The deterioration in labor market performance in most of the industrial countries since the early 1970s remains one of the most serious economic problems confronting policymakers. Even though a broad range of measures has been implemented to tackle this problem, unemployment has continued to rise in a large number of countries, particularly in Europe.
The approach to the problem of unemployment has changed considerably over the past decade. In the period between the two rounds of oil price increases, many countries attempted to promote growth and employment through expansionary macroeconomic policies. Examples of such policies include the easing of monetary and fiscal stance in many countries in 1975, and subsequently the internationally co-ordinated injection of fiscal stimulus in 1978. This approach was abandoned with the widespread re-emergence of inflationary pressures in 1979–80. Although partly attributed to the second round of oil price increases, these inflationary pressures were also widely perceived to be linked to a number of serious imbalances that had built up over the 1970s, and which posed a threat to many countries’ long-term growth prospects. These imbalances included relatively large budget deficits, low profitability, and a number of structural deficiencies in labor markets. Since the beginning of the 1980s, most industrial countries have pursued a strategy aimed at eliminating these imbalances in order to re-establish the conditions for sustained noninflationary growth. This medium-term strategy has, particularly, emphasized inflation control, restoration of profitability, and reductions in government budget deficits and in the public sector’s claim on resources. Many governments have also sought to improve the functioning of markets, particularly those for labor and financial services.
Although the medium-term strategy has been successful in a number of respects, notably in reducing inflation, it has so far been less successful in promoting growth and employment. In Europe, in particular, unemployment rates have continued to rise and averaged 11.2 percent in 1985. Moreover, many observers seem to agree that unemployment in most European countries is unlikely to decline much, if at all, over the balance of the decade. This prospect, together with the observation that employment has increased significantly in recent years in some countries that have pursued relatively expansionary policies, has led some analysts to conclude that, in view of the marked reduction in rates of inflation and the improvement in profitability that has been achieved, fiscal and monetary policies in some European countries should now become more expansionary.
The purpose of this paper is to review recent analytical and empirical research on the determination of employment, to provide a framework for evaluating the merits of alternative policies to cope with unemployment. Particular emphasis is placed on the mechanisms of employment and wage determination described in recent studies. The review has been restricted in two respects. First, the emphasis is predominantly on the macroeconomics of labor markets. The more microeconomic aspects of these markets—for example, the role of minimum wage laws, wage differentials, regional and occupational mismatches, or sectoral growth trends—are assuredly important, but beyond the scope of the present paper. Second, in surveying the empirical evidence, emphasis has been placed on cross-country studies that attempt to explain the pronounced differences in labor market developments among the major industrial countries over the past fifteen or twenty years.
Summary and Conclusions
Unemployment, Employment, and Growth
Unemployment is not solely a European problem, nor is it equally serious in all of the European countries. However, in analyzing international differences in labor market behavior, it is convenient to concentrate on the European countries as a whole or the four major European countries as a group, and to contrast their employment performance with that of the United States and Japan.
The unemployment rate in the European countries averaged 11.2 percent in 1985, having risen in every year since 1973 when it stood at only 3 percent of the labor force. The average unemployment rate in Europe was lower than that of the United States until the beginning of the 1980s. However, since 1983 the U.S. unemployment rate has been reduced from almost 10 percent (on an annual average basis) to about 7 percent in early 1986 whereas Europe’s unemployment rate rose by almost 2 percentage points over the same period. Europe’s poor labor market performance is even more striking in comparison with Japan, which has been able to maintain close-to-full-employment conditions over the past decade.
These differences in unemployment trends mask even larger differences in employment and labor force behavior across countries. Whereas most countries in Europe have experienced relatively moderate increases in their labor forces, the U.S. labor force has continued to follow a strong upward trend (Chart 1). Most of the rise in the U.S. labor force has been absorbed by the rise in employment, which expanded by almost 50 percent between 1965 and 1985. By comparison, in Europe there has been a cumulative increase in employment of only 2.5 percent over the same period. In Japan, employment rose by some 25 percent over these two decades, only marginally less than the increase in the labor force.
Europe, the United States, and Japan: Labor Market Trends, 1965–85
(In millions)
Europe, the United States, and Japan: Labor Market Trends, 1965–85
(In millions)
Europe, the United States, and Japan: Labor Market Trends, 1965–85
(In millions)
In trying to explain these differences in employment growth across countries, the “Keynesian” economist would attribute at least part of the differences in the behavior of employment and unemployment across countries to differences in cyclical developments. According to this view, an obvious place to begin is to examine the role of cyclical developments and whether there have been major differences in countries’ rates of growth of economic activity. And indeed, the employment performance of individual countries suggests that year-to-year fluctuations in employment have, at least to some degree, been correlated with changes in aggregate activity (Chart 2).
Europe, the United States, and Japan: Real GNP and Employment, 1966–85
(Annual changes, in percent)
Europe, the United States, and Japan: Real GNP and Employment, 1966–85
(Annual changes, in percent)
Europe, the United States, and Japan: Real GNP and Employment, 1966–85
(Annual changes, in percent)
A rather different perspective on the relationship between output and employment growth is provided by Chart 3, which suggests that there has been surprisingly little relationship between countries’ long-run growth performance and their ability to create jobs. For example, over the past two decades, France and the Federal Republic of Germany have experienced about the same rates of growth of output as the United States and Canada. And yet employment was hardly higher in 1985 than in 1965 in the case of the two European countries whereas in North America employment expanded at an annual rate of more than 2 percent over the same period. Considering the data for the last decade gives essentially the same result.
Major Industrial Countries: Growth of Employment and Output, 1965–85 and 1975–85
(Compound annual growth rates, in percent)
Major Industrial Countries: Growth of Employment and Output, 1965–85 and 1975–85
(Compound annual growth rates, in percent)
Major Industrial Countries: Growth of Employment and Output, 1965–85 and 1975–85
(Compound annual growth rates, in percent)
The lack of any systematic relationship between countries’ long-run growth and employment performances reflects the fact that output per person employed (labor productivity) or, conversely, the labor intensity of production, has developed quite differently across countries. Thus, over the past 20 years, the labor intensity of production has declined by almost 50 percent in Europe and by 70 percent in Japan, compared with a reduction of only 15 percent in the United States (Chart 4).
Real Wages
The factors that account for these developments in labor productivity (or in the labor intensity of production) are of crucial importance for any analysis of unemployment. One school of thought which is prominent in the recent literature has emphasized the significance of the behavior of real and relative labor costs. According to this “classical” view, developments in labor productivity and in unemployment across countries largely reflect differences in the behavior of labor costs that have been independent of the level of aggregate demand. This contrasts with the Keynesian view mentioned earlier that attributes the rise in unemployment in Europe primarily to insufficient demand at the existing level of real labor costs.
The empirical evidence presented in the section on the determinants of the demand for labor points to a significant negative relationship between real wages and the demand for labor, with the relationship being relatively weak in the short run but increasingly important over two or three years. This relationship reflects the fact that firms reduce their use of labor in response to a rise in real wages, both by increasing the use of those production factors that become relatively less expensive and, because of the impact on profitability, by reducing the level of production. Given this evidence, the relatively rapid growth of real wages in Europe—relative to the rate of growth that would have been warranted from a high-employment perspective—appears to have been a major reason for the sluggish growth of employment. Indeed, whereas output growth in the United States and Europe has been quite similar over the past two decades, real labor costs have increased much less in the United States than in the major European countries (Chart 5). Furthermore, although labor cost increases in Europe decelerated substantially after 1975 they have continued to exceed U.S. labor cost increases.
Europe, the United States, and Japan: Real Wages and Employment, 1965–85
(Compound annual growth rates, in percent)
1 Compensation per employee deflated by GNP deflator.2 Compensation per employee deflated by consumer prices.3 Number of persons employed.
Europe, the United States, and Japan: Real Wages and Employment, 1965–85
(Compound annual growth rates, in percent)
1 Compensation per employee deflated by GNP deflator.2 Compensation per employee deflated by consumer prices.3 Number of persons employed.Europe, the United States, and Japan: Real Wages and Employment, 1965–85
(Compound annual growth rates, in percent)
1 Compensation per employee deflated by GNP deflator.2 Compensation per employee deflated by consumer prices.3 Number of persons employed.The main mechanism through which the rise in real wages has prevented greater employment gains in Europe over the past ten to fifteen years seems to have been a substitution of capital for labor which has lowered the labor intensity of production significantly more than in the United States. Hence, the labor intensity of production, or the productivity of labor, appears to be strongly related to the rate of growth of real labor costs. Of course, developments in labor productivity have also been related to technical progress in a broad sense (total factor productivity), and to changes in the skills of the labor force. However, while developments in these other factors have differed somewhat across countries, they do not appear to have been sufficiently important to fully account for the differences in productivity growth.
There are a number of important caveats with respect to the apparent relationship between differences in employment and labor cost developments across countries. In particular, real labor costs have increased substantially faster in Japan than in other countries, and yet unemployment remains low. There has in fact been a significant slowdown in the rate of growth of employment in Japan, a development which at least in part appears to have reflected the strong growth in real product wages. However, the slowdown in employment growth was accompanied by slower growth of the labor force. The rise in real labor costs above and beyond that accounted for by underlying differences in total factor productivity growth may therefore be interpreted as a normal market response to labor becoming relatively more scarce.
In view of the important role that differences in the developments of real labor costs appear to have played in explaining the observed differences in employment growth across countries, the reasons for these differences in real labor cost developments are clearly of great importance for economic policy. The empirical evidence reviewed in the section on real wage flexibility suggests that there have been significant differences among countries in the responsiveness of real wages to adverse disturbances (such as higher oil prices or slower growth in total factor productivity) that call for a reduction in the rate of growth of real wages if high employment conditions are to be preserved.
In this context, the studies reviewed distinguish between short-run and long-run wage flexibility. A commonly used measure of long-run flexibility is based on the increase in unemployment “required” to reduce the rate of growth of real wages by a given amount. Measures of short-run flexibility usually allow for inertia in nominal wage formation, which may occur, for example, in countries with little or no automatic wage indexation. In such cases, a disturbance which raises prices may in fact reduce actual real wage growth, at least for a period, thereby moderating the effect of the disturbance on unemployment. In most of the empirical literature on wage behavior, Japan is shown as having a high degree of real wage flexibility in both the short and long run, which is consistent with the observation that real consumption wages—defined as nominal wages deflated by consumer prices—in fact decelerated more in Japan after 1975 than in the other major countries (Chart 5). In the United States, real wages are also found to be relatively flexible in the short run as a result of nominal inertia in the wage formation process, but the degree of flexibility seems to decline over time. In Europe, real wages are typically found to be relatively rigid in both the short and the longer run.
In addition to differences in the flexibility of real wages across countries, there also appear to be differences in the underlying momentum of real wage gains. For instance, trend increases in real wages seem to have been stronger in Europe than in the United States. Hence, under similar economic circumstances, real wages have tended to rise significantly faster in Europe than in the United States. This may be a carryover from the 1960s when the scarcity of labor in Europe, together with rapid technological progress, may have justified relatively larger increases in real wages than in the United States.
In contrast to the relatively rapid growth of real labor costs during the 1970s, it is apparent that there has been a substantial deceleration in real labor cost increases in Europe over the last few years. Identification of the factors underlying this deceleration is fundamental to its implications for future employment behavior. If the slower growth in real wages is attributable to an adjustment of the wage bargaining position of workers to a slower trend rate of growth in warranted real wages, this would suggest that an adjustment process leading to an increase in the labor intensity of production and, hence, employment is under way. If, on the other hand, the deceleration of real wage increases is due primarily to the impact of high unemployment, the labor market might also adjust ultimately, but the adjustment process would be longer and could require a high level of unemployment over a long period.
The evidence reviewed in the section on wage flexibility does not suggest that there has been any significant “break” in wage behavior in Europe, either after 1973 or more recently. For example, stability tests of the responsiveness of real wages to unemployment generally have failed to reveal significant changes in wage behavior. Moreover, while some tests suggest that real wage claims have begun to adjust to the slower growth of warranted real wages since 1973, they also suggest that the adjustment process is extremely slow. Thus, while this type of evidence needs to be continuously reassessed, it cannot be assumed that the recent slowdown in real wage growth reflects a more flexible approach to wage setting, and that slower wage increases would necessarily be sustained if unemployment were to return to more normal levels.
Another factor underlying the observed differences in employment and wage developments across countries has been differences in the magnitude of the adverse disturbances that have been experienced over the past two decades. These disturbances have included a secular slowdown in the growth of total factor productivity, rises in commodity and energy prices, exchange rate and terms-of-trade fluctuations, higher payroll taxes, and sharp increases in both indirect and direct taxes. The rise in tax burdens has driven large wedges between the cost of labor to firms and the net earnings of employees, thereby reducing the growth of the warranted net consumption wage. The interaction of these disturbances and the inflexibility of real wages has resulted in substantial gaps in some countries between actual and warranted real wages. The evidence suggests that this problem has been much more severe in Europe than in the United States, reflecting differences both in the degree of real wage flexibility and in the magnitude of the disturbances experienced. The adverse disturbances were relatively severe in Japan as well; however, because of the long-run flexibility of real consumption wages in that country, the negative consequences for employment have been relatively insignificant compared with the European experience.
It is not possible to determine precisely by how much real wages may have moved out of line in any particular country. It is apparent from the evidence discussed later, however, that real wage gaps (between actual and warranted wages) increased significantly in many European countries following the two rounds of oil price increases—periods when unemployment rose sharply—and that they are currently relatively wide. Although some measures of the wage gaps indicate that a narrowing has taken place in recent years, this appears to have owed more to the effects of rising unemployment on wages than to an increased responsiveness of wages to labor market conditions.
Another manifestation of the lack of flexibility of real wages in the face of adverse disturbances has been a sharp rise in the “nonaccelerating inflation rate of unemployment” (the NAIRU) in many European countries during the 1970s and early 1980s. The level of the NAIRU reflects a number of “structural” characteristics of the economy, including the institutional organization of markets (for instance, regulations), demographics, preferences of individuals, and incentives provided through the structure of taxes, transfers, and subsidies. In addition, the NAIRU is directly affected by any disturbance to the economy which reduces the rate of growth of the warranted real wage when real wages are not perfectly flexible. Despite its limitations, developments over time in this unemployment rate convey useful information about the existence of wage rigidities in the face of disturbances.
With the exception of Japan, the NAIRU has increased sharply since the late 1960s in all the major industrial countries.1 The modest increase in the rate in Japan testifies to that country’s high degree of wage flexibility which has largely offset the implications of the severe shocks experienced in the 1970s, particularly the oil price increases. The less pronounced rise in the NAIRU in the United States relative to Europe seems to reflect both differences in the severity of the disturbances experienced and greater real wage flexibility, particularly in the short run. The early 1980s have seen a particularly dramatic rise in the NAIRU in many European countries, primarily as a consequence of rising import prices in the wake of the second oil price shock and the sharp appreciation of the U.S. dollar. Because of the lack of flexibility of consumption wages, the rise in import prices led to an increase in real wages (measured in terms of producer prices) relative to their warranted level. In the United States, the NAIRU rose only modestly in the early 1980s reflecting both a somewhat stronger total factor productivity performance than in the 1970s and the strong dollar which reduced import prices. To some extent, the movements in real exchange rates, the terms of trade, and, hence, the NAIRU since 1980 appear to have reflected the divergences in fiscal stance between the United States and Europe over this period.
When the actual unemployment rate is above the NAIRU—which appears, for example, to be the case in several European countries at present—the interpretation of such a discrepancy is not straightforward. It may reflect either a deficiency of demand—in which case it would point to the existence of Keynesian unemployment—or it may reflect classical unemployment resulting from a gap between the actual and the warranted level of real wages. In the latter case, an unemployment rate above the NAIRU would indicate that an adjustment is in motion whereby real wages are temporarily growing below the warranted rate in order to close the existing wage gap. Such an adjustment process could be very protracted in economies with little wage flexibility. The interpretation of the position of the NAIRU relative to the actual rate of unemployment therefore requires some notion of the magnitude of any remaining wage gap.
In view of the projected improvement in Europe’s terms of trade following the depreciation of the U.S. dollar since early 1985 and the sharp decline in oil prices early in 1986, the NAIRU in European countries might be expected to decline significantly in 1986–87. Such a reduction would indicate a reversal of some of the negative disturbances previously experienced and—to the extent that wage gaps would also narrow—should have a beneficial impact on the actual rate of unemployment. For such a reduction in the NAIRU as well as in the actual rate of unemployment to materialize, it would be important to limit the extent to which the terms-of-trade gains would be reflected in higher growth of real consumption wages.
Conclusions
Notwithstanding the serious difficulties in distinguishing empirically between classical and Keynesian unemployment, most of the evidence surveyed does not support the view that the principal reason for Europe’s poor labor market performance vis-à-vis the United States and Japan has been weak demand, independent of any wage problems. The similarity of the long-run growth performance of Europe and the United States, the relatively strong upward trend in real wages relative to the growth of warranted wages, the evidence of highly rigid real wages, and the pronounced rise in the NAIRU all point to real wage problems as a major factor underlying the rise in unemployment in Europe.
If, indeed, unemployment is primarily of a classical nature, general demand reflation would only reduce unemployment significantly insofar as it might raise the warranted wage or lower actual real wages. Warranted wages might be raised, for example, by encouraging higher investment. However, such an effect would be uncertain; it would depend critically on the kinds of investment that were encouraged and it would probably only be felt after a long period. General demand reflation might also stimulate employment in a situation of classical unemployment to the extent that the rate of inflation would increase, leading to a reduction in real labor costs. The evidence on wage behavior in Europe suggests, however, that higher inflation would be unlikely to change real wages permanently. In any case, governments would be ill-advised to jeopardize the hard-won gains on the inflation front.
It should also be noted that general demand reflation would in any case be much less effective in situations where unemployment is predominantly classical. Relatively large fractions of any fiscal stimulus, for example, would end up in imports or higher inflation rather than in higher real output and employment. As a result, countries would have relatively little to show for the increases in fiscal deficits which can be expected to result from fiscal expansion. More generally, of course, even if some of the rise in unemployment in Europe in recent years may have partly reflected Keynesian factors, it is not obvious that the circumstances prevailing in most countries would justify a shift to a reflationary policy stance. In many countries, the reductions in budget deficits and in underlying inflation rates have not yet been sufficient to warrant a relaxation of policies.
On the evidence presented in this paper, a preferable solution to the unemployment problem would be to seek to raise the labor intensity of production through a moderation of the growth of real labor costs reflecting a change in wage behavior. Simulation studies carried out, for example, by the U.K. Treasury and the European Commission point to positive effects on employment from wage moderation. These studies suggest that such effects can be substantial, in particular when the monetary authorities maintain the same rate of growth of monetary aggregates as they would have aimed at in the absence of wage moderation, and where fiscal policy is adjusted to maintain the same budget deficit in relation to gross national product (GNP) as would otherwise have obtained.
Wage moderation is of course difficult to achieve. Experience shows that deflationary demand management policies are an extremely costly way (in terms of employment and output) to reduce wage increases. Moreover, they do not necessarily affect wage behavior and may result in a movement along the Phillips curve rather than a shift in the curve. A more promising approach would be to combine measures to improve the functioning of labor markets with measures to reduce real labor costs. With respect to the latter, a possibility would be to reduce payroll taxes. Alternatively, direct taxes could be lowered as a quid pro quo for lower wage increases.
It is also important to keep in mind that in the typical European country, because of extremely high tax burdens, the net take-home pay is probably no more than 40 to 50 percent of gross labor costs. Under such circumstances, “wage moderation” cannot necessarily be achieved simply by restraint on the part of wage earners; there may also be a need to reduce some of the many wedges between net earnings and total labor costs, and in particular employers’ contributions to social security schemes. Although several countries have reduced social insurance taxes in recent years, payroll taxes remain high in most European countries. Nevertheless, budgetary constraints mean that the scope for cutting taxes will depend on the strength of the resolve to reduce public expenditure. Even though many governments aim to reduce the public sector’s overall claims on resources, progress in this area has been extremely slow.
Measures to reduce real labor costs should also be combined with structural policies to improve the functioning of labor markets and the flexibility of real wages. The rate of unemployment compatible with stable inflation—the NAIRU—has increased dramatically in Europe in the course of the 1970s and early 1980s. To some extent this has reflected the consequences of the adverse disturbances experienced during this period. However, it has also been due to rigidities in wage formation which have reflected both the existence of geographical and occupational mismatches, and the institutional and organizational setup in each economy. For some time, governments have sought to eliminate such deficiencies by encouraging labor mobility, by implementing job training schemes, by modifying or eliminating wage indexation arrangements, and by a broad range of measures aimed at enhancing the functioning of labor markets. Such structural measures are an essential component of any efforts to raise employment and ensure flexibility in the face of future disturbances.
Even if unemployment in Europe is perceived as being largely a classical phenomenon, the influences which underlie it have changed substantially over time. In the 1970s, unemployment rose significantly in both the United States and Europe. In the United States, where employment continued to rise, a substantial part of the rise in unemployment was attributable to increased structural unemployment reflecting mismatches and demographic factors. In Europe, although structural factors played a role, the rise in unemployment appears to have largely reflected an adjustment in the labor market to the rapid growth in real wages since the late 1960s, together with slower growth of total factor productivity and adverse terms-of-trade movements. While the recession following the first round of oil price increases also contributed to the rise in unemployment both in Europe and the United States, the lack of flexibility of real wages in Europe in the face of the initial import price shock caused a steady deterioration in labor market conditions even during the recovery from the 1974–75 recession.
The first half of the 1980s has witnessed a series of new shocks which necessitated further labor market adjustment, particularly in Europe. The second round of oil price increases, the additional terms-of-trade deterioration caused by the depreciation of the European currencies against the dollar, and the 1980–82 recession all contributed to a further rise in unemployment in Europe, which has continued through 1985. In comparison, because of the strong dollar, the terms of trade improved sharply in the United States in the wake of the second oil price increase. The pronounced differences in policy stance between the United States and Europe in this period thus seem to have contributed to the differences in labor market performance, both because of their short-run impact on aggregate demand and because of their effects on exchange rates and the terms of trade.
Circumstances in the second half of the 1980s are likely to be quite different again and may provide a unique opportunity for tackling Europe’s unemployment problem. In particular, given the sharp decline in oil prices early in 1986, and the significant depreciation of the dollar since early 1985, increases in real consumption wages in Europe in 1986 will translate into much smaller increases in real labor costs measured in terms of producer prices. While this might potentially permit a substantial narrowing of real wage gaps in European countries, it is not obvious that this will occur. For example, pent-up demands for improvements in living standards following several years of sluggish real wage growth may be considerable. Moreover, in view of the limited progress in reducing public expenditure the scope for tax reductions in many countries is smaller than expected, which also may lead to pressures for higher real wage increases as a result of the terms-of-trade improvement. However, it is apparent from the evidence reviewed in this paper that to the extent the terms-of-trade gains are passed on to wage earners, the European countries would miss an important opportunity to achieve a significant improvement in labor market conditions.
Determinants of Demand for Labor
Until the 1970s, the significance of factor prices for employment generally received little attention in debates on macroeconomic policy. Trend increases in employment appeared to closely correspond to the growth in aggregate output, and employment fluctuations over the business cycle appeared to be directly explainable by shifts in aggregate demand. Indeed, in many macroeconomic models, economic activity and employment were frequently linked through relationships that disregarded relative and absolute factor prices.
Following the secular slowdown in rates of growth of output and employment in the early 1970s, and since the marked differences in employment growth across countries could not be explained by differences in economic growth, macroeconomic research increasingly turned its attention to the role of factor prices in influencing the volume of employment at a given rate of output growth, and to the many interactions among factor prices, employment, and growth. As a result, a more balanced approach to the determination of employment has emerged, together with a clearer understanding of the forces that influence the factor intensity of production and the level of production itself.
This section surveys the recent empirical evidence on the role of factor prices in influencing the demand for labor in the major industrial countries; it seeks to obtain a range of “consensus” estimates of the significance of factor prices for employment; and it discusses the direct significance of aggregate demand for employment determination. While the ultimate intention is to throw light on the economy-wide demand for labor, much of the empirical evidence that is available is confined to the manufacturing sectors of the major industrial economies. Economy-wide generalizations should be made with caution unless it can be assumed that the demand for labor in other sectors behaves in a similar way and is influenced by the same factors as in manufacturing.
In general, actual employment will reflect a large number of interactions among the considerations that influence the quantity of labor firms choose to hire, the terms and conditions under which labor is supplied, and general economic circumstances. Because many of these factors are interrelated, it is useful to adopt a systematic approach and to consider in turn the amount of labor firms will choose to hire (the demand for labor), the determinants of the quantity of labor supplied (the supply of labor), and, finally, the interactions between labor demand and supply on the one hand, and goods markets on the other. This section on the demand for labor assumes that firms are able to purchase all the labor they desire at given factor prices, and that there are no constraints on the availability of capital. In the next two sections, the demand for labor is combined with the supply side of the labor market, while interactions with goods markets and issues of capital availability are discussed in the final section.
Even in the partial equilibrium setting assumed here, the demand for labor is potentially influenced by a large number of factors: the cost of labor to firms; the prices of the factors of production that are used in conjunction with labor; conditions in goods markets; and by any regulations that influence the terms and conditions under which labor can be hired. The aggregate demand for labor in the economy as a whole, or in a sector of the economy, is also influenced by the entry and exit of firms, and by changes in the distribution of output and employment among existing firms. Under these conditions it is useful to adopt a number of simplifying assumptions.
Much of the theoretical and empirical work on the demand for labor has been based on the assumptions that employment decisions are made by a large number of identical price-taking firms, and that these decisions reflect objectives such as profit maximization or cost minimization. The profit maximization and price-taking assumptions give rise to the familiar downward-sloping demand curve for labor, measured against the real wage, which can be regarded as approximating firms’ labor demand in the medium run. However, firms’ labor demand decisions may not be well approximated by such a curve in the very short run. This survey distinguishes between two approaches to the demand for labor in the very short run: a Keynesian approach in which firms may be pushed off their medium-run labor demand curve in the very short run, principally as a result of fluctuations in aggregate demand; and a classical approach in which firms may operate off the medium-run labor demand curve mainly as a consequence of the costs associated with rapid adjustment of inputs. After considering these aspects of labor demand, the survey then turns to the factors that influence the demand for labor at a given level of production in the longer run, when all inputs are freely variable.
Studies of the demand for labor prior to the 1970s often did not explicitly consider the impact of intermediate input prices on the demand for labor. Sharp fluctuations in the prices of energy and other raw materials since the early 1970s, however, have elicited considerable interest in the significance of intermediate input prices for the behavior of labor demand. This survey will therefore discuss not only the influences of the costs of labor and of capital on labor demand, but will seek as well to identify the impact of intermediate input prices.
Theoretical Labor Demand Functions
As already mentioned, much empirical work has assumed that employment decisions are taken by a large number of identical firms that maximize profits. These firms are usually assumed to operate with wellbehaved production functions that provide a technological link between output and the inputs of production factors. The production function framework can be either in gross output or value added terms, depending on whether the impact of intermediate input prices on labor demand is explicitly taken into account and on the nature of the production technology.2
Gross output measures in principle the total value of production; it includes, therefore, the contributions of the material inputs that are used in the production process in addition to the contributions by labor and capital. Value added, on the other hand, measures only the contribution to production of labor and capital.3 Because of the large increases in the prices of intermediate inputs that occurred in the 1970s, many of the recent empirical studies of the demand for labor have been conducted in terms of a gross output production function framework (for example, Pindyck (1979), Lipschitz and Schadler (1984), Pindyck and Rotemberg (1983), Symons and Layard (1984)). Some other studies have, however, relied on the more traditional value added framework as a basis for firms’ employment decisions (for example, Bruno and Sachs (1985), Bruno (1985)). When the production function for gross output is separable between value added and intermediate inputs, either approach may be used interchangeably; each can in principle capture the impact of changes in intermediate input prices on the demand for labor. When the function is not separable, the traditional value added approach may be inappropriate (see Appendix II).
The specification of the labor demand function on the basis of a production function requires a hypothesis about the production technology. Two of the most commonly used specifications in the recent literature have been the Cobb-Douglas production function (equations (1) and (2)) and the constant elasticity of substitution (CES) production function (equations (3) and (4)) written here with no technical progress terms and on the assumption of separability between value added and intermediate inputs.4
Here, Qt and Vt refer respectively to gross output and value added while Lt, Kt, and Nt refer to the inputs of labor, capital, and intermediate inputs.
These production functions differ principally in their implications for the ease of substitution between factors of production and the behavior of factor cost shares. In the case of the Cobb-Douglas production function, factor shares are constant; the share of labor in value added is given by θL while the share of value added in gross output is given by θV. The Cobb-Douglas production function also has a relatively high degree of substitutability between factors, as indicated by a (Hicks-Allen) elasticity of substitution of unity.5 In the case of the CES production function, factor shares are no longer necessarily constant, and the elasticity of substitution between labor and capital is not restricted to be equal to unity. The elasticity of substitution between value added and intermediate inputs in the CES function is given by
In principle, the inputs appearing in the production function are the service flows of production factors, labor input, for example, being measured in terms of man-hours. Indeed, much of the international evidence on the demand for labor has been in terms of total hours worked. In general, the empirical literature has not dealt with the choice between employment variation through hours worked and that through changes in the number of persons employed. (For exceptions, see Hart (1984) and Riechel (1986).)
Demand for Labor in the Short to Medium Run
The demand for labor in the short to medium run is typically assumed to reflect the adjustment of labor and intermediate inputs, given a path for the capital stock. It allows therefore for changes in the demand for labor arising from changes in the level of production, and also for changes arising from substitution between labor and intermediate inputs. Optimal adjustment of variable inputs in the medium run is assumed to occur when the marginal product of each input is equated to its marginal cost in real terms. This condition results in the familiar downward-sloping medium-run demand for labor curve shown in Chart 6. Here employment in man-hours (Lt) is measured on the horizontal axis while the real cost of labor
The response of labor demand to a change in the real cost of labor in the medium run (measured in terms of the price of gross output, as in Chart 6) will comprise two kinds of effects: substitution and scale effects. The former captures any substitutions between labor and intermediate inputs that occur at a given level of gross output, in response to a change in the price of labor relative to intermediate input prices. The scale effect captures any changes in the demand for labor that result from firms changing their scale of production in response to a change in costs relative to revenues as implied, for example, by a change in real labor costs.
In the medium run, with the capital stock predetermined, the total response of labor demand to a change in the cost of labor, for example a reduction in real labor costs from A to B in Chart 6, depends on the ease of substitution between labor and intermediate inputs (which determines the size of the substitution effect) and on the share of labor in total costs (which influences the size of the scale effect). Other things being equal, the higher the elasticity of substitution between labor and intermediate inputs and the larger is the share of labor in total costs, the stronger will be the impact of a given change in real labor costs on the demand for labor. Under these conditions a reduction in real labor costs will induce the maximum substitution of labor for intermediate inputs; it will also lead to a large reduction in costs relative to revenues and a relatively large increase in profitability, and hence a significant increase in output and employment.
At least in the manufacturing sectors of the major industrial countries, there do not appear to be strong grounds for believing that there are systematic differences across countries in the ease of substitution between factors.8 In the current framework, therefore, any differences in the medium-run responsiveness of labor demand to changes in labor costs across countries tend to reflect differences in the share of labor in total costs and hence scale effects.
The demand for labor function that results from firms’ optimization problem in the medium run can be written either in gross output terms (equation (5)) or value added terms (equation (6)):9
Here,
In the case of Cobb-Douglas technology, the elasticities of labor demand with respect to a change in the real cost of labor measured in gross output or value added terms are the same. Theoretical estimates of the real labor cost elasticity can easily be computed for the manufacturing sectors of the industrial countries on the basis of the cost share of labor; such estimates give an idea of possible sizes of elasticities that may be uncovered in empirical work. Reflecting differences in cost shares, countries such as the United Kingdom with a relatively large share of labor in costs have relatively high wage elasticities (in absolute value), while Japan with its relatively smaller share has a lower elasticity (Table 1). As noted below, empirical estimates of medium-term real wage elasticities typically fall short of the elasticity estimates implied by Cobb-Douglas technology; this suggests that alternative production function specifications are likely to fit the facts better. Nevertheless, the Cobb-Douglas estimates are useful for illustrative purposes since the properties of this function are relatively simple. The Cobb-Douglas function has also formed the basis for the calculation of many wage-gap measures (see the section on the wage gap).
Theoretical Medium-Run Labor Demand Elasticities with Respect to Real Labor Costs, with Capital Stock Constant
Real labor costs are measured as nominal labor costs deflated by the gross output or value added deflator. These elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function. The elasticities are calculated from (see Appendix III):
Theoretical Medium-Run Labor Demand Elasticities with Respect to Real Labor Costs, with Capital Stock Constant
| Country | Elasticities1 |
|---|---|
| Canada | –3.0 |
| United States | –3.7 |
| Japan | –1.8 |
| France | –3.6 |
| Germany, Fed. Rep. of | –2.4 |
| Italy | –3.2 |
| United Kingdom | –4.8 |
Real labor costs are measured as nominal labor costs deflated by the gross output or value added deflator. These elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function. The elasticities are calculated from (see Appendix III):
Theoretical Medium-Run Labor Demand Elasticities with Respect to Real Labor Costs, with Capital Stock Constant
| Country | Elasticities1 |
|---|---|
| Canada | –3.0 |
| United States | –3.7 |
| Japan | –1.8 |
| France | –3.6 |
| Germany, Fed. Rep. of | –2.4 |
| Italy | –3.2 |
| United Kingdom | –4.8 |
Real labor costs are measured as nominal labor costs deflated by the gross output or value added deflator. These elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function. The elasticities are calculated from (see Appendix III):
In addition to being influenced by real labor costs, the demand for labor in the medium run can depend on the real prices of intermediate inputs, on the given path for the capital stock, and on the level of multifactor productivity (equation (5)). While a higher path for the capital stock or an improvement in technology are normally assumed to raise the demand for labor at a given real wage, by raising the marginal product of labor and hence shifting the labor demand curve outward, the impact of intermediate input prices on labor demand is generally uncertain, unless the production function is assumed to be separable in value added and intermediate inputs and all factors are cooperative in production.10 Under such assumptions, a rise in the real prices of intermediate inputs would tend to reduce the demand for labor in the medium run, even though labor is substituted for intermediate inputs at a given level of gross output. The negative effect comes about because higher intermediate prices would lead firms to reduce the scale of production (due to lowered profitability) and hence reduce the demand for labor, by enough to outweigh the positive substitution effect.
The medium-run labor demand function reflects the quantity of labor firms will wish to hire at different levels of real and relative labor costs. Alternatively, the demand function can be inverted to determine the maximum product wage that would be consistent with high-employment conditions (the “warranted product wage”). The latter approach is particularly useful in relation to the supply side of the labor market (see the next two sections). However the demand for labor is viewed, the size of the elasticity of labor demand with respect to real wages and the impact of other considerations on labor demand have crucial implications for the behavior of employment. In the medium run, changes in the demand for labor depend negatively on product wage growth and payroll taxes, with effects the size of which depends on the real wage elasticity of labor demand; labor demand is higher the more rapid is the growth of technology (total factor productivity) and the faster is the given rate of capital accumulation. Alternatively, at a given path for employment the warranted product wage will grow faster the more rapidly is capital accumulating and the faster is multifactor productivity growing; the warranted wage grows more slowly when payroll taxes increase.
The kind of labor demand function that applies in the medium run may not accurately describe employment decisions in the very short run. Much macroeconomic research has tended to regard very short-run employment fluctuations as reflecting directly shifts in aggregate demand at given real factor prices; it has consequently regarded firms as operating off their medium-run classical labor demand curves in the very short run. More recently, however, an alternative dynamic classical approach has been developed, according to which firms may be off their medium-run labor demand function due to adjustment costs. This alternative approach has significantly different implications from the standard Keynesian approach to the demand for labor in the very short run. According to the Keynesian approach, employment is assumed to respond directly to aggregate demand in the short run; in the classical approach, aggregate demand influences employment in the short run only through its implications for current and expected future factor prices.
Following Sargent (1979), the classical approach retains an exclusive role for factor prices in influencing labor demand in the very short run, and is based on an assumption of increasing costs associated with rapid adjustments in labor and other inputs. Because of such costs, it may be optimal for firms to adjust their input of labor gradually in response to changes in real labor costs and other factors. This typically implies only a weak relationship between employment and labor costs in the very short run, but with effects that increase over time. Under such conditions, firms do not always operate along the medium-run labor demand curve given in Chart 6; their employment decisions, however, will ultimately approach those implied by this demand curve. Starting off at point A in Chart 6, for example, a reduction in real labor costs to
This dynamic classical approach to labor demand can give rise to complicated short-run labor demand functions, in which current employment decisions reflect both past decisions as well as current and expected future movements in real labor costs. Firms’ reactions to changes in labor costs will be influenced by whether those changes are expected to be permanent or temporary, and employment may respond to expected future changes in costs as well as being in the process of responding to past changes. In several studies, it has been assumed for simplicity that the dynamic labor demand function can be represented in a simple form such as given by equation (7) (Bruno (1985), Bruno and Sachs (1985)):
Here employment
While the assumption of increasing costs associated with rapid adjustments of labor inputs retains a role for (current and expected future) real labor costs when firms are off their medium-run labor demand functions, the alternative Keynesian approach to very short-run labor demand decisions has been based on a direct link between fluctuations in aggregate demand and employment decisions; this approach has also in some cases allowed for lags due to costs of adjusting labor inputs, obscuring to some extent the distinction between the classical and Keynesian approaches.
A direct role for aggregate demand in influencing labor demand can arise when firms encounter some form of temporary quantity constraints in goods markets, preventing them from selling as much as they wish to supply. In this case, firms are viewed as being pushed off their medium-run labor demand curves (their “notional” demand for labor curves), leading to an effective demand for labor below the notional or desired demand. At a real cost of labor of
The evidence on each of these approaches to labor demand in the very short run has tended to be regarded as being of crucial importance to the debate over whether unemployment is classical or Keynesian in nature. The disagreement here, however, does not of itself concern whether aggregate demand fluctuations lead to variations in employment in the very short run. Rather, it concerns whether shifts in aggregate demand lead to variations in employment at given factor prices, or whether changes in factor prices, current or expected, are the principal channel by which such fluctuations in aggregate demand lead to changes in employment. According to the Keynesian view, the significance of real wage changes for employment in the very short run is dominated by a direct effect from aggregate demand changes; in the classical view, real wage and other factor price changes are the principal mechanism by which aggregate demand influences employment, even in the very short run.
Demand for Labor at a Given Level of Output
When all inputs are variable, the demand for labor is assumed to be based on the efficient use of all factors of production, including capital. This permits substitution between capital and labor, and between capital and other production factors. The long-run demand for labor schedule can be used to determine the mix of factors that will be employed at a given level of production and, hence, the factor intensity of alternative production levels.
At a given level of value added, the conditional demand for labor function can be written as:
where Rt(1 – D) measures the user cost of capital, net of any subsidies on the use of capital.13 Factor costs are here represented by nominal factor prices. A rise in the nominal cost of labor relative to that of capital will in general reduce the demand for labor at a given level of value added. A rise in labor costs will therefore lower the labor intensity of production and increase the capital intensity of production. Equivalently, at a given level of total factor productivity, a rise in labor costs will raise the productivity of labor while reducing that of capital. Because the level of value added is assumed to be constant, the elasticities discussed here reflect only a substitution between labor and capital.
Table 2 illustrates a set of constant-value-added labor demand elasticities based on Cobb-Douglas technology. Under Cobb-Douglas technology, and as explained in Appendix III, these elasticities can be calculated on the basis only of information on the share of labor in value added. The reported wage elasticities typically provide an upper bound for the estimated values of actual wage elasticities, since it is unlikely in practice that the elasticity of substitution between labor and capital will be as high as unity as implied by Cobb-Douglas technology. These estimates therefore can be regarded as suggesting the maximum impact of factor prices on labor demand; alternatively, since value added is held constant, they can be regarded as an indication of the maximum impact that a change in factor prices would be expected to have on the labor intensity of production.
Theoretical Labor Demand Elasticities with Respect to Nominal Wages Relative to Nominal Capital Costs, with Value Added Constant
The elasticities refer to manufacturing and are based on the twin assumptions of a linearly homogeneous Cobb-Douglas production function and separability between value added and intermediate inputs. Elasticities reflect the average shares of labor in value added over the period 1961–81. The numbers give the percentage fall in the demand for labor which would result from a 1 percent rise in the price of labor relative to capital, at a constant level of value added. The formula for computing the elasticities is given by ηL = (1 – θL) (see Appendix III), where θL = average share of labor in manufacturing value added.
Theoretical Labor Demand Elasticities with Respect to Nominal Wages Relative to Nominal Capital Costs, with Value Added Constant
| Country | Elasticities1 |
|---|---|
| Canada | –0.33 |
| United States | –0.27 |
| Japan | –0.55 |
| France | –0.28 |
| Germany, Fed. Rep. of | –0.42 |
| Italy | –0.32 |
| United Kingdom | –0.21 |
The elasticities refer to manufacturing and are based on the twin assumptions of a linearly homogeneous Cobb-Douglas production function and separability between value added and intermediate inputs. Elasticities reflect the average shares of labor in value added over the period 1961–81. The numbers give the percentage fall in the demand for labor which would result from a 1 percent rise in the price of labor relative to capital, at a constant level of value added. The formula for computing the elasticities is given by ηL = (1 – θL) (see Appendix III), where θL = average share of labor in manufacturing value added.
Theoretical Labor Demand Elasticities with Respect to Nominal Wages Relative to Nominal Capital Costs, with Value Added Constant
| Country | Elasticities1 |
|---|---|
| Canada | –0.33 |
| United States | –0.27 |
| Japan | –0.55 |
| France | –0.28 |
| Germany, Fed. Rep. of | –0.42 |
| Italy | –0.32 |
| United Kingdom | –0.21 |
The elasticities refer to manufacturing and are based on the twin assumptions of a linearly homogeneous Cobb-Douglas production function and separability between value added and intermediate inputs. Elasticities reflect the average shares of labor in value added over the period 1961–81. The numbers give the percentage fall in the demand for labor which would result from a 1 percent rise in the price of labor relative to capital, at a constant level of value added. The formula for computing the elasticities is given by ηL = (1 – θL) (see Appendix III), where θL = average share of labor in manufacturing value added.
Conditional factor demand functions may be considered at a constant level of gross output rather than value added (Pindyck (1979), Pindyck and Rotemberg (1983)) leading to some ambiguity in the interpretation of constant-activity factor demand elasticities (Appendix II). In this case, the labor demand function is stated explicitly in terms of the prices of intermediate inputs, as well as those of capital and labor:
Here
Theoretical Labor Demand Elasticities with Respect to Nominal Wages and Nominal Capital Costs, with Gross Output Constant
The elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function that is separable in value added and intermediate inputs. Whereas the elasticities in Table 2 are calculated at a constant level of value added, the elasticities in this table are calculated at a constant level of gross output, using the following formulae (see Appendix III):
Theoretical Labor Demand Elasticities with Respect to Nominal Wages and Nominal Capital Costs, with Gross Output Constant
| Elasticities With Respect to: 1 | ||
|---|---|---|
| Country | Nominal wage costs | Nominal capital costs |
| Canada | –0.70 | 0.15 |
| United States | –0.67 | 0.12 |
| Japan | –0.80 | 0.25 |
| France | –0.68 | 0.13 |
| Germany, Fed. Rep. of | –0.74 | 0.19 |
| Italy | –0.69 | 0.14 |
| United Kingdom | –0.64 | 0.09 |
The elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function that is separable in value added and intermediate inputs. Whereas the elasticities in Table 2 are calculated at a constant level of value added, the elasticities in this table are calculated at a constant level of gross output, using the following formulae (see Appendix III):
Theoretical Labor Demand Elasticities with Respect to Nominal Wages and Nominal Capital Costs, with Gross Output Constant
| Elasticities With Respect to: 1 | ||
|---|---|---|
| Country | Nominal wage costs | Nominal capital costs |
| Canada | –0.70 | 0.15 |
| United States | –0.67 | 0.12 |
| Japan | –0.80 | 0.25 |
| France | –0.68 | 0.13 |
| Germany, Fed. Rep. of | –0.74 | 0.19 |
| Italy | –0.69 | 0.14 |
| United Kingdom | –0.64 | 0.09 |
The elasticities refer to manufacturing and are based on a linearly homogeneous Cobb-Douglas production function that is separable in value added and intermediate inputs. Whereas the elasticities in Table 2 are calculated at a constant level of value added, the elasticities in this table are calculated at a constant level of gross output, using the following formulae (see Appendix III):
Empirical Evidence on Labor Demand
In reviewing the evidence on the demand for labor, three aspects are of particular interest. The first of these concerns the impact of real wages on labor demand in the medium run (as given by the medium-run real wage elasticity) as well as the impact of intermediate input prices (as given by the medium-run intermediate input price elasticity). The size of these elasticities determines the degree to which labor demand is likely to respond to real product wage and intermediate input price developments in the medium run. The second concerns employment fluctuations in the very short run, and the significance of aggregate demand (Keynesian) versus real wage (classical) considerations in explaining short-run changes in labor demand. Finally, evidence on the significance of factor prices for labor demand at a given level of output and, hence, for the labor intensity of production is of importance in attempting to explain differences in labor market developments across countries.
The Short and Medium Run
The demand for labor in the medium run is assumed to reflect the optimal adjustment of labor (and of intermediate inputs) at a given path of the capital stock. The response of labor demand to a change in real wages (or to a change in the real prices of intermediate inputs) reflects therefore two considerations: substitution possibilities between labor and intermediate inputs, and scale effects.
Table 4 (Columns 3 and 4) shows recent estimates of medium-run real wage elasticities and, where available, elasticities of labor demand with respect to real intermediate input prices.14 The estimates of these elasticities fluctuate somewhat across the studies reviewed. This reflects to some degree slightly different definitions of variables, different proxies for the capital stock, different lag specifications, and differences in the sectors being studied; it also reflects that real product wages in some cases are measured in value added terms and, in others, in gross output terms. For most countries, however, the medium-run real wage elasticities tend to be above unity in absolute value; in some cases they exceed two. The chief exceptions are France and the United States where the elasticities average close to –0.5.15 The estimated elasticities are generally significantly below those that are implied by Cobb-Douglas technology (Table 1), and also disagree with the rankings implied by the latter. Were Cobb-Douglas technology appropriate, countries with a relatively low share of labor in costs (for example, Japan) would be expected to have relatively low real wage elasticities; countries with relatively high labor shares (for example, the United Kingdom) would be expected to have relatively high wage elasticities.
Labor Demand Elasticities with Respect to Labor Costs and Intermediate Input Prices, with Capital Stock Constant
Symons and Layard (1984) report two alternative estimates based on different lag specifications; the second Symons and Layard estimate, when included, refers to what they regard as a preferred estimate.
Lipschitz and Schadler (1984), in contrast to the other studies reported here, do not allow for lagged adjustment.
Labor Demand Elasticities with Respect to Labor Costs and Intermediate Input Prices, with Capital Stock Constant
| Very Short Run | Medium Run | |||
|---|---|---|---|---|
| Real labor costs | Intermediate input prices | Real labor costs | Intermediate input prices | |
| Canada | ||||
| Bruno (1985) | –0.1 | … | –2.4 | … |
| Symons and Layard (1984)1 | –0.5 | –0.1 | –2.6 | –1.8 |
| Layard and Nickell (1984) | … | … | … | … |
| United States | ||||
| Bruno (1985) | 0.1 | … | 0.2 | … |
| Symons and Layard (1984)1 | –0.2 | –0.8 | –1.3 | –4.6 |
| Symons and Layard (1984) | … | … | –0.6 | –3.4 |
| Layard and Nickell (1984) | –0.1 | … | –0.4 | … |
| Japan | ||||
| Bruno (1985) | –0.2 | … | –2.7 | … |
| Symons and Layard (1984)1 | –0.0 | 0.0 | –2.4 | –2.6 |
| Lipschitz and Schadler (1984)2 | … | … | –0.8 | –0.3 |
| Layard and Nickell (1984) | –0.0 | … | –1.1 | … |
| France | ||||
| Bruno (1985) | –0.2 | … | –2.0 | … |
| Symons and Layard (1984)1 | 0.0 | 0.2 | –0.3 | 0.0 |
| Symons and Layard (1984) | … | … | –0.3 | –0.1 |
| Layard and Nickell (1984) | –0.0 | … | –0.6 | … |
| Germany, Fed. Rep. of | ||||
| Bruno (1985) | –0.1 | … | … | … |
| Symons and Layard (1984)1 | 0.2 | 0.1 | –0.4 | –1.2 |
| Symons and Layard (1984) | … | … | –1.8 | –2.1 |
| Layard and Nickell (1984) | –0.1 | … | –1.5 | … |
| United Kingdom | ||||
| Bruno (1985) | –0.1 | … | –1.0 | … |
| Symons and Layard (1984)1 | … | … | –1.8 | –0.4 |
| Lipschitz and Schadler (1984)2 | … | … | –1.8 | –0.3 |
| Layard and Nickell (1984) | –0.1 | … | –0.9 | … |
Symons and Layard (1984) report two alternative estimates based on different lag specifications; the second Symons and Layard estimate, when included, refers to what they regard as a preferred estimate.
Lipschitz and Schadler (1984), in contrast to the other studies reported here, do not allow for lagged adjustment.
Labor Demand Elasticities with Respect to Labor Costs and Intermediate Input Prices, with Capital Stock Constant
| Very Short Run | Medium Run | |||
|---|---|---|---|---|
| Real labor costs | Intermediate input prices | Real labor costs | Intermediate input prices | |
| Canada | ||||
| Bruno (1985) | –0.1 | … | –2.4 | … |
| Symons and Layard (1984)1 | –0.5 | –0.1 | –2.6 | –1.8 |
| Layard and Nickell (1984) | … | … | … | … |
| United States | ||||
| Bruno (1985) | 0.1 | … | 0.2 | … |
| Symons and Layard (1984)1 | –0.2 | –0.8 | –1.3 | –4.6 |
| Symons and Layard (1984) | … | … | –0.6 | –3.4 |
| Layard and Nickell (1984) | –0.1 | … | –0.4 | … |
| Japan | ||||
| Bruno (1985) | –0.2 | … | –2.7 | … |
| Symons and Layard (1984)1 | –0.0 | 0.0 | –2.4 | –2.6 |
| Lipschitz and Schadler (1984)2 | … | … | –0.8 | –0.3 |
| Layard and Nickell (1984) | –0.0 | … | –1.1 | … |
| France | ||||
| Bruno (1985) | –0.2 | … | –2.0 | … |
| Symons and Layard (1984)1 | 0.0 | 0.2 | –0.3 | 0.0 |
| Symons and Layard (1984) | … | … | –0.3 | –0.1 |
| Layard and Nickell (1984) | –0.0 | … | –0.6 | … |
| Germany, Fed. Rep. of | ||||
| Bruno (1985) | –0.1 | … | … | … |
| Symons and Layard (1984)1 | 0.2 | 0.1 | –0.4 | –1.2 |
| Symons and Layard (1984) | … | … | –1.8 | –2.1 |
| Layard and Nickell (1984) | –0.1 | … | –1.5 | … |
| United Kingdom | ||||
| Bruno (1985) | –0.1 | … | –1.0 | … |
| Symons and Layard (1984)1 | … | … | –1.8 | –0.4 |
| Lipschitz and Schadler (1984)2 | … | … | –1.8 | –0.3 |
| Layard and Nickell (1984) | –0.1 | … | –0.9 | … |
Symons and Layard (1984) report two alternative estimates based on different lag specifications; the second Symons and Layard estimate, when included, refers to what they regard as a preferred estimate.
Lipschitz and Schadler (1984), in contrast to the other studies reported here, do not allow for lagged adjustment.
In addition, in all the major industrial countries other than the United Kingdom and France, there appears to be a significant negative impact of intermediate input prices on the demand for labor in the medium run (Table 4). In some cases, the estimated elasticity is of the same magnitude as the real wage elasticity; in the case of the United States it is much larger (in absolute value). It is apparent, as well, that the average intermediate input price elasticities vary considerably across countries.
In contrast to the relatively high medium-term elasticities, empirical estimates of short-run labor demand elasticities seem to point to only a minor role for real wages in influencing the demand for labor within a quarter. The evidence thus appears to confirm the Keynesian view referred to earlier. The estimated short-run wage elasticities (reported in Table 4) are mostly small and in some cases have the “wrong” sign.16 While the elasticities are small, however, they are in most cases different from zero; the average elasticities are typically in the range of –0.1 to –0.2.
In examining these elasticities, it is worth bearing in mind some facts about the original studies. (For a complete description of each study, the reader is referred to the references in the bibliography.) In brief, the Symons and Layard (1984) estimates apply to manufacturing, and estimation is carried out using Ordinary Least Squares, with and without instrumental variables, from the first quarter of 1956 through the fourth quarter of 1980. The first Symons and Layard estimates in the table—the medium-run estimates—come from their Table 2 and are based on instrumental variables, except in the case of the United Kingdom estimates which are based on Table 1 in their paper and hence on Ordinary Least Squares. The first, very short-run estimates recorded in the table come from Table 4 in the Symons and Layard paper, and are based on instrumental variables, except in the case of the United Kingdom for which the very short-run estimates are based on Table 3 in Symons and Layard, and on Ordinary Least Squares estimation techniques. The very short-run elasticities recorded in the table are the impact effects (within a quarter) of changes in real wages and raw material prices.
The Lipschitz and Schadler (1984) study is also based on manufacturing. It uses annual data and applies only to the United Kingdom and Japan. Their estimates are based on the simultaneous estimation of a gross output production function and a labor demand function, at a predetermined capital stock, with allowance for cyclical effects as captured by the inclusion in each equation of a cyclical variable. Lipschitz and Schadler do not allow for lagged adjustment of labor input as a result of adjustment costs, and the medium-run elasticities reported in the table come from their Table 4, and are the coefficients attached to real wages and real raw material prices in their labor demand equations. Note that in the case of the United Kingdom, the real raw material price was lagged one period.
The Bruno (1985) study is also based on manufacturing data and involves the estimation of a labor demand function based on the slow adjustment of actual to desired labor input. Unlike Symons and Layard, and Lipschitz and Schadler, the Bruno estimates are based on the real wage measured in value added terms; raw materials prices are not included in the labor demand functions. The medium-run wage elasticities reported in the table are based on annual data over the period 1961–82 and come from Table 6 in Bruno (1985); they are derived as the ratio of the estimated coefficient on the wage variable to one minus the coefficient on the lagged dependent variable. The very short-run elasticities are derived from the impact effect of wages in the labor demand function, divided by four to obtain a quarterly estimate that can be compared with other studies that are based on quarterly data.
Time Lags in the Adjustment of Labor Demand to Desired Levels, with Capital Stock Constant
Bruno (1985). These numbers are based on a labor demand function specification of the form
Symons and Layard (1984). The first column reports the mean lag on wages in quarters from Table 2 in Symons and Layard; the second column shows the preferred lag lengths (the number of lagged wage terms in their preferred equation; see their Table 5) as described by Symons and Layard. Note, as indicated in Table 4, the Symons and Layard estimates are based on quarterly data, and hence lag lengths are in quarters.
Not significantly different from zero at two decimal places.
Time Lags in the Adjustment of Labor Demand to Desired Levels, with Capital Stock Constant
| Country | Proportion of Adjustment Made over Four Quarters1 | Lag in Response of Employment to Real Wage Cost Changes (In quarters)2 | |
|---|---|---|---|
| Canada | —3 | 5.3 | 4 |
| United States | 0.66 | 4.8 | 1 |
| Japan | 0.38 | 19.8 | 4 |
| France | 0.29 | 2.9 | 1 |
| Germany, Fed. Rep. of | —3 | 5.2 | 3 |
| United Kingdom | 0.59 | … | … |
Bruno (1985). These numbers are based on a labor demand function specification of the form
Symons and Layard (1984). The first column reports the mean lag on wages in quarters from Table 2 in Symons and Layard; the second column shows the preferred lag lengths (the number of lagged wage terms in their preferred equation; see their Table 5) as described by Symons and Layard. Note, as indicated in Table 4, the Symons and Layard estimates are based on quarterly data, and hence lag lengths are in quarters.
Not significantly different from zero at two decimal places.
Time Lags in the Adjustment of Labor Demand to Desired Levels, with Capital Stock Constant
| Country | Proportion of Adjustment Made over Four Quarters1 | Lag in Response of Employment to Real Wage Cost Changes (In quarters)2 | |
|---|---|---|---|
| Canada | —3 | 5.3 | 4 |
| United States | 0.66 | 4.8 | 1 |
| Japan | 0.38 | 19.8 | 4 |
| France | 0.29 | 2.9 | 1 |
| Germany, Fed. Rep. of | —3 | 5.2 | 3 |
| United Kingdom | 0.59 | … | … |
Bruno (1985). These numbers are based on a labor demand function specification of the form
Symons and Layard (1984). The first column reports the mean lag on wages in quarters from Table 2 in Symons and Layard; the second column shows the preferred lag lengths (the number of lagged wage terms in their preferred equation; see their Table 5) as described by Symons and Layard. Note, as indicated in Table 4, the Symons and Layard estimates are based on quarterly data, and hence lag lengths are in quarters.
Not significantly different from zero at two decimal places.
Labor Demand Elasticities with Respect to Labor Costs, Capital Costs, and Intermediate Input Prices, with Value Added or Gross Output Constant
Elasticities were not available separately for intermediate input prices and capital costs. Reported elasticities are based on the constraint that elasticities sum to zero under zero degree homogeneity of the demand functions in factor prices.
Labor Demand Elasticities with Respect to Labor Costs, Capital Costs, and Intermediate Input Prices, with Value Added or Gross Output Constant
| Elasticity of Labor Demand with Respect to: | ||||
|---|---|---|---|---|
| Wages | Capital costs | Intermediate prices | ||
| Canada | 0.66 | 1 | ||
| Pindyck (1979) (gross output) | –0.66 | … | … | |
| United States | 0.65 | 1 | ||
| Pindyck (1979) (gross output) | –0.65 | … | … | |
| Pindyck and Rotemberg (1983) (gross output) | –0.78 | –0.02 | 0.65 (materials) | |
| 0.15 (energy) | ||||
| Japan | 0.46 | 1 | ||
| Pindyck (1979) (gross output) | –0.46 | … | … | |
| France | 0.4 | 1 | ||
| Pindyck (1979) (gross output) | –0.4 | … | … | |
| Germany, Fed. Rep. of | 0.47 | 1 | ||
| Pindyck (1979) (gross output) | –0.47 | … | … | |
| Franz and König (1985) (value added) | –0.97 | –0.01 | –0.17 | |
| United Kingdom | ||||
| Owen (1985) (gross output) | –0.25 | 0.10 | 0.15 | |
| Pindyck (1979) (gross output) | 0.23 | 1 | ||
| –0.23 | ||||
| Nickell (1981) (value added) | –0.19 | 0.19 | … | |
Elasticities were not available separately for intermediate input prices and capital costs. Reported elasticities are based on the constraint that elasticities sum to zero under zero degree homogeneity of the demand functions in factor prices.
Labor Demand Elasticities with Respect to Labor Costs, Capital Costs, and Intermediate Input Prices, with Value Added or Gross Output Constant
| Elasticity of Labor Demand with Respect to: | ||||
|---|---|---|---|---|
| Wages | Capital costs | Intermediate prices | ||
| Canada | 0.66 | 1 | ||
| Pindyck (1979) (gross output) | –0.66 | … | … | |
| United States | 0.65 | 1 | ||
| Pindyck (1979) (gross output) | –0.65 | … | … | |
| Pindyck and Rotemberg (1983) (gross output) | –0.78 | –0.02 | 0.65 (materials) | |
| 0.15 (energy) | ||||
| Japan | 0.46 | 1 | ||
| Pindyck (1979) (gross output) | –0.46 | … | … | |
| France | 0.4 | 1 | ||
| Pindyck (1979) (gross output) | –0.4 | … | … | |
| Germany, Fed. Rep. of | 0.47 | 1 | ||
| Pindyck (1979) (gross output) | –0.47 | … | … | |
| Franz and König (1985) (value added) | –0.97 | –0.01 | –0.17 | |
| United Kingdom | ||||
| Owen (1985) (gross output) | –0.25 | 0.10 | 0.15 | |
| Pindyck (1979) (gross output) | 0.23 | 1 | ||
| –0.23 | ||||
| Nickell (1981) (value added) | –0.19 | 0.19 | … | |
Elasticities were not available separately for intermediate input prices and capital costs. Reported elasticities are based on the constraint that elasticities sum to zero under zero degree homogeneity of the demand functions in factor prices.
The Layard and Nickell (1984) study covers aggregate economy-wide employment and is based on annual data. The wage variable is in value added terms and intermediate input prices do not appear in the labor demand function. The medium-run elasticities reported in the table come from Table 1 in Layard and Nickell (1984); the very short-run elasticities are the impact of real wages on labor demand in the Layard and Nickell equations, divided by four to obtain a quarterly elasticity that is comparable to other elasticities in the table.
The link between the very short-run and the medium-run wage elasticities already discussed is provided by the length of time required for labor demand to adjust to a change in real wages at a given path for the capital stock (Table 5). Adjustment lags are generally relatively long with the most rapid adjustment occurring in France. In view of the perception of a high degree of flexibility in Japanese labor markets, it is surprising that the Japanese lag is significantly longer than elsewhere; the reason may be the high value of the wage elasticity in the medium term. Overall, the evidence suggests that real labor costs do have implications for labor demand in the short run, with effects that are at first relatively small but that increase after some time.
Because a significant proportion of the observed short-run employment fluctuations appears not to be explicable by changes in real labor costs (given the relatively small real wage elasticities), the question arises as to the sources of recorded employment fluctuations in the very short run. In their search for alternative explanations, analysts have attempted to clarify the respective roles of intermediate input prices and of aggregate demand in influencing labor demand in the very short run.
There is as yet very little agreement, and a number of studies have reached essentially conflicting results (Geary and Kennan (1982), Bruno and Sachs (1985), Bruno (1985), Symons and Layard (1984), Layard and Nickell (1985)).17 It appears that the ability of factor prices to account for employment fluctuations in the very short run, and particularly in the 1970s and 1980s, depends to an important degree on whether real intermediate input prices are also included in the labor demand function. For example, Symons and Layard (1984), who include real intermediate input prices in the labor demand function, conclude that changes in factor prices can “explain” employment fluctuations in the very short run, except in the United States. Bruno (1985), who does not include intermediate input prices in the labor demand function, reaches the opposite conclusion (see Appendix II for further discussion of this point).
While the significance of factor prices for labor demand in the very short run thus remains an open question, two pieces of evidence are clear. The first of these is that the size of employment fluctuations over the typical business cycle is usually too large to be accounted for by current real wage movements alone: given estimates of real wage elasticities and the magnitude of real wage movements, employment seems to vary by more than is justified by real wage movements in the short run (see, for example, Lipschitz and Schadler (1984)). The second is that labor demand functions without intermediate input prices do not typically fit well over the last two business cycles.
In Bruno’s (1985) application of the Keynesian approach, which includes aggregate demand as an explanatory variable in the labor demand function, it appears that current factor prices (particularly labor costs) have been of less direct importance in “explaining” the growth of labor demand in the period after 1978 than was the case between 1974 and 1978, particularly in Europe. By contrast, the role of aggregate demand increased significantly in the 1978–82 period compared with 1974–78.
In summary, the disagreement about the appropriate specification of the very short-run labor demand function is not about whether aggregate demand fluctuations in the short run may affect employment. It concerns, rather, the nature of the transmission mechanism that is involved. In the classical interpretation, employment fluctuations over the business cycle are regarded as being explained by movements in current and expected factor prices, which themselves may reflect underlying changes in aggregate demand. In the Keynesian view, aggregate demand is assumed to directly influence the demand for labor.
Demand for Labor at a Given Level of Production
When all production factors including capital are variable, the mix of factors that will be employed to produce a given level of output can be freely varied. If the production function is weakly separable in value added and intermediate inputs, the mix can be considered interchangeably at a constant level of value added or gross output; it can be considered therefore either in terms of the prices of labor and capital only, or including as well the prices of intermediate inputs. When separability does not hold, the prices of intermediate inputs must usually be taken into account.
At a constant level of value added, the assumptions of Cobb-Douglas technology, value added separability, and of an elasticity of substitution between labor and capital of unity, theoretically place an upper bound on the elasticity of labor demand with respect to nominal wages of approximately –0.2 to –0.4, except in the case of Japan for which the upper bound is higher in absolute value (Table 2). Unfortunately, the only empirical estimate available is for the United Kingdom: this is –0.19 (Table 6).18 The estimated constant gross output wage elasticities are generally significantly smaller (in absolute value) than the theoretical elasticities shown in Table 3, which are based on Cobb-Douglas technology. The elasticity of the demand for labor with respect to the nominal wage is typically in the range (at a constant level of gross output) of –0.4 to –0.6.
The results in Table 6, in addition to identifying a role for nominal wages in influencing labor demand, also suggest a role for capital costs, as well as for the prices of intermediate inputs. A disaggregation of inputs into raw materials and energy is found sometimes to give rise to complementarities between capital and energy, hence violating value added separability, and to ambiguous effects of energy prices on labor demand (see Artus (1984) for a discussion of these issues). Because of the relatively small share of energy in total costs in most industrial countries, however, any effect of oil price changes may not be important empirically for macro demand for labor functions (see Bruno and Sachs (1985)).
Some information about the studies is relevant to an interpretation of Table 6. The Pindyck (1979) study is based on the specification and estimation of trans-logarithmic cost functions for the manufacturing sectors of the reported countries. The approach gives rise to non-constant price elasticities, and those reported in the table come from Table 4 in Pindyck; they are evaluated at sample means of cost shares. The study is based on three factors, labor, capital, and energy and on account of non-reporting of data on cost shares it was not possible to obtain estimates of cross-price elasticities.
The Pindyck and Rotemberg (1983) study is based on the specification and estimation of a four-factor, dynamic demand model, using trans-logarithmic cost functions for materials and energy and the estimation of first-order conditions for factors which are assumed to be subject to quadratic adjustment costs. The estimates recorded in the table come from Table 2 in Pindyck and Rotemberg and are referred to, by them, as long-run elasticities. These are calculated at sample means of cost shares. Alternative estimates by Pindyck and Rotemberg (see for example their Table 3) give very similar results, with a marginally lower wage elasticity in absolute value.
The Franz and König (1985) study is based on the specification of a labor demand function for the German economy. The estimates recorded in this table come from Table 17 in Franz and König, and are based on Ordinary Least Squares. To convert these estimates into a long-run format, the coefficients attached to wages, the rental rate and raw material prices have here been divided by one minus the coefficient on the lagged dependent variable.
The Owen (1985) study is based on the United Kingdom’s manufacturing sector and the source for this study was the U.K. Treasury report on the relationship between employment and wages in the United Kingdom. The Owen study is based on three productive factors: labor, capital, and raw materials. Finally, the Nickell (1981) study is also drawn from the U.K. Treasury report and it applies to the U.K. manufacturing sector.
Both the longer-run and shorter-run factor price elasticities implicitly contain information about the ease of substitution between labor and capital. Under the assumption of separability of the production function in value added and intermediate inputs, implied consensus estimates of labor-capital substitution elasticities are given in Table 7. These elasticities are generally found to be below unity, suggesting again that Cobb-Douglas technology may not be the most appropriate assumption for this type of analysis. Substitution possibilities clearly exist and are important, but not to the high degree implied by Cobb-Douglas technology.
Consensus Estimates of Hicks-Allen Elasticities of Substitution Between Labor and Capital1
The estimates recorded are based principally on Artus (1984) and on Lipschitz and Schadler (1984), on the assumption of separability between value added and intermediate inputs. The estimates are taken from Artus (1984), Table 3, and from Table 3 in Lipschitz and Schadler (1984). In addition, the estimates are based on Pindyck (1979) and Griffin and Gregory (1976). All of the estimates refer to the manufacturing sector of the countries concerned. Bruno (1985) quotes an average value for the elasticity of substitution of 0.7. The wage elasticity estimates of Bruno (1985), as reported in Table 4 (medium-run wage elasticities) can be converted into elasticities of substitution by multiplying by the average share of capital in value-added (one third in most countries, slightly higher in Japan). The resulting elasticities fall in the ranges depicted above except in the cases of the United States and United Kingdom.
Consensus Estimates of Hicks-Allen Elasticities of Substitution Between Labor and Capital1
| Canada | 0.5 – 0.8 |
| United States | 0.5 – 0.9 |
| Japan | 0.4 – 0.8 |
| France | 0.5 – 0.8 |
| Germany, Fed. Rep. of | 0.7 – 0.8 |
| United Kingdom | 0.6 – 0.8 |
The estimates recorded are based principally on Artus (1984) and on Lipschitz and Schadler (1984), on the assumption of separability between value added and intermediate inputs. The estimates are taken from Artus (1984), Table 3, and from Table 3 in Lipschitz and Schadler (1984). In addition, the estimates are based on Pindyck (1979) and Griffin and Gregory (1976). All of the estimates refer to the manufacturing sector of the countries concerned. Bruno (1985) quotes an average value for the elasticity of substitution of 0.7. The wage elasticity estimates of Bruno (1985), as reported in Table 4 (medium-run wage elasticities) can be converted into elasticities of substitution by multiplying by the average share of capital in value-added (one third in most countries, slightly higher in Japan). The resulting elasticities fall in the ranges depicted above except in the cases of the United States and United Kingdom.
Consensus Estimates of Hicks-Allen Elasticities of Substitution Between Labor and Capital1
| Canada | 0.5 – 0.8 |
| United States | 0.5 – 0.9 |
| Japan | 0.4 – 0.8 |
| France | 0.5 – 0.8 |
| Germany, Fed. Rep. of | 0.7 – 0.8 |
| United Kingdom | 0.6 – 0.8 |
The estimates recorded are based principally on Artus (1984) and on Lipschitz and Schadler (1984), on the assumption of separability between value added and intermediate inputs. The estimates are taken from Artus (1984), Table 3, and from Table 3 in Lipschitz and Schadler (1984). In addition, the estimates are based on Pindyck (1979) and Griffin and Gregory (1976). All of the estimates refer to the manufacturing sector of the countries concerned. Bruno (1985) quotes an average value for the elasticity of substitution of 0.7. The wage elasticity estimates of Bruno (1985), as reported in Table 4 (medium-run wage elasticities) can be converted into elasticities of substitution by multiplying by the average share of capital in value-added (one third in most countries, slightly higher in Japan). The resulting elasticities fall in the ranges depicted above except in the cases of the United States and United Kingdom.
In order to provide a rough assessment of the significance of relative factor prices for labor demand, Table 8 indicates the change in the cost of labor relative to capital since 1973 in the five major industrial economies. It is evident that the price of labor has risen significantly faster than that of capital in all countries, other than the United States. Using an estimate of –0.2 for the elasticity of labor demand with respect to the relative price of labor (at a constant level of value added) the table also indicates the implied fall in the demand for labor over the period, other things being equal.19 The resulting factor substitution effects range from an employment decline of about 13 percent in the United Kingdom, to an essentially insignificant effect in the United States.
The Relative Price of Labor and Capital and Factor Substitution Effects
Calculated at estimate of labor demand elasticity of –0.2.
The Relative Price of Labor and Capital and Factor Substitution Effects
| Country | Ratio of the Price of Labor to the Price of Capital 1981 (1973 = 100) | Implied Fall in Labor Demand: 1981 over 1973 (Percent)1 |
|---|---|---|
| United States | 104.1 | 0.8 |
| Japan | 119.8 | 3.9 |
| France | 123.9 | 4.8 |
| Germany, Fed. Rep. of | 117.7 | 3.5 |
| United Kingdom | 166.1 | 13.2 |
Calculated at estimate of labor demand elasticity of –0.2.
The Relative Price of Labor and Capital and Factor Substitution Effects
| Country | Ratio of the Price of Labor to the Price of Capital 1981 (1973 = 100) | Implied Fall in Labor Demand: 1981 over 1973 (Percent)1 |
|---|---|---|
| United States | 104.1 | 0.8 |
| Japan | 119.8 | 3.9 |
| France | 123.9 | 4.8 |
| Germany, Fed. Rep. of | 117.7 | 3.5 |
| United Kingdom | 166.1 | 13.2 |
Calculated at estimate of labor demand elasticity of –0.2.
This section has established an important role for the real cost of labor in influencing the demand for labor, especially in the medium run. It has served also to identify the significance of factor prices for the labor intensity of production and the significance of intermediate input prices. As such, the review has served to provide bounds on the “likely” magnitude of elasticities and, hence, on the significance of labor costs in particular and factor prices in general (Table 9). While the empirical evidence suggests that there are some differences in wage elasticities across countries in the short to medium run, such differences appear to be of less significance at a given level of output. Recorded differences in employment behavior across countries therefore reflect more than just differences in real wage elasticities on the demand side of the labor market. They also seem to reflect interactions with the supply side of labor markets as well as with goods markets, subjects discussed in the following sections.
Consensus Views on Labor Demand Elasticities with Respect to Labor Costs
The constant output elasticities are estimated either at a constant level of value added or gross output; they do not therefore include output variations except insofar as value added is allowed to change when gross output is held constant. Because these elasticities do not include output effects, they typically are smaller (in absolute value) than the medium-run and short-run elasticities which are calculated allowing output to change.
Drazen et al. (1985) find an average value for this elasticity across the 10 major OECD countries of –0.2.
Average values of elasticities.
Consensus Views on Labor Demand Elasticities with Respect to Labor Costs
| Medium Run | Very Short Run | |||||
|---|---|---|---|---|---|---|
| Estimated at:1 | Calculated at given path for the capital stock | Calculated at given capital stock and less than complete labor adjustment | ||||
| Country | Constant value added2 | Constant gross output3 | ||||
| Canada | … | –0.7 | –2.4– | –2.6 | –0.1– | –0.5 |
| United States | –0.2– –0.5 | –0.7 | –0.4– | –1.3 | –0.2– | 0.1 |
| Japan | … | –0.5 | –0.8– | –2.7 | –0.0– | –0.2 |
| France | … | –0.4 | –0.3– | –2.0 | –0.0– | –0.2 |
| Germany, Fed. Rep. of | … | –0.7 | –0.4– | –1.8 | –0.1– | 0.2 |
| United Kingdom | –0.2 | –0.2 | –0.9– | –1.8 | –0.1– | –0.1 |
The constant output elasticities are estimated either at a constant level of value added or gross output; they do not therefore include output variations except insofar as value added is allowed to change when gross output is held constant. Because these elasticities do not include output effects, they typically are smaller (in absolute value) than the medium-run and short-run elasticities which are calculated allowing output to change.
Drazen et al. (1985) find an average value for this elasticity across the 10 major OECD countries of –0.2.
Average values of elasticities.
Consensus Views on Labor Demand Elasticities with Respect to Labor Costs
| Medium Run | Very Short Run | |||||
|---|---|---|---|---|---|---|
| Estimated at:1 | Calculated at given path for the capital stock | Calculated at given capital stock and less than complete labor adjustment | ||||
| Country | Constant value added2 | Constant gross output3 | ||||
| Canada | … | –0.7 | –2.4– | –2.6 | –0.1– | –0.5 |
| United States | –0.2– –0.5 | –0.7 | –0.4– | –1.3 | –0.2– | 0.1 |
| Japan | … | –0.5 | –0.8– | –2.7 | –0.0– | –0.2 |
| France | … | –0.4 | –0.3– | –2.0 | –0.0– | –0.2 |
| Germany, Fed. Rep. of | … | –0.7 | –0.4– | –1.8 | –0.1– | 0.2 |
| United Kingdom | –0.2 | –0.2 | –0.9– | –1.8 | –0.1– | –0.1 |
The constant output elasticities are estimated either at a constant level of value added or gross output; they do not therefore include output variations except insofar as value added is allowed to change when gross output is held constant. Because these elasticities do not include output effects, they typically are smaller (in absolute value) than the medium-run and short-run elasticities which are calculated allowing output to change.
Drazen et al. (1985) find an average value for this elasticity across the 10 major OECD countries of –0.2.
Average values of elasticities.
Flexibility of Real Wages
The preceding section established that the evolution of real labor costs has important implications for the quantity of labor that firms wish to employ. However, this in itself is not sufficient to give rise to an unemployment problem. For this to happen, the level of real wages must move out of line with the warranted wage level, which is the wage that would be compatible with high-employment conditions, and which equals the marginal product of labor at full employment. In this context, the flexibility of real wages plays a crucial role in determining the rise in unemployment which can be expected to result from adverse disturbances—such as those experienced during the 1970s and early 1980s—that affect the level of the warranted wage. If real wages adjust fully to a reduction in the marginal product of labor and, hence, in the warranted real wage following an adverse disturbance, employment can be expected to be maintained at a high level. However, if real wages are inflexible in the face of shocks, firms will be forced to bring actual labor productivity back into line with real wages by reducing employment relative to other inputs in production. This may occur through firms either making adjustments in their factor mix or shutting down unprofitable labor-intensive lines of production.
This section introduces the supply side of the labor market and attempts to identify the reasons for the marked differences in real wage developments across countries that have been observed over the past two decades. Particular emphasis is placed on differences in the degree of real wage flexibility. In this regard an important distinction is made between the real product wage which measures the total cost of labor in relation to the prices firms receive for their products, and the real consumption wage which measures the wage received by workers in relation to the prices they pay for consumption goods. The former is the relevant concept from the point of view of labor demand (as discussed in the previous section) while the latter is the relevant concept for labor supply.20 Changes in the terms of trade, payroll taxes, and indirect taxes, create a wedge between movements in these two wage measures. Given the interaction between labor supply and demand in determining real wages and employment, if real consumption wages are rigid in the face of adverse disturbances, such as a deterioration in the terms of trade, this may lead to an increase in product wages (relative to their warranted level) and consequently to a decline in employment.
Theoretical Framework
In order to explain differences in wage behavior across countries, it is necessary to supplement the labor demand theory with a model of the supply side of the labor market. The combined labor supply and demand framework will then be used to develop and clarify a number of interrelated concepts such as real and nominal wage rigidity; the distinction between the real product wage and the real consumption wage; the real wage gap; and the NAIRU (nonaccelerating inflation rate of employment).
Two alternative theoretical frameworks will be discussed. The first is based on a perfectly competitive market-clearing approach, which illustrates the consequences of various disturbances for employment in a well-functioning labor market. The second framework, which allows for rigidities in wage formation, helps to illustrate the consequences of disturbances when real wages are not perfectly flexible. A comparison of the results of these two approaches serves to highlight the importance of real wage flexibility in determining the consequences for employment which result from an adverse disturbance. Based on the empirical evidence, the rigidities approach appears to be a more accurate characterization of the recent functioning of labor markets in the industrial countries, particularly in Europe.
Well-Functioning Labor Markets
Following the analysis in the previous section, the demand for labor function may be expressed by equation (10), where the notation is the same as in the earlier section and where α0 to α5 are positive and represent the structural parameters of the labor demand function.
This equation is based on a gross output production function, hence the inclusion of intermediate input prices, and is applicable to the medium term (the capital stock
On the assumption that wage earners’ choice between work and leisure is based on intertemporal optimization, the supply of labor can be expressed by the following equation:
where the deflator for the wage is the consumer price index which is defined as a geometric average of the prices of home goods (marked up by the indirect tax rate, TI) and imports:
β0 to β4 represent the structural parameters of the labor supply function and are defined to be positive. TP is the personal tax rate net of transfers. In order to allow for intertemporal substitution possibilities, W* represents the “normal” or the expected future consumption wage. Real interest rates (rt) are assumed to influence intertemporal substitution decisions by altering the incentive to work in one period relative to another. Zt represents all remaining (exogenous) factors that influence the supply of labor, including the population of working age, the level of unemployment benefits, and wealth. Current wages and interest rates are assumed to influence the supply of labor with a positive sign; expected future wages are assumed to have a negative impact.21
It is clear from equations (10) and (11) that employment will depend on the behavior of both product wages and consumption wages. It is convenient initially, however, to abstract from differences in the behavior of these wage concepts and focus on the product wage. Changes in the relative price of imports and home goods are therefore ignored at this stage. Combining equations (10) and (11) and setting λ = 1 results in the following expression for the market-clearing product wage, which will be referred to as the warranted real product wage in the remainder of this paper:
The level of employment which corresponds to the warranted real product wage (the equilibrium level of employment) can be computed by substituting
The three equations (10), (11), and (12) may be used to illustrate the consequences for real wages and employment of various disturbances or shocks that lower the marginal product of labor in an economy that is characterized by a well-functioning labor market. For example, a decline in total factor productivity at a given capital stock will reduce the marginal product of labor at all levels of employment (α4 and α5 are positive). This shift of the labor demand function will reduce the warranted and the actual real product wage. The resulting movement along the labor supply function will lead to a lower equilibrium level of employment as well ((β1) is also positive). However, because wages are flexible, the actual real wage declines with the warranted real wage, thereby minimizing the adverse consequences of the shock for employment.
Other disturbances which shift either the labor demand or supply schedules would affect real product wages in a similar fashion. An increase in employment taxes (TW) would constitute a negative shock to labor demand since it would create a wedge between the cost of labor to the firm and its marginal product, which would reduce the warranted real product wage and, hence, the actual real wage. Similarly, a disturbance which reduces labor supply, such as an increase in import prices relative to domestic prices or an increase in the personal tax rate, would tend to raise warranted product and consumption wages as well as actual real wages, and to lower employment. An increase in labor supply would have the opposite effect. In all of these examples, the labor market remains in equilibrium because actual real wages adjust fully to changes in the marginal product of labor.
Labor Markets Characterized by Rigidities
The market-clearing framework (equations (10) to (12)) can be adapted to illustrate the implications of a disturbance which reduces the warranted real product wage in a situation when actual real wages do not adjust commensurately (potential reasons for such rigidities in real wages will be discussed below). In this case, the labor demand and labor supply equations cannot be satisfied simultaneously and employment will be assumed to be determined by the labor demand function alone. If, for example, total factor productivity declines but the actual level of real product wages remains the same, then employment will fall below its equilibrium level. In other words, if real product wages do not adjust to the resulting reduction in the labor productivity schedule, employment will be reduced in order to bring labor productivity back into line with the level of real wages. Thus, real wage rigidity in the face of disturbances which reduce the warranted wage will tend to raise the rate of unemployment.
The rigidities approach is often formalized in the following manner. The demand for labor equation is expressed in proportional growth rate terms as:
Here Ṡ may be interpreted as the warranted rate of growth of real product wages.22 In line with the analysis presented earlier, any factor which affects the marginal product of labor at existing employment levels would be captured by Ṡ. For example, a slowdown in total factor productitivy growth, a rise in the relative price of intermediate inputs, a rise in taxes on employment, or a decline in the stock of capital would be expected to lead to a reduction in Ṡ.23 It again follows that a decline in the rate of growth of the warranted product wage relative to the actual product wage will reduce the quantity of labor demanded. Equation (13) is occasionally written with the change in the unemployment rate as the dependent variable. In that case it follows that a decline in the rate of growth of the warranted product wage relative to the actual product wage will tend to raise the rate of unemployment.
The labor supply hypothesis associated with (13) explicitly recognizes the existence of rigidities in wage formation by replacing the competitive market labor supply function with a real wage adjustment function which applies to a labor market with widespread use of labor contracts (equation (14)). It is assumed that firms can hire any number of workers at the agreed wage.
The wage adjustment function states that nominal wage growth is based on a target growth rate for gross real consumption wages (ṠT) which is translated into nominal terms using the expected rate of consumer price inflation
The two-equation system (13) and (14) then determines the rate of change of real wages and the associated rate of unemployment. These equations can be used to illustrate the consequences of an adverse real disturbance such as a permanent slowdown in total factor productivity growth. In this event, the warranted rate of growth of real product wages (Ṡ) will decline; if target real consumption wage growth remains unchanged the rate of unemployment must increase sufficiently (and permanently) to bring actual product wage increases into line with the lower rate of increase in the warranted product wage. As firms adjust employment in accordance with their labor demand function, the long-run change in unemployment that would result from a permanent reduction in the warranted rate of growth of real product wages, assuming that domestic and foreign prices rise at the same rate, would be given by:24
Here Ṡ1 is the new lower rate of increase of the warranted wage. If the real wage target (ṠT) were to adjust to the new warranted real wage path, no increase in unemployment would need to occur. Equation (15) also suggests that the more responsive wage behavior is to labor market conditions (the higher is the coefficient β), the smaller would be the rise in the rate of unemployment that would be expected to result from a disturbance to the economy which permanently lowers the rate of growth of the warranted wage.
An important implication of equation (15) is that a permanent reduction in the rate of growth of the warranted wage—with no change in target real consumption wage growth—leads to a permanently higher rate of unemployment because the required reduction in wage growth has to be sustained. In the case of a temporary reduction in the growth of warranted wages, a temporary increase in unemployment would be sufficient to bring the rate of growth of actual real wages back into line with that of the warranted real wage.
In order to analyze the implications of terms-of-trade disturbances it is important to note that workers’ wage targets, when measured in product prices, will be affected by changes in the terms of trade. This follows by substituting the expression for the consumer price index into equation (14) to give:
According to (14′) an expected deterioration in the terms of trade, other things being equal, will result in higher wage demands in terms of expected output prices. With an unchanged rate of growth of the warranted product wage unemployment will increase until the rate of increase of actual consumption wages has been reduced by the amount of the terms of trade deterioration; this would bring actual product wage growth in line with warranted product wage growth.
A number of recent studies have attempted to use estimates of the responsiveness of real consumption wages to unemployment to construct measures of real wage rigidity.25 Such measures of real consumption wage rigidity depend in part on the way price expectations are assumed to be formed. Several authors (see, for example, Layard et al. (1984)) have assumed that price expectations can be approximated by a distributed lag on past inflation rates. On this assumption, the wage equation can be expressed as follows:26
Empirical estimates of the coefficients in equation (16) have been used to construct both short-run and long-run measures of wage rigidity (see, for example, Grubb, Jackman, and Layard (1982) and Coe (1985)). A commonly used measure of long-run real wage rigidity is (1/β) which is the increase in unemployment “required” to reduce the rate of growth of real wages by 1 percentage point when wage targets do not change and price expectations are realized. The short-run measure allows for inertia in nominal wage formation, which may occur, for example, in countries where wage indexation is of little importance. In such cases, a disturbance which raises prices may in fact reduce actual real wage growth substantially, at least for a period, thereby moderating the effect on unemployment of the disturbance, even if the parameter β is small. To capture such effects, the measure of short-run real wage rigidity usually is defined as the short-run impact coefficient of inflation in relation to the coefficient of unemployment in the wage equation (α1/β) (see equation (16)).
This framework may also be used to determine the NAIRU—the rate of unemployment that is necessary to stabilize the rate of wage inflation and to bring the growth rate of actual real product wages into line with that of the warranted product wage following an adverse disturbance. In the long run, when price expectations are realized, the actual unemployment rate will tend to converge to the NAIRU. This rate therefore represents the equilibrium unemployment rate for an economy. Any attempt to use traditional demand expansion policies to reduce the actual unemployment rate below the NAIRU will result in accelerating inflation and a squeeze on profits which ultimately must lead to a reversal of this policy.
The NAIRU can be expressed in terms of either the original equilibrium unemployment rate or the actual rate of unemployment.
In the first case, the NAIRU will increase relative to the original equilibrium unemployment rate if the target growth rate of real consumption wages increases, or if the growth of the warranted real consumption wage (Ṡc) decelerates following a disturbance to the economy. The second equation states that the NAIRU will increase relative to the actual unemployment rate if wage increases accelerate, for example in response to a rise in import prices. This equation also implies that when actual unemployment is above the NAIRU, real consumption wages will grow by less than their warranted rate.
These expressions are based on Layard et al. (1984), and are derived from the long-run equilibrium wage and price equations, where the latter is based on the labor demand function (equation (13)). It is important to note that estimates of the NAIRU are dependent on the specification of the underlying wage and price equations as well as the sector of the economy to which they apply. In Coe (1985), for example, the NAIRU is determined principally by trend productivity growth, by trend changes in the terms of trade, and by social insurance taxes paid by firms. In the context of the framework presented in this paper, the NAIRU may be considered to be the natural rate of unemployment.
Two features of this rate which contrast with the natural rate of unemployment as defined by Friedman (1968) should be noted. First, because the underlying wage equations assume adaptive expectations, empirical estimates of the NAIRU are consistent only with a constant inflation path. Second, following a shock which raises the NAIRU above U0, when the actual unemployment rate equals the NAIRU there is an excess supply of labor as indicated by the fact that actual real wages increase by less than the target growth rate of real wages. This contrasts with Friedman’s case where there is no excess supply of labor because of the assumption of an important competitive segment in the labor market.
Studies of Wage Behavior
The market-clearing (equilibrium) and the rigidities (disequilibrium) frameworks for wage determination have both been applied in the literature. However, there have been relatively few studies based on the equilibrium approach and these studies in general have not confirmed the applicability of the approach. For example, Andrews and Nickell (1982) compare the results of the market-clearing and rigidities approaches in explaining the rise in unemployment in the United Kingdom from 1950 to 1979. They conclude that, because the labor supply effects suggested by the equilibrium model appear to be inconsistent with the existing evidence on labor supply behavior, the results from the disequilibrium approach are superior. Studies for the United States by Altonji (1982), Ashenfelter and Card (1982), and Clark and Summers (1982) also fail to support the market-clearing approach to labor market developments based on the intertemporal substitution hypothesis.
Studies of wage determination based on disequilibrium approaches, including the traditional Phillip’s Curve, have generally been more successful than those based on the market-clearing approach. Concentrating on those comparative international studies that attempt to classify countries according to their degree of wage flexibility, the disequilibrium studies reviewed here can be grouped into two broad categories: studies which focus on inertia in nominal wage formation, and those which focus on the response of real wages to changing economic conditions. As a general caveat it is important to note that these studies measure the flexibility of real consumption wages rather than real product wages; the latter is the relevant concept for employment decisions. For most countries, and particularly in the case of Japan, the wage data presented in the next section indicate that real product wages have decelerated less than real consumption wages over the period under review. This would suggest that real product wages in general have been more rigid than real consumption wages.
In one of the first studies of inertia in wage formation, Sachs (1979) defines wage rigidity in terms of the speed with which price shocks feed into nominal wages. If price increases are passed through to wages quickly, real wages are considered to be rigid while nominal wages are considered to be flexible. Sachs’ empirical work for the period 1963 to 1978 suggests that the European countries and Japan have short lags from price changes to nominal wages, implying that these countries are characterized by relatively rigid real wages but flexible nominal wages. Conversely, in the case of the United States, the relatively slow response of nominal wages to price changes suggests that nominal wages are rigid but that real wages are flexible. These apparent differences in wage behavior are consistent with the observed differences in the behavior of employment in the face of the disturbances experienced in the 1970s, and in particular as a result of the first oil price shock. Branson and Rotemberg (1980) also found significantly greater short-run nominal inertia in U.S. wages than in the other major industrial countries. Gordon (1982), using different criteria, including the standard deviation of wages relative to the standard deviation of hours worked, and the relative responsiveness of wages compared with hours worked, concluded that real wages are more flexible in the United States and in Japan, than in Europe.
Grubb, Jackman, and Layard (1983) broadened the focus to include the effects of changes in the rate of unemployment. Their definition of real wage rigidity, which is consistent with the definitions discussed previously, indicates the amount of unemployment that would result from a disturbance to the economy, and which would be necessary to stabilize wage growth. In addition, combining several properties of the wage equation, they define nominal wage rigidity as the long-run measure of real wage rigidity (1/β) multiplied by the average lag in the wage equation ((1 – α1)/α1).
The empirical results (Table 10) point to considerable differences in the measured degree of real and nominal wage rigidity across countries.27 Contrary to the conclusion of some of the earlier studies, which focused solely on nominal inertia, Japan is now found to have the lowest level of real wage rigidity in both the short and the long run. By contrast, in the case of most European countries, real wage rigidity is high, indicating that unemployment needs to increase substantially to moderate the growth of real wages. According to these definitions, the United States is found to have above-average real wage rigidity in the long run. The empirical measures of nominal wage rigidity in general conform to the results of the previous studies: Japan and Europe have fairly flexible nominal wages whereas nominal wages in the United States are relatively rigid.
Estimates of Real and Nominal Wage Rigidity1
The estimates by Coe are based on semiannual data for the entire private economy whereas the estimates for Grubb, Jackman, and Layard are based on annual data only for the manufacturing sector.
In comparing the intercept terms of Coe’s wage equations for the United States and other countries it should be noted that different wage variables are used in these equations. The wage variable for the United States is an hourly earnings measure whereas for other countries a compensation per employee measure is used. This latter measure will be influenced by the strong negative trend in average weekly hours worked which may affect estimates of the intercept term in the wage equation.
Estimates of Real and Nominal Wage Rigidity1
| Estimates of Real Wage Rigidity | ||||||||
|---|---|---|---|---|---|---|---|---|
| Short run | Long run | Estimates of Nominal Wage Rigidity | Estimates of Intercept Term in Wage Equations | |||||
| Country | Coe | Grubb, Jackman, and Layard | Coe | Grubb, Jackman, and Layard | Coe | Grubb, Jackman, and Layard | Coe2 | Grubb, Jackman, and Layard |
| Canada | 0.54 | 0.60 | 1.67 | 1.56 | 0.81 | 0.77 | 5.29 | 5.00 |
| United States | 0.67 | 1.09 | 3.06 | 4.17 | 3.35 | 3.14 | 2.58 | 1.73 |
| Japan | 0.28 | 0.13 | 0.28 | 0.12 | 0.14 | 0.18 | –3.34 | 18.77 |
| France | 1.52 | 0.51 | 3.03 | 0.59 | 4.56 | 0.29 | 2.36 | 8.18 |
| Germany, Fed. Rep. of | 0.58 | 1.54 | 0.61 | 1.30 | 1.16 | 0.36 | 0.88 | 6.38 |
| Italy | 1.48 | 1.17 | 1.48 | 1.14 | 4.44 | 0.61 | 5.84 | 10.23 |
| United Kingdom | 1.94 | 2.39 | 5.82 | 2.38 | 4.85 | 0.78 | 2.13 | 4.00 |
The estimates by Coe are based on semiannual data for the entire private economy whereas the estimates for Grubb, Jackman, and Layard are based on annual data only for the manufacturing sector.
In comparing the intercept terms of Coe’s wage equations for the United States and other countries it should be noted that different wage variables are used in these equations. The wage variable for the United States is an hourly earnings measure whereas for other countries a compensation per employee measure is used. This latter measure will be influenced by the strong negative trend in average weekly hours worked which may affect estimates of the intercept term in the wage equation.
Estimates of Real and Nominal Wage Rigidity1
| Estimates of Real Wage Rigidity | ||||||||
|---|---|---|---|---|---|---|---|---|
| Short run | Long run | Estimates of Nominal Wage Rigidity | Estimates of Intercept Term in Wage Equations | |||||
| Country | Coe | Grubb, Jackman, and Layard | Coe | Grubb, Jackman, and Layard | Coe | Grubb, Jackman, and Layard | Coe2 | Grubb, Jackman, and Layard |
| Canada | 0.54 | 0.60 | 1.67 | 1.56 | 0.81 | 0.77 | 5.29 | 5.00 |
| United States | 0.67 | 1.09 | 3.06 | 4.17 | 3.35 | 3.14 | 2.58 | 1.73 |
| Japan | 0.28 | 0.13 | 0.28 | 0.12 | 0.14 | 0.18 | –3.34 | 18.77 |
| France | 1.52 | 0.51 | 3.03 | 0.59 | 4.56 | 0.29 | 2.36 | 8.18 |
| Germany, Fed. Rep. of | 0.58 | 1.54 | 0.61 | 1.30 | 1.16 | 0.36 | 0.88 | 6.38 |
| Italy | 1.48 | 1.17 | 1.48 | 1.14 | 4.44 | 0.61 | 5.84 | 10.23 |
| United Kingdom | 1.94 | 2.39 | 5.82 | 2.38 | 4.85 | 0.78 | 2.13 | 4.00 |
The estimates by Coe are based on semiannual data for the entire private economy whereas the estimates for Grubb, Jackman, and Layard are based on annual data only for the manufacturing sector.
In comparing the intercept terms of Coe’s wage equations for the United States and other countries it should be noted that different wage variables are used in these equations. The wage variable for the United States is an hourly earnings measure whereas for other countries a compensation per employee measure is used. This latter measure will be influenced by the strong negative trend in average weekly hours worked which may affect estimates of the intercept term in the wage equation.
Applying essentially the same methodology to data for the entire private sector, Coe (1985) arrives at similar conclusions (Table 10). Japan is found to have the lowest degree of real wage rigidity in both the short and the long run whereas the United Kingdom is characterized by the most rigid real wages of all the industrial countries covered by this study. Due to significant nominal inertia, the United States has a relatively low degree of real wage rigidity in the short run. However, in the long run the degree of real wage rigidity is about the same as in Europe.
That the United States appears to suffer from the same degree of real wage rigidity in the longer run as European countries is obviously a controversial result. Indeed, a wide array of information—including differences in labor market regulations and inter-industry wage differentials—points to more flexible labor markets in the United States than in Europe. It is also generally recognized that the average wage data used in these studies are masking greater wage dispersion in the United States than in Europe. In particular, wages paid by new firms or to new workers do seem to be more flexible in the United States as suggested by the recent widespread adoption of two-tier wage contracts. Another important deficiency of this kind of analysis is that the rigidity concepts are based on partial responses which do not fully reflect the complete properties of the wage equations. Indeed, there is one important aspect of the wage equations underlying these measures of rigidity which is not captured. That is the differences in intercept terms which seem to suggest that the trend increases in European wages are higher than in the case of the United States.28
The empirical evidence of wage behavior may thus be summarized as follows. There appears to be a consensus that Japan has significantly less rigid, and therefore more flexible, real wages than other industrial countries. In the United States, real wages do seem to be relatively flexible in the short run. Even though the measures based on partial responses suggest that in the long run, real wages are no more flexible in the United States than in Europe, the estimated wage equations generally point to stronger trend increases in European wages than in U.S. wages. Hence, for the same unemployment and inflation performance, the estimated wage equations indicate that real wages tend to increase significantly faster in Europe. Thus, under similar economic conditions, differences in wage behavior are likely to result in less employment creation in Europe than in the United States.
Stability of Real Wage Behavior
As discussed earlier, there has been a significant deceleration in real labor cost increases in Europe over the past decade, and particularly in the most recent period. The reasons underlying this development have important implications for future employment behavior. If the slower growth in real product wages is attributable to an adjustment of real wage targets to the slower trend growth in warranted real wages or, alternatively, to an increase in the responsiveness of wages to unemployment (that is, an increase in β in equation (14)), this would suggest that an adjustment process leading to an increase in the labor intensity of production is under way. If, on the other hand, the deceleration of real wage increases is solely a reflection of the surge in unemployment, the adjustment process would be much longer and could require a high level of unemployment over a long period.
It is difficult to test for changes in the responsiveness of real wage targets to variations in economic conditions. Wage targets are not directly observable and any empirical work must rely on more or less appropriate proxy variables. Grubb, Jackman, and Layard (1983) attempted first to test for changes over time in the constant term in an equation where the constant term is assumed to capture the real wage target. This test did not reveal any statistically significant changes in the constant terms. In a second test, they assumed that the real wage target is an adaptive function of past real wage developments. This test, which was conducted on data for the area of the Organization for Economic Cooperation and Development (OECD) as a whole, suggested that real wage targets do seem to have begun to adjust to the slower growth of warranted real wages since 1973, but that the adjustment process is very slow. In fact, the results indicate that very little adjustment takes place in the first seven years following a disturbance which reduces the warranted real wage path. Grubb, Jackman, and Layard have also conducted tests of the responsiveness of real wages to unemployment without finding evidence of an increase in responsiveness in the period from 1973 to 1980. In most cases, tests of the stability of the wage equations overall failed to reveal significant changes in wage behavior after 1973.29
Coe (1985) also found that wage equations are stable after 1973. With the exception of those for the United Kingdom, Coe’s equations seem to be stable after 1979 as well. These results to some extent contradict the finding of many analysts that there has been a general tendency for wage equations to overpredict actual real wage developments in recent years. Coe’s results suggest that these “systematic” prediction errors arise because of the inclusion in many of the equations of a non-linear unemployment term which implies that the effect of an additional percentage point increase in the unemployment rate declines as unemployment rises. If unemployment is included linearly, wage equations seem to fit the data more closely, and the deceleration in wages after 1980 is reasonably well explained by the increase in unemployment. Overall, there does not seem to be strong evidence of any “break” in wage behavior, either after 1973 or in the most recent period. The evidence will obviously have to be reassessed as more data become available.
How Much Are Real Wages Out of Line?
The previous two sections suggested that real labor costs play a crucial role in employment decisions and that the flexibility of real wages differs considerably across countries. This section examines how these differences in wage flexibility have interacted with the adverse disturbances that have been experienced during the 1970s and 1980s. The main question addressed concerns the extent to which real wages have moved out of line with the level that would be warranted from a high-employment perspective.
Developments in Real Wages
Wage developments have differed substantially across countries, both before and after the first round of oil price increases. According to internationally comparable data on hourly labor costs for the manufacturing sectors of the major industrial countries, the United States has had the smallest increase in real product wages over the past two decades (Table 11).30 Although increases in real product wages in Europe decelerated significantly after 1975, they continued to exceed labor cost increases in the United States. Real product wages have increased substantially faster in Japan than in the other major countries.
Developments in Real Product Wages, Real Consumption Wages, and Employment in Manufacturing
(Percent changes at annual rates)
Developments in Real Product Wages, Real Consumption Wages, and Employment in Manufacturing
(Percent changes at annual rates)
| Real Product Wages | Real Consumption Wages | Employment (Total hours worked) | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Country | 1965–1974 | 1975–1984 | 1965–1984 | 1965–1974 | 1975–1984 | 1965–1984 | 1965–1974 | 1975–1984 | 1965–1984 |
| Canada | 4.6 | 2.1 | 3.3 | 3.4 | 1.8 | 2.6 | 1.4 | –1.2 | 0.1 |
| United States | 3.1 | 2.6 | 2.8 | 1.4 | 0.5 | 1.0 | 1.3 | –0.2 | 0.6 |
| Japan | 11.7 | 8.0 | 9.8 | 8.7 | 1.5 | 5.1 | 0.9 | 0.4 | 0.7 |
| France | 7.3 | 4.5 | 5.9 | 5.3 | 3.6 | 4.4 | 0.1 | –3.0 | –1.5 |
| Germany, Fed. Rep. of | 6.2 | 3.3 | 4.8 | 6.6 | 2.6 | 4.5 | –1.1 | –2.2 | –1.6 |
| Italy | 6.7 | 3.3 | 5.0 | 7.3 | 2.7 | 4.9 | 0.3 | –1.9 | –0.8 |
| United Kingdom | 5.2 | 2.1 | 3.7 | 4.3 | 2.5 | 3.4 | –1.8 | –3.7 | –2.8 |
Developments in Real Product Wages, Real Consumption Wages, and Employment in Manufacturing
(Percent changes at annual rates)
| Real Product Wages | Real Consumption Wages | Employment (Total hours worked) | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Country | 1965–1974 | 1975–1984 | 1965–1984 | 1965–1974 | 1975–1984 | 1965–1984 | 1965–1974 | 1975–1984 | 1965–1984 |
| Canada | 4.6 | 2.1 | 3.3 | 3.4 | 1.8 | 2.6 | 1.4 | –1.2 | 0.1 |
| United States | 3.1 | 2.6 | 2.8 | 1.4 | 0.5 | 1.0 | 1.3 | –0.2 | 0.6 |
| Japan | 11.7 | 8.0 | 9.8 | 8.7 | 1.5 | 5.1 | 0.9 | 0.4 | 0.7 |
| France | 7.3 | 4.5 | 5.9 | 5.3 | 3.6 | 4.4 | 0.1 | –3.0 | –1.5 |
| Germany, Fed. Rep. of | 6.2 | 3.3 | 4.8 | 6.6 | 2.6 | 4.5 | –1.1 | –2.2 | –1.6 |
| Italy | 6.7 | 3.3 | 5.0 | 7.3 | 2.7 | 4.9 | 0.3 | –1.9 | –0.8 |
| United Kingdom | 5.2 | 2.1 | 3.7 | 4.3 | 2.5 | 3.4 | –1.8 | –3.7 | –2.8 |
While the implications of real wage developments cannot be assessed independently of developments in the warranted real wage, the observed differences in product wage growth between the United States and Europe appear to confirm the negative relationship between real labor costs and employment discussed earlier. In Japan, while the unemployment rate has remained low, there also has been a significant slowdown in the rate of growth of manufacturing employment. This development seems, at least in part, to have been related to the strong growth in real product wages. However, to the extent the slowdown in employment growth in Japan also reflected slower growth of the labor force, the relatively strong rise in real wages may be interpreted as a normal market response to labor becoming relatively more scarce.
The same general ordering of performances holds for increases in gross real consumption wages, which in most cases have fallen below the growth rates of real product wages, reflecting terms of trade losses and increases in indirect taxes. The rate of increase of real consumption wages also decelerated significantly in the period 1975–84 compared with the previous decade, and again the deceleration was more pronounced in Europe than in the United States. However, the gap between U.S. and European consumption wage increases in the latter period was even larger than in the case of real product wages. Japan recorded a particularly sharp deceleration in real consumption wage increases in the 1975–84 period.
Shocks Versus Rigidities
With reference to the market-clearing real wage equation developed earlier (equation (12) in the previous section), it is clear that the major industrial countries have experienced a number of severe disturbances or shocks during the period considered which have had important implications for warranted wages. Some of the more important shocks have been the rise in oil prices, the secular decline in total factor productivity growth, the rise in real interest rates, and the sharp rise in taxes on employment in many countries. Direct and indirect tax rates have also increased dramatically over the period, further reducing the scope for increases in net real consumption wages relative to real product wages. From the selection of indicators in Table 12 it is evident that the disturbances have varied substantially across countries, and that in general they have been more severe in Europe and Japan than in the United States.
Selected Indicators of Disturbances in Major Industrial Countries
(Compound annual percentage rates of change)
Ratio of employers’ contributions to social security schemes to compensation of employees for the total economy.
Selected Indicators of Disturbances in Major Industrial Countries
(Compound annual percentage rates of change)
| Total Factor Productivity Growth | Changes in Terms of Trade | Employers’ Contributions to Social Security Schemes1 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Country | 1961–73 | 1974–81 | 1982–85 | 1961–73 | 1974–81 | 1982–85 | 1965 | 1970 | 1975 | 1980 |
| Canada | … | … | … | 0.3 | –0.5 | –1.2 | 2.0 | 3.1 | 3.8 | 3.7 |
| United States | 1.8 | 0.5 | 1.3 | –0.2 | –2.8 | 6.0 | 3.3 | 3.8 | 4.9 | 5.5 |
| Japan | … | … | … | –0.9 | –6.9 | 2.3 | 4.6 | 5.3 | 6.1 | 7.0 |
| France | 4.3 | 1.0 | 1.3 | 1.0 | –2.3 | 2.0 | 18.2 | 19.1 | 20.5 | 22.2 |
| Germany, Fed. Rep. of | 3.0 | 1.6 | 1.3 | 1.9 | –3.3 | 1.1 | 9.3 | 10.3 | 12.3 | 13.0 |
| Italy | … | … | … | –1.0 | –3.9 | 2.7 | 16.2 | 17.3 | 18.9 | 17.6 |
| United Kingdom | 2.2 | 0.0 | 2.4 | — | 0.4 | –0.3 | 3.9 | 4.4 | 6.0 | 6.1 |
Ratio of employers’ contributions to social security schemes to compensation of employees for the total economy.
Selected Indicators of Disturbances in Major Industrial Countries
(Compound annual percentage rates of change)
| Total Factor Productivity Growth | Changes in Terms of Trade | Employers’ Contributions to Social Security Schemes1 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Country | 1961–73 | 1974–81 | 1982–85 | 1961–73 | 1974–81 | 1982–85 | 1965 | 1970 | 1975 | 1980 |
| Canada | … | … | … | 0.3 | –0.5 | –1.2 | 2.0 | 3.1 | 3.8 | 3.7 |
| United States | 1.8 | 0.5 | 1.3 | –0.2 | –2.8 | 6.0 | 3.3 | 3.8 | 4.9 | 5.5 |
| Japan | … | … | … | –0.9 | –6.9 | 2.3 | 4.6 | 5.3 | 6.1 | 7.0 |
| France | 4.3 | 1.0 | 1.3 | 1.0 | –2.3 | 2.0 | 18.2 | 19.1 | 20.5 | 22.2 |
| Germany, Fed. Rep. of | 3.0 | 1.6 | 1.3 | 1.9 | –3.3 | 1.1 | 9.3 | 10.3 | 12.3 | 13.0 |
| Italy | … | … | … | –1.0 | –3.9 | 2.7 | 16.2 | 17.3 | 18.9 | 17.6 |
| United Kingdom | 2.2 | 0.0 | 2.4 | — | 0.4 | –0.3 | 3.9 | 4.4 | 6.0 | 6.1 |
Ratio of employers’ contributions to social security schemes to compensation of employees for the total economy.
Grubb, Jackman, and Layard (1983) have calculated the implications of the import price rises and productivity declines across countries for the period 1973 to 1980 relative to the period 1960 to 1972 (Table 13). Their estimates suggest that the warranted rate of growth of real product wages fell by about 1¼ percent per annum more in the European Community than in the United States in the latter part of the 1970s. Additional evidence on changes in total factor productivity growth (Table 12) also suggests that Europe has experienced a more serious slowdown in underlying productivity growth than the United States.
Reduction in the Rate of Growth of the Warranted Real Wage, 1973–80 Compared with 1960–721
Calculated as the average annual growth rate in the period 1973–80 minus the average annual growth rate in the period 1960–72, in percentage points. Negative shocks are expressed as positive numbers.
Reduction in the Rate of Growth of the Warranted Real Wage, 1973–80 Compared with 1960–721
| Country | Total Reduction | Reduction Due to Slower Labor Productivity Growth | Reduction Due to Higher Import Prices |
|---|---|---|---|
| Canada | 2.15 | 1.53 | 0.62 |
| United States | 1.72 | 1.28 | 0.44 |
| Japan | 6.13 | 5.26 | 0.87 |
| France | 2.15 | 1.79 | 0.36 |
| Germany, Fed. Rep. of | 2.21 | 1.61 | 0.60 |
| Italy | 4.75 | 3.77 | 0.98 |
| United Kingdom | 1.75 | 1.09 | 0.66 |
| EC average | 2.95 | 1.86 | 1.09 |
| OECD average | 2.63 | 1.77 | 0.86 |
Calculated as the average annual growth rate in the period 1973–80 minus the average annual growth rate in the period 1960–72, in percentage points. Negative shocks are expressed as positive numbers.
Reduction in the Rate of Growth of the Warranted Real Wage, 1973–80 Compared with 1960–721
| Country | Total Reduction | Reduction Due to Slower Labor Productivity Growth | Reduction Due to Higher Import Prices |
|---|---|---|---|
| Canada | 2.15 | 1.53 | 0.62 |
| United States | 1.72 | 1.28 | 0.44 |
| Japan | 6.13 | 5.26 | 0.87 |
| France | 2.15 | 1.79 | 0.36 |
| Germany, Fed. Rep. of | 2.21 | 1.61 | 0.60 |
| Italy | 4.75 | 3.77 | 0.98 |
| United Kingdom | 1.75 | 1.09 | 0.66 |
| EC average | 2.95 | 1.86 | 1.09 |
| OECD average | 2.63 | 1.77 | 0.86 |
Calculated as the average annual growth rate in the period 1973–80 minus the average annual growth rate in the period 1960–72, in percentage points. Negative shocks are expressed as positive numbers.
These data suggest that the interactions between rigidities and disturbances have played an important role in the differences in wage developments across countries. For example, taking into account differences in overall labor market conditions, the greater real wage flexibility in Japan relative to Europe is consistent with the more pronounced deceleration in real wage growth in Japan after 1975. By implication, these interactions therefore also help to explain the differences in employment behavior, in rates of growth of labor productivity, and in capital intensity across countries.
The complex relationships between real wages and employment may be summarized as follows (Table 14). The faster increase in real labor costs imply that Europe is on a slower employment growth/faster labor productivity growth path than the United States. This was true also prior to 1973. In the period 1960 to 1973, for quite similar real output growth performances, Europe experienced much less employment growth than the United States (0.2 percent annually versus 2.0 percent), much faster labor productivity growth (4.4 percent versus 2.0 percent), and a much faster increase in the capital-labor ratio (6.2 percent annually in Germany, for example, versus 2.3 percent in the United States). During the 1960s, the relatively rapid growth in European labor costs was essentially consistent with the high rate of growth of total factor productivity and the slow growth of the European labor force (indeed, to judge from the inflows of foreign workers, labor was in excess demand in this period). During the 1970s, however, the rate of increase of actual wages became incompatible with the maintenance of high employment conditions. Although European wage increases did decelerate significantly, because of the existence of rigidities in wage formation, they did not decelerate by as much as the rate of growth of warranted wages, which decelerated more than in the United States because of the magnitude of the shocks experienced.
Employment, Output, Labor Productivity, and Capital Intensity, 1960–81
(Average annual changes in percent)
1973–79.
1973–78.
Employment, Output, Labor Productivity, and Capital Intensity, 1960–81
(Average annual changes in percent)
| Civilian Employment | Real GDP | Real GDP per Worker | Capital-Labor Ratio | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Country | 1960–73 | 1973–81 | 1960–73 | 1973–81 | 1960–73 | 1973–81 | 1960–73 | 1973–81 | ||
| United States | 2.0 | 2.0 | 4.0 | 2.1 | 2.0 | 0.2 | 2.3 | 1 | 0.7 | 1 |
| Japan | 1.3 | 0.7 | 10.3 | 3.6 | 9.1 | 2.9 | 10.8 | 1 | 6.9 | 1 |
| France | 0.7 | 0.1 | 5.6 | 2.4 | 4.9 | 2.3 | 4.8 | 4.7 | ||
| Germany, Fed. Rep. of | 0.2 | –0.4 | 4.5 | 2.1 | 4.2 | 2.5 | 6.2 | 4.7 | ||
| Italy | –0.4 | 0.8 | 5.3 | 2.4 | 5.6 | 1.7 | –5.1 | –2.7 | 2 | |
| United Kingdom | 0.2 | –0.8 | 3.1 | 0.5 | 2.9 | 1.4 | … | … | ||
1973–79.
1973–78.
Employment, Output, Labor Productivity, and Capital Intensity, 1960–81
(Average annual changes in percent)
| Civilian Employment | Real GDP | Real GDP per Worker | Capital-Labor Ratio | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Country | 1960–73 | 1973–81 | 1960–73 | 1973–81 | 1960–73 | 1973–81 | 1960–73 | 1973–81 | ||
| United States | 2.0 | 2.0 | 4.0 | 2.1 | 2.0 | 0.2 | 2.3 | 1 | 0.7 | 1 |
| Japan | 1.3 | 0.7 | 10.3 | 3.6 | 9.1 | 2.9 | 10.8 | 1 | 6.9 | 1 |
| France | 0.7 | 0.1 | 5.6 | 2.4 | 4.9 | 2.3 | 4.8 | 4.7 | ||
| Germany, Fed. Rep. of | 0.2 | –0.4 | 4.5 | 2.1 | 4.2 | 2.5 | 6.2 | 4.7 | ||
| Italy | –0.4 | 0.8 | 5.3 | 2.4 | 5.6 | 1.7 | –5.1 | –2.7 | 2 | |
| United Kingdom | 0.2 | –0.8 | 3.1 | 0.5 | 2.9 | 1.4 | … | … | ||
1973–79.
1973–78.
Indicators of the Real Wage Problem
Several indicators have been proposed to measure the extent of the real wage problem. One type of indicator attempts to quantify the discrepancy between the actual wage level and the warranted real wage—the wage gap. Another type attempts to measure the extent to which unemployment “needs” to increase following a disturbance to the system in order to bring the growth of actual real wages into line with that of the warranted real wage. The latter is usually referred to as the nonaccelerating inflation rate of unemployment.
The Simple Wage Gap Concept
A relatively straightforward method to measure the gap between the actual and the warranted level of real wages is based on the hypothesis (or judgment) that if the share of wages in value added is constant over time, the level of employment will be adequate or appropriate. The implication is that real product wage increases should correspond to the rate of increase of labor productivity, and that real consumption wage increases should correspond to productivity increases adjusted for changes in the terms of trade.
The Organization for Economic Cooperation and Development regularly updates and publishes measures based on this concept. Chart 7 shows how actual real product wages and productivity have evolved between 1970 and 1984 in North America, Japan, and Europe. The data confirm the marked difference between the growth of labor costs in Europe and North America over this period. Both productivity and real wage increases in Europe have generally been much higher than in the United States. However, because European real wage increases have exceeded the growth of productivity, particularly immediately before and after the first oil price shock, a significant gap arose between productivity and labor costs, a problem which appears to have been avoided in the United States.
Industrial Countries: Productivity and Real Labor Costs, 1970–84
(Indices, 1970 = 100)
Source: Organization for Economic Cooperation and Development, Employment Outlook, September 1985.1 Austria, Denmark, France, Finland, the Federal Republic of Germany, Italy, Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom.2 Output per person employed.
Industrial Countries: Productivity and Real Labor Costs, 1970–84
(Indices, 1970 = 100)
Source: Organization for Economic Cooperation and Development, Employment Outlook, September 1985.1 Austria, Denmark, France, Finland, the Federal Republic of Germany, Italy, Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom.2 Output per person employed.Industrial Countries: Productivity and Real Labor Costs, 1970–84
(Indices, 1970 = 100)
Source: Organization for Economic Cooperation and Development, Employment Outlook, September 1985.1 Austria, Denmark, France, Finland, the Federal Republic of Germany, Italy, Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom.2 Output per person employed.A surprising implication of this concept is that when applied to Japan, it points to a substantial and growing real wage gap, an observation which is hard to reconcile with Japan’s labor market performance. The OECD relates this development to Japan’s initially high level of profitability which allowed employers to absorb the fall in the profit share without employment losses (OECD Economic Outlook, June 1985). As discussed above, the strong growth of Japanese wages also may have reflected the fact that labor became relatively scarce.
There are a number of additional problems with the simple wage gap concept. The most important problem is the way in which cyclical or induced variations in productivity can influence the measured wage gap. For example, to the extent a demand-induced recession is accompanied by a decline in productivity growth, the measure would indicate a growing wage gap in a situation where demand rather than wages would have caused the imbalance. Moreover, if there is an imbalance initially between the actual and the warranted level of real wages, the process of closing down the least efficient units of production and shedding labor would lead to a reduction of the wage gap as employment falls. Indeed, the absence of a wage gap measured in this way is no guarantee that the wage level is appropriate from a high-employment perspective. An illustration of this point is the fact that Europe’s wage gap appears to have been reduced since 1980. With unemployment continuing to rise, this development has probably owed more to rationalization and labor shedding than to a reduction of the real wage problem. Another difficulty with the simple wage gap concept is that it is strongly influenced by the choice of base year. For example, if the wage gaps in Chart 7 were recomputed with base year 1973 instead of 1970, Europe’s gap would disappear by 1984, a result which clearly would be counterintuitive given the current state of the labor markets in most European countries.
Alternative Wage Gap Concepts
To overcome some of the problems connected with the simple wage gap measure, several authors have formalized the concept by relating it to an explicit production technology and by making adjustments for cyclical variations in productivity. These studies are based on the principle that with market clearing and competitive factor pricing, the real product wage will be equal to the marginal product of labor, at least over the medium run. The wage gap concept then attempts to measure the difference between actual real product wages and the marginal product of labor at some constant and high level of employment.
Sachs (1979 and 1983), Bruno and Sachs (1985), and Bruno (1985) have calculated this type of wage gap on the assumption that the underlying production technology is Cobb-Douglas. Because a Cobb-Douglas production function is characterized by a fixed ratio of labor’s marginal product to its average product, the problem of measuring the warranted real wage level becomes a question of calculating average productivity at full employment.31 However, the measure would still have to be normalized by some base-year benchmark. The cyclical adjustment of average productivity has been carried out using alternative approaches. In Bruno (1985), the adjustment is made first on the assumption that the average growth in labor productivity from 1960 to 1973 and from 1973 to 1983 represents the respective “full-employment” trends in productivity over the two periods. Using 1965 to 1969 as a benchmark (with the wage gap assumed to be zero), this approach results in the estimates shown in Table 15.
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-I
(Percentages over 1965–69 average)
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-I
(Percentages over 1965–69 average)
| Country | 1965 | 1970 | 1973 | 1976 | 1979 | 1981 | 1982 | 1983 |
|---|---|---|---|---|---|---|---|---|
| Canada | –1.7 | 1.5 | –1.4 | 4.6 | 0.9 | 1.5 | 1.8 | 2.0 |
| United States | 1.2 | –1.3 | 3.1 | 0.6 | 4.0 | 5.0 | 5.3 | 4.9 |
| Japan | 2.3 | 4.1 | 9.8 | 21.5 | 24.1 | 23.4 | 20.2 | 16.4 |
| France | 0.3 | –3.8 | –0.3 | 4.9 | 2.6 | 2.7 | 4.1 | … |
| Germany, Fed. Rep. of | 1.7 | 1.9 | 8.0 | 14.0 | 14.6 | 17.1 | 13.3 | 9.6 |
| Italy | 3.8 | 4.2 | 10.9 | 17.8 | 9.6 | 6.5 | 4.8 | 2.9 |
| United Kingdom | –1.5 | 1.5 | 3.1 | 8.1 | 9.3 | 14.3 | 13.9 | 13.9 |
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-I
(Percentages over 1965–69 average)
| Country | 1965 | 1970 | 1973 | 1976 | 1979 | 1981 | 1982 | 1983 |
|---|---|---|---|---|---|---|---|---|
| Canada | –1.7 | 1.5 | –1.4 | 4.6 | 0.9 | 1.5 | 1.8 | 2.0 |
| United States | 1.2 | –1.3 | 3.1 | 0.6 | 4.0 | 5.0 | 5.3 | 4.9 |
| Japan | 2.3 | 4.1 | 9.8 | 21.5 | 24.1 | 23.4 | 20.2 | 16.4 |
| France | 0.3 | –3.8 | –0.3 | 4.9 | 2.6 | 2.7 | 4.1 | … |
| Germany, Fed. Rep. of | 1.7 | 1.9 | 8.0 | 14.0 | 14.6 | 17.1 | 13.3 | 9.6 |
| Italy | 3.8 | 4.2 | 10.9 | 17.8 | 9.6 | 6.5 | 4.8 | 2.9 |
| United Kingdom | –1.5 | 1.5 | 3.1 | 8.1 | 9.3 | 14.3 | 13.9 | 13.9 |
The findings based on the Bruno-Sachs-I technique suggest that after a rise in the wage gap in the early 1970s and again after the two rounds of oil price increases, the wage gap has tended to narrow in most European countries after 1982. These results indicate that wage gaps increased significantly more in Europe than in North America following the oil shocks. Among the European countries, on this basis, the United Kingdom and the Federal Republic of Germany stand out with large remaining gaps by 1983. Again, Japan’s large wage gap by 1979, which has subsequently narrowed significantly, contrasts sharply with its unemployment record.
Bruno and Sachs have also applied an alternative adjustment of productivity, involving a direct correction for the effects of the unemployment rate on productivity growth.32 This correction is important in periods such as the late 1970s and early 1980s when most economies were not operating at full employment and, hence, actual labor productivity growth may have deviated from full-employment labor productivity growth. Indeed, the resulting estimates, Bruno-Sachs-II, suggest much larger wage gaps than Bruno-Sachs-I, particularly in the 1980s (Table 16). Moreover, on this basis wage gaps in Europe exhibit much less tendency to narrow after 1982. In France and the United Kingdom wage gaps continue to increase while in the Federal Republic of Germany the wage gap decreases more slowly. However, in Japan there is on this basis a more marked narrowing of the wage gap after 1981.
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-II
(Percentages over 1965–69 average)
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-II
(Percentages over 1965–69 average)
| Country | 1965 | 1970 | 1973 | 1976 | 1979 | 1981 | 1982 | 1983 |
|---|---|---|---|---|---|---|---|---|
| Canada | –1.9 | 1.9 | –0.5 | 3.3 | 0.8 | 2.2 | 2.9 | 3.5 |
| United States | 0.2 | 0.1 | 6.0 | 2.9 | 6.8 | 8.1 | 8.6 | 8.4 |
| Japan | 2.2 | 4.3 | 10.1 | 18.2 | 20.7 | 19.8 | 16.6 | 12.7 |
| France | 0.0 | –3.4 | –0.4 | 7.9 | 10.7 | 14.3 | 17.4 | … |
| Germany, Fed. Rep. of | 2.0 | 1.5 | 7.2 | 13.0 | 15.3 | 19.1 | 15.9 | 12.9 |
| Italy | 2.3 | 6.4 | 15.4 | 19.5 | 11.8 | 9.1 | 7.6 | 5.9 |
| United Kingdom | –2.0 | 2.2 | 4.6 | 11.0 | 16.4 | 24.1 | 25.0 | 26.4 |
Cyclically Adjusted Wage Gap Measures: Bruno-Sachs-II
(Percentages over 1965–69 average)
| Country | 1965 | 1970 | 1973 | 1976 | 1979 | 1981 | 1982 | 1983 |
|---|---|---|---|---|---|---|---|---|
| Canada | –1.9 | 1.9 | –0.5 | 3.3 | 0.8 | 2.2 | 2.9 | 3.5 |
| United States | 0.2 | 0.1 | 6.0 | 2.9 | 6.8 | 8.1 | 8.6 | 8.4 |
| Japan | 2.2 | 4.3 | 10.1 | 18.2 | 20.7 | 19.8 | 16.6 | 12.7 |
| France | 0.0 | –3.4 | –0.4 | 7.9 | 10.7 | 14.3 | 17.4 | … |
| Germany, Fed. Rep. of | 2.0 | 1.5 | 7.2 | 13.0 | 15.3 | 19.1 | 15.9 | 12.9 |
| Italy | 2.3 | 6.4 | 15.4 | 19.5 | 11.8 | 9.1 | 7.6 | 5.9 |
| United Kingdom | –2.0 | 2.2 | 4.6 | 11.0 | 16.4 | 24.1 | 25.0 | 26.4 |
Bruno (1985) has provided an explanation for the recorded narrowing of the wage gap measures. By decomposing the changes in the wage gap into changes in the consumption wage, the terms of trade and productivity, his calculations suggest that the narrowing of the gap is attributable mainly to slower wage growth. It is unclear, however, whether this deceleration in wage growth has been due to a shift in wage behavior or whether it reflects the rise in unemployment. If, as suggested by the analysis in the previous section, it is the latter, then the recorded narrowing of the wage gap does not necessarily constitute any solution to the real wage problem.
The Bruno-Sachs measures discussed so far are based on Cobb-Douglas technology which assumes unitary elasticity of substitution between capital and labor and, hence, a constant share of labor in value added. If the elasticity of substitution is less than unity, a rise in real wages would result in a rising share of labor in value added even if wages grow in line with the warranted real wage. Observed increases in the share of labor would therefore not necessarily be an indication of the size of any real wage problem. Because of the relative scarcity of labor in Japan, together with the rapid accumulation of capital, such an hypothesis is clearly more plausible if the real wage gap concept is to be valid also for that country.
To test this hypothesis, it is necessary to estimate the production technology explicitly, for example by using a CES production function approach, which does not restrict the elasticity of substitution between labor and capital to be equal to unity. Because of data constraints, such tests have been limited to the manufacturing sector. The most prominent examples are the studies of Artus (1984) and Lipschitz and Schadler (1984). Bruno and Sachs (1985) and Bruno (1985) also discuss this approach in terms of a sensitivity analysis but do not estimate a production function.
The measure developed by Artus is based on the assumptions that factor proportions may change over time, that the rate of disembodied technical progress may also vary, and that the manufacturing sector faces a given supply of labor. Despite numerous data problems the approach permitted the successful estimation of elasticities of substitution between capital and labor for the seven major industrial countries. The estimated elasticities of substitution for these countries lie in the range 0.5 to 0.8, which is less than the unitary elasticity underlying the Cobb-Douglas specification. These results are also consistent with the studies of labor demand surveyed earlier. The estimated coefficients are subsequently used to compute the wage level which would be warranted to fully employ the manufacturing labor force. Two alternative models are tested, model A which is a standard two-factor CES production function involving separability between value added and intermediate inputs, and model B which is a non-separable, three-factor production function, with complementarity between energy and capital.
The resulting wage gap measures are shown in Chart 8. In Italy, Canada, and the United States, the actual real wage level in the early 1980s appears to be relatively close to the warranted level. By contrast, in France, the Federal Republic of Germany, the United Kingdom, and Japan, the wage level was found to be a significant obstacle to a return to high employment conditions. Qualitatively, these results confirm the findings of Bruno and Sachs (1985); in particular, the CES approach also identifies Japan as a country with a real wage problem.
Major Industrial Countries: Deviation of Actual from Warranted Real Wage Rates, 1955–82
(In percent)
Source: Artus (1984).Note: A positive number indicates that the actual real wage rate (deflated by the relevant value added deflator) exceeds the real wage rate consistent with the chosen high-employment norm.
Major Industrial Countries: Deviation of Actual from Warranted Real Wage Rates, 1955–82
(In percent)
Source: Artus (1984).Note: A positive number indicates that the actual real wage rate (deflated by the relevant value added deflator) exceeds the real wage rate consistent with the chosen high-employment norm.Major Industrial Countries: Deviation of Actual from Warranted Real Wage Rates, 1955–82
(In percent)
Source: Artus (1984).Note: A positive number indicates that the actual real wage rate (deflated by the relevant value added deflator) exceeds the real wage rate consistent with the chosen high-employment norm.Lipschitz and Schadler (1984) also applied a CES function to Japanese and U.K. data, assuming separability between value added and intermediate inputs. Mainly because of differences in the assumed growth of the manufacturing labor force and the rate of technical progress, their estimates of wage gaps for Japan and particularly for the United Kingdom are significantly higher than those of Artus (Chart 9).
Japan and the United Kingdom: Comparison of Estimated Wage Gaps in Lipschitz and Schadler (1984) and Artus (1984)
(Indices, 1963 = 100)
Sources: Artus (1984); Lipschitz and Schadler (1984).Note: The wage gap is measured as an index of the actual over the warranted real wage level.
Japan and the United Kingdom: Comparison of Estimated Wage Gaps in Lipschitz and Schadler (1984) and Artus (1984)
(Indices, 1963 = 100)
Sources: Artus (1984); Lipschitz and Schadler (1984).Note: The wage gap is measured as an index of the actual over the warranted real wage level.Japan and the United Kingdom: Comparison of Estimated Wage Gaps in Lipschitz and Schadler (1984) and Artus (1984)
(Indices, 1963 = 100)
Sources: Artus (1984); Lipschitz and Schadler (1984).Note: The wage gap is measured as an index of the actual over the warranted real wage level.Several authors have used the wage gap concept to explain changes in unemployment. Attempts have also been made to use wage gaps to account for differences in unemployment and employment behavior across countries. The method used to construct measures of the wage gap implies that an increase in the wage gap will reduce employment if firms are operating on their medium-run labor demand schedules. Bruno and Sachs (1985) found such a negative relationship between the growth of total manhours worked in manufacturing and the wage gap in the major industrial countries.33 They compared the increase in the average wage gap from the 1960s to the 1970s with the difference between the average annual growth rate of total manhours worked in the 1960s and in the 1970s. While the growth of manhours worked in manufacturing slowed everywhere in the 1970s, the smallest decelerations occurred in Canada, the United States, and France, that is, in those countries in which the increase in the wage gap was the smallest. The greatest deceleration in the growth of hours worked occurred in Japan where the increase in the wage gap was the largest. There was also a significant deceleration in the growth of hours worked in the Federal Republic of Germany and the United Kingdom where there were sizable increases in the wage gap.
The relationship between changes in the wage gap and the unemployment rate is complicated because induced changes in labor force behavior must also be taken into account. Wage gap measures are calculated on the basis of the estimated full-employment labor force rather than the actual labor force. Changes in employment conditions due to changes in the wage gap are likely to encourage workers either to enter or drop out of the labor force. Nevertheless, Bruno and Sachs find a significant positive relationship between the wage gap and the unemployment rate in Japan, the Federal Republic of Germany, the United Kingdom, and Canada, when the lagged unemployment rate is used as a proxy for slow adjustment in the labor market. There is no significant relationship for the United States and France. However, when Bruno and Sachs add an aggregate demand variable to the relationship they find a positive relationship between the wage gap and unemployment for France as well.
In summary, the notion of a wage gap can perform a useful diagnostic function. However, in practice, the difficulties associated with the choice of underlying production technology, the cyclical adjustment of productivity, the choice of base year, and the cyclical sensitivity of wage behavior, make it extremely difficult to quantify the wage gap precisely, and to estimate by how much real wages are out of line. Nevertheless, the evidence suggests that the progressive widening of wage gaps in the European countries, with the exception of Italy, from 1971–82 is a major factor in the sharp rise in unemployment in those countries over this period. In 1983, the latest year for which data are available for most countries, the wage gap continued to increase in the United Kingdom, while it declined in the Federal Republic of Germany and Italy. However, the estimates of labor demand functions already discussed suggest that the beneficial impact of this narrowing of the wage gap on employment in these latter two countries may only appear after a considerable lag. It can also be concluded with a reasonable degree of assurance that real wages in most European countries are still significantly out of line with the level that would be warranted from a high employment perspective. Unless the adverse disturbances experienced over the last ten to twenty years are reversed, any sizable reduction in unemployment would therefore have to be based on a significant reduction in labor costs.
The NAIRU
Another manifestation of the problem of rigid real wages in the face of adverse disturbances to the economy is the sharp rise in the nonaccelerating inflation rate of unemployment in many European countries during the 1970s and early 1980s. The definition of the NAIRU and its relationship with the other concepts discussed in this section, were discussed earlier. It may be recalled from that discussion that the NAIRU can be expressed in terms of the original equilibrium unemployment rate of the economy and any gap between target real wage growth and warranted real wage growth which emerges after a disturbance enters the economic system.
The level of the NAIRU will reflect a number of “structural” characteristics of the economy, including the institutional organization of goods and factor markets (including regulations), demographics, preferences of individuals and incentives provided through the structure of taxes, transfers, and subsidies. In addition, the NAIRU is directly affected by disturbances when real wages are not flexible—it indicates the rate of unemployment necessary to prevent an acceleration in wage inflation and, hence, to bring about a reduction in the growth of real wages, once disturbances enter the economic system. A rise in the NAIRU, and a corresponding increase in the actual unemployment rate, will force actual real wage growth into line with the reduced path of warranted real wages following an adverse disturbance. When the disturbance ends, the NAIRU therefore tends to return to the level of the original equilibrium unemployment rate. An increase in the responsiveness of real wages to unemployment would also tend to reduce the NAIRU. To the extent it can be quantified, the NAIRU has important implications for policy purposes. Any attempt to use traditional demand expansion policies to reduce the actual unemployment rate below the NAIRU will result in accelerating inflation and a squeeze on profits which ultimately must lead to a reversal of this policy.
Empirical estimates of developments in the NAIRU are presented in Table 17. Differences in the estimates across alternative studies reflect differences in the specification of the underlying wage and price equations and serve as an important reminder of the imprecise nature of such calculations, particularly in the case of point estimates for any particular period.34 In addition, estimates of the NAIRU which apply to a short period of time can reflect any adverse disturbances such as terms of trade losses which occur during the period even if they are transitory, thereby overstating the level of unemployment toward which the economy can be expected to converge in the long run. Nevertheless, developments over time in the NAIRU do convey useful information about the need for wage flexibility in the face of disturbances.
Estimates of the Non-Accelerating Inflation Rate of Unemployment
These unemployment rates which are reported in Coe have been converted to a standardized basis by the OECD.
Coe’s estimates of the NAIRU, and the actual unemployment rate estimates for 1981–83 include only the first half of 1983.
Estimates of the Non-Accelerating Inflation Rate of Unemployment
| NAIRU | Actual Unemployment Rate1 | |||
|---|---|---|---|---|
| Country | Time Period | Coe | Layard et al. | |
| Canada | 1968–70 | 4.0 | 4.8 | |
| 1971–75 | 7.0 | 6.0 | ||
| 1973 | 6.4 | |||
| 1974 | 6.5 | |||
| 1976–80 | 8.5 | 7.7 | ||
| 1979 | 6.2 | |||
| 1981–832 | 7.5 | 9.9 | ||
| United States | 1961–69 | 4.8 | ||
| 1967–70 | 3.0 | 4.0 | ||
| 1970–73 | 6.0 | |||
| 1971–75 | 6.0 | 6.0 | ||
| 1974–81 | 6.8 | |||
| 1976–80 | 6.0 | 6.8 | ||
| 1979 | 7.2 | |||
| 1981–832 | 6.5 | 8.8 | ||
| Japan | 1971–75 | 1.0 | 1.4 | |
| 1976–80 | 1.5 | 2.1 | ||
| 1981–832 | 2.0 | 2.3 | ||
| France | 1967–70 | 2.5 | 1.8 | |
| 1971–75 | 3.5 | 2.7 | ||
| 1976–80 | 3.0 | 5.3 | 5.2 | |
| 1981–832 | 8.0 | 6.9 | 8.3 | |
| Germany, Fed. Rep. of | 1967–70 | 1.0 | 1.0 | |
| 1971–75 | 1.5 | 1.8 | ||
| 1976–80 | 3.0 | 3.7 | 3.6 | |
| 1981–832 | 8.0 | 5.3 | 6.3 | |
| Italy | 1967–70 | 4.5 | 5.4 | |
| 1971–75 | 7.0 | 5.8 | ||
| 1976–80 | 6.5 | 8.9 | 7.1 | |
| 1981–832 | 6.5 | 7.7 | 8.9 | |
| United Kingdom | 1967–70 | 1.0 | 2.3 | |
| 1971–75 | 7.5 | 3.0 | ||
| 1976–80 | 7.5 | 4.6 | 5.4 | |
| 1981–832 | 6.0 | 9.5 | 10.5 | |
| European Community | 1966–70 | 2.6 | ||
| 1971–75 | 5.3 | |||
| 1976–80 | 5.3 | |||
| 1981–83 | 7.6 | |||
These unemployment rates which are reported in Coe have been converted to a standardized basis by the OECD.
Coe’s estimates of the NAIRU, and the actual unemployment rate estimates for 1981–83 include only the first half of 1983.
Estimates of the Non-Accelerating Inflation Rate of Unemployment
| NAIRU | Actual Unemployment Rate1 | |||
|---|---|---|---|---|
| Country | Time Period | Coe | Layard et al. | |
| Canada | 1968–70 | 4.0 | 4.8 | |
| 1971–75 | 7.0 | 6.0 | ||
| 1973 | 6.4 | |||
| 1974 | 6.5 | |||
| 1976–80 | 8.5 | 7.7 | ||
| 1979 | 6.2 | |||
| 1981–832 | 7.5 | 9.9 | ||
| United States | 1961–69 | 4.8 | ||
| 1967–70 | 3.0 | 4.0 | ||
| 1970–73 | 6.0 | |||
| 1971–75 | 6.0 | 6.0 | ||
| 1974–81 | 6.8 | |||
| 1976–80 | 6.0 | 6.8 | ||
| 1979 | 7.2 | |||
| 1981–832 | 6.5 | 8.8 | ||
| Japan | 1971–75 | 1.0 | 1.4 | |
| 1976–80 | 1.5 | 2.1 | ||
| 1981–832 | 2.0 | 2.3 | ||
| France | 1967–70 | 2.5 | 1.8 | |
| 1971–75 | 3.5 | 2.7 | ||
| 1976–80 | 3.0 | 5.3 | 5.2 | |
| 1981–832 | 8.0 | 6.9 | 8.3 | |
| Germany, Fed. Rep. of | 1967–70 | 1.0 | 1.0 | |
| 1971–75 | 1.5 | 1.8 | ||
| 1976–80 | 3.0 | 3.7 | 3.6 | |
| 1981–832 | 8.0 | 5.3 | 6.3 | |
| Italy | 1967–70 | 4.5 | 5.4 | |
| 1971–75 | 7.0 | 5.8 | ||
| 1976–80 | 6.5 | 8.9 | 7.1 | |
| 1981–832 | 6.5 | 7.7 | 8.9 | |
| United Kingdom | 1967–70 | 1.0 | 2.3 | |
| 1971–75 | 7.5 | 3.0 | ||
| 1976–80 | 7.5 | 4.6 | 5.4 | |
| 1981–832 | 6.0 | 9.5 | 10.5 | |
| European Community | 1966–70 | 2.6 | ||
| 1971–75 | 5.3 | |||
| 1976–80 | 5.3 | |||
| 1981–83 | 7.6 | |||
These unemployment rates which are reported in Coe have been converted to a standardized basis by the OECD.
Coe’s estimates of the NAIRU, and the actual unemployment rate estimates for 1981–83 include only the first half of 1983.
With the exception of Japan and France, the NAIRU rose significantly in the 1970s in the major industrial countries, confirming the interactions between disturbances and rigidities discussed earlier. Grubb, Layard, and Symons (1984), in a study of eight European Community countries, concluded that the increase in the NAIRU in these countries during the 1970s was attributable in equal part to higher relative import prices, lower productivity growth, and the importance of the time trend in the estimated wage equations. They suggest the hypothesis that the latter may represent an increase in the natural rate of unemployment resulting from rising social welfare payments in real terms. If real wage targets had adjusted to the change in the rate of growth of warranted real wages implied by the changes in productivity growth and the terms of trade, or if wage settlements had been more sensitive to unemployment, the NAIRU would have risen significantly less in these countries.
The early 1980s witnessed a further dramatic rise in the NAIRU in most of the European countries, primarily as a consequence of rising import prices in the wake of the second oil price shock and the appreciation of the dollar. The NAIRU in the United States rose only modestly in 1982–83 reflecting a better productivity performance than in the 1970s and the strong dollar which reduced import prices. As discussed earlier, to some extent these movements in real exchange rates, the terms of trade, and, hence, the NAIRU since 1980 seem to have reflected the divergence in fiscal stance between the United States and Europe.
Cross-country comparisons of estimated NAIRUs are problematic because the estimated wage equations underlying the calculations perform significantly better for some countries than for others. Nevertheless, the fact that the NAIRU has, in general, risen so much more in Europe than in Japan and the United States over the last 15 years, and especially since 1980, is consistent with the evidence on the differences in real wage behavior across countries discussed above. In particular, the modest increase in Japan’s NAIRU testifies to this country’s high degree of real wage flexibility which has offset the implications of the severe shocks Japan experienced since the early 1970s, particularly the oil price increases. The less pronounced rise in the NAIRU in the United States relative to Europe can be attributed, at least in part, to the differences in the severity of the disturbances experienced as well as to greater real wage flexibility, particularly in the short run.
Finally, with respect to Table 17 it should be noted that in many European countries the actual unemployment rate is currently above the most recent estimates of the NAIRU. It needs to be stressed that the interpretation of such a discrepancy, and its implications for economic policy, are not straightforward. Some analysts have interpreted this gap as being attributable to Keynesian unemployment and have advised the use of fiscal stimulus to close it.
Discrepancies between the NAIRU and the actual unemployment rate can stem from a variety of factors, however, and must be interpreted with care. For example, a transitory disturbance may end and still leave a sizable wage gap which would result in the actual unemployment rate being above the NAIRU, while the economy is in the process of adjusting to that disturbance. A deficiency in aggregate demand may also result in the actual unemployment being higher than the NAIRU. It follows from these examples that a discrepancy between this rate and the actual unemployment rate in itself does not contain information about the reasons for such a discrepancy and does not necessarily indicate the existence of Keynesian unemployment (see the next section). The potential implications of a positive disturbance, such as the sharp decline in oil prices in early 1986, illustrate this point. To the extent that the actual unemployment rate is above the NAIRU because the level of the real wage is too high, an improvement in the terms of trade, while it takes place, will tend to reduce actual unemployment toward the NAIRU by reducing real product wages relative to real consumption wages. If such a shock were the initial indication of a permanently lower rate of increase of oil prices and correspondingly a higher trend rate of growth of the warranted real consumption wage, the resulting reduction of the NAIRU should also have a beneficial impact on actual unemployment. However, such effects would not materialize if target real wage growth were adjusted upward in line with the improvement in warranted real wage growth.
Interactions Between Goods Markets and Labor Markets
While differences in the growth rates of aggregate output have not been sufficient alone to account for differences in employment behavior across countries over the long run, it is apparent that they have been important for employment performance in some periods. In particular, the slow growth of output in Europe in the period since 1982, compared with that of the United States and Japan, has been regarded as at least part of the explanation for the recent rise in Europe’s unemployment, even if slow output growth was of less importance over the decade of the 1970s as a whole (see Blanchard et al. (1985)). In view of such arguments, much of the recent comparative economic research on labor markets has attempted to explain differences in employment behavior in terms of real wage factors (classical factors) being of importance for employment in some periods and aggregate demand factors (Keynesian factors) being of importance in other periods. Hence, for example, while Bruno and Sachs (1985) attribute much of the rise in European unemployment in the 1970s to classical factors, they regard a significant part of the increase since 1982 as being due to Keynesian factors. The conclusion that weak demand has been behind the recent rise in Europe’s unemployment has formed the basis for some of the recent recommendations that Europe now should begin to pursue more expansionary financial policies to reduce unemployment (Bruno (1985)).
In practice, however, explaining unemployment is not as simple as the dichotomy between classical and Keynesian factors might suggest. The distinction between these causes of unemployment ignores, in particular, that there are important interactions between labor and goods markets, and that supply and demand factors are often difficult to disentangle at a macro-economic level. In general equilibrium, employment is influenced by many factors with numerous interactions between labor and goods markets, which implies that both real wages and aggregate demand will tend to play a role in any account of employment behavior.
This section attempts to clarify the distinction between classical and Keynesian unemployment and why it is difficult to distinguish between these kinds of unemployment, taking into account a number of interactions between labor and goods markets. It then reviews simulation exercises which have sought to assess the general equilibrium consequences of wage moderation under alternative policy assumptions. In such simulations, wage moderation is found to have important implications not just for the level of employment at a given rate of growth of output, but also for the rate of growth of output.
Classical Versus Keynesian Unemployment
Traditionally, the distinction between classical and Keynesian approaches to unemployment has been in terms of the relative significance of real wages compared with aggregate demand as an explanation of unemployment behavior.35 It is useful to illustrate this distinction by Chart 10, which reproduces the medium-run labor demand curve of Chart 6. The supply of labor (the labor force) is assumed to be constant at D. Point C represents a structural rate of unemployment determined by factors such as the rate of flow of workers into the labor market as well as regional and occupational mismatches; B is the maximum amount of labor firms wish to employ at the current real wage (E). The difference between points B and C is classical unemployment. Both points lie on the same labor demand curve, implying that BC represents the consequences of a “wage gap” EF. If actual labor demand only amounts to A, the difference between points A and B represents Keynesian unemployment; given the position of the labor demand curve, this kind of unemployment reflects a deficient level of aggregate demand; employment is lower than implied by the wage level. Because firms are constrained in goods markets by low demand, they are “pushed off their medium-run labor demand curves.36
Here the essential difference between classical and Keynesian unemployment depends on whether unemployment reflects insufficient aggregate demand at a given level of real labor costs or a level of real labor costs that is too high to be consistent with high employment, independently of the level of aggregate demand. Of course, at any point in time, a given level of unemployment may consist partly of classical and partly of Keynesian unemployment; it will also, in general, reflect structural influences on the rate of unemployment.
Keynesian unemployment is in some cases identified as reflecting a situation where the actual unemployment rate exceeds the NAIRU. It is apparent however, from the earlier discussion that the amount of Keynesian unemployment is unrelated to the NAIRU, which only indicates the unemployment rate that would be necessary to stabilize the rate of inflation and to force the rate of growth of actual wages into line with the rate of growth of the warranted wage. That is to say, the NAIRU does not contain information about the level of actual real wages relative to their warranted level.
The distinction between classical and Keynesian unemployment has important policy implications, as noted earlier. (See, for example, Blanchard et al. (1985).) Most importantly, whether a given amount of unemployment is classical or Keynesian is regarded as having implications for whether adjustments in real wage costs are a necessary accompaniment to higher employment in the short to medium run. If unemployment is classical, it can by its nature only be eliminated by lower real labor costs (or by a higher warranted real wage); if unemployment is Keynesian and reflects insufficient demand at a given level of wage costs, it can be eliminated by higher demand with no fall in wage costs. These distinctions do not in themselves rule out the possibility that classical unemployment could be reduced by higher demand; this would be the case if higher demand would lead to a reduction in real wages or an increase in warranted wages. The distinctions merely serve to highlight whether a change in real wages is regarded to be a necessary part of the adjustment to a lower level of unemployment.
Sources of Classical and Keynesian Unemployment
Because classical unemployment reflects a level of real labor costs that is too high, it can arise from any factor that causes real labor costs (to firms) to be above the levels that would be consistent with high employment. In general, there are many factors that can raise the level of real labor costs above a high-employment level. As indicated earlier, the warranted level of real labor costs may fall relative to the actual level of wage costs on account of lower capital accumulation, lower total factor productivity growth, higher payroll taxes, or, under some conditions, as a consequence of higher prices for intermediate inputs. Alternatively, assuming that suppliers of labor base their wage demands on the real consumption wage, net of taxes, the real labor costs that are relevant for firms’ hiring decisions (in product terms) may rise relative to their warranted level due to wage push factors stemming from higher income taxes or from increases in the prices of consumption goods relative to firms’ output prices. If wage contracts are set in nominal terms, real labor costs may also rise due to an unanticipated reduction in the rate of inflation.37 A real labor cost problem (classical unemployment) can therefore arise due to fluctuations in aggregate demand or supply. In addition, a recession might raise the prices of imports relative to home goods (raising product wages when consumption wages are sticky). This could lead to classical unemployment.
Conversely, while Keynesian unemployment is defined to reflect demand deficiency at a given level of labor costs, it could itself arise due to a real labor cost problem. For example, if real labor costs are too high, the impact of lower employment is likely to lead to lower household income and spending which could create Keynesian unemployment. A labor cost problem may also lead to lower investment spending and to lower aggregate domestic demand overall. At the same time, a labor cost problem could affect external competitiveness and, hence, reduce exports and raise imports.
It is apparent from this discussion that both classical and Keynesian unemployment may be influenced by the same factors, leading to obvious difficulties in distinguishing between them. Indeed, many of the disturbances in the 1970s have had implications for employment that are capable of being understood in both classical and Keynesian terms. For example, the disturbances of the 1970s such as lower growth in total factor productivity or higher prices of intermediate inputs, have had implications for aggregate demand as well as for supply. Moreover, the tight policies necessary to reduce inflation in the early 1980s may have increased unemployment through both classical and Keynesian mechanisms.
Even if it were possible empirically to distinguish between classical and Keynesian unemployment at a given point in time, it would also be possible for one kind of unemployment to lead to a rise in the other, after a period. For example, by depressing demand, a situation of classical unemployment might give rise to Keynesian unemployment; and a prolonged period characterized by Keynesian unemployment might reduce the warranted real wage due to depressed investment spending, ultimately giving rise to classical unemployment.
Classical and Keynesian Unemployment and the Flexibility of Labor Costs
Classical unemployment ultimately arises due to insufficient flexibility of real labor costs in the face of disturbances that call for lower real wages if high-employment conditions are to be maintained. Hence, the sources of this type of unemployment are ultimately related to the degree of real wage flexibility and to the disturbances which economies have experienced.
The degree of flexibility of real labor costs also has significant implications for the ability of expansionary policies to reduce unemployment, if such unemployment is classical in origin. In the classical case, demand reflation will only raise employment if it serves either to raise the warranted real wage or reduce actual real labor costs. If both actual and warranted real wages are rigid, demand reflation will be unsuccessful in eliminating classical unemployment.
The evidence discussed earlier is not generally favorable to the view that the substantial part of unemployment in Europe reflects weak demand, independently of any wage problems. To the extent that this view is correct, demand reflation would only succeed in reducing unemployment insofar as it might raise the warranted wage, for example through an increase in investment spending or through a cyclical rise in labor productivity. In view of the experience of the late 1970s, neither of these mechanisms is likely to guarantee success in practice, however. As an alternative, Marris (1985) has suggested that macro policies should seek to expand enough to raise the rate of inflation in goods markets. With wage indexation having been abolished in many countries, such a policy might help to reduce real labor costs, thereby stimulating employment. While this might work in the very short run (assuming money illusion on the part of wage earners) it is unlikely, however, to result in permanent reductions in real wages. In any case, such policies are unlikely to be adopted by European governments in view of the hard-won progress in reducing inflation since the early 1980s. Hence, to the extent unemployment is classical, the authorities are faced with the reality that it is unlikely to be susceptible to elimination by traditional demand reflation.
Capital-Constrained Unemployment
Notwithstanding the scope for reducing classical unemployment through policies that encourage wage moderation or of reducing Keynesian unemployment by demand reflation, at any given time the capital stock may not be sufficiently large to permit full employment, regardless of the level of real labor costs (Blanchard et al. (1985)). This possibility has been emphasized in particular by Giersch (1981) in the context of the experiences of a number of European countries in the 1970s. Such a view is based on the observation that the level of unemployment which would be consistent with a given rate of capacity utilization has tended to increase in all industrial countries since the 1960s, particularly in the European countries, and that remaining unused capacity may be insufficient to absorb unemployed labor. According to Giersch, because of the low levels of investment during the 1970s, as well as the possibility of accelerated capital obsolescence associated with the two rounds of oil price increases, together with the tendency for investment to be of a capital-deepening rather than capital-widening nature, it may not be possible for European firms to move to more labor-intensive production techniques in the short run. What seems to be required under these conditions are not just policies that serve to induce wage moderation or raise demand, but policies which will promote an expansion of the capital stock, for example through investment subsidies. If capital is indeed scarce in Europe, the most important implication seems to be that any effort to reduce unemployment—whether through wage moderation or demand reflation—would require some period of time before the results could be expected to materialize.
In practice, it is difficult to assess the significance of the capital scarcity argument. However, three considerations suggest that while potential shortages of capital may place some constraints on the speed of employment growth, they are unlikely to be an overriding constraint. In the first place, it is apparent that there continues to be some unused capacity in Europe, even if capacity utilization measures are not wholly reliable. Secondly, the labor demand estimates already discussed imply that there is some scope for substituting labor for capital, at least in the medium run. Under such conditions, a policy that reduced the price of labor relative to capital could be expected to result in additional and more labor-intensive production, even at the existing stock of capital. Finally, an expansion of more labor-intensive investment would of course be encouraged by a reduction in real labor costs. To the extent that it would be necessary also to stimulate investment through subsidies, it would be essential to allocate these in a way which did not encourage labor saving investments.
Model Simulations
To conclude the review of the significance of real wages for the determination of employment, this section reviews some recent simulation studies based on models of the whole economy. The advantage of using a macroeconometric model of the whole economy, with appropriately specified labor demand and supply equations, is that a number of interactions between goods and factor markets can be taken into account in the assessment of the potential role for real wage moderation. Such interactions are ignored in the single equation (partial equilibrium) models discussed earlier.
Changes in real wages can be expected to affect both aggregate demand and supply schedules in goods markets. The supply effect is relatively straightforward. A reduction in real wages will increase profits per unit of output, thereby leading to an increase in supply (potential output) and an increase in labor demand. As time passes, employment will also increase as firms substitute labor for other inputs in the production process. In the very short run, the aggregate demand effect may be ambiguous, reflecting a number of considerations, but once the adjustments associated with the change in real wages begin to work themselves through, a reduction in real wages will usually lead to an increase in aggregate demand.
There are three main channels of influence of a reduction in real wages on aggregate demand at a given level of employment: (1) a redistribution of income from wage earners to firms which may reduce private consumption and residential construction; (2) an improvement of international competitiveness; and (3) a possible reduction in inflation and interest rates and, hence, an increase in the real value of wealth. In the period immediately following a decline in real wages, consumption will typically fall in response to the reduction in personal disposable income, while investment and exports will rise as a result of improved profitability and better international competitiveness. The simulations discussed below suggest that which effect will dominate in this initial period will depend upon the stance of monetary and fiscal policies.
As the adjustments set in motion by the wage reduction progress, consumption will begin to increase, reinforcing the faster growth of investment and exports. Consumption growth accelerates because the decline in prices associated with lower wages can be expected to reduce interest rates and increase the real value of wealth. The increase in employment and wage income resulting from substitution effects will also stimulate consumption. As output expands more rapidly and production techniques become more labor intensive, employment growth is likely to accelerate further with favorable consequences for consumption and aggregate demand.
The stance of fiscal and monetary policies during this adjustment period will have a strong influence on the magnitude of the resulting increase in employment and output, and on the composition of the resulting change in aggregate demand. This is because the stance of policy will determine the impact on interest rates and exchange rates resulting from a change in wages. As mentioned above, the stance of policies may be of particular importance in the very short run in determining whether output will rise or fall as a result of a reduction in real wage growth.
The remainder of this section will examine the evidence on the consequences for employment and activity of a change in real wages on the basis of simulation studies published by the Treasury in the United Kingdom and by the European Commission.
United Kingdom Treasury
The United Kingdom Treasury (1985) has recently released a study of the effects of real wages on employment in the U.K. economy, which contained the results of several simulations with the Treasury model. Two main conclusions emerge from these simulations. First, real wage moderation would lead to significant increases in employment even in the short run. Second, the accompanying stance of fiscal and monetary policy is a significant determinant of the nature and size of the change in output and, hence, in employment as a result of real wage moderation. These conclusions have been corroborated by the results of similar simulation exercises conducted with other models of the U.K. economy.38
Changes in real wages alter employment in the Treasury model in a manner similar to the general framework discussed earlier. Private sector employment is modeled in a cost minimization framework. There are separate equations for manufacturing and non-manufacturing private sector employment. For a given level of output, firms in both sectors adjust their inputs of labor, capital, and intermediate inputs so as to minimize the cost of production. The elasticity of private sector employment with respect to a reduction in wages, at a constant level of output, is 0.15 (0.25 in manufacturing and 0.1 in the remainder of the private sector). The long-run elasticity of private sector employment with respect to output, with unchanged relative factor prices, is unity. Most of the effect of a change in output on employment is realized within two years.39
To simulate the impact of wage moderation on the U.K. economy the wage equation in the Treasury model was overwritten and nominal wages were reduced sufficiently to lower the level of real wage costs by 2 percentage points relative to a baseline path. Since prices decline in response to the reduction in unit labor costs the resulting decline in the level of nominal wages is actually greater than 2 percent. Such a change in real wage behavior might occur if real wage targets were to suddenly adapt to existing economic conditions, or if there were an appropriate change in the NAIRU, for example as a result of a positive disturbance to the economy. Underlying this approach to analyzing the effects of real wage moderation is the important assumption that the other coefficients in the model do not change in response to the assumed change in wage behavior.
The results point to a negative elasticity of employment with respect to real wages of ½ to ¾ after three years, depending upon the assumptions made about the policy setting. These estimates include the effects of induced changes in output on employment as well as the substitution between labor and other factors of production. In fact, most of the increase in employment associated with a real wage reduction in these simulations is due to output effects (between 60 percent and 75 percent depending upon the simulation). The higher elasticity estimate corresponds to a policy which adjusts interest rates to maintain an unchanged path for money supply and adjusts personal income tax rates to maintain an unchanged ratio of the Public Sector Borrowing Requirement (PSBR) to nominal GDP (an unchanged nominal framework). The smaller elasticity estimate corresponds to a policy of fixed nominal interest rates and tax rates (fixed rates policy). In the latter case, because real interest rates are higher during the adjustment period, the reduction in real wages results in a smaller increase in output and, hence, employment.
The dynamic implications of a reduction of real wages under the two policy regimes are summarized in Table 18. The results are expressed as the difference between the simulated values and the baseline solution. With an unchanged nominal framework, a reduction in real labor costs results in a very marginal increase in output and employment in the first year. The output effect is simulated to increase significantly in the second year and more or less stabilizes in the third and fourth year. This leveling off of output is explained by the fact that the shock in question is a permanent reduction in the level rather than in the growth rate of real wages. Employment increases significantly by the second year due to a combination of substitution effects and higher activity. By the fourth year, the reduction in the level of real labor costs of 2 percent has raised the level of output by 0.9 percent and the level of employment by 1.4 percent.
United Kingdom: Simulated Effects of Lower Real Wages, with Unchanged Nominal Framework
(Percent changes from baseline levels)
A negative sign indicates an improvement in competitiveness.
United Kingdom: Simulated Effects of Lower Real Wages, with Unchanged Nominal Framework
(Percent changes from baseline levels)
| Year | GDP | Employment | Prices | Average Earnings | Real Earnings | Real Wage Costs | Exchange Rate | Competitiveness1 | Short-term Interest Rates | ||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Retail | Producer | Gross pay | Take-home pay | ||||||||
| 1 | 0.1 | 0.1 | –1.0 | –0.9 | –3.0 | –2.0 | –1.7 | –2.0 | –0.9 | –3.7 | –1.2 |
| 2 | 0.8 | 0.7 | –1.5 | –1.4 | –3.5 | –2.0 | –0.9 | –2.0 | 0.0 | –3.4 | –0.4 |
| 3 | 0.9 | 1.3 | –1.9 | –2.0 | –4.0 | –2.1 | –0.2 | –2.0 | 0.6 | –3.4 | –0.2 |
| 4 | 0.9 | 1.4 | –1.8 | –1.6 | –3.6 | –1.8 | –0.2 | –2.0 | 0.4 | –3.2 | –0.3 |
A negative sign indicates an improvement in competitiveness.
United Kingdom: Simulated Effects of Lower Real Wages, with Unchanged Nominal Framework
(Percent changes from baseline levels)
| Year | GDP | Employment | Prices | Average Earnings | Real Earnings | Real Wage Costs | Exchange Rate | Competitiveness1 | Short-term Interest Rates | ||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Retail | Producer | Gross pay | Take-home pay | ||||||||
| 1 | 0.1 | 0.1 | –1.0 | –0.9 | –3.0 | –2.0 | –1.7 | –2.0 | –0.9 | –3.7 | –1.2 |
| 2 | 0.8 | 0.7 | –1.5 | –1.4 | –3.5 | –2.0 | –0.9 | –2.0 | 0.0 | –3.4 | –0.4 |
| 3 | 0.9 | 1.3 | –1.9 | –2.0 | –4.0 | –2.1 | –0.2 | –2.0 | 0.6 | –3.4 | –0.2 |
| 4 | 0.9 | 1.4 | –1.8 | –1.6 | –3.6 | –1.8 | –0.2 | –2.0 | 0.4 | –3.2 | –0.3 |
A negative sign indicates an improvement in competitiveness.
The changes in the composition of aggregate demand are consistent with the mechanisms described previously. The main point to emphasize about these results is the pivotal role played by the decline in the price level and the adjustments to policies implied by the constant nominal framework. The decline in prices brought about by the reduction in wages raises the real value of wealth which stimulates consumption expenditures. The decline in prices also reduces the demand for money which in the context of an unchanged money supply leads to lower interest rates and hence stronger investment and consumption. The resulting increase in output reduces the PSBR, which triggers a personal income tax cut so as to keep the ratio of the PSBR to nominal GDP constant. It is these responses of interest rates and personal income tax rates which cause the output and employment response to be significantly stronger in this scenario than in the “fixed rates” policy scenario.
In the “fixed rates” scenario (Table 19) the sources of growth in demand are quite different. Investment and net exports are the driving forces whereas the level of consumption is lower throughout the period. Compared with the “unchanged nominal framework” scenario, investment spending is weaker and the improvement in net exports is due to a fall in imports stemming from the decline in consumption. The principal reason for these differences