 -.19 .0
.0 -.34 .0
wage curve: column (1) (5) (6)
- 21 -
.0 -.18 -.28
.0 -.12 -.19
.0 -.24 -.38
REFERENCES Berg, P.J.C.M. van den, Gelanff, G.M.M and Okker, V.R. (1988), "The FreiaKompas Model for The Netherlands: A quarterly Macroeconomic Model for the Short and Medium Term", Economie Modelling, 5, 170-236. Blanchard, O.J. and Diamond, P. (1992), "The Flow Approach to Labor Markets", American Economie Review, 82, 354-359. Blanchflower, D.G. and Oswald, A.J. (1989), "The Wage Curve", NBER Working Paper 3181. Boswijk, H.P. (1992), Cointegration, Identification Thesis, Thesis Publishers, Amsterdam. Butter, F.A.G. den and Wijngaert, R.F. van Labour Conflicts in the Netherlands", Economy, 8, 527-541.
and
Exogeneity,
Ph.D.
de (1992), "Wage Inflation and European Journal of Politica!
Dickey, D.A. and Fuller, W.A. (1981), "Likelihood Ratio Statistics for regressive Time Series With a Unit Root", Econometrica, 49, 1057-1072.
Auto-
Engle, R.F. (1982), "Autoregressive Conditional Heteroskedasticity with mates of the Variance of the UK Inflation", Econometrica, 50, 987-1007.
Esti-
Fase, M.M.G., Kramer, P. and Boeschoten, W.C. (1990), "Morkmon E, het DNB Kwartaal Model voor Nederland", Monetaire Monografieën 11, DNB, Amsterdam Graafland, J.J. (1990a), Thesis, Rotterdam.
Persistent
Unemployment,
Wages and
Hysteresis,
Ph.D.
Graafland, J.J. (1990b), "Long-Term Unemployment in The Netherlands", Central Planning Bureau, Research Memorandum, 66, The Hague. Graafland, J.J. (1991), "Effecten van marginale belasting loonvorming", Maandschrift Economie, 55, 442-455. Graafland, J.J. (1992), 140, 501-514.
"From Phillips
Curve
to
Wage
en
Curve",
premiedruk De
op
Economist,
Graafland, J.J. and Huizinga, F.H. (1990), "Modelling a Wage Equation for The Netherlands", Paper presented at the Ecozoekdag, Groningen. Graafland, J.J. and Verbruggen (1993), "Macro Against Sectoral Wage Equations for The Netherlands", Central Planning Bureau, Research Memorandum, 103. Groene, J.H., Siebrand, J.C. and Windt, N. van der (1984), "RASMUS-I, An Annual Model of the Dutch Economy", Discussion Paper Series 8407/G, Institute for Economie Research, Erasmus University, Rotterdam. Godfrey, L.G. (1979), trika, 66, 67-72.
"Testing
the Adequacy
of
Time
Series Models",
Biome-
Hartog, J. (1980), Tussen vraag en aanbod, Stenfert-Kroese, Leiden. Hassink, JLJ. and Broersma, L. (1993), "Labour Demand and Job-To-Job Movement: Macro Consequences as a Result from Microeconomic Behaviour", Discussion Paper Tinbergen Institute, TI 93-94, Amsterdam. Houweling, A., Huizinga, F.H., Okker, V.R., Stratum, R.M.G. van and Verbrug-
- 22 -
gen (1992), FKSEC, A Macro - econometrie Stenfert-Kroese, Leiden
Model for
the Netherlands,
Jarque, CM. and Bera, A.K. (1980), "Efficiënt Tests for Normality, dasticity and Serial Independence of Regression Residuals", Letters, 6, 255-259.
CBP,
HomoskeEconomics
Keus, P, Liefferink, P and Opstal, R. van (1991), "Aparte vergelijkingen voor contractloon en incidenteel loon in FKSEC", Interne Notitie, Centraal Planbureau, Den Haag. Knoester, A. (1983), "Stagflation and the Inverted International Evidence", De Economist, 131, 229-239.
Haavelmo
Effect:
Some
Knoester, A. and Windt, N. van der (1987), "Real Wages and Taxation on Ten OECD Countries", Oxford Bulletin of Economics and Statistics, 49, 151-169 Kuipers, S.K., Jongbloed, B.W.A., Kuper, G.H. and Sterken, E. (1990), "CESAM: The CCSO Annual Model of the Dutch Economy", Economie Modelling, 7, 202-250. Layard, R. and Nickell, 53, S121-S169.
S.J. (1986),
"ünemployment
in
Britain",
Lever, M.H.C. (1992), "Insider-Outsider Effects in Union Wage Research Memorandum 92-24, Limburg University, Maastricht.
Economica,
Determination",
Lindbeck, A. and Snower, D.J. (1986), "Wage Setting, ünemployment and Insider Outsider Relations", Journal of Political Economy, 76, 235-239. Lipsey, R.G. (1960), "The Relation Between ünemployment and the Change of Money Wage Rates in the United Kingdom, 1862-1957: A Further Analysis", Economica, 27, 1-31. Mulder, CM. (1990), "Langdurige werkloosheid tistische Berichten, 75, 132-136.
in Nederland",
Economisch
Sta-
Oswald, A.J., (1982), "The Microeconomic Theory of the Trade Union", Economie Journal, 92, 567-595. Ours, J.C, van and Ridder, G. (1991), "Cyclical Variation in Vacancy Duration and Vacancy Flows: An Empirical Analysis", European Economie Review, 35, 1143-1155. Phelps, E.S. (1968), "Money-Wage Dynamics Journal Political Economy, 76, 678-711.
and
Labor-Market
Equilibrium",
Phillips, A.W. (1958), "The Relation Between ünemployment and the Change of Money Wage Rates in the United Kingdom, 1862-1957", Economica, 25, 283-299. Phillips, P.CB. (1987), "Time Series Regressions With a Unit Root", trica, 55, 277-301.
Econome-
Phillips, P.CB. and Perron, P. (1988), "Testing Series Regressions", Biometrika, 75, 335-346.
in
for
a
Unit
Root
Time
Sargan, JJ). (1964), "Wages and Prices in the United Kingdom: A Study in Econometrie Methodology", reprinted in D. Hendry and K. Wallis (eds), Econometrics and Quantitative Economics, Oxford, pp. 275-314.
- 23 -
Verbruggen, J.P. (1985), "Een vier-sectorenmodel voor de Nederlandse mie", Discussion Paper 8502, Ministry of Economie Affairs, The Hague.
econo-
J*
•
- 24 -
>
DATA APPENDIX: DEFINITIONS AND SOURCES The quarterly data are mainly taken from the OECD, Main Economie Indicators (MEI). Yearly data, which were later interpolated were mainly obtained from the Dutch Central Planning Bureau (CPB), Lange Reeksen (LR). Interpolation was done by means of a third order polynomial function, cf. Hassink and Broersma (1993). The Dutch Central Bureau of Statistics is abbreviated to CBS. w:
index of hourly rates in manufacturing source: OECD, MEI.
pc:
price index of total consumption source: OECD, MEI.
py:
price index of total output source: OECD, MEI.
h:
measure of labour productivity of the firm sector, based on a weighted average of actual labour productivity, with weights 0.3 in period t, 0.5 in period 0.5 in period t-\ and 0.2 in period t-2. Interpolated. source: CPB, LR.
tp:
social premiums of employers and employees taken together, as percentage of gross wage costs of the firm sector. Interpolated. source: CPB, LR.
ur:
unemployment as percentage of the total labour force ( = employment and unemployment) source: OECD, MEI.
stur:
unemployment rate with a duration shorter than one year, defined as u r - u r 4 , where ur4 is the unemployment rate with a duration longer than one year. source: ur4: Ministry of Social Affairs and Employment.
vr:
vacancies as percentage of total labour force. The data were taken from the OECD, MEI. To correct for the decline of the share of vacancies notified to the Public Employment Offices, we used corrected vacancy data for the eighties. These data are corrected using Information from the CBS vacancy surveys. The CBS vacancy surveys are from October 1980, 1981, 1982, 1983, September 1984, January 1986, 1987, 1988. By interpolating, we calculated the average share of notified vacancies for the years 1980-1987. We assumed that for the period 1961-1979, this share was equal to the share of 1980.
Fue:
we used yearly data on vacancy flows to the Public Employment Office of the period 1971-1978 from Hartog (1980) to calculate average vacancy durations (duration = stock/flow). For 1980-1987, we used CBS vacancy survey data and applying the method described in van Ours and Ridder (1991). The average vacancy duration of 1979 by interpolating the yearly durations of 1978 and 1980. By interpolating the yearly data, we calculated quarterly duration data. Finally, we calculated vacancy flows as the quotiënt of vacancy stocks and vacancy durations. From 1987 to 1991, actual data on vacancy flows were obtained from
- 25 -
the CBS, Sociaal Economische
Maandstatistiek.
mm:
defined as \empmfgt-empmfgt_1\lemp, where emp is total employment in thousand persons of wage earners and salaried employees. Yearly data from 1971-1987, are obtained from the OECD, Labour Force Statistics. Quarterly data from the CBS for 1984-1987 were used to determine quarterly fluctuations in employment. This quarterly pattem was then imposed on the yearly data. Data for 1988-1991 were directly observed from CBS, Sociaal Economische Maandstatistiek. Empmfg is the employment in the manufacturing sector, taken from the CBS, National Account Statistics and interpolated. Cf. Hassink and Broersma (1993).
Feu:
Flow of persons out of employment into unemployment, due to firing. Or the inflow of persons in unemployment insuran.ce (WW). Interpolated source: Sociale Verzekeringsraad, Kroniek van de sociale verzekeringen, 1992.
- 26 -