The Chancellor and the Treasury have recently argued that the current high levels of unemployment in the UK are the result of real wages which are "too high'. Research Fellow Robert Gordon, speaking at a March 22 lunchtime meeting, argued that the data no longer support this explanation of unemployment. The lunchtime meeting was followed by an afternoon workshop devoted to an exploration of the relationship between real wages and unemployment. The Employment Market Research Unit of the Department of Employment provided financial support for the meeting and workshop.

Gordon noted that recent discussion of macroeconomic policy in Europe has been dominated by the belief that real wage growth has not shown enough downward flexibility in response to the worldwide deceleration of productivity growth since 1973. Wage and price adjustment is conventionally thought to be quite different in Europe and the US, and this difference seems confirmed by the stark contrast between recovery in the US and relative stagnation in Europe.

This "classical' unemployment hypothesis has gained wide acceptance in Europe, Gordon noted, and represents an intellectual obstacle to stimulative aggregate demand policy. If unemployment has risen in response to excessive real wage growth, then an expansion of aggregate demand would merely cause inflation instead of higher employment and output. The case for the classical unemployment hypothesis has been made by Michael Bruno and Jeffrey Sachs. They demonstrated that since the late 1960s the real wage in European manufacturing has risen relative to the average product of labour, i.e there has been an increase in the "wage gap'.

Gordon argued that an examination of the wage-gap data for Europe and the US does not support the classical unemployment hypothesis, at least in the manufacturing sector. In the early 1980s the US manufacturing sector has exhibited a more adverse wage gap than any country in Europe (with some minor exceptions). Gordon's research shows that there is no cross-country correlation between the increases in unemployment and the increase in the manufacturing wage gap since the late 1960s. The wage-gap theory may have been appropriate to explain the experience of the 1970s, when wage rates increased faster than productivity and wage gaps for European countries exhibited a uniform "hump'. These European wage gaps have now declined or disappeared, yet high levels of unemployment remain. Gordon argued that wage-gap explanation is therefore obsolete.

Gordon's research decomposed the quarterly growth rate of wages into various components, each of which can exert its effects with substantial lags. Gordon then investigated whether differences exist across countries in the wage and price adjustment process.
This approach allows explicit modelling of the influence on wage behaviour of supply shocks, the productivity slowdown and tax changes. The wage equation can be used to identify the role of real-wage stickiness and to determine whether this "stickiness' differs across countries.

Gordon found evidence of real wage stickiness between 1973-77 in eight countries. The effect was insignificant for the US but substantial for Europe and Japan. There was a reversal in the effect in 1978-83 in France, Germany, the Netherlands, Norway and Sweden. The overall effect of real-wage stickiness for OECD Europe is positive for 1973-77 but roughly zero for 1978-83.

Gordon noted that wage and price inflation is now stable in the UK and the rest of Europe. One might draw from this the pessimistic conclusion that the level of unemployment sufficient to maintain a stable rate of inflation has risen sharply, and that European economies are now operating at this level. If this were so, policies to stimulate aggregate demand would only reignite inflation.

The European situation, however, could be interpreted in a different fashion. If a high level of unemployment persists for long periods, it may cease to exert downward pressure on the rate of inflation. The Phillips curve relationship between the rate of inflation and level of unemployment would therefore break down.

Gordon argued that the Phillips curve seemed to disappear in exactly this fashion in the US between 1929-41 and in the UK between 1923-38. In the US prices went down when unemployment was rising, and went up when unemployment fell. The rate of change of prices did not continue to fall in response to a decade of high unemployment, as the Phillips curve would suggest. In the UK unemployment remained at 10% without any apparent tendency for a continuing deceleration of wage and price changes. Analysis of the data from these periods might suggest that the unemployment rate at which the economy can be operated at a stable rate of inflation - the NAIRU - was 15% in the US and 10% in the UK. The analysis would also have suggested that aggregate demand expansion to reduce unemployment below these figures would be unsustainable. The experience of the post-war period shows that much lower rates of unemployment were possible, despite this statistical evidence. This, Gordon argued, may also be true today and may explain the strong US growth after the tax reductions.

CEPR Associate Programme Director Grayham Mizon chaired the afternoon workshop. Richard Layard (LSE) presented the first paper, entitled "The Causes of British Unemployment.' This was based on joint work with Steve Nickell (Oxford and CEPR). Their model allowed both "classical' unemployment, due to wages which were too high, and "Keynesian' unemployment, due to deficient aggregate demand. Employment was assumed to depend on both real aggregate demand and real wages. Markets were not perfectly competitive, and firms set prices as a mark-up on their wage costs. The mark-up is lower if wage inflation is increasing. Unions, on the other hand, set nominal wages as a mark-up on prices. The wage mark-up is lower if there is more unemployment or if "push' factors are operating less strongly. In the short run, Layard argued, unemployment depends on the "push' factors (social security, employment protection, unions, taxes, relative import prices, etc.) which determine the real wage targets of workers, as well as aggregate demand. The empirical results presented by Layard suggested that the rise in unemployment between 1975-79 was mostly due to "push' factors, such as real wages. The increase after 1980 could be attributed primarily to the fall in aggregate demand.

Participants questioned the specification of the aggregate demand factors, which included the ratio of the adjusted budget deficit to potential GDP, competitiveness and world trade. Michael Beenstock (City University and CEPR) questioned the appropriateness of the NIESR adjusted budget deficit series. Geoffrey Dicks (LBS) suggested that competitiveness might not be an appropriate index of the stance of aggregate demand policy. Robert Gordon pointed out that the productivity slowdown of the 1970s was not accounted for; Layard responded that proxies for it were not significant in the labour demand equation.

The second paper, "The Relationship Between Employment and Wages', was presented by John Odling-Smee (HM Treasury). He discussed simulations of the effects of permanently lower nominal wages, using the Treasury model. Nominal wages were an endogenous variable, determined within the Treasury model; the simulations were carried out by overriding the equation in the model which explained this variable. The Treasury simulations indicated that both output and employment would go up, with the increases in employment higher than those in output. The lags were long, however; it took three years for an appreciable increase. The magnitude of the effects also depended on the fiscal and monetary stance following the cut in nominal wages. With "tight' policies of unchanged tax and interest rates, the increases in output and employment would be substantially lower than with an unchanged level of nominal aggregate demand. These simulations were seen by participants as confirming the Layard thesis that both aggregate demand and real wage influenced employment.

In a short presentation, Robin Bladen-Hovell (Manchester) presented provisional simulation results on a similar exercise with the NIESR model. Lower wages were shown to lead to lower employment. This result occasioned widespread disbelief among participants, and was attributed to the absence of any direct real-wage effect in the model equation determining employment.

Wages are explained within the Treasury model and are therefore "endogenous' to the model. Both simulations discussed at the workshop forced wages to assume a particular value, instead of allowing the model to generate its own level of wages. This was accomplished by removing the equation explaining wages from the model. Grayham Mizon argued that this was an inappropriate procedure. Valid and useful information could be obtained from simulations in which values of variables "exogenous' to the model are chosen so as to produce a particular value of an endogenous variable. Exogenous variables are by definition explained outside the model, and can be varied by the investigator without violating the logic of the model itself. Simply "fixing' the value of an endogenous variable is inappropriate, because it violates the logic by which the model was constructed. Participants also questioned the mechanism which might bring about lower wages within the Treasury model. The model itself could not throw much light on this, since it does not account for the supply side of the labour market.

Robert Gordon (Northwestern University and CEPR) presented the third paper, entitled "Wage-Price Dynamics and the Natural Rate of Unemployment in Europe, Japan and North America'. He discussed in greater detail the research which formed the background to his lunchtime remarks.

Robert Solow (MIT) argued that Gordon's estimate of the 1983 difference between real wage and productivity growth (wage gap) for the US was too high. Gordon replied that this might be the result of the appreciation of the dollar, which limited price increases in US manufacturing. Michael Artis (Manchester and CEPR) mentioned evidence that showed asymmetries in price responses to appreciations and depreciations of exchange rates. David Stanton (Department of Employment) questioned whether Gordon's estimates of cross-country differences in adjustment processes were significant, given the standard errors of the estimated parameters. Richard Layard argued that wages would be affected by labour conditions in the economy as a whole, and not merely in the manufacturing sector.

Grayham Mizon noted in his concluding remarks that the workshop papers had provided useful evidence concerning the relationship between wages and unemployment. The discussion had suggested to him that what was now required was a thorough econometric evaluation and comparison of the alternative models. This would require "reducing' the large-scale models to sub-models involving only their labour markets.