Journal of Policy Modeling

26 (2004) 95112

Public capital and long-term labour market

performance in Belgium
Gerdie Everaert , Freddy Heylen
SHERPPA, Ghent University, Hoveniersberg 24, B-9000 Gent, Belgium
Received 1 June 2002; received in revised form 1 July 2003; accepted 1 November 2003

Abstract
This paper investigates the long-term output and labour market effects of public capital in Belgium within a broad model explaining private sector output and costs, private
employment and capital formation, wage bargaining, price setting and aggregate demand.
Model simulations show strong positive effects of public capital on private output and capital
formation. Public capital and private employment, however, are found to be substitutes. A
clear negative effect of public capital on employment remains, also after taking into account
the effects of public investment on aggregate demand, productivity and wage formation.
As to policy, this paper supports the case for an increase in public investment spending,
complemented by structural labour market reform.
2004 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.
JEL classification: E62; J23; O40
Keywords: Public capital; Public investment; Employment; Economic growth

1. Introduction
Following Aschauer (1989a), a huge literature has studied the macroeconomic
effects of the decline in public investment in most OECD countries since the
1970s. Most of this literature has concentrated on the consequences for output and
productivity. More than 10 years later, the results are still controversial. Aschauer

Corresponding author. Tel.: +32-9-264-7878; fax: +32-9-264-8996.

E-mail address: gerdie.everaert@rug.ac.be (G. Everaert).

0161-8938/$ see front matter 2004 Society for Policy Modeling.

doi:10.1016/j.jpolmod.2003.11.002

96

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

(1989a, 1989b) found highly positive and significant effects of infrastructure on

productivity growth in the US and the G7. In more recent work, however, many
authors have re-examined these results, often within another theoretical framework
or using another methodology. Some confirm Aschauers conclusions, others reject
them (see Sturm, Kuper, & Haan, 1998, for a survey).
This paper studies the effects of public capital formation on the labour market.
We focus on Belgium, which like many other European countries is characterised
by relatively rigid labour markets and high unemployment. For two reasons this
research may be important. First, extensive studies of the labour market effects of
public investment are still very scarce. Second, despite the previous, it has often
been recommended that in the fight against European unemployment, investment
in public infrastructure should be raised (see e.g., the European Commissions
White Paper of 1993 or Drze & Malinvaud, 1994). Our main goal in this paper is to evaluate the relevance of this recommendation. Focusing on Belgium, we
investigate whether or not public capital formation improves labour market performance in general, and employment in particular. To answer this question, we have
developed and estimated a dynamic structural model for the Belgian economy explaining private sector output and costs, private employment and unemployment,
private capital formation, wage bargaining, price setting and aggregate demand.
The model allows for three channels of influence of public capital on employment:
(i) direct complementary or substitution effects for given output, (ii) indirect effects
on real wages, due to changes in labour productivity and/or the unemployment rate,
(iii) indirect effects caused by changes in aggregate demand and the output level.
Due to space limitations, we will concentrate in this paper on the main characteristics of the model and on the most important estimated long-run equilibrium
relations and parameters. A complete and detailed description can be found in
an underlying working paper (see Everaert & Heylen, 2000). Once estimated, we
use the model to simulate private sector performance under an alternative public
investment policy. More specifically, we analyse the economys evolution if a very
strong decline in public investment during the period 19821989 had not occurred.
It should be emphasised that the perspective of this paper is broader than what
has usually been done in the literature. First, as we have mentioned before, existing
research has paid relatively little attention to the influence of public capital on
employment. Second, relevant findings on the employment effects of public capital
generally concern only the direct complementary or substitution effects. Estimating
private sector cost functions and/or factor share or input demand equations, most
studies find that (i) public capital reduces private sector costs, (ii) public capital
acts as a substitute for intermediate inputs and often also labour, (iii) public and
private capital are complements (see again Sturm et al., 1998). In this paper we
define and simulate all relevant channels of influence.
The remainder of the paper is as follows. Section 2 discusses our model.
Section 3 presents the estimates of the model for Belgium using annual data for
19651996. To detect the effects of public capital on economic growth and the
labour market, Section 4 reports the results of our simulation experiment. We find

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

97

that public investment stimulates economic growth, confirming Aschauers public

capital hypothesis. Surprisingly, however, public investment also generates more
unemployment. Section 5 discusses the policy implications of our findings.

2. The model
Our model relies heavily on Layard and Nickell (1986). It describes a representative, imperfectly competitive firm, operating in a small open economy. The firm
decides on its required inputs (labour, capital and materials) to produce a certain
level of output, taking factor prices, technology and the public capital stock as
given. The output level is determined by demand, which is itself a function of the
prices set by the firm on the domestic market and abroad. To maximise its profits,
the firm determines these prices as mark-ups on marginal cost. As for costs, the
cost of capital and the price of materials are exogenous to the small open economy
that we consider. Wages, however, are endogenous. They are bargained between
unions and employers. The following subsections briefly describe production, input prices, output price setting and aggregate supply and demand.
2.1. Production technology and factor demand
Imagine a small open economy where output is produced by N identical private
firms and an (exogenous) government sector. We consider one of these firms. Its
real gross output (X) is produced by services from private capital (K), employed
labour (E), imported intermediate inputs (materials, M) and public capital (KG).
The production function is of the form
+ + +

X = F(A, K, E, KG, M )

(1)

where A reflects technology. The firm minimises total cost TC = PE E + PK K +

PM M subject to the production function under (1). This results in a cost function
of the type
+ +

C = C(X, P , KG, A)

(2)

with P the vector of variable factor prices, P = [PE , PK , PM ]. Public capital and
technology are treated as unpaid fixed inputs. As an adequate approximation of
the cost function under (2), we consider the translog flexible functional form. For
details we refer to Everaert and Heylen (2000). As is well-known from the literature, estimation of this cost function reveals relevant information on a number of
important variables and parameters. To start with, using Shephards lemma, factor
demand equations for variable inputs (E, K, M) can be derived. Furthermore, under
a number of standard conditions the cost function is dual to the production function
and all relevant characteristics of the latter can be derived from the cost function.
A first important characteristic of the production function that can be derived from

98

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

the estimated translog cost function is the elasticity of substitution between inputs.
Second, one can extract from the estimated cost function the willingness-to-pay for
S , i.e., the reduction in the firms
public capital by calculating the shadow price PKG
minimised total cost due to one additional unit of public capital. Very important
for our purpose also are the elasticities of variable input factors with respect to
public capital (i,KG , with i = E, K, M). These elasticities are the result of two
effects. First, an increased supply of productive public infrastructures may imply
that a given output X can be produced at lower cost. This is the case where the
shadow price of public capital is positive. For given factor cost shares, more public
capital then implies a reduced need for any private factor of production. Next,
however, an increase in public capital may change optimal factor cost shares. The
cost share of some private inputs may rise, the share of others may fall. In the end,
for a complementary relationship to show up between public capital and a private
factor of production, the bias of public capital on that factors cost share must be
positive and (more than) compensate the negative total costs effect. Otherwise,
the relationship will be substitutive. Finally, one can compute from the estimated
cost function the elasticity of output with respect to each variable input factor i
(X,i ) and with respect to public capital (X,KG ). Observing these, one can calculate
returns to scale in the production function.
2.2. Input price formation
When the firm decides on the volume of inputs to be used, input prices PM , PK
and PE are given. PM is the price of imported intermediate inputs, expressed in
domestic currency. It is assumed exogenous to the small open economy considered.
The nominal user cost of capital is a function of the real long-term interest rate (R),
the depreciation rate of private capital (k ) and the price of capital goods (Pinv ).
Each of these determinants is assumed to be exogenous:
PK = (R + k )Pinv

(3)

The nominal user cost of labour is a function of the nominal gross wage (W) and
exogenous taxes on labour to be paid by the firm (tbs ).
PE = (1 + tbs )W

(4)

The nominal gross wage has been bargained ex ante with the union. In Everaert
and Heylen (2000) we maximise a standard Nash bargaining function to obtain a
conventional theoretical long-run wage equation, similar to equations presented
by Manning (1993) and Nixon and Urga (1999). The long-run nominal wage
equation that we estimate in the next section allows for a little more flexibility in
the parameters. It is described by Eq. (5):
ln(W) = 0 + (1 p )ln(PX ) + p ln(Pc ) + q ln(Q) u ln(u)
+ b ln(Brr ) tbs ln(1 + tbs ) + ln() + z ln(z)

(5)

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

99

Y
, Y = X M and i > 0
(5 )
E
The bargained nominal gross wage is proportional to the firms output price (PX , to
be discussed below) and to consumer prices (Pc ). Given automatic wage indexation
to consumer prices in Belgium, we expect that in the long-run p = 1. Furthermore, the nominal wage responds positively to the productivity of labour (Q), the
(exogenous) unemployment benefit replacement ratio (Brr ), the firms profit per
unit of value added () and the (exogenous) relative bargaining power of the union
(z). Note that the productivity of labour is defined in (5 ) as real value added Y per
worker. A higher aggregate unemployment rate (u) and higher taxes on labour to
be paid by the firm (tbs ) lead to lower wages. Taxes are exogenous.
with Q =

2.3. Price setting

The representative firm sells its real output (X) on both the domestic market
(Xd ) and abroad (Xf ) at prices Pd and Pf , respectively. Prices are set to maximise
the (nominal) profit function:
n = Pd Xd + Pf Xf C(X, P, KG, A)

(6)

subject to X being equal to Xd + Xf and to the demand equations for Xd and Xf .

Demand rises in real aggregate domestic spending and real world imports, respectively, and falls in the firms relative price at home and abroad. Our assumption that
the firm has market power, implies the well-known mark-up conditions for price
setting, described by (7) and (8). In these equations C/X indicates marginal cost,
whereas d and f denote the absolute value of the price elasticity of demand for
output at home and abroad.
C
d C
Pd =
(7)
= d
, with d > 1
d 1 X
X
Pf =

f C
C
= f
,
f 1 X
X

with f > 1

(8)

Theoretically, there are good reasons for the price elasticity of demand and thus
the mark-up on marginal cost to depend on the business cycle. The sign of this
relation is unclear, however (Layard, Nickell, & Jackman, 1991). We come back
to this in the empirical section. Once Pd and Pf are known, the firms (weighted
average) output price PX can be defined as an identity:
Xd
Xd
PX = Pd + (1 )Pf , with : =
(9)
=
X
Xd + X f
2.4. Aggregate supply and demand
To move to the aggregate economy, we basically follow Layard and Nickell
(1986). Given the assumption of identical firms, aggregate wages and prices will

100

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

be equal to the level determined by the representative firm (W, Pd , Pf , PX ). Other

variables like private sector output, value added, employment, profits and input
volumes are at the aggregate level equal to N times their individual firm size. It
should be emphasised that from now on all our variables (X, E, K, etc.) refer to the
aggregate level.
Eq. (10) describes the equilibrium of supply and demand for real aggregate private output. Aggregate supply X follows from the optimal behavior of firms identified earlier. Firms produce output in response to demand, given relative prices.
The RHS of Eq. (10) is real aggregate demand for output. EXPd stands for real
aggregate domestic demand, Xf IM is net exports of final products:
X = EXPd + (Xf IM)

(10)

Eq. (11) identifies real aggregate domestic demand as the sum of (endogenous) real
private consumption (Cp ), (exogenous) real non-wage government consumption
(Cg ) and (endogenous) real aggregate gross investment (I). Real private consumption in Eq. (12) is assumed mainly (i.e., in the long-run) to be a function of real
household disposable income (Ydis /Pc ). Nominal disposable income (Ydis ) contains
both after-tax labour income and capital income. It is described in Eq. (13). Tax
rates are respectively tw and tc and exogenous. The main endogenous determinants
of labour income are the private sector gross wage bill (EW) and unemployment
benefits (uLs Brr W), with Ls being aggregate labour supply, u the aggregate unemployment rate and Brr the unemployment benefit replacement rate. Both Ls and Brr
are exogenous, u is explained in Eq. (14). A third component of labour income
are government wages (EWg ), which are exogenous. Nominal capital income (Yc )
is composed of dividends and interest income on private and public long-term
bonds. Dividends are a fixed fraction (%1 ) of aggregate after-tax business profits,
(1 tbd )n . As to interest income, we assume that the long-term nominal interest
rate (Rn ) is equal for private and public bonds and exogenous. Furthermore, the
stock of bonds (BH ) is driven by the assumption that households invest a fixed
fraction (%2 ) of their savings in bonds (Eq. (16)). Finally for private consumption,
Eq. (17) specifies the consumer price level as a weighted average of the domestic
output price (Pd ) and the (exogenous) price of imported goods and services (PIM ).
The rate of indirect taxes is tind . It is also exogenous:
EXPd = Cp + Cg + I



Ydis
Cp = Cp
Pc
Ydis = Yd0 + (1 tw )[(EW + EWg ) + uLs Brr W] + (1 tc )Yc
with : u =

Ls E Eg
Ls

Yc = %1 (1 tbd )n + Rn BH1

(11)
(12)
(13)
(14)
(15)

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

101

BH = BH1 + %2 (Ydis Cp Pc )



Xd
IM
Pc = (1 + tind )
Pd +
Pim
Xd + IM
Xd + IM

(16)

I = Ik + Ig + Ih

(18)

K = (1 k )K1 + Ik



Pd
, EXPd
IM = IM
Pim



Pf
Xf = X
,
EXP
w
P

(19)

(17)

(20)
(21)

Eq. (18) defines real aggregate gross investment as the sum of real capital formation by the private business sector (Ik ), the government (Ig ) and households (Ih ).
Household and government investments are taken to be exogenous. Private firms
choose investment in Eq. (19) to obtain their optimal capital stock (K). The latter
follows from the firms cost minimisation process described in Section 2.1.1 As
mentioned before, k is exogenous. Eq. (20) explains real aggregate imports of
final goods and services (IM) as a function of real domestic demand EXPd and the
relative price Pd /PIM . Imports are expected to rise in both arguments. The model
for the demand side is closed by assuming in Eq. (21) that real exports are determined by (exogenous) real world imports (EXPw ) and by the price of exports (Pf )
relative to the (exogenous) price of foreign competitors on the world market (P ).
The demand effects of changes in public investment can now briefly be described. An immediate effect on aggregate output runs via (18), (11) and (10). Secondary effects are caused by the influence of changed output on (un)employment,
which further affects wage formation, productivity and prices. Via real household
disposable income and consumption on the one hand, and competitiveness (relative
prices) and net exports on the other, aggregate demand is again affected. Furthermore, there may be a role for the way in which changes in public investment are
financed. This is evident in the case of changes in taxes, which are prominent in
Eqs. (13), (15) and (17). In the case of bond financing, however, the financing
decision has no immediate influence in our model. This may seem strange given
the role of bonds (BH ) in Eqs. (15) and (16). The reason is double. First of all,
households may substitute government bonds for private sector bonds. Second,
foreigners may buy the newly issued government bonds. Obviously, one might
argue that in the case of bond financing, taxes may have to be raised in the future.
Rational consumers may anticipate this and change their current behavior. Admittedly, our model does not capture these effects. On the other hand, one should
recognise that a lot depends on the effects of changes in government investment on
1 See our brief discussion of the cost function in Section 2.1 from which factor demand equations
can be derived.

102

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

future output and government income for given tax rates. If future output is strong,
future taxes may be avoided! As we shall see in Section 4.2, this expectation may
not be unrealistic. It also brings us to the final equations in our model.
2.5. Government budget balance
Three identities in (22) describe the government budget balance as the difference between nominal revenues (T) and nominal expenditures (G). The former
are mainly determined by the sum of taxes on gross wages in the business sector,
taxes on the unemployed, indirect taxes on nominal domestic demand, direct taxes
on business profits and taxes on households capital income. Expenditures mainly
are government wages, public non-wage consumption, public investment, unemployment benefits and interest payments. T0 captures other revenue categories,
e.g., taxes on wages earned in the government sector. G0 captures other spending
categories, e.g., transfers other than unemployment benefits:
Budget = T G
T = T0 + (tbs + tw )WE + tw Brr W(Ls E Eg )
tind
(22)
+
(Pc (Cp + Cg ) + Pinv I) + tbd  + tc Yc
1 + tind
G = G0 + Wg Eg + Pc Cg + Pinv Ig + Brr W(Ls E Eg ) + Rn GD1
with GD (=GD1 Budget) denoting nominal gross government debt.

3. Empirical analysis
Our empirical analysis relies on annual data from 1965 to 1996 for Belgium.
Our data are taken from the OECD, the National Bank of Belgium and the Belgian
Federal Planning Bureau. Data sources are described in detail in Everaert and
Heylen (2000). Note that as a proxy for technology (A) we use cumulated data for
patents granted by the US Patent and Trademark Office (for details and justification,
see Everaert & Heylen, 2001).
The empirical implementation of the stochastic long-run equilibrium relations
described in Section 2 requires a model of disequilibrium adjustment. Using
three-stage least squares,2 we have simultaneously estimated separate error-correction models for each of the four sectors of the economy, i.e., factor demand and
price setting by firms,3 wage bargaining, total domestic demand and the external
2 Endogenous variables are instrumented using all exogenous variables and once lagged endogenous
variables.
3 The dynamic specification of the cost function, which allows for a consistent derivation of a set
of interrelated factor demand equations in general error-correction form, is taken from Urga (1996)
and Allen and Urga (1999).

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

103

sector, conditioning on the exogenous variables and the endogenous variables determined in the other sectors of the economy. Everaert and Heylen (2000) provide
a detailed explanation of our dynamic empirical model and all results. Due to
space limitations we concentrate in this paper on the relevant estimated long-run
equilibrium relations and on a number of important parameters. For all these relations the null hypothesis of no cointegration can be rejected at the 5% level of
significance.
3.1. Dynamic cost function and dynamic factor demands
For parameter estimates of the dynamic cost function and related dynamic
factor share equations, we refer to Everaert and Heylen (2000). The cost function
is theoretically and empirically well-behaved. Furthermore, the duality conditions
hold, which implies that we can use the parameter estimates of the cost function
and related factor share equations to uncover the characteristics of the underlying
production function. A first important characteristic is that all variable input factors
are found to be substitutes (see again Everaert & Heylen, 2000, for details).
Table 1 reports the long-run elasticities of output with respect to labour, private
capital and materials, a measure of returns to scale (), the shadow price of public
capital and the long-run elasticities of variable inputs with respect to public capital,
all evaluated at the sample mean. The statistically significant positive shadow price
S ) shows that public capital is cost saving. Our estimates imply
of public capital (PKG
that an increase in the public capital stock by 1 euro reduces long-run private sector
cost by 0.24 euro. This positive shadow price is an important element underlying
the negative impact of public capital on private employment. Another element (not
shown) is the negative bias of public capital on the optimal labour cost share. As
a result, an increase in the public capital stock by 1%, for a given output, reduces
private sector employment by about 0.32%, indicating a substitutive relationship
(see E,KG ). On the other hand, the private capital cost share is increasing in public
capital. In net terms, public and private capital are complements. A 1% increase
in the public capital stock raises private sector capital by 0.33% (see E,KG ). This
effect is not accurately measured though.
Evaluated at the sample mean, the production function is characterised by increasing returns over all inputs, including public capital ( = 1.40). The individual
Table 1
Characteristics of the production functiona
Output elasticities, cost flexibility
and returns to scale
X,E
X,M
X,K
a

0.51 (0.03)
0.45 (0.02)
0.13 (0.01)

X,KG
c,X

Shadow price and elasticities

with respect to public capital
0.31 (0.05)
0.92 (0.03)
1.40 (0.07)

S
PKG
c,KG

0.24 (0.05)
0.28 (0.05)

Evaluated at the sample mean. Standard deviations in parentheses.

E,KG
M,KG
K,KG

0.32 (0.17)
0.46 (0.18)
0.33 (0.29)

104

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

elasticities of output with respect to private inputs (X,E , X,K , X,M ) are very much
in line with expectations. Furthermore, the obtained positive and significant elasticity of output with respect to public capital (X,KG = 0.31) supports Aschauers
public capital hypothesis. It also confirms the result obtained by Everaert and
Heylen (2001), using a different methodology.
3.2. Wage bargaining and price setting
To estimate the wage Eq. (5), we use union membership in percent of the
labour force as a measure of union bargaining power (z). Data are mainly from
Ebbinghaus and Visser (1990). The results are reported in Table 2. Since estimating
an unrestricted wage function yields no significant effects for taxes to be paid by
the firm (tbs ) and for the profit rate (), we only report results imposing tbs =
= 0. Note though that tbs has a strong short-run impact on the bargained wage
(not shown). Next, the observation that p is not statistically different from 1, is
consistent with wages being automatically indexed to consumer prices. However,
we do not impose p = 1 because this slightly deteriorates the fit of the model.
Furthermore, the coefficient on labour productivity is significantly smaller than
1, indicating that over the sample period productivity growth is not fully
Table 2
Parameter estimates (3SLS) and residual diagnostics for price setting, wage bargaining, aggregate
consumption and the external sector (19651996)a
Wage bargaining (estimated long-run relation)
ln(Wt )ln(zt ) = 2.564(1.260) + 0.708 ln(Pc,t )(0.196) + 0.292 ln(Px,t )() + 0.743(0.092) ln(Qt )
0.112(0.035) ln(ut ) + 0.262(0.068) ln(Brr,t ) + 1.030(0.112)
Price setting (estimated long-run relation)
96
= (1.464
Pd,t
(0.063) + 3.403bct (0.797) 0.008oilt (0.002) )(Ct /Xt ) + 0.321(0.048) d93
Aggregate consumption function (estimated long-run relation)
= 251, 491
Cp,t
(155,864) + 0.822(0.024) (Ydis,t /Pc,t )
The coverage rate of exports over final imports (estimated long-run relation)
ln(Xf /IM)t = 7.518(1.099) 0.197(0.081) ln(Pd /Pim )t 0.425(0.169) ln(Pf /P )t 0.413(0.071)
ln(EXPd )t
-ln(Wt )
R2

-Pd ,t

-Cp,t

-ln(Xf /IM)t

0.88
0.77 [0.68]
4.60 [0.33]

0.90
1.77 [0.41]
2.84 [0.58]

0.61
0.57 [0.75]
4.93 [0.30]

specification)b

and residual diagnostics (dynamic

R2
0.97
Normality
2 (2) = 0.52 [0.77]
Serial correlation
2 (4) = 3.71 [0.45]

a Due to space limitations we only report the estimated long-run equations. For all these relations
the null hypothesis of no cointegration can be rejected at the 5% level of significance. For detailed
results on the dynamic specification and cointegration tests, we refer to Everaert and Heylen (2000).
Standard errors in parentheses, P-values between square brackets.
b JarqueBera (JB) test for residual normality and BoxPierce (BP) test for fourth-order serial
correlation.

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

105

reflected in wages. The remaining coefficients all have signs confirming theoretical
predictions. Unemployment appears to exert only a moderate downward pressure
on real wages, though. Union membership in contrast has a strong impact on the
bargained wage. The estimated equilibrium domestic price equation is
Pd = (d0 + d1 bc + d2 oil)

C
X

The mark-up (d ) on marginal cost (C/X, which is estimated jointly with the
cost function) in the long-run specification is a function of the business cycle (bc)
and the evolution of oil prices (oil).4 The motivation for including oil prices in the
mark-up stems from the observation that firms appear to be reluctant to transmit
oil price shocks fully into domestic prices. The results in Table 2 show that in the
long-run, when the business cycle is neutral and oil prices are at their average value
(bc = oil = 0), firms set domestic prices 46% higher than the marginal production
cost. The positive coefficient on bc implies that the mark-up is pro-cyclical. The
significant negative coefficient on oil confirms that oil shocks are not fully reflected
in domestic prices, but partly absorbed by firms profits.5
Although graphical inspection reveals a close relationship between export prices
(Pf ) and marginal cost, we are unable to detect a stable causal relation running
from marginal cost to export prices. Rather, export prices exert some feedback on
marginal cost. This observation suggests that, in contrast to the theory in Section 2,
Belgian exporting firms are price takers on the world market. In order to safeguard
profitability (marginal) cost has to adjust to changes in prices on the export market
(e.g., through changes in labour productivity). Although deviations appear to be
long lasting, a tentative exploration of the relation between Pf and P indeed shows
a close connection in the long-run. In what follows, export prices are therefore
assumed to be exogenously determined on the world market.
3.3. Aggregate consumption and net exports
Estimation of the private consumption function reveals a long-run marginal
propensity to consume equal to 0.82. As for international trade, to eliminate the
impact of the steady increase in openness of the Belgian economy, we estimate the
ratio of exports over imports of final products (i.e., Xf /IM) instead of Eqs. (20)
and (21) separately. Both higher domestic demand, higher domestic prices and
higher export prices reduce exports relative to final imports. Higher import prices
4 The business cycle variable bc is the percentage deviation of business sector GDP from its
HodrickPrescott trend; oil is the percentage deviation from the average oil price over the sample
period.
5 Note that we have also included a level-shift dummy (d 93 ), being 1 in the period 19931996, in
96
the cointegrating relationship for domestic prices. The reason is that the price indices used, were no
longer available from 1993 onwards. The extension of the series over 19931996 has been done at the
National Bank of Belgium with a slightly less sophisticated methodology, resulting in a possible break
in the series in 1993.

106

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

and higher prices on the world market have the opposite effect. Real foreign expenditures have no significant long-run effect on Xf /IM and have therefore been
dropped from the cointegrating vector.

4. System simulations
This section explores the impact of changes in the public capital stock on private sector performance, especially (un)employment, by simulating the model of
Section 3 under an alternative public investment policy. We analyse how the Belgian economy would have evolved if a strong decline in public investment in
19821989 had not taken place. Section 4.1 briefly goes into the ability of the estimated model to capture the endogenous variables actual evolution. Section 4.2
discusses the effects of the alternative public investment policy.
4.1. Benchmark simulation
Everaert and Heylen (2000) compare the actual evolution in total cost (C),
the cost shares of labour, capital and materials, real output (X), the productivity
of labour (Q), the nominal gross wage (W), the consumer price level (Pc ), the
unemployment rate (u) and the government budget with these variables evolution
obtained from the dynamic simulation of the model over the period 19701996,
taking the evolution of the exogenous variables as given.6 In general, our model
is well able to capture the movements in the endogenous variables.
4.2. Alternative public investment policy
To fight its huge debt and deficits the Belgian government adopted two longlasting fiscal consolidation programs, in particular in 19821987 and in 19921996.
A strong cut in public investment was an important part of the 19821987 consolidation program (see Fig. 1). Real public investment in Belgium fell back from
4.63% of (business sector) GDP in 1981 to less than 1.75% in 1989. Consolidation
in the 1990s left public investment largely unaffected. In this section we analyse
how the economy would have evolved if the strong decline in public investment
in 19821989 had not occurred and if to finance this more bonds had been
issued. The alternative public investment series (Ig sim) is obtained by fixing the
share of public investment in (business sector) GDP on its 1981 level during the
period 19821989. From 1990 onward, the simulated investment rate is again allowed to vary starting from its higher 1989 level with the actual rate (see
6 The model is simulated over the period 19701996 since no data are available for the 1960s for
some variables in some identities. Due to lack of data, %2 is in the simulation assumed to be equal to
1. Data for %1 are provided by the Belgian Federal Planning Bureau and taken as given. Further, tc is
fixed at 15%.

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

107

Fig. 1. Real public investment (in % of business sector GDP) and real public capital stock (in billions)
under the alternative investment policy, 19701996.

Fig. 1). From the alternative investment series, we then calculate the corresponding
public capital stock (Kg sim). In 1996 this hypothetical capital stock would have
been 28.5% higher than observed in reality.
Obviously, an interesting question would be to assess the effects of alternative
ways to finance the simulated rise in public investment. We keep this for future
research. As for now, it will be important to keep an eye on the long-run consequences of bond financing for the government debt and deficit. If these went off
the rails, it would clearly not be possible to call our simulation results realistic
(because of their unsustainability).
Inserting Ig sim and Kg sim, the model is dynamically simulated over the period
19701996. Table 3 compares the simulation results for key endogenous variables
with these variables evolution according to the benchmark simulation. We concentrate on the long-run effects, i.e., the deviation from the benchmark in 1996.
A number of interesting observations stand out. (i) Although public capital is cost
saving, a shock to public investment does not lower the volume of private sector
cost. In contrast, 1996 costs are 2.7% higher than the benchmark. (ii) The cost
shares of labour and materials decrease slightly, in favour of a higher cost share
of private capital. Given the small increase in total costs, the 1996 private capital
stock is 13.2% larger. (iii) Compared to the benchmark, 1996 real private sector
output is 5.8% higher. (iv) Private sector labour productivity is 13.3% higher. (v)
Table 3
Long-run effects of the alternative public investment policy (difference between alternative and benchmark simulation in 1996)
Total cost (C)
Output (X)
Private capital (K)
Private employment (E)
Materials (M)
Labour productivity (Q)

2.7
5.8
13.2
6.9
1.9
13.3

Nominal wage (W)

6.8
Private consumption (Cp )
4.7
Investment/output (I/X)
4.0a
Consumer price (Pc )
0.2
Cost share of labour
1.3a
Cost share of materials
0.3a

Cost share of capital

1.6a
Unemployment rate
3.8a
Profit rate
3.4a
Government budget 0.9a

a These data are differences in percentage points. All other data are differences between levels in
percentage.

108

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

Table 4
Long-run effects of the alternative public investment policy on unemployment (difference between
simulated alternative outcome and benchmark, 1996)

Unemployment rate
(in percentage points)

Production channel

Demand channel

Wage channel

Total effect

+3.3

0.13

+0.7

+3.8

The average annual growth rate of nominal wages is 0.4% points higher, resulting in a 6.8% higher nominal wage in 1996. Since consumer prices are largely
unaffected, the increase in the nominal wage implies a similar increase in the real
wage. (vi) Given the moderate increase in cost, the decrease in the cost share of
labour combined with the increase in wages implies a 6.9% decrease in private
employment. As a result, the 1996 unemployment rate is 3.8% points higher. (vii)
Higher output being produced with fairly stable costs, the profit rate rises strongly
(+3.4% points).
Our results for the labour market are surprising. Public investment programmes
as such seem to undermine employment, at least in Belgium. Table 4 shows the
three sub-effects that are behind the simulated unemployment rise: (i) substitution
effects for given output and factor prices (production channel), (ii) effects from
changes in aggregate demand and the level of output (demand channel) and (iii)
effects from changes in real wages (wage channel).
The production channel undermines employment. First, given the positive
shadow price of public capital, total cost (for given output, wages and prices)
in 1996 decreases with 5%. Second, for given total cost, higher public capital
exerts a negative bias on the cost share of labour. In total, substitution of public
capital for labour accounts for a 3.3% points increase in the unemployment rate
(for further details, see Everaert & Heylen, 2000). Demand effects are due to the
fact that higher public capital raises aggregate spending and output. In the first
place, there is the increase in public investment itself. Furthermore, the effects of
higher public capital propagate through the system. Private investment, calculated
from changes in the private capital stock, as well as private consumption, affected
by household disposable income and consumer prices, contribute to further output
increases. The total increase in 1996 output amounts to 5.8%. (Exports are not
affected since they are sold on the world market at an exogenous price.) The direct
effect of this aggregate demand increase on employment is unfavourable since the
cost share of labour falls in output. Indirectly, however, given a cost flexibility of
output of 0.92 (see c,X in Table 1), the increase in private output implies a rise
in 1996 total cost of about 5.2%. For given cost shares, this generates an increase
in private employment. Combining both effects, a small positive net demand effect on private employment remains (+0.3%). The unemployment rate decreases
by 0.13% points. Finally, there is the wage channel. Real wages are influenced
by an increase in the public capital stock through changes in labour productivity,
changes in the unemployment rate and changes in domestic prices. Prices being

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

109

Fig. 2. Government budget balance (in % of business sector GDP), simulation under the alternative
investment policy (19701996).

largely unaffected (see Table 3), we further disregard them. Due to higher output
being produced with less labour, 1996 labour productivity is 13.3% higher than
in the benchmark. This partially feeds through into higher wages (+9.9%). Despite a fall in unit labour costs, employment is negatively affected because the
growth in real wages induces an increase in the relative cost of labour and substitution of cheaper capital and materials.7 Since the production channel and the rise
in labour productivity raise the unemployment rate, the full impact of wages on
(un)employment depends on how unions react. Given the estimated low responsiveness of real wages to unemployment, rising unemployment offsets only part
of the rise in real wages, leaving a net increase of 7.0%. This 7.0% wage increase
accounts for a 0.7% points rise in the unemployment rate.
A final and very important result in Table 3 concerns the governments financial
balance. As can be seen, the simulated bond financed rise in public investment does
not undermine the governments budget in the long-run. Fig. 2 provides a detailed
picture. Mainly thanks to higher economic growth and tax receipts, in 1996 the
increase in the deficit compared to the benchmark is only 0.9% points of GDP
(coming from a maximum of 2.61% points in 1989). Extrapolation of the evolution
of the budget suggests a decrease in the deficit, relative to the benchmark, from
about 1998 onwards. Clearly, for the reliability of the above mentioned simulation
results and for the assumptions on private consumption that we have made in
Section 2, this result is reassuring.

5. Conclusion and policy implications

In this paper we develop and estimate a structural model to analyse the
long-run impact of public capital on private sector performance, in particular
7 Note that the substitution of capital and materials for labour for a given output in the second
round further raises labour productivity. As a result, the unemployment rate will increase further.

110

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

(un)employment, in Belgium. The estimates show that services from public capital
significantly reduce private sector cost, for given output and wages. An increase in
the public capital stock with 1 euro reduces long-run private cost with 0.24 euro.
The output elasticity of public capital implied by these estimates equals 0.31,
which confirms Aschauers public capital hypothesis as well as our earlier results
(Everaert & Heylen, 2001), although a totally different methodology has been
used. With respect to private sector inputs, public capital and labour are found to
be substitutes, public capital and private capital complements. Model simulations
show that the basic negative relationship between public capital and employment
is not altered once the effects of a change in public capital on aggregate demand
and wages are taken into account, on the contrary.
As to policy, our research strongly supports the case for an increase in public
capital spending, although not unconditionally. The main benefits are obvious.
We find at least for Belgium that public capital contributes significantly
to higher output and productivity growth. In the light of current concerns, e.g.,
about the growth gap between the EU and the US or about the future retirement
burden, it is clearly important to strengthen the economys productive capacity.
Public capital spending should not, however, be raised unconditionally. This paper has also demonstrated the need for complementary measures. First of all, in
contrast to suggestions by, e.g., the European Commission (1993), higher public
capital spending may cause (more) unemployment. Part of the rise in unemployment is due to upward pressure on real wages prompted by an increase in labour
productivity. Part is also due to the often observed weak responsiveness of Belgian
wages to rising unemployment itself (see also Abraham, De Bruyne, & Auwera,
2000; Heylen & Van Poeck, 1995). If policy makers want higher public investment spending to contribute to employment, complementary measures to enhance
wage moderation and wage flexibility are (again) called for. In this respect, our
results support recommendations by, e.g., Calmfors (1998) or Wyplosz (2001).
Furthermore, labour market performance may benefit from a careful selection of
public investment projects. Our empirical analysis relies on a broad concept of
public capital. It does not imply that all investment projects bring about substitution of labour. For example, investment in training facilities or mobility for the
unemployed may be beneficial for employment. So may infrastructure investment
aimed at structural conversion of weak regions. Further research on the labour
market effects of sub-categories of public capital is required here. Complementary public investment and labour market policy is important also from a different
perspective. It has often been argued that labour market reform is unlikely to succeed in an unfavorable macroeconomic environment (e.g., Allsopp & Vines, 1998;
Wyplosz, 2001). The growth stimulating effect of public investment can be used
to raise the short-run benefits of structural reform and consequently its probability
of success.
The second reason for not unconditionally increasing public investment results
from existing budgetary constraints. Given the enduring need in Belgium to reduce
the government debt ratio, as well as the restrictions imposed by the EMU Stability

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

111

and Growth Pact, there is no room for deficits anymore, not even temporary ones.
On the one hand, one may regret this. Our analysis confirms that an increase in
government investment strengthens the productive capacity of the economy, with
no permanent worsening of the government budget. Furthermore, it has been shown
in various studies that fiscal consolidation has a higher probability of success when
government investment is increased rather than cut (e.g., Alesina & Perotti, 1996;
Heylen & Everaert, 2000). On the other hand, as long as this narrow framework
exists, it will force policy makers to carefully balance the costs and benefits of their
spending decisions and to improve the composition of their expenditures. Higher
public investments are still possible, but they should be financed from budgetary
margins created by reductions elsewhere (e.g., declining interest payments or other
transfers). Even then, the long-run benefits of public investment should still be
balanced against the benefits of faster debt reduction.

Acknowledgments
The authors thank Jakob de Haan, Gerd Hansen, Glenn Rayp and Rudi Vander Vennet for helpful suggestions. They have also benefited from comments by
participants at the 13th Annual EALE Conference (September 2001, Jyvskyl)
and participants at the conference on Macroeconomic Transmission Mechanisms:
Empirical Applications and Econometric Methods (May 2000, Copenhagen). Any
remaining errors are ours.

References
Abraham, F., De Bruyne, K., & Van der Auwera, I. (2000). Challenges for successful wage setting
in Europe. In H. Ooghe, et al. (Eds.), The economic and business consequences of the EMU
(pp. 181209). Norwell: Kluwer Academic Publishers.
Alesina, A., & Perotti, R. (1996). Fiscal adjustments in OECD countries: Composition and
macro-economic effects. NBER Working Paper, No. 5730. National Bureau of Economic Research.
Allen, C., & Urga, G. (1999). Interrelated factor demands from dynamic cost functions: An application
to the non-energy business sector of the UK economy. Economica, 66, 403413.
Allsopp, C. J., & Vines, D. (1998). The assessment: Macroeconomic policy after EMU. Oxford Review
of Economic Policy, 14(3), 123.
Aschauer, D. A. (1989a). Is public expenditure productive. Journal of Monetary Economics, 23, 177
200.
Aschauer, D. A. (1989b). Public investment and productivity growth in the Group of Seven. Federal
Reserve Bank of Chicago Economic Perspectives, 13, 1725.
Calmfors, L. (1998). Macroeconomic policy, wage setting and employmentwhat difference does the
EMU make? Oxford Review of Economic Policy, 14(3), 125151.
Drze, J., & Malinvaud, E. (1994). Growth and employment: the scope of a European initiative. European Economic Review, 38, 489504.
Ebbinghaus, B., & Visser, J. (1990). Trade unions in Western Europe since 1945. London: Macmillan.
European Commission. (1993). Growth, competitiveness, employment: The challenges and ways forward into the 21st century (White Paper). Bulletin of the European communities. Supplement 6/93,
Luxembourg.

112

G. Everaert, F. Heylen / Journal of Policy Modeling 26 (2004) 95112

Everaert, G., & Heylen, F. (2000). Public capital and labour market performance in Belgium. Working Paper. Faculty of Economics and Business Administration, Ghent University, No. 00/98 (url:
http://fetew.rug.ac.be).
Everaert, G., & Heylen, F. (2001). Public capital and productivity growth: Evidence for Belgium,
19531996. Economic Modeling, 18, 97116.
Heylen, F., & Everaert, G. (2000). Success and failure of fiscal consolidation in the OECD: A multivariate analysis. Public Choice, 105, 103124.
Heylen, F., & Van Poeck, A. (1995). National labour market institutions and the European Economic
and Monetary Integration process. Journal of Common Market Studies, 33, 573595.
Layard, R., & Nickell, S. (1986). Unemployment in Britain. Economica, 53, 121169.
Layard, R., Nickell, S., & Jackman, R. (1991). Unemployment. Oxford: Oxford University Press.
Manning, A. (1993). Wage bargaining and the Phillips curve: The identification and specification of
aggregate wage equations. Economic Journal, 103, 98118.
Nixon, J., & Urga, G. (1999). Investment, growth and unemployment. Modelling the supply side of the
UK economy. In C. Driver & T. Temple (Eds.), Investment, growth and employment: Perspectives
for policy (pp. 221248). London: Routledge.
Sturm, J. E., Kuper, G. H., & de Haan, J. (1998). Modelling government investment and economic
growth on a macro level: A review. In S. Brakman, H. van Ees, & S. K. Kuipers (Eds.), Market
behaviour and macroeconomic modelling. London: MacMillan.
Urga, G. (1996). On the identification problem in testing the dynamic specification of factor-demand
equations. Economics Letters, 52, 205210.
Wyplosz, C. (2001). Economic growth and the labour markets: Europes challenge. In J. Smets & M.
Dombrecht (Eds.), How to promote economic growth in the Euro area (pp. 152176). Cheltenham:
Edward Elgar.