- Email: [email protected]
Capital shortages and asymmetries in UK unemployment
STRUCTURAL CHANGE AND ECONOMIC DYNAMICS Structural Change and Economic Dynamics 9 ( 1998) 189-204
ELSEVIER
Capital shortages and asymmetries in UK unemployment Philip Arestis *, Iris Biefang-Frisancho
Mariscal
Department of Economics, University of East London, Longbridge Road, Dagenham, Essex. RM8 ZAS, UK
Abstract The paper is concerned with the determination of wages and unemployment in the UK. Particular emphasis is put on the asymmetric effect of capital stock on employment. It is argued that capital scrapping in response to the two oil price shocks, combined with subsequent sluggish growth in capital, may be responsible for the rise of the NAIRU and the persistence of unemployment. Furthermore, since during the recovery investment responded slowly, capital shortage may to some extent be responsible for the rise in long-term unemployment. The theoretical model also suggests that workers are concerned about their relative position in the wage hierarchy. Consequently, nominal wages adjust asymmetrically in response to changes in the NAIRU. The empirical analysis is concerned with testing the above hypotheses, by using quarterly data for the UK from 1966 until 1994. We use cointegration analysis for the determination of wages, unemployment and the ratio of the long-term unemployed. An asymmetric error correction model tests for different speeds of adjustment in response to changes in the NAIRU. 0 1998 Elsevier Science B.V. All rights reserved. Keywords;
UK; Wages; Unemployment;
JEL Classification:
C13;
C30;
C32;
Nairu; Capital
shortages;
Cointegration
530; 531; 564
1. Introduction The persistence of high unemployment has been one of the most puzzling developments of the past twenty years or so. In the UK, unemployment averaged 2.1% between 1966 and 1973, and since 1974 it has risen to an average of 7.5%. The prevailing view of the persistence of unemployment and of the continuous rise in the NAIRU is that explanations and solutions can only be found on the supply side and not on the demand side of the economy (Layard et al., 1991; Lindbeck, 1993; OECD, 1994; Summers, 1990). Most economists regard the problem of job creation * Corresponding author. Tel: +44-181-590-7000; fax: +44-181-849-3549; e-mail: [email protected] 0954-349X/98/$19.00 0 1998 Elsevier Science B.V. All rights reserved. I’II SO954-349X(97)00025-8
190
P. Ares&
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
mainly as a matter of encouraging more employment with the existing stock of capital. However, any long-run analysis should consider that the capital stock may vary. The relationship between capital stock and employment may be a simple explanation which suggests that investment in new productive capacity increases the number of jobs, while the destruction of existing capacity may destroy jobs (Rowthorn, 1995). It may also be that the fall in investment encourages asymmetric responses in the labour market and thus the persistence of unemployment. This paper is concerned with the determination of aggregate wages and unemployment in the UK. We try to explain the persistence of unemployment by beginning with theories that concentrate on supply side failures in the labour market (Arestis and Skott, 1993). The model is extended by examining the relationship between capital stock and employment, a procedure that we have started recently in Arestis and Biefang-Frisancho Mariscal (1998b). It is argued that the fall in investment over the last twenty years or so is one of the major elements behind the fall in employment and that only a substantial rise in productive capital may reduce unemployment. It is also argued that capital scrapping during the two oil price shocks, combined with low investment subsequently, may have caused long-term unemployment. This is not to say that low skill or other worker characteristics are not important, but that investment-enhancing policies may have prevented longterm unemployment to some extent. Furthermore, we introduce a new dimension to the discussion of money wage determination and unemployment by incorporating non-linearities in the relationship. There are various reasons for expecting that wages adjust with different speed to disequilibrium errors (Acemoglu and Scott, 1994; Keynes, 1936; Holzer and Montgomery, 1993). The asymmetric error correction model applied here, accounts for the different speeds of adjustment in response to deviations of the actual unemployment rate from the attractor, the NAIRU (Granger and Lee, 1989; Granger and Swanson, 1996). The objective of the empirical investigation is to test each of the hypotheses put forward. Firstly, we want to find out whether capital shortage increases the NAIRU and the extent to which it may do so. Secondly, we test whether a fall in investment creates long-term unemployment, and thirdly, we consider whether the speed of adjustment in nominal wages depends on unemployment being above or below the NAIRU. The paper begins with the underlying theoretical model, which is followed by a section which deals with the empirical analysis. The final section summarises and concludes.
2. The theoretical model In this section we extend the wage model put forward in Arestis and Skott (1993) and Arestis and Biefang-Frisancho Mariscal ( 1998a,b), itself based on efficiency wage models (Shapiro and Stiglitz, 1984; Bowles and Boyer, 1990). The fundamental idea is that conflict in the work place arises over labour productivity (e) and the
P. Are&,
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
191
real wage (W/P). This is summarized in Eq. (1):
WP=f(e 1
(1)
1
where the sign under a variable gives the direction of the partial derivative. Labour productivity depends on the cost of job loss (c) and socio-economic factors (zJ (Arestis and Biefang-Frisancho Mariscal, 1998a), so that:
(2’)
e = e’(c,zl)
with the cost of job loss conventionally determined by the real wage (W/P), the expected real wage (IV/P), the chance of unemployment (U), and the expected income from unemployment (w”,/Pe) which allows us to rewrite Eq. (2’) as:
e= W/P), 4
(2)
A
where z includes z1 and the remaining elements in c. Efficiency wage theories are concerned with the relationship between work effort and the real wage. In fact, wages are negotiated in a decentralized bargaining process which determines the nominal wage on the basis of workers’ expected real wage (V/P). Combining Eqs. ( 1) and (2), we may establish a direct relationship between the negotiated nominal wage (IV,, J, the expected real wage and the socio-economic factors (z) as: W t+1
=P”UW”IP”h
4
(3)
We next discuss briefly the socio-economic elements which refer to the conflict over income distribution and worker militancy (Rowthorn, 1977, 1995), pay norms and fairness of pay between groups of workers (Keynes, 1936; Hicks, 197.5) and the level and composition of unemployment (Blanchard and Summers, 1986; Lindbeck and Snower, 1986; Nickell, 1987). Conflict in the labour market is measured by the aspiration gap (n*-ZI”) which shows the extent to which the target profit share (n*) differs from the negotiated profit share (P). The latter depends on the claims of the foreign (F) and government (G) sectors and demand conditions in the labour market proxied by unemployment (U). The target profit share is determined by the claims of the government and foreign sectors and demand conditions in the product market (@). A positive aspiration gap implies that workers’ actual wage is below the one they anticipated in wage negotiations. In response they may threaten to withdraw or reduce work effort as a means of enforcing wage claims. Workers are the more successful in demanding higher wages, the lower the rate of unemployment is and the more militant (X) they are. Fairness of pay between groups of workers relates to the wage structure, where workers resist money wage cuts for fear of a decline in their relative position in the wage hierarchy (Keynes, 1936). A wage system with well-established wage differentials is regarded as fair and it becomes what is expected to be (Hicks, 1975).
192
P. Arestis, I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
The real expected wage (IV/F) is thought to capture this historical element that affects both the effort function and the negotiated wage. Hysteresis theories are concerned with the composition of unemployment and the effect different groups in the pool of unemployed have on real wage determination. The long-term unemployed (LU) effectively drop out of the labour market, so that they do not exert downward pressures on wage claims themselves, but allow insiders to have a major impact. The rising inflow into short-term unemployment (AU) threatens insiders’ jobs and they may give in on wage objectives. We may now summarize the socio-economic elements and thus write the z-vector as: z=z(17*
-ll”,F,T,X,U,
We/P=, W;/P”,AU,LU)
(4)
Substituting Eq. (4) into Eq. (3), we may arrive at: W,+, ==V(n+
-~,~,F,T,X,~,~W”IP”~,~W~lP”~,AU,LUl +
+++p
+
+
+
(5)
Writing the model in log-linear form, Eq. (6) follows: w, + 1
=pe + a, (w; -p”)
+ CLZ(We-p”)
+
cI3 [
--a,[p-w+Ip]+C(5X+Clgti+Clgtz~
p” - We + lp” + @“I +h(u)
(6)
where lower case letters denote logarithms and where all coefficients are greater than zero. The function h(u) comprises unemployment and hysteresis effects. The vector ti stands for the claims of the foreign and government sectors on private output and contains real import costs and tax variables. The expressions in square brackets describe the aspiration gap as discussed above, without the tax and import cost variables since they are already included in the ti variable. The first square bracket describes the target profit share as a mark-up over real wage costs in relation to the demand conditions in the product market, proxied by capacity utilization. At this stage, we incorporate one of the hypotheses we are particularly interested in, namely the effect of capital on wages and employment (Malinvaud, 1986; Sneessens and Dreze, 1986; Bean, 1989; Rowthorn, 1995; van de Klundert and van de Schaik, 1990). The fundamental idea of the capital shortage hypothesis is that big adverse shocks, as for example the two oil price shocks, may persistently increase unemployment due to a fall in investment. The oil price shocks increased the claims of the foreign sector on national income and caused accelerated inflation. In order to bring inflation under control, restrictive demand policies were introduced. As a result, capacity utilization fell and firms scrapped spare capital.’ After the shocks were reversed, inflationary pressures were reduced and employment rose. However, employment did not return to previous levels due to a lack of productive capital. As a consequence, the economy is locked into a situation of high unemployment
1 In fact, the growth
rate of gross capital
was on average
-0.4%
between
1974 and 1981.
P. Arestis, I. B.-F. Mar&al 1 Structural Change and Economic Dynamics 9 (1998) 189-204
193
combined with normal capacity utilization. 2 It follows from this discussion that capacity utilization is explained by capital stock and the level of economic activity as this is proxied by the unemployment rate (Rowthorn, 1995). Substituting capital stock and unemployment for @ in Eq. (6) and assuming that expectations are formed as in Layard et al. (1991), i.e., as: As;, r = s, + As, and performing some simple algebraic transformations, AAw=cc,Ip+a,Alp-a,(~-p)-cr,A(w-p)+a,(w, +a,x+cr,ti+cr,Ati+a,,k+h(u)
(7) we arrive at Eq. (8): -p)+ol,A(w,
-p) (8)
The relationship in Eq. (8) can be interpreted as follows: nominal wage acceleration depends on real wages, productivity, unemployment, real benefits, tax and import costs, capital stock, the expected changes in these variables and a variable capturing militancy. Most of the variables included are standard (Arestis and Skott, 1993; Arestis and Biefang-Frisancho Mariscal, 1994; Wren-Lewis, 1992; Hall and Henry, 1987; Layard et al., 1991), except for the capital stock. An increase in the capital stock allows real wages to rise, as capacity utilization falls and firms’ ability to mark-up is limited (Rowthorn, 1995). The long-run real wage equation can be obtained by letting A(w-p), A(w,-p), Alp, Ati, Au and Ak equal to zero to give: (w-P)=B~(wu-P)+B2x-P3~+84~~+85ti+Bsk+B,lp
(9)
where lu denotes the ratio of the long-term unemployed in the pool of the unemployed. As suggested by membership theories, the long-term unemployed are not able to price themselves into the labour market because of a decline in human capital, low motivation for active job search, and a stigmatizing perception of the potential employer. When the target and the negotiated profit shares are equal, where the latter is derived as a residual in the bargaining process, unanticipated inflation is zero and the “equilibrium” unemployment rate can be derived (Rowthorn, 1977). Assuming that firms mark-up prices over labour costs, an aspiration gap of zero implies that the real wage (per unit of output) acceptable to firms is equal to the bargained wage. Furthermore, assuming productivity neutrality with respect to unemployment (Blanchard, 1988) an assumption that is widely backed by empirical results (Manning, 1992; Elmeskov, 1993; National Institute, 1995; Arestis and BiefangFrisancho Mariscal, 1994), and equating firms’ target (feasible) real wage share with the bargained real wage share, we may derive the non-accelerating inflation rate of *Note that high levels of unemployment and “normal” levels of capacity utilization indicate a limited substitutability between capital and labour during this period. Consequently, the argument that the slowdown in capital accumulation (from an average of 3.5% in the pre-shock period to 1.4% since 1974) may have no adverse implications for employment provided the reduction in capital accumulation is associated with the adoption of more labour intensive techniques than would otherwise have been, is not valid here.
194
P. Arestis. I. B.-F, Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
unemployment
(NAIRU)
as follows:
#=Yl(w,--p)+y,x+y,ti-y,k+y,lu
(10)
The NAIRU is an increasing function of real benefits, tax and import costs, and worker militancy, and declines with the increase in the capital stock. A positive relationship between real wages and unemployment benefit implies that a fall in the latter results in a decline in unemployment, due to its labour-cost-reducing impact. However, a cut in wage compensation may have other effects that outweigh the favourable cost effects. In response to a fall in real benefits, the unemployed may reduce searching time, and they may have to accept jobs for which they are less suitably qualified than they would accept if the cost of search was lower. The consequent inefficiencies in the labour market may have an adverse effect on the unemployment rate and it may be an empirical question as to which of the effects is more important. We return to this issue in Section 3. An increase in taxes or import costs will, under given demand conditions, increase conflict over income shares and will raise inflationary pressure. The more wage earners resist a fall in their income share, the lower are the tax and import effects on the long-run unemployment rate. Worker militancy affects the bargained real wage and ceterisparibus income distribution, where the latter determines the level of employment through its supply and demand side effects.
3. Empirical investigation3 All estimations reported in this section are undertaken on the computer package MICROFIT4.0, except for the estimation of the cointegrating relationship for the ratio of the long-term unemployed, which was estimated by PCGIVE8.0, since it appears not to be possible to obtain directly the long-run static relationship and the necessary diagnostics in MICROFIT4.0. The first part of the empirical section deals with a discussion on the data. This is followed by the determination of the cointegrating relationships corresponding to Eqs. (9) and (10). The effect of capital shortage on the long-term unemployed is tested in the sub-section that follows, with the third sub-section concerned with the error correction models dictated by the analysis and empirical results in the rest of this section. 3.1. Variable dejinitions and data The model is estimated using seasonally adjusted data over the period 196641 to 1994Q4 which was determined strictly by the availability of data. The data come 3All estimations reported in this section are undertaken on the computer package MICROFIT4.0, except for the estimation of the cointegrating relationship for the ratio of the long-term unemployed, which was estimated by PCGIVE8.0, since it appears not to be possible to obtain directly the long-run static relationship and the necessary diagnostics in MICROFIT4.0.
P. Arestis. I B.-F. Mariscal 1 Structural
Change and Economic Dynamics 9 (1998) 189-204
195
from various issues of Economic Trends: UK Economic Accounts (A and B), Department of Employment Gazette, OECD Business Sector Database, and Annual Abstract of Statistics. We define and discuss briefly each variable in turn that is used in the estimations. The variable “average earnings” was calculated as the ratio of total wages and salaries over total employment, where the former excludes force’s pay. “Average earnings” were deflated by the consumer price index. This index may be appropriate to determine workers’ real consumption wage as an indication of workers’ living standard. Another variable that is part of the profit function in that it determines wage costs, is labour productivity. It is calculated as the ratio of output GDP to total employment. Furthermore, it is argued in the previous section that worker’s productivity depends on income from work and receipts from unemployment. It appears to be impossible to capture appropriately on an aggregate level income from unemployment, since there is a complex system of different types of benefit payments, depending on family size and other circumstances. The variable which was used here is the index of the weekly flat rate unemployment benefit. One of the underlying ideas of the model is that wages are the result of conflict over income shares. Workers are the more successful in achieving wage targets the more militant they are. It is notoriously difficult to measure militancy. We used the number of strikes as a proxy, knowing that a strike is only one (extreme) form of militancy and also knowing that only a fraction of actual strikes are registered. This proxy is only useful if we can assume that the ratio between registered strikes, actual strikes and other smaller conflicts is constant (Paldam, 1989). Other proxies of militancy such as the union mark-up by Layard and Nickel1 (1986), measures of trade union density, or number of days lost are also known to be deficient (Hall and Henry, 1987; Paldam, 1989; Purdy and Zis, 1974). Additionally, it was stressed that conflict over income distribution increases when private sector income falls due to increased demands from the government or from the foreign sectors. We experimented with different variables to capture this effect (such that we included the income, direct, and employers’ tax rates and the real price of imports weighted by the share of imports, or, the real exchange rate separately). The variable that turned out to be most useful is defined as the sum of the direct tax rate as a proportion of the wage bill, employers’ tax rate and the difference between the consumer and producer price index. This variable proxies the gap between the consumer and product wage (Layard et al., 1991). Any increase in this wedge generates the potential for conflict in the labour market. In the theoretical model it was suggested that the long-term unemployed effectively dropped out of the labour market. In line with other studies, we used the ratio of those who are unemployed for more than a year to total unemployment (National Institute, 1995). The unemployment rate was calculated as total unemployment over labour force. Furthermore, it was proposed that, assuming that the substitutability between capital and labour is limited, the decline in investment in the 70s and 80s may have caused capital shortage unemployment in the subsequent years. It is difficult to estimate this effect particularly over the suggested sample period. It is apparent that in the 80s and 90s capital equipment per person is higher than in the 60s and 70s. Using capital stock data, this change in technology requirements cannot be captured satisfactorily. By not explicitly taking this change into
196
P. Ares&
I. B.-F. Mar&al
/ Structural Change and Economic Dynamics 9 (1998) 189-204
account, we may underestimate the necessary rise in capital to increase employment.4 The only source we know that provides quarterly data on fixed gross capital stock is the OECD.’ 3.2. Cointegrating
vectors for real wages and unemployment
We estimate a vector autoregressive regression (VAR) model for the determination of real wages and unemployment as described by Eqs. (9) and (lo), applying the maximum likelihood test and estimation procedures of Johansen ( 1988).6 The Akaike Information Criterion and the Schwarz-Bayesian Criterion (Ltithkepohl, 1991) suggest that the optimal lag length for the VAR is two. The likelihood ratio test of the deletion of the deterministic trend and a dummy for the first quarter of 1975, where real wages were exceptionally high, rejects the null hypothesis with a chi-square statistic (CHSQ) of 36.7 for 12 degrees of freedom. Based on the variable set in Eq. (9), the results of the maximal eigenvalue test, the trace test, the Hannan-Quinn and Schwarz-Bayesian criteria suggest two cointegrating vectors. We proceeded with the identification of the two cointegrating vectors, applying the Pesaran and Shin (1994) approach which relies on imposing non-homogeneous restrictions on the two cointegrating vectors.’ The result is the following two cointegrating relationships: (w-p)=Ip-0.17u+0.010x+0.800k+0.4901u
(lla)
U=0.100x-0.200/z+
(llb)
1.2OOlu
where all variables are in logarithms. The likelihood ratio test of the validity for the overidentifying restriction is CHSQ(6) = 10.17 with a significance level of 0.12, denoting the validity of the restricted system. The first equation describes the average real wage (w-p), determined by labour productivity (fp), unemployment (u), strike activity (x), gross capital stock (k) and long-term unemployment (ZU),defined as the ratio of those who are unemployed for more than 52 weeks to the total number of unemployed. The second equation explains unemployment as a function of strike activity, capital stock, and long-term unemployment. We find proportionality between labour productivity and real wages, suggesting that gains in labour productivity are completely absorbed by increases in real wages, without affecting unemployment. The effect of militancy on wages, Eq. (1 la), and 4We have not treated technical change and capital accumulation separately. This is because in a study of this type that uses highly aggregated data, such a distinction would not pick up any differences that might exist. Proper account for the effect of such a difference should be undertaken in studies which use considerable disaggregation. sA detailed description on the calculation of stocks of fixed capital is provided by the OECD in “Methods Used by OECD Countries to Measure Stocks of Fixed Capital ~ National Accounts: Sources and Methods”, No. 2, Paris 1993. 6The results of the Dickey-Fuller tests suggest that all variables included here are integrated of order one; they are not reported here, but can be obtained from the authors upon request. ‘This approach tests the cointegrated system for general non-linear restrictions and thus uses a generalization of the results by Johansen (1992) and Phillips (1991) for the linear case.
P. Ares&
I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
191
unemployment, Eq. (1 lb), is uncontroversial. As we know, particularly from the 1970s organized labour in the UK reacted to supply side shocks with increased strike activity, thereby causing a decline in output and wage gains above labour productivity growth. Investment in productive capacity affects both real wages and unemployment. At a given level of economic activity, a higher capital stock implies higher capacity utilization, thereby reducing firms’ ability to increase prices and allowing real wages to rise. Furthermore, investment in new capacity increases employment. The long-term unemployed do not exert downward pressure on real wages and their effect on the NAIRU is considerable. Interestingly enough, the tax and import wedge do not affect the cointegrating vectors. The lack of empirical evidence of tax and import effects is due to them being more likely short-run effects. For example, an increase in import cost in relation to the domestic price level improves competitiveness and raises wage pressure by making consumption goods more expensive relative to value added output. If a rise in competitiveness enables firms to mark-up prices to compensate for the higher cost in the long-run, this would imply that they could increase their domestic profit share indefinitely (Layard et al., 1991). Equally, a permanent effect of competitiveness on wage setting implies that workers can resist a decline in real wages indefinitely. A similar argument would apply to increases in taxation and benefits. Our estimates of the NAIRU, which on the whole seem to be in line with those of other studies, are shown in Table 1. The levels of the estimated NAIRU over the full sample period and the period since 1982 are 5.9% (6.0%) and 7.5% (9.4%), where the actual unemployment rates are given in brackets. From 1988 onwards, the estimated NAIRU and the actual unemployment rates are 6.2% and 8.4%, respectively. The estimates are closely related to the actual unemployment rate. It is, however, evident that unemployment has been above its long-run level over the last fifteen years or so. Table 2 quantifies the factors that affected the changes in the NAIRU over particular subperiods. The effect of the contributory variables on the NAIRU can be explained by the changes in the average values of each variable. More concretely, the differences were calculated with the help of the formula: ANAIRU = f3AZ,where Table 1 Period
u%
Source
1969-1973
2.5
1974-1980
3.8
1981-1987
10.1
1988-1990
6.8
other studies Eq. (llb) other studies Eq. (llb) other studies Eq. (llb) other studies Eq. (llb)
NAIRU% (11) (13) (15) (5)
1.665.6 4.6 4.5-7.3 5.3 5.229.9 8.5 3.5-8.1 7.1
Notes: All the information in Table 1, except for the NAIRU estimates of Eq. (llb), are taken from Cromb (1993). The entry “other studies” refers to the total number (in brackets) of other sources which report the NAIRU range referred to in the last column.
198
P. Ares&
I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
Table 2 Period (1982-94)/( (1974-94)/( (1982-94)/(
1966-81) 1966-73) 1974-81)
1U
k
x
3.9 3.3 3.2
-0.4 -0.5 -0.8
-0.8 -0.2 -0.3
A refers to the difference of x between two periods, x refers to the average value of the variable under consideration and 6’refers to the corresponding coefficient as estimated in Eq. (1 lb). The periods are determined as follows: the pre- first oil price shock period (1966-1973), the pre- and oil price shock periods (1966-1981), the period of the two shocks (1974-1981), and the shock and after-shock periods (197441994). For all periods, the rise in the NAIRU is overwhelmingly attributed to the rise in long-term unemployment as the coefficients of Eq. (1 lb) already indicated. Over all periods, long-term unemployment contributed more than 3 percentage points to the increase in the NAIRU. Growth in capital stock reduced the NAIRU by 0.8 percentage points when the after-shock and shock periods are compared. The dramatic fall in militancy since 1982 reduced the NAIRU by 0.8 percentage points in comparison with the 1960s and 1970s which were characterized by high strike frequency. These results raise two questions: why is it that the growth in capital stock contributes so little to reducing the NAIRU, and why is it that changes in the NAIRU are principally explained by long-term unemployment? We suggested earlier that it may be that long-term unemployment and capital shortage are directly related. The argument is that the consequences of capital scrapping in the two shock periods were a rapid increase in unemployment. During the recovery, capacity was constrained so that only a fiction of the displaced workers were re-employed and as the installation of additional capaicty takes a consderable period of time, thsoe workers who remained unemployed during the recovery became the stigmatized and disencouraged long-term unemployed, as suggested by hysteresis models. If this was the case, it could very well be that some of the effect measured by the capital stock, was actually contributed by the long-term unemployment rate. It is thus important that we study the latter more closely. 3.3. The cointegrating relationship for long-term unemployment Big adverse shocks like the oil price shocks cause capital scrapping, or at least a significant slowdown in investment. Since there is a considerable time lag between the rise in unemployment during a recession, the recovery and, eventually, the installation of additional capital, that would allow a return to previous employment levels, displaced workers that could not find employment at the beginning of the recovery must become long-term unemployed. In this section we estimate the effect of a shortage in capital on long-term unemployment. The specification of long-term unemployment is related to the one in Layard et al.
P. Arestis, I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
199
( 199 1). Accordingly, long-term unemployment is described by the level of unemployment since the greater the unemployment level is, the harder it is to get back to work and the longer the unemployment stretch. This assumption is based on the premise that job creation is insufficient to satisfy demand, or, in other words, outflow is less considerable than inflow. This seems to be the case for the UK. However, initially, the ratio of the long-term unemployed falls when unemployment rises. We also include the percentage change in the unemployment rate and the capital stock. We started with the estimation of an autoregressive distributed lag model. The lag structure was determined on the basis of F-tests.’ The resulting relationship is: 8.638
+
0.490~
-
0.836k -
0.074Au
(12)
lU= (4.39)
(14.9)
Estimation period: RESET(1,90)=2.5 ADF(4)=
-4.2
(4.4)
1967Q4-199444 NORM(2)=4.0
(2.6) WALD( 3) = 293
AR( 5,86) = 0.6
HET(34,56)=0.9
DF= -4.3
All coefficients are correctly signed and significant. The significance of the explanatory variables was tested with the WALD test and the residuals were tested for fourth-order autocorrelation (AR), heteroscedasticity (HET), autoregressive conditional heteroscedasticity (ARCH), functional form misspecification (RESET) and normality (NORM) and none of the tests was significant at the usual 5% significance level. The augmented Dickey-Fuller (ADF) and the Dickey-Fuller (DF) tests, indicate that this is a cointegrating relationship. ’ Capital stock is a significant variable with a reasonably high effect on long-term unemployment, giving some support to our hypothesis. It was also considered that increased mobility should reduce unemployment duration (Benoit-Guilbot, 1994). This effect was measured by a mismatch variable, which however did not turn out to be significant. We tested for the effect of real unemployment benefits on long-term unemployment and this was never significant, either. The duration of unemployment benefits could also be an important factor in the determination of long-term unemployment. However, we could not test for this due to the unavailability of time series data. But it should be noted that a relevant OECD study (OECD, 1991) does not find an apparent overall correlation between the level or the duration of unemployment benefits on long-term unemployment.”
‘For further details see PC-GIVE version 8.0. 9All variables except for Au, which is I(O), are Z( 1). The inclusion of an I(0) variable in a set of Z( 1) variables does not affect the cointegrating relationship (Charemza and Deadman, 1992). This variable was included due to previous economic reasoning. “Worker characteristics such as education, age and skill are important factors in the determination of long-term unemployment (Portugal and Addison, 1995). These factors are not included in this study, largely due to the absence of appropriate time series data.
200
P. Ares&
I. B.-F Mariscal/
Structural
Change and Economic Dynamics 9 (1998) 189-204
3.4. Error correction models
Following Engle and Granger (1987), we estimate the error correction models corresponding to the long-run relationships for nominal wage acceleration, the unemployment rate and the long-term unemployment rate. The novelty of the short-run wage equation is that we employed the asymmetrical error correction model by Granger and Lee (1989) and Granger and Swanson (1996), who suggest that adjustment to the long-run can be thought of as being captured by different speeds back to the attractor, which in our case is the NAIRU. The theoretical model suggests that workers are concerned about their relative position in the wage structure and resist nominal wage cuts. Nominal wages are sticky downwards, while upward wage adjustments will be readily sought after. In order to test this hypothesis, the error correction term (ECM) of the long-run unemployment equation was plit into its positive and negative deviations from the mean. With respect to the above hypothesis, we would expect that the coefficient of adjustment is significantly lower when unemployment is above its equilibrium level (positive deviation) than when it is below it (negative deviation). In order to test for the equality of the coefficients, a Wald test (WALD IQ) was used along with the ECM denoting the negative deviations (ecmu-) and the ECM with the positive deviations (ecmu+). The estimated error correction models are reported in Table 3. The error correction terms are correctly signed and significant in all three models. None of the models seem to have obvious diagnostic problems. We begin our discussion of the empirical results of Table 3 with the nominal wage acceleration as described by Eq. (S), remembering, however, the long-run proportionality between real wages and productivity. We find that real wage growth, real unemployment benefit growth, unemployment acceleration as well as the growth of taxes and import cost affect nominal wage acceleration. We could not find a significant effect of capital stock growth on nominal wage acceleration, suggesting that additional investment does not mitigate inflationary pressures in the short-run. The Wald statistic rejects the null hypothesis of equality of the coefficients of the ecmu+ and the ecmu- with a CHI ‘( 1) = 0.82 and a significance level of 0.36 indicating that there is asymmetry in the adjustment process. The coefficient on ecmu+ is, as expected, significantly lower than the coefficient on ecmu-. We may therefore conclude that there is downward stickiness in nominal wage adjustment when unemployment is above its long-run level.” The short-run equation that corresponds to the NAIRU in Eq. (1 lb) is only determined by its own dynamics, the error correction terms and a dummy for 1974, which captures the effects of the three-day working week at the beginning of that year. The growth rate of the ratio of long-term unemployed moves together with unemployment growth. Increases in unemployment growth reduce long-term unem-
“We also estimated a symmetric error correction model for wage acceleration. The effects of the individual variables remained stable, however, the diagnostics were slightly better in the asymmetric model.
P. Ares& Table 3 Dynamic
I. B.-F Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
201
relationships
Estimation variable
period
1967Q2-1994Q4 AA(w-Ip)
1967Q3-1994Q4 AU
1967Q331994Q4 Alu
constant
-0.111
-0.135
-0.015
(2.9)
(3.5)
ecmu, _ 1
(4.9) - 0.020
ecmu,?
,
(4.1) -0.0117
ecmu,
3
(2.1) - 0.024 (3.1)
ecmw,_s ecmw, _ 4
-0.052
- 0.084 (2.9)
NW,-A-z
(7.0) - 0.227 (10.2) 0.093
WJ-d-4
(3.3) 0.097
ecmlu, _ 1
(3.5)
AU,-1
0.876 (18.4) 0.178
1.023 (10.1)
(1.9) AAu, _ i
-0.169 (5.1)
AAu,_,
-0.251
AAu,_,
(2.3) -0.367 (3.6)
NW-P-k~Ip),-, A(w-p-&),-s
-0.930 (11.0) -0.343
Ati,_,
(4.2) 0.072
D74.1
(1.9) 0.157 (5.4)
R2 AR RESET NORM( 2) HET WALD IQ
0.64 1.6 (497) 1.2 (1100) 2.1 2.3 (1109) 0.8
0.78 2.0 (4100) 1.4 (1103) 3.8 2.8 (1108)
0.65 1.9 (4102) 1.2 (1105) 4.0 2.6
Note: R2 is the adjusted coefficient of determination and the meaning of the diagnostic tests was explained in the text. The values in brackets under the coefficients denote the t-statistic and the values in brackets following the diagnostics denote the degrees of freedom. All tests are F-tests, except for the test for normality and the Wald IQ tests, which are chi-square tests with two and one degrees of freedom, respectively.
202
P. Are&,
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
ployment. Adjustments a year.
to the long-run relationship
are achieved after just over
4. Summary and conclusions We have presented an aggregate wage model for the UK which combines wage theories that are concerned with explaining failures in the supply side of the labour market. This model is extended by incorporating demand side aspects; more precisely, the capital shortage hypothesis. The model provides us with long- and short-run wage and unemployment equations, which we proceed to estimate. The empirical section begins by identifying two cointegrating vectors, one describing real average wages and the other the NAIRU. It also finds it necessary to investigate the possibility of a cointegrating relationship for long-run unemployment, which is estimated, before the appropriate error correction models are investigated. We find that real wages are explained by labour productivity, long-term unemployment, the unemployment rate, worker militancy and the capital stock. As there is homogeneity between real wages and labour productivity, the NAIRU is determined by long-term unemployment, worker militancy and capital stock. Tax wedge effects, import cost effects and real benefits are not found to be relevant in the long-run, but they prove to be important in the error correction models. For the long-run analysis, we decompose the impact of each explanatory variable on the NAIRU. We find that the greatest effect comes from long-term unemployment and to a minor degree from capital shortage and militancy. Since we think that capital shortage may particularly affect long-term unemployment, we test for the effect of capital shortage on long-term unemployment and find a considerable and significant effect. For the error correction model of nominal wage determination, we find significant asymmetric responses. Nominal wages adjust when actual unemployment is above the NAIRU much more slowly than when unemployment is below the NAIRU, which confirms our theoretical priors. The policy implications of our findings are that shortages in demand may have persistent effects on employment. Although, programmes that improve the skills of displaced workers can be important, policies that aim to enhance investment may be necessary. These policies should aim at reducing long-term real interest rates, producing a high rate of return on productive investment and creating a stable financial environment to support the real sector of the economy.
Acknowledgements
We are grateful to the editors of Structural Change and Economic Dynamics, especially to Harald Hagemann, to an anonymous referee and to Jim Marley and Peter Howells for their helpful comments. Also to Gary Young of the NIESR for providing data on a number of variables.
P. Ares&
I. B.-F. Mar&al 1 Structural Change and Economic Dynamics 9 (1998) 189-204
203
References Acemoglu, D., Scott, A., 1994. Asymmetries in the cyclical behaviour of UK labour markets. The Economic Journal, 1303-l 323. Arestis, P., Skott, P., 1993. Conflict, wage relativities and hysteresis in UK wage determination. Journal of Post Keynesian Economics, 3655386. Arestis, P., Biefang-Frisancho Mariscal, I., 1994. Wage determination in the UK: Further empirical results using cointegration. Applied Economics, 417-427. Arestis, P., Biefang-Frisancho Mariscal, I., 1998a. Conflict elements in UK wage determination: Keynesian, Post-Keynesian and New Keynesian theoretical and empirical considerations. In: Rotheim, R. (Ed.), New Keynesian Economics and Post Keynesian Alternatives. Routledge, London. Arestis, P., Biefang-Frisancho Mariscal, I., 1998b. Conflict, effort and capital stock in UK wage determination. Empirica, in press. Bean, C., 1989. Capital shortage and persistent unemployment. Economic Policy, 12-53. Benoit-Guilbot, O., 1994. Why are there so many unemployed in the EU? In: Benoit-Guilbot, O., Gallie, D. (Eds.), Long-term Unemployment. Pinter Publishers, London. Blanchard, O.J., 1988. Comment on Nickell: The supply side and macroeconomic modelling. In: Bryant, R.C. et al. (Eds.), Empirical Economic for Independent Economies. Brookings Institution, Washington. Blanchard, O.J., Summers, L.H., 1986. Hysteresis and the European unemployment problem. In: Fischer, S. (Ed.), NBER Macroeconomics Annual 1986. MIT Press, Cambridge, MA. Bowles, S., Boyer, R., 1990. A wage-led employment regime: Income distribution, labour discipline and aggregate demand in welfare capitalism. In: Marglin, S., Schor, J. (Eds.), The Golden Age of Capitalism. Clarendon Press, Oxford. Charemza, W.W., Deadman, D.F., 1992. New Directions in Econometric Practice. Edward Elgar, Aldershot. Cromb, R., 1993. A survey of recent econometric work on the NAIRU. Journal of Economic Studies, 27-51. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: Representation, estimation and testing. Econometrica, 251-276. Elmeskov, J., 1993. High and persistent unemployment: Assessment of the problem and its causes. Economics Department Working Paper No. 132. OECD, Paris. Granger, C.W.J., Lee, T.H., 1989. Investigation of production, sales and inventory relationships using multicointegration and non-symmetric error correction models. Journal of Applied Econometrics, 145-159. Granger, C.W.J., Swanson, N., 1996. Future developments in the study of cointegrated variables. Oxford Bulletin of Economics and Statistics, 5377553. Hall, S.G., Henry, S.G.B., 1987. Wage models. National Institute Economic Review, 70-75. Hicks, J., 1975. The Crisis in Keynesian Economics. Basil Blackwell, Oxford. Holzer, H.J., Montgomery, E.B., 1993. Asymmetries and rigidities in wage adjustments by firms. Review of Economics and Statistics, 3977408. Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 231-254. Johansen, S., 1992. Identifying restrictions of linear equations. mimeo, Institute of Mathematical Statistics, University of Copenhagen. Keynes, J.M., 1936. The General Theory of Employment, Interest and Money. Macmillan, London. van de Klundert, T.C.M.J., van de Schaik, A.B.T.M., 1990. Unemployment persistence and loss of productive capital: A Keynesian approach. Journal of Macroeconomics, 3633380. Layard, R., Nickell, S. Jackman, R., 1991. Unemployment. Oxford University Press, Oxford. Layard, R., Nickell, S., 1986. Unemployment in Britain. Economica, S12llS169. Lindbeck, A., 1993. Unemployment and Macroeconomics. MIT Press, Cambridge, MA. Lindbeck, A., Snower, D.J., 1986. Wage setting, unemployment and insider-outsider relations. American Economic Review, 2355239. Ltithkepohl, H., 1991, Introduction to Multiple Time Series Analysis. Springer-Verlag, Berlin.
204
P. Arestis, I B.-F. Mar&al
1 Structural
Change and Economic Dynamics 9 (1998) 189-204
Malinvaud, E., 1986. The rise of unemployment in France. Economica, S1977S217. Manning, A., 1992. Productivity growth, wage setting and the equilibrium rate of unemployment. Discussion Paper No. 63, Centre for Economic Performance, London School of Economics, London. National Institute, 1995. National Institute Model 12. Nickell, S., 1987. Why is wage inflation in Britain so high? Oxford Bulletin of Economics and Statistics, 103-129. OECD, 1991. Employment Outlook. OECD, Paris. OECD, 1994. The OECD Jobs Study, Part 1. OECD, Paris. Paldam, 1989. A wage structure theory of inflation, industrial conflicts and trade unions. Scandinavian Journal of Economics, 63-81. Pesaran, M.H., Shin, Y., 1994. Identification of and testing for cointegrating relations with general non-homogeneous restrictions. mimeo, Department of Applied Economics. Phillips, P.C.B., 1991. Optimal inference in cointegrated systems. Econometrica 59, 2833306. Portugal, P., Addison, T. 1995. Short- and Long-term Unemployment: a Parametric Model with TimeVarying Effects, Oxford Bulletin of Economics and Statistics, pp. 2055227. Purdy, D.L., Zis, G., 1974. Trade unions and wage inflation in the UK: A reappraisal. In: Laidler, D. Purdy, L.D. (Eds.), Inflation and Labour Markets. Manchester University Press, Manchester. Rowthom, R.E., 1977. Conflict, inflation and money. Cambridge Journal of Economics, 215-239. Rowthorn, R.E., 1995. Capital formation and unemployment. Oxford Review of Economic Policy, 26639. Shapiro, C., Stiglitz, E., 1984. Equilibrium unemployment as a worker discipline device. American Economic Review, 433-444. Sneessens, H.R., Dreze, J.H., 1986. A discussion of Belgian unemployment, combining traditional concepts and disequilibrium econometrics. Economica, S89-S119. Summers, L., 1990. Understanding Unemployment. MIT Press, Cambridge, MA. Wren-Lewis, S., 1992. Nominal inertia and Keynesian effects. NIESR Discussion Paper No. 174.
ELSEVIER
Capital shortages and asymmetries in UK unemployment Philip Arestis *, Iris Biefang-Frisancho
Mariscal
Department of Economics, University of East London, Longbridge Road, Dagenham, Essex. RM8 ZAS, UK
Abstract The paper is concerned with the determination of wages and unemployment in the UK. Particular emphasis is put on the asymmetric effect of capital stock on employment. It is argued that capital scrapping in response to the two oil price shocks, combined with subsequent sluggish growth in capital, may be responsible for the rise of the NAIRU and the persistence of unemployment. Furthermore, since during the recovery investment responded slowly, capital shortage may to some extent be responsible for the rise in long-term unemployment. The theoretical model also suggests that workers are concerned about their relative position in the wage hierarchy. Consequently, nominal wages adjust asymmetrically in response to changes in the NAIRU. The empirical analysis is concerned with testing the above hypotheses, by using quarterly data for the UK from 1966 until 1994. We use cointegration analysis for the determination of wages, unemployment and the ratio of the long-term unemployed. An asymmetric error correction model tests for different speeds of adjustment in response to changes in the NAIRU. 0 1998 Elsevier Science B.V. All rights reserved. Keywords;
UK; Wages; Unemployment;
JEL Classification:
C13;
C30;
C32;
Nairu; Capital
shortages;
Cointegration
530; 531; 564
1. Introduction The persistence of high unemployment has been one of the most puzzling developments of the past twenty years or so. In the UK, unemployment averaged 2.1% between 1966 and 1973, and since 1974 it has risen to an average of 7.5%. The prevailing view of the persistence of unemployment and of the continuous rise in the NAIRU is that explanations and solutions can only be found on the supply side and not on the demand side of the economy (Layard et al., 1991; Lindbeck, 1993; OECD, 1994; Summers, 1990). Most economists regard the problem of job creation * Corresponding author. Tel: +44-181-590-7000; fax: +44-181-849-3549; e-mail: [email protected] 0954-349X/98/$19.00 0 1998 Elsevier Science B.V. All rights reserved. I’II SO954-349X(97)00025-8
190
P. Ares&
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
mainly as a matter of encouraging more employment with the existing stock of capital. However, any long-run analysis should consider that the capital stock may vary. The relationship between capital stock and employment may be a simple explanation which suggests that investment in new productive capacity increases the number of jobs, while the destruction of existing capacity may destroy jobs (Rowthorn, 1995). It may also be that the fall in investment encourages asymmetric responses in the labour market and thus the persistence of unemployment. This paper is concerned with the determination of aggregate wages and unemployment in the UK. We try to explain the persistence of unemployment by beginning with theories that concentrate on supply side failures in the labour market (Arestis and Skott, 1993). The model is extended by examining the relationship between capital stock and employment, a procedure that we have started recently in Arestis and Biefang-Frisancho Mariscal (1998b). It is argued that the fall in investment over the last twenty years or so is one of the major elements behind the fall in employment and that only a substantial rise in productive capital may reduce unemployment. It is also argued that capital scrapping during the two oil price shocks, combined with low investment subsequently, may have caused long-term unemployment. This is not to say that low skill or other worker characteristics are not important, but that investment-enhancing policies may have prevented longterm unemployment to some extent. Furthermore, we introduce a new dimension to the discussion of money wage determination and unemployment by incorporating non-linearities in the relationship. There are various reasons for expecting that wages adjust with different speed to disequilibrium errors (Acemoglu and Scott, 1994; Keynes, 1936; Holzer and Montgomery, 1993). The asymmetric error correction model applied here, accounts for the different speeds of adjustment in response to deviations of the actual unemployment rate from the attractor, the NAIRU (Granger and Lee, 1989; Granger and Swanson, 1996). The objective of the empirical investigation is to test each of the hypotheses put forward. Firstly, we want to find out whether capital shortage increases the NAIRU and the extent to which it may do so. Secondly, we test whether a fall in investment creates long-term unemployment, and thirdly, we consider whether the speed of adjustment in nominal wages depends on unemployment being above or below the NAIRU. The paper begins with the underlying theoretical model, which is followed by a section which deals with the empirical analysis. The final section summarises and concludes.
2. The theoretical model In this section we extend the wage model put forward in Arestis and Skott (1993) and Arestis and Biefang-Frisancho Mariscal ( 1998a,b), itself based on efficiency wage models (Shapiro and Stiglitz, 1984; Bowles and Boyer, 1990). The fundamental idea is that conflict in the work place arises over labour productivity (e) and the
P. Are&,
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
191
real wage (W/P). This is summarized in Eq. (1):
WP=f(e 1
(1)
1
where the sign under a variable gives the direction of the partial derivative. Labour productivity depends on the cost of job loss (c) and socio-economic factors (zJ (Arestis and Biefang-Frisancho Mariscal, 1998a), so that:
(2’)
e = e’(c,zl)
with the cost of job loss conventionally determined by the real wage (W/P), the expected real wage (IV/P), the chance of unemployment (U), and the expected income from unemployment (w”,/Pe) which allows us to rewrite Eq. (2’) as:
e= W/P), 4
(2)
A
where z includes z1 and the remaining elements in c. Efficiency wage theories are concerned with the relationship between work effort and the real wage. In fact, wages are negotiated in a decentralized bargaining process which determines the nominal wage on the basis of workers’ expected real wage (V/P). Combining Eqs. ( 1) and (2), we may establish a direct relationship between the negotiated nominal wage (IV,, J, the expected real wage and the socio-economic factors (z) as: W t+1
=P”UW”IP”h
4
(3)
We next discuss briefly the socio-economic elements which refer to the conflict over income distribution and worker militancy (Rowthorn, 1977, 1995), pay norms and fairness of pay between groups of workers (Keynes, 1936; Hicks, 197.5) and the level and composition of unemployment (Blanchard and Summers, 1986; Lindbeck and Snower, 1986; Nickell, 1987). Conflict in the labour market is measured by the aspiration gap (n*-ZI”) which shows the extent to which the target profit share (n*) differs from the negotiated profit share (P). The latter depends on the claims of the foreign (F) and government (G) sectors and demand conditions in the labour market proxied by unemployment (U). The target profit share is determined by the claims of the government and foreign sectors and demand conditions in the product market (@). A positive aspiration gap implies that workers’ actual wage is below the one they anticipated in wage negotiations. In response they may threaten to withdraw or reduce work effort as a means of enforcing wage claims. Workers are the more successful in demanding higher wages, the lower the rate of unemployment is and the more militant (X) they are. Fairness of pay between groups of workers relates to the wage structure, where workers resist money wage cuts for fear of a decline in their relative position in the wage hierarchy (Keynes, 1936). A wage system with well-established wage differentials is regarded as fair and it becomes what is expected to be (Hicks, 1975).
192
P. Arestis, I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
The real expected wage (IV/F) is thought to capture this historical element that affects both the effort function and the negotiated wage. Hysteresis theories are concerned with the composition of unemployment and the effect different groups in the pool of unemployed have on real wage determination. The long-term unemployed (LU) effectively drop out of the labour market, so that they do not exert downward pressures on wage claims themselves, but allow insiders to have a major impact. The rising inflow into short-term unemployment (AU) threatens insiders’ jobs and they may give in on wage objectives. We may now summarize the socio-economic elements and thus write the z-vector as: z=z(17*
-ll”,F,T,X,U,
We/P=, W;/P”,AU,LU)
(4)
Substituting Eq. (4) into Eq. (3), we may arrive at: W,+, ==V(n+
-~,~,F,T,X,~,~W”IP”~,~W~lP”~,AU,LUl +
+++p
+
+
+
(5)
Writing the model in log-linear form, Eq. (6) follows: w, + 1
=pe + a, (w; -p”)
+ CLZ(We-p”)
+
cI3 [
--a,[p-w+Ip]+C(5X+Clgti+Clgtz~
p” - We + lp” + @“I +h(u)
(6)
where lower case letters denote logarithms and where all coefficients are greater than zero. The function h(u) comprises unemployment and hysteresis effects. The vector ti stands for the claims of the foreign and government sectors on private output and contains real import costs and tax variables. The expressions in square brackets describe the aspiration gap as discussed above, without the tax and import cost variables since they are already included in the ti variable. The first square bracket describes the target profit share as a mark-up over real wage costs in relation to the demand conditions in the product market, proxied by capacity utilization. At this stage, we incorporate one of the hypotheses we are particularly interested in, namely the effect of capital on wages and employment (Malinvaud, 1986; Sneessens and Dreze, 1986; Bean, 1989; Rowthorn, 1995; van de Klundert and van de Schaik, 1990). The fundamental idea of the capital shortage hypothesis is that big adverse shocks, as for example the two oil price shocks, may persistently increase unemployment due to a fall in investment. The oil price shocks increased the claims of the foreign sector on national income and caused accelerated inflation. In order to bring inflation under control, restrictive demand policies were introduced. As a result, capacity utilization fell and firms scrapped spare capital.’ After the shocks were reversed, inflationary pressures were reduced and employment rose. However, employment did not return to previous levels due to a lack of productive capital. As a consequence, the economy is locked into a situation of high unemployment
1 In fact, the growth
rate of gross capital
was on average
-0.4%
between
1974 and 1981.
P. Arestis, I. B.-F. Mar&al 1 Structural Change and Economic Dynamics 9 (1998) 189-204
193
combined with normal capacity utilization. 2 It follows from this discussion that capacity utilization is explained by capital stock and the level of economic activity as this is proxied by the unemployment rate (Rowthorn, 1995). Substituting capital stock and unemployment for @ in Eq. (6) and assuming that expectations are formed as in Layard et al. (1991), i.e., as: As;, r = s, + As, and performing some simple algebraic transformations, AAw=cc,Ip+a,Alp-a,(~-p)-cr,A(w-p)+a,(w, +a,x+cr,ti+cr,Ati+a,,k+h(u)
(7) we arrive at Eq. (8): -p)+ol,A(w,
-p) (8)
The relationship in Eq. (8) can be interpreted as follows: nominal wage acceleration depends on real wages, productivity, unemployment, real benefits, tax and import costs, capital stock, the expected changes in these variables and a variable capturing militancy. Most of the variables included are standard (Arestis and Skott, 1993; Arestis and Biefang-Frisancho Mariscal, 1994; Wren-Lewis, 1992; Hall and Henry, 1987; Layard et al., 1991), except for the capital stock. An increase in the capital stock allows real wages to rise, as capacity utilization falls and firms’ ability to mark-up is limited (Rowthorn, 1995). The long-run real wage equation can be obtained by letting A(w-p), A(w,-p), Alp, Ati, Au and Ak equal to zero to give: (w-P)=B~(wu-P)+B2x-P3~+84~~+85ti+Bsk+B,lp
(9)
where lu denotes the ratio of the long-term unemployed in the pool of the unemployed. As suggested by membership theories, the long-term unemployed are not able to price themselves into the labour market because of a decline in human capital, low motivation for active job search, and a stigmatizing perception of the potential employer. When the target and the negotiated profit shares are equal, where the latter is derived as a residual in the bargaining process, unanticipated inflation is zero and the “equilibrium” unemployment rate can be derived (Rowthorn, 1977). Assuming that firms mark-up prices over labour costs, an aspiration gap of zero implies that the real wage (per unit of output) acceptable to firms is equal to the bargained wage. Furthermore, assuming productivity neutrality with respect to unemployment (Blanchard, 1988) an assumption that is widely backed by empirical results (Manning, 1992; Elmeskov, 1993; National Institute, 1995; Arestis and BiefangFrisancho Mariscal, 1994), and equating firms’ target (feasible) real wage share with the bargained real wage share, we may derive the non-accelerating inflation rate of *Note that high levels of unemployment and “normal” levels of capacity utilization indicate a limited substitutability between capital and labour during this period. Consequently, the argument that the slowdown in capital accumulation (from an average of 3.5% in the pre-shock period to 1.4% since 1974) may have no adverse implications for employment provided the reduction in capital accumulation is associated with the adoption of more labour intensive techniques than would otherwise have been, is not valid here.
194
P. Arestis. I. B.-F, Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
unemployment
(NAIRU)
as follows:
#=Yl(w,--p)+y,x+y,ti-y,k+y,lu
(10)
The NAIRU is an increasing function of real benefits, tax and import costs, and worker militancy, and declines with the increase in the capital stock. A positive relationship between real wages and unemployment benefit implies that a fall in the latter results in a decline in unemployment, due to its labour-cost-reducing impact. However, a cut in wage compensation may have other effects that outweigh the favourable cost effects. In response to a fall in real benefits, the unemployed may reduce searching time, and they may have to accept jobs for which they are less suitably qualified than they would accept if the cost of search was lower. The consequent inefficiencies in the labour market may have an adverse effect on the unemployment rate and it may be an empirical question as to which of the effects is more important. We return to this issue in Section 3. An increase in taxes or import costs will, under given demand conditions, increase conflict over income shares and will raise inflationary pressure. The more wage earners resist a fall in their income share, the lower are the tax and import effects on the long-run unemployment rate. Worker militancy affects the bargained real wage and ceterisparibus income distribution, where the latter determines the level of employment through its supply and demand side effects.
3. Empirical investigation3 All estimations reported in this section are undertaken on the computer package MICROFIT4.0, except for the estimation of the cointegrating relationship for the ratio of the long-term unemployed, which was estimated by PCGIVE8.0, since it appears not to be possible to obtain directly the long-run static relationship and the necessary diagnostics in MICROFIT4.0. The first part of the empirical section deals with a discussion on the data. This is followed by the determination of the cointegrating relationships corresponding to Eqs. (9) and (10). The effect of capital shortage on the long-term unemployed is tested in the sub-section that follows, with the third sub-section concerned with the error correction models dictated by the analysis and empirical results in the rest of this section. 3.1. Variable dejinitions and data The model is estimated using seasonally adjusted data over the period 196641 to 1994Q4 which was determined strictly by the availability of data. The data come 3All estimations reported in this section are undertaken on the computer package MICROFIT4.0, except for the estimation of the cointegrating relationship for the ratio of the long-term unemployed, which was estimated by PCGIVE8.0, since it appears not to be possible to obtain directly the long-run static relationship and the necessary diagnostics in MICROFIT4.0.
P. Arestis. I B.-F. Mariscal 1 Structural
Change and Economic Dynamics 9 (1998) 189-204
195
from various issues of Economic Trends: UK Economic Accounts (A and B), Department of Employment Gazette, OECD Business Sector Database, and Annual Abstract of Statistics. We define and discuss briefly each variable in turn that is used in the estimations. The variable “average earnings” was calculated as the ratio of total wages and salaries over total employment, where the former excludes force’s pay. “Average earnings” were deflated by the consumer price index. This index may be appropriate to determine workers’ real consumption wage as an indication of workers’ living standard. Another variable that is part of the profit function in that it determines wage costs, is labour productivity. It is calculated as the ratio of output GDP to total employment. Furthermore, it is argued in the previous section that worker’s productivity depends on income from work and receipts from unemployment. It appears to be impossible to capture appropriately on an aggregate level income from unemployment, since there is a complex system of different types of benefit payments, depending on family size and other circumstances. The variable which was used here is the index of the weekly flat rate unemployment benefit. One of the underlying ideas of the model is that wages are the result of conflict over income shares. Workers are the more successful in achieving wage targets the more militant they are. It is notoriously difficult to measure militancy. We used the number of strikes as a proxy, knowing that a strike is only one (extreme) form of militancy and also knowing that only a fraction of actual strikes are registered. This proxy is only useful if we can assume that the ratio between registered strikes, actual strikes and other smaller conflicts is constant (Paldam, 1989). Other proxies of militancy such as the union mark-up by Layard and Nickel1 (1986), measures of trade union density, or number of days lost are also known to be deficient (Hall and Henry, 1987; Paldam, 1989; Purdy and Zis, 1974). Additionally, it was stressed that conflict over income distribution increases when private sector income falls due to increased demands from the government or from the foreign sectors. We experimented with different variables to capture this effect (such that we included the income, direct, and employers’ tax rates and the real price of imports weighted by the share of imports, or, the real exchange rate separately). The variable that turned out to be most useful is defined as the sum of the direct tax rate as a proportion of the wage bill, employers’ tax rate and the difference between the consumer and producer price index. This variable proxies the gap between the consumer and product wage (Layard et al., 1991). Any increase in this wedge generates the potential for conflict in the labour market. In the theoretical model it was suggested that the long-term unemployed effectively dropped out of the labour market. In line with other studies, we used the ratio of those who are unemployed for more than a year to total unemployment (National Institute, 1995). The unemployment rate was calculated as total unemployment over labour force. Furthermore, it was proposed that, assuming that the substitutability between capital and labour is limited, the decline in investment in the 70s and 80s may have caused capital shortage unemployment in the subsequent years. It is difficult to estimate this effect particularly over the suggested sample period. It is apparent that in the 80s and 90s capital equipment per person is higher than in the 60s and 70s. Using capital stock data, this change in technology requirements cannot be captured satisfactorily. By not explicitly taking this change into
196
P. Ares&
I. B.-F. Mar&al
/ Structural Change and Economic Dynamics 9 (1998) 189-204
account, we may underestimate the necessary rise in capital to increase employment.4 The only source we know that provides quarterly data on fixed gross capital stock is the OECD.’ 3.2. Cointegrating
vectors for real wages and unemployment
We estimate a vector autoregressive regression (VAR) model for the determination of real wages and unemployment as described by Eqs. (9) and (lo), applying the maximum likelihood test and estimation procedures of Johansen ( 1988).6 The Akaike Information Criterion and the Schwarz-Bayesian Criterion (Ltithkepohl, 1991) suggest that the optimal lag length for the VAR is two. The likelihood ratio test of the deletion of the deterministic trend and a dummy for the first quarter of 1975, where real wages were exceptionally high, rejects the null hypothesis with a chi-square statistic (CHSQ) of 36.7 for 12 degrees of freedom. Based on the variable set in Eq. (9), the results of the maximal eigenvalue test, the trace test, the Hannan-Quinn and Schwarz-Bayesian criteria suggest two cointegrating vectors. We proceeded with the identification of the two cointegrating vectors, applying the Pesaran and Shin (1994) approach which relies on imposing non-homogeneous restrictions on the two cointegrating vectors.’ The result is the following two cointegrating relationships: (w-p)=Ip-0.17u+0.010x+0.800k+0.4901u
(lla)
U=0.100x-0.200/z+
(llb)
1.2OOlu
where all variables are in logarithms. The likelihood ratio test of the validity for the overidentifying restriction is CHSQ(6) = 10.17 with a significance level of 0.12, denoting the validity of the restricted system. The first equation describes the average real wage (w-p), determined by labour productivity (fp), unemployment (u), strike activity (x), gross capital stock (k) and long-term unemployment (ZU),defined as the ratio of those who are unemployed for more than 52 weeks to the total number of unemployed. The second equation explains unemployment as a function of strike activity, capital stock, and long-term unemployment. We find proportionality between labour productivity and real wages, suggesting that gains in labour productivity are completely absorbed by increases in real wages, without affecting unemployment. The effect of militancy on wages, Eq. (1 la), and 4We have not treated technical change and capital accumulation separately. This is because in a study of this type that uses highly aggregated data, such a distinction would not pick up any differences that might exist. Proper account for the effect of such a difference should be undertaken in studies which use considerable disaggregation. sA detailed description on the calculation of stocks of fixed capital is provided by the OECD in “Methods Used by OECD Countries to Measure Stocks of Fixed Capital ~ National Accounts: Sources and Methods”, No. 2, Paris 1993. 6The results of the Dickey-Fuller tests suggest that all variables included here are integrated of order one; they are not reported here, but can be obtained from the authors upon request. ‘This approach tests the cointegrated system for general non-linear restrictions and thus uses a generalization of the results by Johansen (1992) and Phillips (1991) for the linear case.
P. Ares&
I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
191
unemployment, Eq. (1 lb), is uncontroversial. As we know, particularly from the 1970s organized labour in the UK reacted to supply side shocks with increased strike activity, thereby causing a decline in output and wage gains above labour productivity growth. Investment in productive capacity affects both real wages and unemployment. At a given level of economic activity, a higher capital stock implies higher capacity utilization, thereby reducing firms’ ability to increase prices and allowing real wages to rise. Furthermore, investment in new capacity increases employment. The long-term unemployed do not exert downward pressure on real wages and their effect on the NAIRU is considerable. Interestingly enough, the tax and import wedge do not affect the cointegrating vectors. The lack of empirical evidence of tax and import effects is due to them being more likely short-run effects. For example, an increase in import cost in relation to the domestic price level improves competitiveness and raises wage pressure by making consumption goods more expensive relative to value added output. If a rise in competitiveness enables firms to mark-up prices to compensate for the higher cost in the long-run, this would imply that they could increase their domestic profit share indefinitely (Layard et al., 1991). Equally, a permanent effect of competitiveness on wage setting implies that workers can resist a decline in real wages indefinitely. A similar argument would apply to increases in taxation and benefits. Our estimates of the NAIRU, which on the whole seem to be in line with those of other studies, are shown in Table 1. The levels of the estimated NAIRU over the full sample period and the period since 1982 are 5.9% (6.0%) and 7.5% (9.4%), where the actual unemployment rates are given in brackets. From 1988 onwards, the estimated NAIRU and the actual unemployment rates are 6.2% and 8.4%, respectively. The estimates are closely related to the actual unemployment rate. It is, however, evident that unemployment has been above its long-run level over the last fifteen years or so. Table 2 quantifies the factors that affected the changes in the NAIRU over particular subperiods. The effect of the contributory variables on the NAIRU can be explained by the changes in the average values of each variable. More concretely, the differences were calculated with the help of the formula: ANAIRU = f3AZ,where Table 1 Period
u%
Source
1969-1973
2.5
1974-1980
3.8
1981-1987
10.1
1988-1990
6.8
other studies Eq. (llb) other studies Eq. (llb) other studies Eq. (llb) other studies Eq. (llb)
NAIRU% (11) (13) (15) (5)
1.665.6 4.6 4.5-7.3 5.3 5.229.9 8.5 3.5-8.1 7.1
Notes: All the information in Table 1, except for the NAIRU estimates of Eq. (llb), are taken from Cromb (1993). The entry “other studies” refers to the total number (in brackets) of other sources which report the NAIRU range referred to in the last column.
198
P. Ares&
I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
Table 2 Period (1982-94)/( (1974-94)/( (1982-94)/(
1966-81) 1966-73) 1974-81)
1U
k
x
3.9 3.3 3.2
-0.4 -0.5 -0.8
-0.8 -0.2 -0.3
A refers to the difference of x between two periods, x refers to the average value of the variable under consideration and 6’refers to the corresponding coefficient as estimated in Eq. (1 lb). The periods are determined as follows: the pre- first oil price shock period (1966-1973), the pre- and oil price shock periods (1966-1981), the period of the two shocks (1974-1981), and the shock and after-shock periods (197441994). For all periods, the rise in the NAIRU is overwhelmingly attributed to the rise in long-term unemployment as the coefficients of Eq. (1 lb) already indicated. Over all periods, long-term unemployment contributed more than 3 percentage points to the increase in the NAIRU. Growth in capital stock reduced the NAIRU by 0.8 percentage points when the after-shock and shock periods are compared. The dramatic fall in militancy since 1982 reduced the NAIRU by 0.8 percentage points in comparison with the 1960s and 1970s which were characterized by high strike frequency. These results raise two questions: why is it that the growth in capital stock contributes so little to reducing the NAIRU, and why is it that changes in the NAIRU are principally explained by long-term unemployment? We suggested earlier that it may be that long-term unemployment and capital shortage are directly related. The argument is that the consequences of capital scrapping in the two shock periods were a rapid increase in unemployment. During the recovery, capacity was constrained so that only a fiction of the displaced workers were re-employed and as the installation of additional capaicty takes a consderable period of time, thsoe workers who remained unemployed during the recovery became the stigmatized and disencouraged long-term unemployed, as suggested by hysteresis models. If this was the case, it could very well be that some of the effect measured by the capital stock, was actually contributed by the long-term unemployment rate. It is thus important that we study the latter more closely. 3.3. The cointegrating relationship for long-term unemployment Big adverse shocks like the oil price shocks cause capital scrapping, or at least a significant slowdown in investment. Since there is a considerable time lag between the rise in unemployment during a recession, the recovery and, eventually, the installation of additional capital, that would allow a return to previous employment levels, displaced workers that could not find employment at the beginning of the recovery must become long-term unemployed. In this section we estimate the effect of a shortage in capital on long-term unemployment. The specification of long-term unemployment is related to the one in Layard et al.
P. Arestis, I. B.-F. Mar&al / Structural Change and Economic Dynamics 9 (1998) 189-204
199
( 199 1). Accordingly, long-term unemployment is described by the level of unemployment since the greater the unemployment level is, the harder it is to get back to work and the longer the unemployment stretch. This assumption is based on the premise that job creation is insufficient to satisfy demand, or, in other words, outflow is less considerable than inflow. This seems to be the case for the UK. However, initially, the ratio of the long-term unemployed falls when unemployment rises. We also include the percentage change in the unemployment rate and the capital stock. We started with the estimation of an autoregressive distributed lag model. The lag structure was determined on the basis of F-tests.’ The resulting relationship is: 8.638
+
0.490~
-
0.836k -
0.074Au
(12)
lU= (4.39)
(14.9)
Estimation period: RESET(1,90)=2.5 ADF(4)=
-4.2
(4.4)
1967Q4-199444 NORM(2)=4.0
(2.6) WALD( 3) = 293
AR( 5,86) = 0.6
HET(34,56)=0.9
DF= -4.3
All coefficients are correctly signed and significant. The significance of the explanatory variables was tested with the WALD test and the residuals were tested for fourth-order autocorrelation (AR), heteroscedasticity (HET), autoregressive conditional heteroscedasticity (ARCH), functional form misspecification (RESET) and normality (NORM) and none of the tests was significant at the usual 5% significance level. The augmented Dickey-Fuller (ADF) and the Dickey-Fuller (DF) tests, indicate that this is a cointegrating relationship. ’ Capital stock is a significant variable with a reasonably high effect on long-term unemployment, giving some support to our hypothesis. It was also considered that increased mobility should reduce unemployment duration (Benoit-Guilbot, 1994). This effect was measured by a mismatch variable, which however did not turn out to be significant. We tested for the effect of real unemployment benefits on long-term unemployment and this was never significant, either. The duration of unemployment benefits could also be an important factor in the determination of long-term unemployment. However, we could not test for this due to the unavailability of time series data. But it should be noted that a relevant OECD study (OECD, 1991) does not find an apparent overall correlation between the level or the duration of unemployment benefits on long-term unemployment.”
‘For further details see PC-GIVE version 8.0. 9All variables except for Au, which is I(O), are Z( 1). The inclusion of an I(0) variable in a set of Z( 1) variables does not affect the cointegrating relationship (Charemza and Deadman, 1992). This variable was included due to previous economic reasoning. “Worker characteristics such as education, age and skill are important factors in the determination of long-term unemployment (Portugal and Addison, 1995). These factors are not included in this study, largely due to the absence of appropriate time series data.
200
P. Ares&
I. B.-F Mariscal/
Structural
Change and Economic Dynamics 9 (1998) 189-204
3.4. Error correction models
Following Engle and Granger (1987), we estimate the error correction models corresponding to the long-run relationships for nominal wage acceleration, the unemployment rate and the long-term unemployment rate. The novelty of the short-run wage equation is that we employed the asymmetrical error correction model by Granger and Lee (1989) and Granger and Swanson (1996), who suggest that adjustment to the long-run can be thought of as being captured by different speeds back to the attractor, which in our case is the NAIRU. The theoretical model suggests that workers are concerned about their relative position in the wage structure and resist nominal wage cuts. Nominal wages are sticky downwards, while upward wage adjustments will be readily sought after. In order to test this hypothesis, the error correction term (ECM) of the long-run unemployment equation was plit into its positive and negative deviations from the mean. With respect to the above hypothesis, we would expect that the coefficient of adjustment is significantly lower when unemployment is above its equilibrium level (positive deviation) than when it is below it (negative deviation). In order to test for the equality of the coefficients, a Wald test (WALD IQ) was used along with the ECM denoting the negative deviations (ecmu-) and the ECM with the positive deviations (ecmu+). The estimated error correction models are reported in Table 3. The error correction terms are correctly signed and significant in all three models. None of the models seem to have obvious diagnostic problems. We begin our discussion of the empirical results of Table 3 with the nominal wage acceleration as described by Eq. (S), remembering, however, the long-run proportionality between real wages and productivity. We find that real wage growth, real unemployment benefit growth, unemployment acceleration as well as the growth of taxes and import cost affect nominal wage acceleration. We could not find a significant effect of capital stock growth on nominal wage acceleration, suggesting that additional investment does not mitigate inflationary pressures in the short-run. The Wald statistic rejects the null hypothesis of equality of the coefficients of the ecmu+ and the ecmu- with a CHI ‘( 1) = 0.82 and a significance level of 0.36 indicating that there is asymmetry in the adjustment process. The coefficient on ecmu+ is, as expected, significantly lower than the coefficient on ecmu-. We may therefore conclude that there is downward stickiness in nominal wage adjustment when unemployment is above its long-run level.” The short-run equation that corresponds to the NAIRU in Eq. (1 lb) is only determined by its own dynamics, the error correction terms and a dummy for 1974, which captures the effects of the three-day working week at the beginning of that year. The growth rate of the ratio of long-term unemployed moves together with unemployment growth. Increases in unemployment growth reduce long-term unem-
“We also estimated a symmetric error correction model for wage acceleration. The effects of the individual variables remained stable, however, the diagnostics were slightly better in the asymmetric model.
P. Ares& Table 3 Dynamic
I. B.-F Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
201
relationships
Estimation variable
period
1967Q2-1994Q4 AA(w-Ip)
1967Q3-1994Q4 AU
1967Q331994Q4 Alu
constant
-0.111
-0.135
-0.015
(2.9)
(3.5)
ecmu, _ 1
(4.9) - 0.020
ecmu,?
,
(4.1) -0.0117
ecmu,
3
(2.1) - 0.024 (3.1)
ecmw,_s ecmw, _ 4
-0.052
- 0.084 (2.9)
NW,-A-z
(7.0) - 0.227 (10.2) 0.093
WJ-d-4
(3.3) 0.097
ecmlu, _ 1
(3.5)
AU,-1
0.876 (18.4) 0.178
1.023 (10.1)
(1.9) AAu, _ i
-0.169 (5.1)
AAu,_,
-0.251
AAu,_,
(2.3) -0.367 (3.6)
NW-P-k~Ip),-, A(w-p-&),-s
-0.930 (11.0) -0.343
Ati,_,
(4.2) 0.072
D74.1
(1.9) 0.157 (5.4)
R2 AR RESET NORM( 2) HET WALD IQ
0.64 1.6 (497) 1.2 (1100) 2.1 2.3 (1109) 0.8
0.78 2.0 (4100) 1.4 (1103) 3.8 2.8 (1108)
0.65 1.9 (4102) 1.2 (1105) 4.0 2.6
Note: R2 is the adjusted coefficient of determination and the meaning of the diagnostic tests was explained in the text. The values in brackets under the coefficients denote the t-statistic and the values in brackets following the diagnostics denote the degrees of freedom. All tests are F-tests, except for the test for normality and the Wald IQ tests, which are chi-square tests with two and one degrees of freedom, respectively.
202
P. Are&,
I. B.-F. Mariscal/ Structural Change and Economic Dynamics 9 (1998) 189-204
ployment. Adjustments a year.
to the long-run relationship
are achieved after just over
4. Summary and conclusions We have presented an aggregate wage model for the UK which combines wage theories that are concerned with explaining failures in the supply side of the labour market. This model is extended by incorporating demand side aspects; more precisely, the capital shortage hypothesis. The model provides us with long- and short-run wage and unemployment equations, which we proceed to estimate. The empirical section begins by identifying two cointegrating vectors, one describing real average wages and the other the NAIRU. It also finds it necessary to investigate the possibility of a cointegrating relationship for long-run unemployment, which is estimated, before the appropriate error correction models are investigated. We find that real wages are explained by labour productivity, long-term unemployment, the unemployment rate, worker militancy and the capital stock. As there is homogeneity between real wages and labour productivity, the NAIRU is determined by long-term unemployment, worker militancy and capital stock. Tax wedge effects, import cost effects and real benefits are not found to be relevant in the long-run, but they prove to be important in the error correction models. For the long-run analysis, we decompose the impact of each explanatory variable on the NAIRU. We find that the greatest effect comes from long-term unemployment and to a minor degree from capital shortage and militancy. Since we think that capital shortage may particularly affect long-term unemployment, we test for the effect of capital shortage on long-term unemployment and find a considerable and significant effect. For the error correction model of nominal wage determination, we find significant asymmetric responses. Nominal wages adjust when actual unemployment is above the NAIRU much more slowly than when unemployment is below the NAIRU, which confirms our theoretical priors. The policy implications of our findings are that shortages in demand may have persistent effects on employment. Although, programmes that improve the skills of displaced workers can be important, policies that aim to enhance investment may be necessary. These policies should aim at reducing long-term real interest rates, producing a high rate of return on productive investment and creating a stable financial environment to support the real sector of the economy.
Acknowledgements
We are grateful to the editors of Structural Change and Economic Dynamics, especially to Harald Hagemann, to an anonymous referee and to Jim Marley and Peter Howells for their helpful comments. Also to Gary Young of the NIESR for providing data on a number of variables.
P. Ares&
I. B.-F. Mar&al 1 Structural Change and Economic Dynamics 9 (1998) 189-204
203
References Acemoglu, D., Scott, A., 1994. Asymmetries in the cyclical behaviour of UK labour markets. The Economic Journal, 1303-l 323. Arestis, P., Skott, P., 1993. Conflict, wage relativities and hysteresis in UK wage determination. Journal of Post Keynesian Economics, 3655386. Arestis, P., Biefang-Frisancho Mariscal, I., 1994. Wage determination in the UK: Further empirical results using cointegration. Applied Economics, 417-427. Arestis, P., Biefang-Frisancho Mariscal, I., 1998a. Conflict elements in UK wage determination: Keynesian, Post-Keynesian and New Keynesian theoretical and empirical considerations. In: Rotheim, R. (Ed.), New Keynesian Economics and Post Keynesian Alternatives. Routledge, London. Arestis, P., Biefang-Frisancho Mariscal, I., 1998b. Conflict, effort and capital stock in UK wage determination. Empirica, in press. Bean, C., 1989. Capital shortage and persistent unemployment. Economic Policy, 12-53. Benoit-Guilbot, O., 1994. Why are there so many unemployed in the EU? In: Benoit-Guilbot, O., Gallie, D. (Eds.), Long-term Unemployment. Pinter Publishers, London. Blanchard, O.J., 1988. Comment on Nickell: The supply side and macroeconomic modelling. In: Bryant, R.C. et al. (Eds.), Empirical Economic for Independent Economies. Brookings Institution, Washington. Blanchard, O.J., Summers, L.H., 1986. Hysteresis and the European unemployment problem. In: Fischer, S. (Ed.), NBER Macroeconomics Annual 1986. MIT Press, Cambridge, MA. Bowles, S., Boyer, R., 1990. A wage-led employment regime: Income distribution, labour discipline and aggregate demand in welfare capitalism. In: Marglin, S., Schor, J. (Eds.), The Golden Age of Capitalism. Clarendon Press, Oxford. Charemza, W.W., Deadman, D.F., 1992. New Directions in Econometric Practice. Edward Elgar, Aldershot. Cromb, R., 1993. A survey of recent econometric work on the NAIRU. Journal of Economic Studies, 27-51. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: Representation, estimation and testing. Econometrica, 251-276. Elmeskov, J., 1993. High and persistent unemployment: Assessment of the problem and its causes. Economics Department Working Paper No. 132. OECD, Paris. Granger, C.W.J., Lee, T.H., 1989. Investigation of production, sales and inventory relationships using multicointegration and non-symmetric error correction models. Journal of Applied Econometrics, 145-159. Granger, C.W.J., Swanson, N., 1996. Future developments in the study of cointegrated variables. Oxford Bulletin of Economics and Statistics, 5377553. Hall, S.G., Henry, S.G.B., 1987. Wage models. National Institute Economic Review, 70-75. Hicks, J., 1975. The Crisis in Keynesian Economics. Basil Blackwell, Oxford. Holzer, H.J., Montgomery, E.B., 1993. Asymmetries and rigidities in wage adjustments by firms. Review of Economics and Statistics, 3977408. Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 231-254. Johansen, S., 1992. Identifying restrictions of linear equations. mimeo, Institute of Mathematical Statistics, University of Copenhagen. Keynes, J.M., 1936. The General Theory of Employment, Interest and Money. Macmillan, London. van de Klundert, T.C.M.J., van de Schaik, A.B.T.M., 1990. Unemployment persistence and loss of productive capital: A Keynesian approach. Journal of Macroeconomics, 3633380. Layard, R., Nickell, S. Jackman, R., 1991. Unemployment. Oxford University Press, Oxford. Layard, R., Nickell, S., 1986. Unemployment in Britain. Economica, S12llS169. Lindbeck, A., 1993. Unemployment and Macroeconomics. MIT Press, Cambridge, MA. Lindbeck, A., Snower, D.J., 1986. Wage setting, unemployment and insider-outsider relations. American Economic Review, 2355239. Ltithkepohl, H., 1991, Introduction to Multiple Time Series Analysis. Springer-Verlag, Berlin.
204
P. Arestis, I B.-F. Mar&al
1 Structural
Change and Economic Dynamics 9 (1998) 189-204
Malinvaud, E., 1986. The rise of unemployment in France. Economica, S1977S217. Manning, A., 1992. Productivity growth, wage setting and the equilibrium rate of unemployment. Discussion Paper No. 63, Centre for Economic Performance, London School of Economics, London. National Institute, 1995. National Institute Model 12. Nickell, S., 1987. Why is wage inflation in Britain so high? Oxford Bulletin of Economics and Statistics, 103-129. OECD, 1991. Employment Outlook. OECD, Paris. OECD, 1994. The OECD Jobs Study, Part 1. OECD, Paris. Paldam, 1989. A wage structure theory of inflation, industrial conflicts and trade unions. Scandinavian Journal of Economics, 63-81. Pesaran, M.H., Shin, Y., 1994. Identification of and testing for cointegrating relations with general non-homogeneous restrictions. mimeo, Department of Applied Economics. Phillips, P.C.B., 1991. Optimal inference in cointegrated systems. Econometrica 59, 2833306. Portugal, P., Addison, T. 1995. Short- and Long-term Unemployment: a Parametric Model with TimeVarying Effects, Oxford Bulletin of Economics and Statistics, pp. 2055227. Purdy, D.L., Zis, G., 1974. Trade unions and wage inflation in the UK: A reappraisal. In: Laidler, D. Purdy, L.D. (Eds.), Inflation and Labour Markets. Manchester University Press, Manchester. Rowthom, R.E., 1977. Conflict, inflation and money. Cambridge Journal of Economics, 215-239. Rowthorn, R.E., 1995. Capital formation and unemployment. Oxford Review of Economic Policy, 26639. Shapiro, C., Stiglitz, E., 1984. Equilibrium unemployment as a worker discipline device. American Economic Review, 433-444. Sneessens, H.R., Dreze, J.H., 1986. A discussion of Belgian unemployment, combining traditional concepts and disequilibrium econometrics. Economica, S89-S119. Summers, L., 1990. Understanding Unemployment. MIT Press, Cambridge, MA. Wren-Lewis, S., 1992. Nominal inertia and Keynesian effects. NIESR Discussion Paper No. 174.