The focus of Gregor\ and Duncan (1979) was on explaining increased unemployment for the aggregate economy from the mid 1970s Their explanadon
38. Technical progress was calculated as ihe fined values of a total factor producin-ity senes regressed against a time pohnomial of degree 6.
argument was that productivity movements were not consistent witli real wage movements, and that the price of labour relative to capital fell at times when the
demand for labour fell. This was argued to suggest that the emphasis on labour-capital substitution, resulting f r o m increased real wages, by Jonson et al (1978), Gruen (1978), Snape (1979) and Holmes (1979) was misplaced. Funher, the relation between output and productivity movements were contrarN- to those implied by the existence of a
downward sloping marginal product curve for labour. Critiques of Gregory and Duncan by Andersen (1979), Blandy (1979). Gruen (1979), Higgins (1979) and Snape (1979) argued that issues of aggreganon. interpretation of the data and the appropriate implicit labour demand specificanon meant that Gregor>' and Duncan's (1979) conclusions were not clear cut.
Phipps (1983) undertook a disaggregated study of labour demand for the
Australian economy and found differences between models of the demand for labour at the aggregate level and for component sectors, including manufacturing. Using a
sample from 1962-63 to 1976-77 with the wage and quandty specification, Phipps (1983) found that while the coefficient on the real wage was not well determined at the aggregate level, there were significant and negative coefficients on real wages for manufacturing, mining, construcnon, and electricity and gas. The wage elasdcity of labour demand for manufacturing, conditional on output, was -0.52, while the elasticity with respect to output was 0.98. The differences in the results between the two levels of aggregation suggest that there is information to be gained from disaggregation. This result was similar to that of Johnston e: al (1978) who. estimating over the whole economy, found that wages were not significant in a specification that included output of the economy.
Newell and Symons (1985) in their sur\'ey of labour demand in OECD countries obtained a significant inverse relationship between the demand for labour and the
Study including Australia.^^ Recent work in this area has been by Russell and Tease (1988) who made use of the wage and quantity specification for the aggregate economy for 1969(3) to 1987(4). They found a significant inverse reladonship between the
demand for labour and the product wage, with a long run elasdcity of labour d e m a n d of -0.61, condidonal on the level of output. In explaining the recessions of 1974-75 and
1982-83 they found that "... the large rise in real wages ... had an i m p o n a n t bearing on the fall of employment in those years" (p. 16). They also argue that falls in real wages since 1983 have been i m p o n a n t to the growth of employment in the aggregate
economy.
S y m o n s (1985b) estimated the demand for labour for Australian manufacturing between 1969(3) and 1981(3), based on the short-run model of profit maximising labour demand presented in (2.22) and imposing constant returns to scale. Significant and
negative coefficients were obtained for the product wage and material prices. The long- run elasticity of labour demand with respect to the product wage was -0.91, and the elasticity for material prices was -0.48. The wage elasticity was not significandy different when estimated over sub-samples and the model passed tests of stability. Analysis with the model indicated that between 1970 and 1981 the negative impact on e m p l o y m e n t of increases in the product wage were not offset by the positive effect on labour demand of increases in the capital stock. Material price declines had a small positive effect on the demand for labour.
R i m m e r (1989) estimated the demand for labour in Australian manufacturing, specifically addressing specifically the role of the wage explosions of 1973-74 and 1981-82. A specification of labour demand based on CES technology is used for a sample f r o m 1962-63 to 1985-86. The parameters determining capital intensity and the degree of substitution were allowed to vary over the sample period. It was postulated that the wage explosions were the major cause of changes in the sector and expected
39. Bean et al (1987) estimated the demand for labour using a specification ihat included demand variables. The results are discussed in Chapter Five.
that they would reduce the level of labour intensity in production, and increase the degree of substitution between capital and labour. The share of labour in value added was the dependent variable and the technical progress was modelled as a function of a linear trend, deviations of actual around potential output and a constant. The other major explanatory variable was product wages.
The results of estimation indicated that contrary to expectations labour intensity increased over the sample period and the degree of substitution between labour and capital fell. This result held when allowance for disequilibrium adjustment was made using d u m m y variables for 1974-75 and 1982-83. Explanations of the unexpected results were based on the effects of compositional changes in the Textiles, Clothing and Footwear and Transport equipment industries arising f r o m increased border protection in these industries increasing increasing labour intensity at the disaggregate level being large enough to also do so to a significant extent at the aggregate level.
2.6.3 Policy and the D e m a n d for Labour
The last two sub-sections have reviewed an empirical debate that, while having taken at least two distinct forms, has lasted since the publication of Keynes' General Theorv. The issue appeared initially to be a simple one: whether real wages rose or fell with employment. In this section, discusses this aspect of the debate and reviews the empirical evidence following the cyclical approach.
Drobny (1988) and Michie (1987) are recent examples of the interpretation of Keynes' work which concentrate on the relationship between wages and the business c y c l e . a p p r o a c h is to test for contemporaneous correlations between real or product wages and proxies of the business cycle (usually output or employment). Michie arsued that:
40. Bils (1987) c o n c e n t r a t e s on the relation between marginal cost and prices over the business cycle. Sheen ( 1 9 9 0 ) an interpretation of Australian data following this m e t h o d o l o g y .
the cyclical wage literature is concerned not with a full causal system of wages and product prices via a simultaneous system, but simply with the net result of forces which are at work on the direction of m o v e m e n t of these variables, and with the resulting cyclical behaviour of product wages ... the issue in question is how the wage moves in relation to cyclical output (p. 39)
Michie (1987) examined the cyclical relationship between output and the
product wage between 1950 and 1980 for the UK, USA, France, Canada, Germany and Japan. The data is transformed as deviations f r o m a 5 year moving average. The
(detrended) product wage is then regressed on (detrended) output. Michie (1987) argued that the disparate nature of the results, including a sensitiveness to the inclusion of additional variables and lags, indicated no clear pattern between countries of an inverse relationship between real wages and output."^^ Michie (1987) also used hours worked as a proxy variable for the business cycle. This specification was close to a simple labour demand schedule - although product wages were the dependent variable (in a staristical sense).
As part of the study and to demonstrate the fragility of the coefficient signs in the regression Michie (1987) estimated a model with (detrended) output and (detrended) employment as the explanatory variables of (detrended) real wages. This yielded an inverse relationship between employment and wages for all six countries, with the coefficient on hours worked significant at the 5 per cent level for four countries; however, the sign on output variables changed to be positive and declined in
significance when jointly estimated with employment and real wages as the dependent variable. The model estimated is a rewritten version of the wage and quantity
specification, presented in equation (2.21a). Michie's results are in fact consistent with those that would be expected under the neoclassical framework.
41. There is some ambivalence in the interpretation of ihe results obtained by Michie. Negative coefficients were obiained for 5 of the 6 countries when employed persons was used as the dependent variable (including the most significant of the regressions), and all but the U S A exhibited an inverse relationship between product wages and hours worked (3 of the 5 c o u n t n e s were significant at the 5 per cent level while the U S A was not). This is also noted in Symons (1989).
Models similar to that estimated by Michie (1987) were estimated using output, hours worked and nominal wages deflated by product prices for Australian
manufacturing for 1960-61 to 1984-85 and are presented in Table 2.1.^2 y ^ g ^^gd in e s n m a n o n are detrended as d e v i a n o n s f r o m five year moving averages.
The results indicate an inverse relanonship between the product wages and both output (column (1)) and employment (column (2)). The results are sensitive to the
specificanon of the model. Output is significant and negarively signed in column (1), as is employment in column (2). The joint estimation of employment and output presented in column (3) affects the sign and significance of output and employment, with the significance of output in column (3) lower than that obtained in column (1).
An altemanve method of examining the correlarion between real wages and economic activity (using output as the measure of activity) is to take first differences of the logs of the variables and to regress one on the other, including a trend variable as a second explanator)' variable to take account of any trend behaviour remaining after differencing. The results of estimating such a model using manufacturing data for 1954- 55 to 1984-85 are presented in Table 2.2. They indicate a significant inverse
relationship between output and product wages.
The other form of the empirical debate has been to determine the role of real wages in determining the demand for labour. As is pointed out by Symons (1989) in his review of the book on the subject by Michie (1987), it is not clear that the question as to whether real wages do or do not move with the trade cycle is an interesting one. The sign of the wage/employment relationship depends on the source of shocks to the economy, and policy is really interested in the size of the wage effect on employment. The later policy question can only be answered by a consideration of the existence and stabilitv of a labour demand function.
42. E s u m a u o n reproduced the techniques applied by Michie as closely as possible. Estimation was undertaken using natural numbers as opposed to logarithms.
Table 2.1; Regression Results using Variables Identified by Michie
Dependent variable - deviations of product wage f r o m 5 year moving average
Lag (1) (2) (3) Output t+1 -0.63 [-2.81* -0.42 [-0.9] t -0.62 [ - 2 . 4 ] ' 0.08 [0.2] t-1 0.25 [1.1] -0.22 [-0.5] E m p l o y m e n t t+1 -0.30 [-1.6] 0.09 [0.2] t -0.75 [-3.9]* -0.89 [-1.2] t-1 0.37 [1.9] 0.56 [1.3] Trend -0.007 [-4.0]^ -0.004 [-1.4] -0.006 [-2.0]* Constant 0.27 [5.0]" 0.14 [2.1]* 0.2 [1.9] 0.6 0.8 0.8 Source: Appendix A. " significant at five per cent.
Table 2.2: Relationship between product wages and output
Dependent variable - first difference of log of product wages
Output -0.51 [ - 2 . 9 ] ' Trend -0.002 [-2.0]* Constant 0.08 [3.7]* R2 0.26 Autocorrelation D u r b m - W a t s o n 1.75 L M x ' ( l ) 0.2 Normality Jarque-Bera x~(2) 0.15 Source: Appendix A. * significant at five per cent.
Symons (1989) made four further observations concerning estimating the demand for labour. First, genuine measures of capital stock and technical progress should be used, as simple trend terms are not sufficient to capture their effects. This point was illustrated in Symons and Layard (1984). Second, the product wage should be correctly measured, using the appropriate deflator. If a gross measure of output prices is used, the price of materials should be included in the model. The relevance of deflators of the nominal wage was illustrated in Nickell and Symons (1990). Third, the dynamic structure of the model should be taken into account, suggesdng that lagged variables should be at least included in inidal specification. As Symons (1989) shows, a
contemporaneous correlation between the demand for labour and wages can be either positive or negadve, and srill be consistent with an underlying inverse relationship between the demand for labour and wages. Fourth, a relationship between the demand for labour can be idennfied in esnmation when there is sufficient variability in the data. Thus, many of the difficulties of idennfying an inverse relationship using data prior to the late 1960s can be attributed to the trend like growth of employment and wages up to that period.
2.7 Summary
This chapter has presented a range of models of the demand for labour within by the Keynesian and neoclassical frameworks. The models range from the "strong" Keynesian model, in which markets are in disequilibrium and prices play no direct pan in the
labour market, to the neoclassical model in which Keynesian aggregate demand has no role.
Three basic models of labour demand were presented in the chapter. First, the invened production funcnon, in which output, conditioned by the capital stock (or a trend variable) determines the demand for labour. This model is common to all
frameworks. The second was the wage and quandty model which could be interpreted in a number of wavs and in which labour demand was determined by output and
product wages. The third model was of labour demand derived f r o m neoclassical profit maximisation, in which demand played no part, and the demand for labour was
determined by the product wage (and material prices) and conditioned on the capital stock.
In addressing the empirical debate, the chapter described two approaches to the relationship between wages and employment. The first was concerned with the
relationship between contemporaneous movements in wages and employment. The second involved estimating the demand for labour and the role of demand factors and the real wage.