driven by the fact that the composition of value added differ in the two areas. While manufacturing is more important in the North–Center, both agriculture and
public services are more important in the South. These differences in productive structure can influence unemployment also via the price wedge in opposite
directions in the North and in the South.
12
While average tax rates have increased during the 1980s and 1990s, the real price of imported energy has significantly declined. On the other hand, real social
transfers per head have increased both in the North–Center and in the South. The uncovered asymmetries in the long run association of social transfers per head, the
real price of energy and the tax wedge with regional unemployment suggest that the positive association of regional unemployment differentials with higher real
social transfers per head and with the lower real prices of energy has prevailed over the negative association between these differentials and the increase in the tax
wedge.
4. Wages and unemployment
Poor employment performance in the South during the 1980s and the 1990s is not necessarily the exclusive outcome of negative idiosyncratic shocks taking
place during the same period, but could also be the lagged response of firms to negative shocks that occurred much earlier. It is tempting to extend to the Italian
South a recent interpretation of high European unemployment in the 1980s and 1990s, that views it as the result of the lagged response of capital to the
Aappropriation pushB of the 1970s, when union pressure increased significantly the labor share. In a nutshell, the story goes as follows. With a putty-clay technology,
capital is difficult to adjust in the short run, but in the long run both the substitution of capital to labor and the development of more capital intensive
technologies can reduce employment and increase the profit share to values even
Ž .
higher than before the appropriation push Caballero and Hammour, 1998 . A
12
Let the price wedge P
W
be P
W
s a c q b p
y p
Ž .
i i i
i m
c s eq p y p
i i
Ž .
where i indicates the region North, South , c is the competitiveness, p is the price of imported
i m
energy and materials, p is a world price index, p is the price index of the region, and e stands for
i
the exchange rate. a and b are proportional to the external exposure and to the dependence of the
i i
economy on imported materials, respectively. If EerE p and E p
rE p are small, then EP
W
rE p f b
m m
i m
i
y a E p rE p . Therefore, EP
W
rE p G 0 if E p rE p
F b ra . The smaller is the dependence on
i i
m i
m i
m i
i
Ž .
Ž .
imported energy and materials b with respect to the openness to foreign markets a , the higher is
i i
the probability of finding EP
W
rE p -0. If this is the case, a rise in p
should imply a reduction in
i m
m
local unemployment.
Table 6 Percentage changes in labor productivity, labor cost, labor share, and capitalroutput ratio
Productivity Labor cost
Labor share Capitalroutput
1970 – 1979 North–West
3.36 2.92
y0.43 2.25
North–East 3.10
2.79 y0.30
0.24 Center
3.55 3.83
0.27 1.44
South–West 3.04
4.23 1.15
5.59 South–East
3.48 4.90
1.36 3.10
North–Center 3.40
3.20 y0.20
1.58 South
3.16 4.40
1.20 4.82
1980 – 1994 North–West
2.69 1.40
y1.26 0.44
North–East 2.54
1.10 y1.40
0.03 Center
2.33 1.22
y1.08 0.73
South–West 2.15
0.53 y1.58
0.97 South–East
2.69 1.17
y1.48 0.36
North–Center 2.53
1.28 y1.23
0.48 South
2.31 0.72
y1.56 0.78
closely related story is that, while adverse labor supply shifts have been important for the increase of unemployment during the 1970s, wage moderation and rising
capital shares can be consistent with high and rising unemployment in the 1980s Ž
. only in the presence of negative labor demand shocks Blanchard, 1998 .
To illustrate the Italian experience, Table 6 shows the percentage changes in the labor share and in labor productivity in the private sector between 1970 and 1979
and between 1979 and 1994 for the different geographical areas.
13
Furthermore, Fig. 5 shows that the divergence of regional unemployment rates was accompanied
by nationwide convergence of labor costs, productivity, and capitalroutput ratios. As shown by Table 6, the labor share increased during the 1970s in the South and
remained nearly constant on average in the Northern and Central regions. During the same period, the unemployment rate increased by close to 3.5 percentage
13
It is well known that the relationship between real wages and the labor share depends on the production technology. When the technology is Cobb Douglas, the labor share is constant and does not
vary with changes in labor costs. When the technology is CES, however, the labor share s varies with
L
the capital–output ratio according to
´
s s1y Ak
Ž .
L t t
where k is the capital–output ratio and A is a measure of disembodied technical progress. See Ž
. Bentolila and Saint Paul 1998 . In this case, real wages can affect the labor share only by affecting the
capital–output ratio. Even when higher wages lead to a higher k, the labor share increases only if labor Ž
. and capital are complements in production ´ -0 .
G. Brunello
et al.
r Labour
Economics 8
2001 103
– 129
118
Fig. 5. Regional convergence of unemployment, labor cost, productivity, capitalroutput ratio over the period 1970–1994. Solid lines are the result of a robust Ž
. Ž
. Ž
. line fitting Tuckey, 1977 . Lines with negative positive slope indicate convergence divergence . North–Central and Southern regions are represented by
circles and triangles, respectively.
Ž points in the South and by about 2 percentage points in the North–Center see
. Table 1 . Interestingly, the increase in the labor share in the South was accompa-
nied by positive employment growth, that limited the expansion of the rate of unemployment. During the 1980s and 1990s, the labor share fell faster in the
southern than in other regions and unemployment boomed to its currently very high levels. While employment growth remained positive in the North–Center, it
turned negative in the South where, at the end of the 15-year-period, employment was lower than in 1979. Since labor force growth remained high, unemployment
had to increase.
The labor share in the South increased relatively to the national average during most of the 1970s and started to decrease in the early 1980s. The increase in the
Ž .
1970s was the result of the rapid growth of relative to the North wages, which Ž
. was not compensated by relative labor productivity growth. As discussed in
Ž .
detail by Faini 1993 , an important event that hit specifically the South at the end of the 1960s was the elimination of institutional rules allowing for the existence of
Ž .
regional wage differentials in union contracts gabbie salariali . Following this
Ž event, only partially compensated by specific payroll tax breaks
fiscalizzazione .
degli oneri sociali e sgraÕi contributiÕi , relative labor costs surged in the South and the labor share increased. As unemployment started to increase, the share
partially declined. The rapid increase of the capital stock in the South during this period was induced not only by the increase in labor costs, but also by political
reasons. Therefore, the productivity of newly installed capital in the South was generally lower than in the North–Central regions and, while regional productivity
converged at a nationwide level, productivity differentials remained important, and even increased, between the North–Center and the South.
The arguments above do not explain why regional unemployment differentials persist. From a theoretical standpoint, differences in regional unemployment can
increase and persist when local wages are not sensitive to local economic conditions. This is the case when wage setting is relatively centralized and wage
determination is influenced mainly by the economic conditions prevailing in the leading economic areas of the economy. When a negative shock increases
Ž .
unemployment in a backward region the South , wages are not significantly affected and higher unemployment can persist in the absence of other adjustment
Ž .
factors. When the same shock hits the leading region the North–Center , how- ever, wages are negatively affected and the decline in wage growth can contribute
to reduce unemployment.
14
When regions characterized by economic asymmetries are politically integrated, as in Germany and in Italy, national wage formation can be dominated by the
14
Ž .
A detailed theoretical setting is presented by Brunello et al. 2000 .
economic interests of the leading regions.
15
In this case, the failure of regional wages to respond to regional local conditions in some areas of the country can
exacerbate unemployment differentials by eliminating an important adjustment mechanism. We empirically investigate this possibility by pooling regional data
for 18 Italian regions during the period 1970–1994 and by estimating the following empirical model
16
Dlnw s a D q b T q dDln u q hDlnPM q g lnw
Ý
i t i
i i t
t i ty1
i
q u ln u q s ln u
q ´ 3
Ž .
N ty1 i ty1
i t
where w is the real gross wage in region i, u
is the unemployment rate
i t N t
prevailing in the North–Center, T is a linear trend, D is a set of regional
i
dummies, and ´ is the error term. Since changes in the current regional
i t
unemployment rate are likely to be endogenous, we use 2SLS estimates. Our selected instruments are the lagged change in regional unemployment, changes in
the national rate of unemployment and the first lag of a cyclical indicator, obtained as the residual from fitting regional real GNP on a quadratic trend.
17
If the leading region hypothesis is correct, we expect to find that, conditional on the local unemployment rate, unemployment in the North–Center attracts a
significant coefficient. In the extreme version of the Aleading areaB hypothesis, local unemployment does not affect local wages, which depend only on unemploy-
ment in the leading area. Our estimates are presented in Table 7, where we show the long run elasticities of the real wage both to regional unemployment, h
, and
wu
to unemployment in the leading area, h . When we consider all the regions, both
wun
regional unemployment and unemployment in the North–Center attract a signifi- cant and negative coefficient in the wage equation. Moreover, the long run wage
elasticity of regional unemployment is about half as large as the long run wage elasticity of unemployment in the leading area. This finding confirms that unem-
ployment in the North–Center plays a key role in Italian regional wage determina- tion. When we focus on the southern labor market, however, we find no evidence
that regional unemployment significantly affect regional wages after controlling
15
Ž .
See Saint Paul 1997 for a discussion.
16
Ž .
In previous work Brunello et al., 2000 we have investigated the question by using aggregate data. Our evidence can be summarized with the following two points: 1. aggregate wage setting in Italy
depends only on the rate of unemployment prevailing in the Northern and Central areas of the country. Southern unemployment does not affect wage pressure. 2. There is evidence of a long-run cointegrating
relationship between unemployment in the Northern and Central areas, the tax wedge, the real interest rate and a measure of union power.
17
Since both Dln PM , T and ln u are common to all the regions in the sample, we correct the
t N ty1
standard errors by using the option cluster in the routine regress of Stata 5.0.
Table 7 Elasticity of real wages to regional and North–Center unemployment in the private sector
Variable All regions
Southern regions Dln u
y0.002 0.017
i
DlnPM 0.026
0.025
t
ln u y0.023
y0.032
Ny 1
T 0.002
0.003 lnw
y0.138 y0.137
iy 1
ln u y0.012
y0.024
iy 1 2
R 0.33
0.35 N
450 200
obs
h y0.087
y0.175
wu
h y0.167
y0.236
wu n
Dependent variable: Dlnw. Sample period: 1970–1994. One asterisk and two asterisks when the coefficients are significant at the 10 and at the 5 level of confidence, respectively. White consistent
standard errors.
for unemployment in the leading area. This result clearly supports the view that wage setting is not affected by southern unemployment.
18
5. Labor and firm mobility
