Jacques R. Artus
A comparative study of wages in manufacturing in Europe, North America, and Japan
The failure of the major European countries to create new jobs during the past decade, with the resulting emergence of double digit unemployment rates, is often blamed on unduly high wages. Observers point out that, especially in manufacturing, the real wage rate defined from the employer’s standpoint—that is, including all labor costs and divided by value added rather than by consumer prices—tended to grow faster than labor productivity. This led to a rise in the share of labor costs in value added and a decline in the share of incomes going to capital (see Chart 1). By contrast, the rise in labor’s share of value added was much more limited in the United States and Canada and their success with job creation much better. The share of labor also rose significantly in Japan, but from a very small base.
Chart 1Actual labor share of value added in manufacturing, 1955–821
Source: IMF data.
1Labor shares of value added at factor cost, net of inventory appreciation.
The evidence concerning the rise in the labor share (from a base level associated with a satisfactory employment situation), while suggestive of excessively high wages, is hardly conclusive. In particular, the increased share of labor costs in Europe could equally well have been warranted by long-run changes in production techniques or in the structure of the manufacturing sector (shifting, for example, from capital-intensive to labor-intensive techniques or industries), in the relative scarcity of labor vis-à-vis capital, or in the relative price of energy. Similarly, the accompanying unemployment could have been quite unrelated to real wage levels; it could have been the result of higher turnover rates and increased regional and skill mismatches, or of a deficiency of aggregate demand. Before jumping to conclusions that have major implications for policy, one needs to give the evidence further scrutiny.
This article is based on an empirical study that sought to determine whether recent changes in the shares of labor in value added were caused by the factors listed above, and would therefore not impede a return to “high” employment, or by rigidities in labor markets that kept the real wage rate above the marginal productivity of labor at high employment. Actual real wages in manufacturing in Canada, France, Germany, Italy, the United Kingdom, and the United States were compared with estimates of the “warranted” wages—that is, the real wages calculated to be consistent with high employment of labor, given the existing capital stock. The comparison showed that in France, Germany, and the United Kingdom, the real wage rate was indeed higher than the level compatible with high employment after about 1970. In Canada and the United States, this was not the case. For Japan and Italy, the results were less conclusive.
In considering these results, it is important to take the limitations of the study into account. First, it was narrowly focused on the manufacturing sector, where labor market rigidities are particularly prone to preventing an adjustment of real wage rates. In most other sectors, including agriculture and services, competitive forces normally play a greater role in determining wage rates, and they contain a high proportion of self-employed, which blurs the distinction between labor and capital costs. Nevertheless, it would be desirable to consider also these other sectors, particularly given the growing importance of the services sector in industrial countries. Second, the study viewed the capital stock as given (“exogenous”); no attempt was made to explain investment. Thus the possibility of a vicious circle in which a disequilibrium real wage rate reduced investment, which in turn reduced the growth of the warranted real wage rate, was outside the scope of the study. Third, goods prices, including prices of manufactures, intermediate inputs used in the manufacturing sector, and consumer goods, were also viewed as exogenous. This last assumption implied that the exchange rate was given—a point with wider implications elaborated on later.
Method of analysis
Two models were used as a basis for the estimates of the warranted share of labor and the corresponding warranted real wage rate. Model A assumed a constant-elasticity-of-substitution production function relating capital and employed labor (in man-hours) to value added. The innovative element in this model was that it made allowances not only for changes in the rate of “disembodied” productivity growth—that is, in the growth of real value added that did not appear to be directly related to measured increases in labor or capital inputs—but also for changes in the relative weights of labor and capital. The recognition that these weights could, and indeed probably did, change in the longer run as production techniques and the structure of the manufacturing sector evolved was considered important for the validity of the results.
Model B extended the production function to incorporate energy as a specific factor. While it seemed safe to assume that the prices of raw materials did not affect the capital-labor ratio, it was not sensible to rule out the effects of energy prices. Specifically, the model allowed for the possible complementarity between energy and capital; as a result, a marked increase in the relative price of energy could lead entrepreneurs to increase their demand for labor, and decrease that for the composite “capital-energy” input.
For both models it was assumed that, on average, labor was paid its marginal product during the second half of the 1950s and during the 1960s, periods for which there was no reason to expect that the real wage rate was out of equilibrium (except in Japan where it was possibly too low). Under this critical constraint, the parameters of the two production functions were estimated for 1955-82. The estimated parameters were then used to calculate the warranted labor shares and real wages, not only for the second half of the 1950s and for the 1960s, but also for the 1970s and early 1980s when the equality between the real wage rate and the marginal product of employed labor may not have held, and when evidence suggested that the high-employment labor supply was considerably larger than the actual amount of labor used.
Focusing first on model A, the estimates for the simple production function showed that the rate of disembodied productivity growth increased during the 1960s, but fell back during the 1970s and early 1980s in all countries except Germany (where the rate was constant throughout the period). These results corroborated other findings that a significant part of the decline in the last decade in the rate of growth of output per man-hour was not accounted for by a lower rate of capital accumulation. This conclusion held even when the rate of capital accumulation was adjusted downward to take account of the rise in pollution abatement investment and the premature obsolescence of capital stock resulting from the two waves of energy price increases. Cross-country differences in the rate of growth of disembodied productivity were quite stable. For the past three decades, this rate tended to be about 3 percentage points higher in Japan and 2 percentage points higher in France and Italy than in the United States, Canada, and the United Kingdom. Germany, by maintaining a constant rate, moved from the low-growth group during the 1960s and early 1970s to the middle group during the last ten years.
Apart from these findings, which had obvious implications for the warranted growth of real wage rates, the estimates also showed that the relative weight on the capital stock tended to increase. This was not surprising. The tendency for a gradual increase in the capital-intensity of production techniques has been evident for a very long time, and this, ceteris paribus, would call for a rise in the share of capital costs in value added and therefore a decline in the labor share. To put it differently, the warranted growth rate of real wages was less than the growth rate of output per man-hour. However, the data suggested that the tendency was interrupted during 1971-82 in Japan. A possible reason was that the structure of production changed markedly in Japan during this period as industries requiring a high level of raw materials and energy imports, many of which were also capital intensive, were de-emphasized.
The elasticity of substitution between labor and capital was lower than one in all countries (a result that matched the findings of most other studies). The important implication of this low substitutability was that a rise in the capital/labor ratio would tend to increase the warranted share of labor in value added. When the ratio rose rapidly, as it did in the 1960s, this effect could be more important than the effect of the gradual increase in the relative weight of capital, and the warranted labor share could then increase. When investment was low, as in the 1970s and early 1980s, the inverse could happen.
The estimated values of most of the parameters of model B were in line with the corresponding estimates in model A. For example, the relative weight of the composite capital-energy input increased gradually, while the elasticity of substitution between labor and the capital-energy input was less than one. An interesting difference, however, was that the reduction in the rate of disembodied productivity growth between the 1960s and the later period was smaller in the second model, but generally not much smaller. Thus, even when the reduction in the use of energy that was achieved during the past ten years was explicitly taken into account—as it was in model B—there was still a sizable unexplained reduction in the rate of growth of disembodied productivity. Consequently, the increase in energy prices and the following process of economizing on the use of energy could not be blamed for a large part of the decline in disembodied productivity growth. A further interesting result was that the data supported the hypothesis that an increase in the price of energy did often warrant some rise in the price of labor relative to the price of capital.
Warranted, actual wages
Chart 2 shows the estimated wage gaps corresponding to the two models, with the gaps expressed as the deviations of actual from warranted real wage rates as a percent of the latter. For France, Germany, and the United Kingdom, as well as Japan, the two models yielded a wage gap of 12-16 percent for the early 1980s. For the United States and Italy, model A produced a gap of about 5 percent, while model B showed no significant gap. For Canada, model A suggested no gap, while model B suggested that the real wage rate was, if anything, low.
Chart 2Deviation of actual from warranted real wage rate, 1955-821
Source: IMF data.
1 A positive number indicates that the actual real wage rate (measured in terms of the relevant value-added deflator) exceeds the real wage rate that is consistent with the chosen high - employment norm.
These results were broadly similar to those derived by simply looking at the deviations of the actual labor shares from their historical averages during the second half of the 1950s and the 1960s as a measure of disequilibrium in wage costs. Nevertheless, some of the differences were far from negligible.
For example, for the United Kingdom and, especially, for Italy the warranted labor share was estimated to have been higher during the 1970s and early 1980s than during the preceding period, so that the simple approach based on an assumed constant warranted share exaggerated the magnitude of the disequilibrium in recent years. For Japan, the estimates based on the simple approach were found to be even more misleading. The actual labor share rose rapidly in the first half of the 1970s, mainly in 1974-75, and then stabilized, suggesting that Japan adjusted much better to the second wave of oil price rises than to the first. What the results of the study showed, however, was that the early jump was not a severe problem; it started when the actual share was significantly below the warranted share, it was largely related to the presence of temporary labor hoarding (i.e., a delay in firing labor), and there was a gradual increase in the warranted share during that period because of the rapid increase in the capital/labor ratio. By contrast, during the late 1970s and early 1980s labor hoarding was slowly reduced and the stability of the actual share hid a deterioration of the underlying position. Moreover, the warranted share stopped rising because the growth of the capital/labor ratio decelerated sharply.
Further differences with the simple approach were apparent in the results derived from model B. Mainly for the United States, Canada, and Italy, the hypothesis that energy and capital were complementary led to results that were more favorable to labor. Namely, the estimate of the warranted real wage rate was higher and consequently the estimate of the wage gap was lower—in Canada it was even negative. One of the main reasons for this finding was the relatively large increase in the share of energy costs in the total cost of manufacturing production in those three countries during the past ten years. The reasons for this large increase differed. In the United States, the rise in the price of energy relative to the price of labor and capital was especially large; in Canada, the decline in energy use was less than in other countries; while in Italy, manufacturing activities were particularly energy (and oil) intensive. There was also a relatively large increase in the share of energy costs in Japan, but it was more easily absorbed than in the preceding three countries because of the small initial labor share. Nonetheless, the results of model B were somewhat more favorable to labor even in Japan.
Given the sizable margin of error in the data on some of the important variables used to derive the above results, they were tested (using a systematic sensitivity analysis) and found to be relatively robust to variations in the estimates of the gross capital stocks and the high-employment labor inputs.
In all countries except the United Kingdom, the adjustment to the deceleration in the growth of the warranted real wage rate in terms of consumer prices was considerable, with the rate of growth of the actual real wage rate declining by 2 percentage points or more from 1956-69 to 1979-82. So, it would be an exaggeration to say that the rate of growth of the real wage rate was rigid. In fact, there was a great deal of flexibility, but not always enough. In Japan and the United Kingdom, for instance, the actual real wage rate grew in terms of consumer prices at a rate of 2-3 percent during 1979-82 instead of declining in line with the warranted rate. For France and Germany, the other two countries where there is currently a large gap, the growth of the actual real wage rate was kept nearly in line with the low growth of the warranted rate in recent years. The study suggested that the problem for these two countries was not so much a systematic tendency for inertia in the adjustment of real wages to supply shocks, as a failure to reverse the unwarranted increases in real wages of the early and mid-1970s. At least for France, these increases had as much to do with the wage explosion of the early 1970s as did the supply shocks of the mid-1970s.
The study showed that, at least as far as manufacturing was concerned, there are strong reasons to believe that in France, Germany, and the United Kingdom the real wage rate has been too high in recent years, in the sense of being incompatible with high employment. In contrast, it has not been too high in the United States and Canada, although there may of course be some problems in specific industries. There was no evidence of a real wage problem in Italy but poor data prevent firm conclusions (see box). For Japan, the large increase in the labor share observed in the 1970s did not appear to have been fully warranted by concomitant changes in the factors considered in the study. At the same time, the initial labor share was so small that this increase may have been less of a problem than it was in Europe.
The estimates of the warranted real wage rates on which these conclusions were based must be regarded as tentative, for at least three reasons. First, it was difficult to measure the actual flows of labor and capital services and, a fortiori, the high-employment labor supply in manufacturing. The sensitivity analysis showed that the order of magnitude of the estimates is relatively robust to plausible variations in the values taken by these variables; nevertheless, the resulting uncertainty is far from negligible. Second, the estimates suffered from a number of country-specific problems. Finally, an aggregate production function for a whole economic sector is an inherently crude empirical tool because the conditions necessary for aggregation over firms and industries are never fully satisfied, particularly if workers are not paid their marginal products.
The author’s more detailed discussion of these issues is contained in his article, “The Disequilibrium Real Wage Rate Hypothesis—An Empirical Evaluation,” IMF Staff Papers, June 1984. For France and Italy, national accounting figures on nominal value added included inventory appreciation—which accrues neither to fixed capital nor to labor. For France, the figures were adjusted by using data on inventory appreciation for the nonagricultural economy; for Italy, an even rougher adjustment was made. Estimates of the shares in value added of capital and labor, particularly for Italy, may thus be subject to sizable error. Flows of labor and capital services: following a standard approach, this study used series on man-hours worked (except for the United States where only series on man-hours paid were available) and on the gross capital stock as proxies for the unavailable data on the actual flows of labor and capital services. But even these proxies entailed measurement problems; the study attempted to adjust for these, but some distortions, especially in the capital stock data, may well have remained. The main problem with the measurement of the actual flow of labor services was that the available data on man-hours did not reflect the level of education and technical expertise of the work force. In the present context, however, the author concludes that the problem is likely not to be too severe because those changes occur only gradually.
While these limitations mean that the estimates are far from precise, they do not imply that the estimated wage gaps for France, Germany, and the United Kingdom reflect statistical artifacts. There is more uncertainty for Japan because of the lack of a fully adequate base period, but it is doubtful that this problem can completely explain the measured gap. A major factor confirming that the order of magnitude of the gaps is right is the evolution of unemployment. Unemployment can be classical (caused by an unduly high real wage rate), structural (caused by turnover and by regional and skill mismatches), or Keynesian (caused by a deficiency of aggregate demand). All of these types can also occur together. Thus, one should not expect a close cross-country correlation between the overall unemployment rate and the size of the wage gap. However, by early 1984 the unemployment rate in France, Germany, and the United Kingdom was between 5 and 7 percentage points above our estimate of the structural rate for the early 1980s, and this gap does not seem to be declining. In contrast, the unemployment rate was only 2 to 3 percentage points above the structural rate in the United States and Canada, and the spread was decreasing from month to month. In Japan, the economic and social system is such that the rise in unemployment has been quite moderate; nevertheless, employment in manufacturing fell 4 percent from early 1974 to 1982. This is striking, partly because the total labor force increased by 9 percent during this period and partly because, as recently as the 1960s, employment in manufacturing was rising three times faster than the growth of the labor force.
These results are particularly worrisome when viewed against the background of the developments in exchange rates of recent years. The extremely sharp appreciation of the U.S. dollar and the Canadian dollar vis-à-vis the other major currencies over the past four years is likely to have had important effects on the profitability of manufacturing exports—decreasing export profitability in the United States and Canada and increasing export profitability in the other countries. It is thus very disquieting that, despite these developments, it is still in Europe, and possibly in Japan, that there is a real wage problem for the manufacturing sector as a whole, rather than in the United States and Canada.