RDP 8806: Employment, Output and Real Wages 3. Employment, Output and Wages: The Evidence from Previous Studies

The sharp rise in inflation and the deterioration of employment growth in the early 1970s stimulated an extensive debate in Australia on the role played by the increase in real wages. The simultaneous rise in inflation and unemployment in the Western economies at this time also focused international attention on this issue. This section will outline the debate about the real wage-employment nexus in Australia. A summary is provided in Table 1.

TABLE 1 SAMPLE OF AUSTRALIAN EMPIRICAL WORK ON EMPLOYMENT FUNCTIONS
Author(s) Approach Relevant Findings
1970s Studies
Higgins and Fitzgerald (1973) Employment is derived by inverting an estimated production function. Employment is largely determined by output, capacity utilisation and the capital stock.
Freebairn (1977) Surveyed the Australian empirical literature. The long-run elasticity of employment with respect to output in a number of studies ranged between 0.65 and 0.70. He suggested that the real wage elasticity was around −0.5.
Sheehan (1978) Used an earlier equation reported by Gregory and Sheehan (1973) to examine the role of output and the real wage in the fall in employment in 1974–1977. The equation predicted employment well between 1974 and 1977 despite the absence of a real wage term. The addition of a real wage term added little to the equation.
Johnston, Campbell and Simes (1978) Regressed employment, unemployment and investment equations on measures of capacity utilisation and the real wage overhang (which is essentially the gap between the growth in real wages and productivity). The rate of unemployment – but not employment – was significantly related to the real wage overhang. The finding of a significant, positive relationship between investment and the real wage overhang was, however, cited as strong evidence of capital for labour substitution.
Coghlan (1978); Dixon, Parmenter and Sutton (1978) and Jonson, Battellino and Campbell (1978) Conducted simulations of macroeconomic models to trace the effects of movements in real wages. Jonson, et.al. used results of the RBA76 project, but did not conduct full-model simulations. Results suggested that movements in real wages have a significant, negative correlation with employment.
Gregory and Duncan (1979) Examined the post-1974 relationship between output and employment to see if it behaved in a way consistent with a neoclassical interpretation of the effect of a rise in real wages. Productivity did not behave in a way consistent with neoclassical interpretation. The authors concluded, therefore, that the key to growth in employment at that time was stronger output.
1980s Studies
Schelde-Andersen (1980) Cross-section analysis for a range of industries. Both output and real wages play an important part in determining the level of output. Real wage and output elasticities were of the same order of magnitude but of opposite sign.
Symons (1985) Estimated a standard neoclassical demand for a labour function for the manufacturing sector. The data could not reject such a specification. The real wage elasticity was significant and large, varying between −0.75 and −1.07.
Pissarides (1987) Estimated a multi-equation model of the Australian labour market. The real wage parameter was negative and significant in the employment equation with an elasticity of −0.79.
EPAC (1988) Reported the results of simulated falls in real unit labour costs (RULC) in a number of macro-econometric models. Results suggested that a 2 per cent fall in RULC would lead to a 1.5 per cent rise in employment.

The fall in employment in 1974/75 roughly coincided with a slowing in GDP growth and a rise in real wages. Consequently, much of the early debate focused on whether the slowing of activity or the rise in real wages caused the fall in (and subsequent weakness of) employment. The resolution of this issue was important because of its implications for policy. If the weakness in employment was largely “Keynesian” (i.e., due to deficient demand), then the problem could be alleviated by the appropriate expansion of policy. If, on the other hand, excessive real wage growth had induced the weakness in employment (i.e., unemployment was “Classical”), then a reduction in real wages was the appropriate policy to stimulate employment (and output).

The issue was clouded by the lack of definitive empirical evidence at the time. This was probably due to the lack of substantial variation in real wages prior to 1973. Initially, many studies failed to find a significant role for real wages in labour demand decisions. As a result, a number of authors advocated expansionary policies (rather than wage restraint) as a solution to the fall in employment in the mid 1970s. Among these were: Hughes (1977), Sheehan (1978), Sheehan, Derody and Rosendale (1979) and Gregory and Duncan (1979). Sheehan (1978) noted that an employment equation estimated by Gregory and Sheehan (1973) predicted employment reasonably well between 1973 and 1977 even though there was no real wage term in the equation. A real wage term, when added to the specification, was only marginally significant. Sheehan argued that a cut in real wages would not stimulate employment. This result was based on a simple stylized model of the Australian economy. The model reached this conclusion because the contractionary effect of a cut in real wages (via lower household expenditure) offset the positive influence of lower real wages on labour demand decisions.

Sheehan, et. al. (1979) surveyed the literature more widely and found little evidence of a significant relationship between employment and real wages. They cited evidence from a number of early labour market studies, including Higgins and Fitzgerald (1973), Valentine (1975), Gregory and Sheehan (1973) and Clark (1976); none of these found a significant role for real wages in the employment decision. Gruen (1979) and Holmes (1979) argued that Sheehan, et. al. were highly selective and that a more extensive survey would not support their conclusion.

Gregory and Duncan (1979a) examined the post-1974 relationship between output and employment. They concluded that the relationship had not behaved in a way consistent with a neo-classical interpretation of the effect of a rise in real wages. (In such a model, a rise in the real wage will lead to a substitution of capital for labour and a rise in measured productivity.) Gregory and Duncan found that productivity growth actually slowed following the rise in real wages in the early 1970s. This led them to conclude that the key to growth in employment at that time was stronger output growth.

However, while a body of economists persisted with the view that real wages were not a significant factor in the fall in employment in the mid 1970s, there was a growing body of opinion which argued in the opposite direction. These opinions were especially apparent in the statements of the “official sector” and the OECD.

As an example, work undertaken by the OECD (1977) developed the concept of the “real wage overhang” to describe the surge in real wages, relative to productivity, in many countries around that time. The OECD argued that its examination of the experiences of member countries showed an important role being played by real wages in the weakening in employment. Australia was seen as a clear example of the phenomenon. In its 1978 survey of Australia, the OECD noted that while the contraction in output in Australia in the mid 1970s was milder than most other OECD countries, the fall in employment was much more severe. This was attributed to the large gap between growth in real wages and productivity in Australia; which, in 1975 was the largest of any OECD country.

These arguments were also set out by The Reserve Bank (1977), The Commonwealth Treasury (see Budget Statement No.2 1977/78) and in Commonwealth submissions to various National Wage Cases.

In Australia, Johnston, Campbell and Simes (1978) found that the rate of unemployment was significantly related to a measure of the real wage “overhang”, (which is essentially the gap between growth in real wages and productivity); indeed, the development of this “overhang” caused the usual cyclical relationship between unemployment and activity to break down. Johnston, et. al. went on to estimate an employment equation and found that a measure of the real wage overhang had the expected negative sign but was not significant. They did, however, find a significant positive relationship between real wages and investment and interpreted this as strong evidence of capital for labour substitution (in response to high real wages).

Other studies also found an important role for real wages in the employment decision. Freebairn (1977), in a survey of the Australian literature at the time, found a significant role for both output and real wages in employment equations. The long-run elasticity of employment with respect to output in a number of the studies surveyed ranged between 0.65 and 0.70. The long-run real wage elasticity exhibited substantial variation between studies but Freebairn suggested an elasticity of around −0.5. Corden (1978), examined the downturn of the mid-1970s and argued that slowing in activity could not entirely explain the fall in employment; real wage increases were equally as important.

The results derived from a range of macro-econometric models also tended to support the view that real wages had a significant effect on employment. Coghlan (1978), using the NIF-7 model of the Australian economy, found that if money wages had grown at below trend between 1974 and 1977 and if the real wage overhang had been eliminated, employment and economic activity would have been significantly higher than that experienced over that period. (It should be noted that Coghlan placed a number of caveats on these results.) Estimates from the RBA76 project, reported in Jonson, Battellino and Campbell (1978), showed that if the real wage rose above the (imposed) marginal product of labour then employment would fall. Dixon, Parmenter and Sutton (1978) found that a 1 per cent fall in real wages in the ORANI model would lead to a rise in employment of around 1/2 of 1 per cent. Several problems with this simulation have been noted, however, including the lack of a link between real wages and aggregate demand and the fact that the elasticity of substitution between capital and labour is imposed rather than estimated.

The papers discussed above highlight the conflicting views in the mid 1970s on the relationship between real wages and employment. However, as the decade unfolded, the data allowed sharper estimates of the effects of real wages on employment and studies increasingly found an important link between these variables. It would be reasonable to conclude that, at the end of the decade, the balance of opinion was swinging to the view that real wages and employment are significantly related[3].

Studies conducted in the 1980s added support to this conclusion. Schelde-Andersen (1980), using cross-section evidence for a range of industries, found that both output and real wages played an important part in determining the level of employment. Schelde-Andersen found that the output and real wage elasticities were of the same order of magnitude but of opposite sign.

Symons (1985) estimated a standard neo-classical demand function for labour in the manufacturing sector. Symons found that the data fitted such a specification reasonably well, with a large and significant effect evident for real wages on employment. Although the real wage elasticity varied across time periods it was consistently large, varying between −0.75 and −1.07. As in earlier studies, adjustment to real wage changes was reasonably slow, with the mean lag in the preferred specification being 5.3 quarters.

Pissarides (1987), in a more general study, found similar results, with the long-run elasticity of employment with respect to the real wage estimated at −0.79.

A more recent study by EPAC (1988) reports the results of simulated falls in real unit labour costs in a number of Australian macro-econometric models. The results, in general, suggest that a 2 per cent fall in real unit labour costs would lead to a 1.5 per cent rise in employment in the medium term; implying an employment/real wage elasticity of −0.75.

Footnote

Hughes (1980) and Sheehan (1980) continued to oppose this view. [3]