a simple wage-price model

In order to describe the process of aggregate wage and price determination in the absence of indexation, let us consider a variant of a general wage-price model recently proposed by Tobin (1972a). The model used here has two features that are helpful for the subsequent analysis: (1) as it is representative of much of the recent and on-going empirical work on aggregate wage and price determination, estimates of most of the key parameters are available; and (2) it contains most of the channels or mechanisms by which wage indexation might conceivably affect the rate of inflation.

The model contains four equations—a wage adjustment equation, a price adjustment equation, an equation for price expectations, and an equation for normal capacity utilization, as presented below:

In the above model, W is the money wage rate (or average hourly earnings), P is the price level, P^e is the expected price level, U is the unemployment rate, $\dot{U}$ is the rate of change of the unemployment rate, γ is the trend rate of growth of labor productivity (real output per man-hour), R is the rental cost of capital services, PM is an index of import prices (in domestic currency), X is the capacity utilization rate, Xⁿ is the normal (average) value of the utilization rate, and the superscript ^* denotes the percentage change in a variable; for example,

The wage adjustment equation (1) is a form of the well-known Phillips curve. The rate of change of money wages is related, inter alia, to the unemployment rate and to the rate of change of the unemployment rate, where these two variables are proxies for the level and rate of change of excess demand in the labor market (see Lipsey, 1960). The percentage change in expected prices enters equation (1) to allow wage bargains to be struck in real rather than in nominal terms, while the trend rate of growth of labor productivity is intended to capture the equilibrium path of money wages over time (when there is no excess demand for labor).⁹

The price adjustment equation (2) relates the rate of price change to the rate of change of variable costs and to the level of excess demand in the product market, where the proxy for this latter variable is the deviation of capacity utilization from its normal value. Equation (3) expresses the expected rate of inflation as a weighted average of past actual rates of inflation, and equation (4) uses an equivalent, weighted average formula to express the normal (or expected) capacity utilization rate.

For the subsequent analysis of wage indexation, a few important characteristics of the above wage-price model should be noted. First, the presence of $\stackrel{*}{W}$ in the price equation and of $\stackrel{*}{P}e$ in the wage equation implies that there is a two-way relationship between wage changes and price changes.¹⁰ In turn, this implies that any change in the exogenous variables (U, $\dot{U}$, γ, $\stackrel{*}{R}$, $\stackrel{*}{\mathit{PM}}$, or X) will initiate a wage-price spiral; however, this spiral will not be explosive (will be dampened) so long as the product of the coefficients on $\stackrel{*}{P}e$ and $\stackrel{*}{W}$ in equations (1) and (2) is less than unity—that is, so long as (a₁b₁) < 1. Second, the model suggests that there will be a long-run (permanent) trade-off between the rate of wage (or price) inflation and the unemployment rate, again so long as (a₁b₁) is less than unity. These characteristics of the model can be illustrated by solving equations (1) and (2) for the long-run equilibrium (steady-state) values of $\stackrel{*}{W}$ and $\stackrel{*}{P}$. These two reduced form equations are most conveniently generated by dropping all time subscripts and by invoking the equilibrium conditions that $\dot{U}=0$, $\stackrel{*}{P}=\stackrel{*}{\mathit{P}}e$, and Xⁿ = X, which yields:

Note that in both equations (5) and (6) the coefficient on the unemployment rate will be infinite if (a₁b₁) = 1; hence, there would be no long-run trade-off between the rate of wage (or price) inflation and the unemployment rate. In fact, the model implies in more general terms that there will not be a stable, long-run equilibrium rate of wage (or price) inflation if (a₁b₁) = 1, for in such a case, the denominator in both equations (5) and (6) would be zero.¹¹ Finally, note that a comparison of equations (1) and (2) with equations (5) and (6) reveals that the long-run effect of a change in each of the exogenous variables on the rate of wage (or price) inflation will be greater than the short-run effect by the factor 1/(1 – a₁b₁). Thus, for example, if a₁ and b₁ were each equal to 0.75, the long-run effect on the inflation rate of (say) an increase in import prices would be about 2.3 times greater than the short-run effect.

the impact of wage indexation on inflation

Given the above wage-price model, it can now be seen how wage indexation might be expected to affect the rate of inflation. In general terms, wage indexation can affect the rate of inflation in any of the three following ways: (1) by altering the values of any of the explanatory variables in equations (1) through (4); (2) by altering the values of any of the coefficients in these same four equations; or (3) by introducing a new, or replacing an existing, explanatory variable in equations (1) through (4). Fortunately, many of these possible effects of wage indexation on inflation can probably be safely ignored here (on a priori grounds), since they are likely to be of negligible importance. For example, it seems doubtful that wage indexation would have any direct effect on the rate of change of either import prices or capital goods prices, or on the productivity of labor. In fact, the wage indexation literature suggests that it would be most fruitful to concentrate on the probable impact of wage indexation on (1) the money wage response to actual price changes (that is, on the a₁ coefficient), (2) the role of price expectations in the wage determination process, and (3) the relationship between aggregate demand changes and price changes, including the time lags in this relationship.

the money wage response to price changes

The first and most obvious channel by which wage indexation might affect the rate of inflation is through its effect on the money wage response to price changes. As illustrated by equations (1) and (3), when there is no wage indexation, money wages will respond to expected price changes, so that the money wage response to actual price changes will depend on two parameters—the response of money wages to expected price changes (a_i) and the response of expected price changes to actual price changes (Σα_i). Unfortunately, since price expectations are generally not observable, all that can be observed from empirical data is the money wage response to actual price changes (a₁Σ_αi) and not the two individual components of that relationship. Therefore, in the first part of this section, it is assumed that actual price changes equal expected price changes (that $\stackrel{*}{P_t^e}={\stackrel{*}{P}}_t,$), so that attention can be focused on the probable effect of wage indexation on the a_i coefficient. Later in this section, this assumption can be relaxed to permit an analysis of the effect of wage indexation on the rate of inflation by means of the substitution of actual for expected price changes in equation (1), this time assuming that the a₁ coefficient is the same with indexation and without indexation.

Obviously, the extent to which wage indexation affects the money wage response to price changes depends in large part on the form or type of wage indexation under consideration. Contrary to popular opinion, wage indexation as applied in practice has seldom provided for full wage compensation for all price changes. As such, the discussion below deals with the inflationary implications of several alternative forms of wage indexation.¹²

Full wage indexation

As a starting point, let us consider what can be regarded as a polar case of wage indexation. Suppose that a base year (call it year t) is selected for both the money wage index W and the aggregate price index P and that from year t on, it is mandated that W change by 1 per cent for every 1 per cent change in P. Further, assume that all other determinants of $\stackrel{*}{W}$ and $\stackrel{*}{P}$ are left unaffected by this wage indexation policy, so that price changes are not the sole determinant of money wage changes.¹³ In other words, it will be assumed in terms of the above wage-price model that full wage indexation means that a₁ = 1 and that all exogenous variables and other coefficients in equations (1) through (4) remain the same.

Given these assumptions, it is clear that wage indexation will be inflationary, in both the short and long run, if the wage response to price changes with full indexation is greater than the wage response to price changes in the absence of indexation.¹⁴ Equivalently, since a₁ with indexation is by assumption equal to unity, the required condition for an inflationary impact is that a₁ without indexation must be less than unity. The only unknown parameter is therefore the wage response to price changes in the absence of indexation.

Upon first thought it would appear that there should be no problem in identifying the a₁ parameter (without indexation), given that there exists a voluminous econometric literature on aggregate wage determination and given that almost all of these estimated equations contain estimates of the money wage response to price (cost of living) changes.¹⁵ Upon further reflection, however, it becomes apparent that the empirical wage studies may not provide the required information for two reasons. First, since the parameter of interest is the wage response to price changes in the absence of indexation, it is clear that estimates of a₁ taken from equations fitted over time periods in which there was indexation (for at least some years) will be inappropriate here because the estimated wage response to price changes will reflect in part the influence of wage indexation. This means that, ideally, one should exclude from the sample countries in which there have been national wage indexation policies in effect for much of the postwar period and countries in which a significant proportion of private labor contracts contained cost of living escalator clauses.¹⁶ In the first category, one should probably include Belgium, Denmark, Finland, France, Israel, Italy, the Netherlands, Norway, and Switzerland, among others. Canada, the United Kingdom, and the United States would probably fall in the second category.¹⁷ However, if these countries are excluded from consideration, very few countries with estimated wage equations are left from which to obtain information on the a₁ coefficient. In other words, there will not be an adequate number of observations for the period without indexation to permit a reliable comparison of wage behavior, in periods with and without indexation.

Second, the problem arises as to which time period to use in examining the wage response to price changes, because the empirical literature suggests that the a₁ parameter is quite unstable over time (see Cargill and Meyer, 1974; Goldstein, 1972 and 1974; and Gordon, 1971a). Thus, quite different conclusions could be reached about the size of the a₁ parameter and also about the inflationary impact of indexation, depending on the time period selected.

Despite the above-mentioned caveats, some limited information on the wage response to price changes in the absence of indexation might be gleaned from estimates of the a_x parameter for countries without pervasive indexation. Estimates are presented for the United States, the United Kingdom, Canada, Japan, and the Federal Republic of Germany in Table 1.

Table 1.

Econometric Estimates of the Money Wage Response to Price Changes

¹In the vast majority of the studies listed, the estimate of a₁ refers to the estimated coefficient on ${\stackrel{*}{P}}_t$ or ${\stackrel{*}{P}}_{t-1}$ in the money wage equation. In those cases where a distributed lag was used for price changes, the estimate refers to the sum of the estimated coefficients on the price change variables.

²Q = quarterly data, SA = semiannual data, and A = annual data.

Table 1.

Econometric Estimates of the Money Wage Response to Price Changes

Investigator	Estimate Of a11	Time Period	Data 2
United States
Gordon (1972)	0.70	1954–70	Q
	0.57	1954–68	Q
	0.42	1954–66	Q
Flanagan (1973)	0.37	1951–68	Q
Heien and Popkin (1972)	0.51	1948–65	A
Hirsch (1972)	0.77	1953–68	Q
Perry (1970)	0.33–0.43	1953–68	Q
Turnovsky and Wachter (1972)	0.35	1949–69	SA
De Menu and Enzler (1972)	0.57	1948–68	Q
Hymans and Shapiro (1970)	0.42	1954–68	Q
Black and Kelejian (1970)	0.62	1948–65	Q
Spitäller (1975)	0.58	1960–72	SA
Schultze and Tryon (1965)	0.76–0.88	1948–60	Q
Simler and Tella (1968)	0.19	1948–57	Q
	0.51	1953–64	Q
Rowley and Wilton (1973)	0.18	1948–60	Q
Nordhaus (1972b)	0.87	1955–71	A
United Kingdom
Lipsey (1960)	0.69	1923–39 and 1948–57	A
Flanagan (1973)	0.48	1951–68	Q
Goldstein (1974)	0.16	1954–67	Q
	0.91	1954–71	Q
Ashenfelter and Pencavel (1975)	0.18–0.22	1951–70	Q
Smith (1968)	0.44–0.52	1948–67	Q
Spitäller (1975)	1.0	1960–72	SA
Nordhaus (1972b)	0.61	1955–71	A
Lipsey and Parkin (1970)	0.48	1948–68	Q
Hines (1964)	0.58	1921–61	A
Artus (1975)	1.05	1956–72	SA
Canada
Turnovsky (1972)	0.90	1949–69	SA
Bodkin and others (1967)	0.43–0.51	1953–65	Q
Spitäller (1975)	0.85	1960–72	SA
Kaliski (1964)	0.33	1946–58	A
Zaidi (1969)	0.50	1947–62	A
Choudhry and others (1969)	0.31	1949–66
Rowley and Wilton (1973)	0.24–0.28	1953–65	Q
Helliwell and others (1969)	0.77	1952–65	Q
Japan
Nordhaus (1972b)	1.26	1955–71	A
Spitäller (1975)	1.0	1960–72	SA
Toyoda (1972)	0.48	1956–68	Q
Germany, Federal Rep.
Bodkin and others (1967)	1.23	1954–65	Q
Boelaert (1973)	0.50	1955–69	A
Spitäller (1975)	0.56	1960–72	SA

¹In the vast majority of the studies listed, the estimate of a₁ refers to the estimated coefficient on ${\stackrel{*}{P}}_t$ or ${\stackrel{*}{P}}_{t-1}$ in the money wage equation. In those cases where a distributed lag was used for price changes, the estimate refers to the sum of the estimated coefficients on the price change variables.

²Q = quarterly data, SA = semiannual data, and A = annual data.

View Table

Table 1.

Econometric Estimates of the Money Wage Response to Price Changes

Investigator	Estimate Of a11	Time Period	Data 2
United States
Gordon (1972)	0.70	1954–70	Q
	0.57	1954–68	Q
	0.42	1954–66	Q
Flanagan (1973)	0.37	1951–68	Q
Heien and Popkin (1972)	0.51	1948–65	A
Hirsch (1972)	0.77	1953–68	Q
Perry (1970)	0.33–0.43	1953–68	Q
Turnovsky and Wachter (1972)	0.35	1949–69	SA
De Menu and Enzler (1972)	0.57	1948–68	Q
Hymans and Shapiro (1970)	0.42	1954–68	Q
Black and Kelejian (1970)	0.62	1948–65	Q
Spitäller (1975)	0.58	1960–72	SA
Schultze and Tryon (1965)	0.76–0.88	1948–60	Q
Simler and Tella (1968)	0.19	1948–57	Q
	0.51	1953–64	Q
Rowley and Wilton (1973)	0.18	1948–60	Q
Nordhaus (1972b)	0.87	1955–71	A
United Kingdom
Lipsey (1960)	0.69	1923–39 and 1948–57	A
Flanagan (1973)	0.48	1951–68	Q
Goldstein (1974)	0.16	1954–67	Q
	0.91	1954–71	Q
Ashenfelter and Pencavel (1975)	0.18–0.22	1951–70	Q
Smith (1968)	0.44–0.52	1948–67	Q
Spitäller (1975)	1.0	1960–72	SA
Nordhaus (1972b)	0.61	1955–71	A
Lipsey and Parkin (1970)	0.48	1948–68	Q
Hines (1964)	0.58	1921–61	A
Artus (1975)	1.05	1956–72	SA
Canada
Turnovsky (1972)	0.90	1949–69	SA
Bodkin and others (1967)	0.43–0.51	1953–65	Q
Spitäller (1975)	0.85	1960–72	SA
Kaliski (1964)	0.33	1946–58	A
Zaidi (1969)	0.50	1947–62	A
Choudhry and others (1969)	0.31	1949–66
Rowley and Wilton (1973)	0.24–0.28	1953–65	Q
Helliwell and others (1969)	0.77	1952–65	Q
Japan
Nordhaus (1972b)	1.26	1955–71	A
Spitäller (1975)	1.0	1960–72	SA
Toyoda (1972)	0.48	1956–68	Q
Germany, Federal Rep.
Bodkin and others (1967)	1.23	1954–65	Q
Boelaert (1973)	0.50	1955–69	A
Spitäller (1975)	0.56	1960–72	SA

¹In the vast majority of the studies listed, the estimate of a₁ refers to the estimated coefficient on ${\stackrel{*}{P}}_t$ or ${\stackrel{*}{P}}_{t-1}$ in the money wage equation. In those cases where a distributed lag was used for price changes, the estimate refers to the sum of the estimated coefficients on the price change variables.

²Q = quarterly data, SA = semiannual data, and A = annual data.

An examination of the estimates of a₁ in Table 1 suggests that: (1) the vast majority of the estimates of a₁ are less than unity, which implies, ceteris paribus, that full wage indexation is likely to be inflationary; (2) the estimates of a₁ show considerable variation within each country, suggesting either that this parameter is quite unstable over time or that estimates of a₁ are quite sensitive to relatively minor differences in the specification of the aggregate wage equation; and (3) for those countries in which there are quite a few estimates of a₁ the estimates for the more recent time period (including the post-1968 period) are closer to unity than those for earlier periods. In fact, estimates of the aggregate wage equation for the 1965–74 period would probably yield estimates of a₁ that are not significantly different from unity.¹⁸ If this were correct, it would mean that implementation of full wage indexation in 1975 would produce little, if any, inflationary impact in most countries, at least in the short run.

the role of price expectations in wage determination

As suggested by wage equation (1), one of the determinants of the rate of wage inflation in the absence of wage indexation is the expected rate of inflation. With indexation, the role of expectations about the inflation rate would presumably be greatly reduced.²¹ In fact, one of the supposed advantages of indexation is that it will remove the uncertainty associated with future changes in prices and therefore that labor, product, and financial contracts can then be negotiated exclusively with respect to “real” variables.²² This in effect means that with full wage indexation, wage equation (1) can be rewritten with the actual rate of inflation $\left(\stackrel{*}{P}\right)$ replacing the expected rate of inflation $\left(\stackrel{*}{P^e}\right)$.²³

The above discussion suggests that, given some value for a₁ full wage indexation will be inflationary, anti-inflationary, or inflation-neutral, ceteris paribus, as the actual inflation rate exceeds, falls short of, or equals the expected inflation rate, respectively. Some of the proponents of wage indexation claim that full wage indexation will be anti-inflationary because, in the absence of indexation, workers overestimate the future rate of inflation and then succeed in embodying these erroneous inflation forecasts in their wage contracts. The opponents of wage indexation take the precisely opposite position that workers usually under estimate the actual future inflation rate, so that the implementation of wage indexation would be inflationary. Finally, the prevailing view among neoclassical economists is that while in the short run workers may underestimate or overestimate the inflation rate, in the long run their expectations will be correct (in the long run, $\stackrel{*}{P_t^e}={\stackrel{*}{P}}_t$), so that wage indexation will be inflation-neutral for any given value of a₁.

Unfortunately, economic theory does not offer very much guidance about the formulation of price expectations. As illustrated by equation (3), economists have in general assumed that people form their price expectations as some weighted average of past actual rates of inflation. Empirical work on price expectations has been very limited, however (in large part due to the lack of good survey data on price expectations), with the result that it is not really known whether workers (or employers) have underestimated or overestimated future rates of inflation.²⁴ Thus, the effects of wage indexation on inflation by means of the substitution of actual for expected rates of inflation remains essentially an open question. It might be noted, nevertheless, that if workers consistently overestimated the future rate of inflation and if they were powerful enough to obtain compensation for these price forecasts in their wage contracts, then they would have little to gain from indexation. Since much of the pressure for indexation has come from labor groups, one might therefore conjecture either that workers feel they have in the past underestimated the actual rate of inflation or that workers value the “insurance” that indexation provides against inflation even if it leaves unaffected their average increase in real wages.

A further factor to consider on the question of wage indexation and price expectations is the potential impact of the announcement of indexation on expectations. Some writers (see the references in Bhatia, 1974) have argued that the adoption of wage indexation will be regarded by the public as a clear signal that the government has abandoned its battle against inflation and has instead opted for “living” with inflation. For example, Day (1964, p. 2) has stated that indexing implies “a deliberate acceptance of inflation by the monetary authorities … and it is a clear admission of the expectation of defeat.” Similarly, Collier (1969) has stated that it is disastrous for the government to admit that inflation is inevitable, as this reinforces the price rise. In brief, what these writers are implying is that indexation will increase the public’s price expectations, and this in turn will ultimately increase the actual rate of inflation.²⁵

In evaluating the above argument, it is necessary to distinguish between economy-wide indexation (where all financial, labor, and product contracts are indexed) and partial indexation (where only some assets or labor contracts are indexed). If everything except currency in circulation is indexed on the actual rate of inflation, then an increase in the expected rate of inflation should be of little consequence, since only actual inflation rates matter. On the other hand, if (say) only wage contracts are indexed, then the adoption of wage indexation could lead to an increase in the expected rate of inflation and this in turn would affect the actual rate of inflation in those markets that are not indexed. The basic question, however, of whether in fact the announcement of indexation would actually increase the expected rate of inflation is again an open one, because price expectations are generally not observable. In particular, the relationship between government policy decisions and price expectations is an almost completely unresearched area. Thus, an agnostic position on this issue seems unavoidable.

aggregate demand changes, price changes, and time lags

Recently, Friedman (1974) and Laidler (1974) have proposed yet another mechanism by which indexation can affect the rate of inflation—or, more accurately, how indexation can improve the government’s ability to deal with inflation. It is their position that past attempts to use aggregate demand policy to control inflation have been largely unsuccessful due to the long time lag between decreases in aggregate demand and decreases in the rate of inflation. More specifically, Friedman claims that when aggregate demand falls in an unanticipated manner, producers respond by first reducing output and/or accumulating inventories, next by reducing employment, and only finally by lowering prices. Similarly, laid-off workers first wait to be recalled to their former jobs, next seek new jobs elsewhere, and only finally moderate their money wage demands. He estimates that the total time delay between a change in monetary growth (demand policy) and a change in the rate of inflation is about two years in both the United Kingdom and the United States. During the interim period, it is common for both unemployment and the rate of inflation to be high, with the result that public pressure leads to a reversal of aggregate demand policy and, hence, to little if any reduction in the rate of inflation. Under this situation, it is the unwillingness of the public to accept the temporary side effects of anti-inflationary policy that prevents the government from reducing inflation.

With indexation, Friedman and Laidler argue that the time delay between aggregate demand changes and price changes will be reduced—that is, the time lags in equations (1) and (2) will be shortened.²⁶ The time lag is presumed to be reduced because with indexation it will no longer be necessary to reduce price expectations in order to reduce the future actual rate of inflation and because indexation will make it easier for the public to recognize changes in the actual rate of inflation. A capsule statement of this position is given by Laidler (1974, p. 541):

Without escalator clauses, deficient demand leads to unemployment; this decreases the inflation rate which in turn decreases inflationary expectations so that future wage and price inflation rates are also cut. Each step in the process takes time, but the use of escalator clauses “short circuits” the expectation adaption stage and hence speeds up the whole process.

A second related argument for indexation is that full wage indexation will make it easier politically for the government to reduce inflation because it will reduce the employment and output losses associated with aggregate demand reductions. This point has been most fully developed by Morley (1974) within the context of a simple one-sector model of the economy where all prices are rising (or falling) at the same rate. Morley uses this model to contrast the effects of anti-inflationary, demand-reducing policy under full wage indexation with those in the absence of indexation. First, consider the case without indexation and assume that the government attempts to reduce the inflation rate by adopting a restrictive demand policy in a period in which past actual inflation rates have been high. Morley argues that when aggregate demand is reduced and prices fall, the expected inflation rate will for a time exceed the actual inflation rate. Workers will therefore be demanding, and obtaining, salary increases which turn out afterward to be too high. Consequently, businessmen will have to lay off workers because their selling prices will not be high enough to meet the salary demands of workers—that is, the money wage will exceed the value of labor’s marginal product. Thus, the level of unemployment will rise and the equilibrium level of employment will be restored only when workers revise their expectations so that the equilibrium level of real wages is restored. By analogy, antirecessionary, aggregate demand-increasing policy will be associated with a temporary increase in the level of employment above the equilibrium level.

Now consider the case with full wage indexation. Under indexation, workers will not need to build any inflation forecasts into their wage demands because any real wage losses which they experience due to inflation will be fully restored to them through cost of living adjustment. Therefore, Morley argues, it will not be possible for workers to overestimate the future rate of inflation and to demand an excessive money wage, which is what causes the employment losses under stabilization policy. Rather, with indexation, it is argued that a fall in the actual rate of inflation induced by restrictive demand policy will be automatically matched by a fall in the rate of increase of money wages. Thus, it is claimed that indexation will permit the authorities to move the economy from a higher to a lower rate of inflation without causing unemployment to rise above its equilibrium level because under indexation the real wage will not deviate from its equilibrium level.

Critical evaluation of both the Friedman-Laidler and Morley hypotheses is severely hampered by the lack of empirical studies comparing wage-price flexibility and employment behavior under indexation with that in the absence of indexation.²⁷ Without such empirical evidence, it is difficult to know whether the adjustment of prices and money wages to aggregate demand changes would occur as smoothly and as rapidly under indexation as these authors suggest. Further, both the Friedman-Laidler and Morley arguments have been framed within the context of simple models which embody some rather restrictive assumptions. For example, Friedman implicitly assumes that with indexation producers will be more inclined to reduce their selling prices in response to a decline in demand because such a price decline will lead to a compensating decline in money wages, thereby leaving unaffected the real wage. Such an argument, however, makes no distinction between output price, which is the relevant price variable for the producer’s employment decision, and the consumer price index, which is the relevant price variable for the worker’s employment and wage-bargaining decisions. More specifically, since a single employer’s price change is likely to have a negligible influence on the cost of living, he cannot be certain that his price cuts will be matched by a fall in the money wages of his employees. In fact, given the existence of imported products, indirect taxes, rents, and so on in the cost of living (and given that the producer has no control over the prices of these items), it is questionable whether a decline in the price of (say) all manufactured goods would be associated with a corresponding decline in either the cost of living index or the index of money wage rates. In short, even under full wage indexation, producers cannot be certain about their real wage costs.

Similarly, Morley’s analysis of employment behavior in the absence of indexation rests on the assumption that producers will lay off workers whenever the money wage exceeds the value of labor’s marginal product. As such, it gives no role or influence to the fixed transaction costs associated with hiring, training, and firing workers—all of which drive a wedge between the money wage and the value of labor’s marginal product—and which transform labor into a quasi-fixed factor of production in the short run (see Oi, 1962). Finally, neither Friedman nor Laidler pays much attention to the issue of time lags in the payment of compensation for cost of living changes under indexation—a technical detail, but one which will have an important effect on how fast money wages adjust to price changes and, hence, to aggregate demand changes. In sum, while both the Friedman-Laidler and Morley arguments for indexation have considerable intuitive appeal, many questions remain to be answered before one can be confident about indexation’s positive effect on the government’s ability to deal with inflation.

summarizing the effect of wage indexation on inflation

One of the dominant conclusions that emerges from the preceding analysis is that it is not possible to draw any meaningful conclusions about the effect of wage indexation on inflation without specifying explicitly both what type of wage indexation is being considered and what is being assumed about wage behavior in the absence of wage indexation. Further, even when this is done, very few definitive conclusions can be drawn because existing empirical wage studies are generally not suitable for inferring either what would have happened in the absence of indexation or what would happen if wage indexation were to be introduced.

Despite the large degree of uncertainty surrounding the inflationary implications of wage indexation, the analysis presented in this section suggests the following tentative conclusions:

(1) Full wage indexation is more likely to be inflationary if implemented during a period of low inflation (say, the 1960s) than during a period of high inflation (say, the 1970s) because the wage response to price changes in the absence of indexation is apt to be closer to unity when inflation is high than when it is low.

(2) Both partial wage indexation (a₁ < 1) and threshold wage indexation (a₁ = 1 only when $\stackrel{*}{P}>$ some constant) are likely to be less inflationary (and perhaps even anti-inflationary) than full wage indexation, but they are more likely to lead to worker dissatisfaction in the long run, especially if the a₁ coefficient is set too low or the threshold rate of price increase is set too high.

(3) A wage-price spiral will result from changes in any exogenous determinant of wage (or price) inflation whether indexation is present or not. However, the size of the wage-price spiral (which depends in large part on the coefficients a₁ and b₁) may well be affected by the implementation of wage indexation. Further, threshold wage indexation may help to insulate wage determination from small, temporary price changes.

(4) In cases where workers have erroneously high expectations about the future inflation rate and where workers and labor unions are powerful enough to embody these expectations in their wage contracts, the implementation of wage indexation, ceteris paribus, may well prove to be anti-inflationary by substituting actual inflation rates for expected inflation rates. Of course, where employers hold the balance of power and have erroneously low expectations about the future inflation rate, the above conclusion will be reversed.

(5) In firms and/or industries where producers are price-setters rather than price-takers, wage indexation may improve the government’s ability to deal with inflation by reducing the time lag between aggregate demand changes and price changes. However, even under full wage indexation, producers will not have complete assurance that their price cuts will be matched by a corresponding decline in money wages.

(6) When the government attempts to reduce inflation by implementing restrictive aggregate demand policy, full wage indexation may decrease, ceteris paribus, the rise in unemployment by preventing workers from seeking money wage increases (based on expected rates of inflation) that turn out later to be too high. The slower workers are to adapt their price expectations to the actual rate of inflation, the more likely this is to occur.

the variability of real wages over time

Although workers, employers, and government policymakers are no doubt primarily concerned with the mean rate of increase of real wages over time, they may also have an interest in the variability of real wage changes. Taking a cue from portfolio theory, one might surmise that at any given mean rate of change of real wages, all three groups would prefer a lower variance in the rate of change of wages if for no other reason than that a highly variable rate of change (of real wages) will make it more difficult to plan long-term consumption, production, and stabilization decisions. The question then arises as to what effect, if any, wage indexation is likely to have on the variability of real wages over time.

It may be recalled from Section I that in the absence of indexation it will be necessary for workers and employers to base their wage bargains on the expected rate of inflation over the contract period. In some cases (perhaps most) the inflation forecasts of one or both parties will be erroneous—with the result that the aggrieved party (workers when $\stackrel{*}{P_t}>\stackrel{*}{P_t^e}$ or employers when $\stackrel{*}{P_t}<\stackrel{*}{P_t^e}$) will attempt to obtain compensation for its forecast error in the next wage bargain. Thus, unless $\stackrel{*}{P_t^e}=\stackrel{*}{P_t}$, the next wage bargain will be a function not only of real variables (labor productivity and the excess demand for labor) and the expected rate of inflation but also of the size of the “inflation correction” factor.²⁸ Clearly then, the more variable this inflation correction factor is over time, the more variable, ceteris paribus, the time path of real wage changes will be. Wage indexation, or more appropriately full wage indexation, will probably reduce the variability of real wage changes over time because it will obviate the need for an inflation correction factor as well as for inflation forecasts. However, so long as workers and employers are able to obtain full correction for previous errors in forecasts, the mean rate of increase of real wages will be the same as with full wage indexation.

Let us consider a concrete example. Assume first that a one-year wage contract is signed which provides for a 10 per cent increase in money wages. Assume further that labor and management have agreed to this wage bargain on the assumption that prices will increase by 5 per cent over this one-year period so that, in effect, 50 per cent of the wage bargain represents real factors (say, increased productivity) and 50 per cent represents monetary factors. Now assume that the actual inflation rate proves to be 10 per cent. If inflation expectations and projected productivity do not change, and if labor can obtain full correction for its past forecast error, the next wage bargain will provide for a 15 per cent increase in money wages (where the 15 per cent equals 5 per cent for productivity, 5 per cent for expected inflation, and 5 per cent for inflation correction). This time assume that the actual inflation rate proves to be 3 per cent. If the same assumptions are retained (with the exception that management is now the recipient of the inflation correction), the third money wage bargain will be for 8 per cent. Finally, assume that the actual inflation rate in the third period is 5 per cent. Calculating the increase in real wages for each period, it can be seen by inspection that the three increases will be zero, 12 percent, and 3 per cent—a mean rate of increase in real wages of 5 per cent over the three-year period. It might also be noted that one could generate the same time path of money and real wage changes by dropping the inflation correction factor and by replacing the assumption $\stackrel{*}{P_t^e}=5$ per cent in each period with the assumption $\stackrel{*}{P_t^e}={\stackrel{*}{P}}_{t-1}\mathrm{.}$.

Now repeat the same exercise, assuming full wage indexation—that is, assume that the money wage bargain in each period equals 5 per cent for projected productivity plus the actual rate of inflation over the period. In this case, the increase in real wages will obviously be 5 per cent in each period, so that the mean rate of increase over the three-year period will be identical to that without indexation. The variance of real wage changes will, however, be much smaller than in the case without indexation; in the above example, this variance under full indexation would be zero.

While the preceding exercise is admittedly a stylized one based on rather restrictive assumptions, it illustrates that full wage indexation is likely to lead to greater stability of real wages over time. As such, full wage indexation is also likely to lead to less variability of employment over time, since in classical employment theory the demand for employment is a negative function of the real wage. It should be noted, however, that so long as the expected real wage equals the actual real wage in the long run, the level of employment will be the same with full indexation as without it, other things being held equal.

To the extent that the variability of real wages is reduced by indexation, the mean rate of increase of real (and money) wages may in turn also be affected, although the outcome in this case is difficult to predict since it depends, inter alia, on the characteristics of the utility functions of workers and employers (about which only very limited indirect information is available). For illustrative purposes, assume that the utility of workers varies positively with the real wage and negatively with the variance of real wages, and also assume that the corresponding utility function for employers varies negatively with the real wage but positively with the variance of real wages (that is, assume that workers are risk averters and employers are risk lovers). In such a case, if the variance of real wages falls, the mean level of real wages (or its rate of increase) should fall as well so as to keep the utility of both parties unchanged. In other words, workers will be willing and, just as importantly, will be required to accept a lower real wage to pay for the insurance that indexation provides against variability of real wages. On the other hand, under the seemingly more plausible assumption that employers are also risk averters, both parties will be “better off” when the variance of real wages is reduced; hence, no change in the mean level of real wages should be forthcoming. In the cases where either (but not both) workers or employers are risk neutral, the results appear to be indeterminate if no additional assumptions are made about the properties of the two utility functions. Unfortunately, there have been no attempts (to the author’s knowledge) to test empirically the relationship between changes in aggregate money (or real) wages and the variability of wages so that the practical importance of the above issue remains unanswered.²⁹

the duration of labor contracts

As mentioned previously, one of the presumed advantages of indexation is that it removes the uncertainty associated with future changes in the price level. In other words, when there is no indexation, workers and employers will be assuming a risk when they enter into labor contracts denominated in nominal terms. Further, under the assumption that it will be more difficult to forecast prices accurately in period t + 1 than in period t, this risk will be greater the longer the contract period is. Thus, in situations where inflation rates are high and variable and where workers and employers are risk averters, one would expect labor contracts to be of relatively short duration (probably less than one year).

With the introduction of full wage indexation, one can expect the time duration of most labor contracts to increase because both workers and employers will be protected against unforeseen price changes.³⁰ For example, it has been reported that if it were not for the semiannual cost of living adjustment in most wage contracts, labor unions in Denmark would insist on one-year contracts rather than the two-year or three-year contracts commonly found in practice (see Ulman and Flanagan, 1971, pp. 125–26). To the extent that indexation increases the length of contracts, it is likely to generate some cost savings for employers and workers since wages will not have to be renegotiated as frequently as in the absence of indexation.

aggregate strike activity and industrial peace

Another important factor to consider in assessing the case for or against wage indexation is its probable impact on aggregate strike activity in the economy. Even a casual reading of the literature on industrial strike activity suggests a number of channels by which indexation might be expected to affect the number and duration of strikes in an economy. For the purposes of this paper, it will suffice to mention three such channels. First, although econometric modeling of aggregate strike activity is still in its infancy, there seems to be growing support for the hypothesis that strike frequency (and duration) varies inversely with both the rate of change of real wages (over, say, the preceding two-year period) and the unemployment rate.³¹ Thus, ceteris paribus, if indexation leads to a larger increase in real wages, it should in turn lead to lower strike activity. Second, to the extent that indexation leads to a lengthening of the duration of labor contracts, it should, ceteris paribus, reduce the number of strikes because, at least in countries where labor contracts are legally binding on both parties, strikes can occur only when existing contracts expire. If the number of strikes occurring in a given year (5) is thought of as being equal to the number of labor contracts negotiated in that year (NL) times the probability that a negotiation will end in a strike (P_s)—that is, S = NL · Ps—then indexation can reduce S by reducing NL. Third, and finally, if it is assumed that some strikes occur because labor and management cannot agree on the expected future rate of inflation, then indexation, by eliminating the need for such inflation forecasts, can remove one potential area of disagreement from the negotiations and hence, ceteris paribus, can reduce the number of strikes.

The above analysis suggests that even if wage indexation leads to a larger increase in both money and real wages than would occur in the absence of indexation, the indexation-induced increase in real wages may produce some offsetting benefit to employers in the form of more favorable and more stable labor relations. Laidler (1974), for example, has recently put forward the view that the increasing tendency for U.S. labor contracts to include cost of living escalator clauses has surely contributed to the lack of industrial strife that has accompanied accelerating inflation in the United States. In a similar vein, Wallich (1971) has suggested that “no strike pledges” might be obtained from employees whose labor contracts provide for full cost of living protection. It should be noted, however, that until there is greater certainty about the effect of wage indexation on the rate of change of money (and real) wages, it cannot be predicted with very much confidence what effect wage indexation will have on aggregate strike activity.

the functional distribution of income

Up to this point, the discussion has centered primarily on the probable impact of wage indexation on the rate of change of the price of labor services—that is, on the rate of change of money (or real) wages. Even if, however, it could be unambiguously demonstrated that wage indexation would lead to a larger increase in the rate of change of money wages, it would not necessarily follow that labor’s share in industry (or national) income would increase. This is because increases in money wages lead, ceteris paribus, to decreases in employment and because labor’s share in national income reflects both the price of labor services and the quantity of labor employed.³² More specifically, given some increase in money wages, labor’s share in income will rise, fall, or remain constant as the elasticity of substitution between labor and capital is less than, greater than, or equal to unity, respectively. Further, following Hicks (1963), the elasticity of employment with respect to money wages (the elasticity of the derived demand for labor) will be greater as the elasticity of substitution between labor and capital and the price elasticity of demand for the final product become greater. Also, in cases where the price elasticity of demand exceeds the elasticity of substitution, the elasticity of the derived demand for labor will be larger as labor’s share becomes larger.³³

The above discussion suggests that, in order to predict the likely effect of wage indexation on the functional distribution of income between labor and capital, it will be necessary to know the size of the elasticity of substitution, as well as the effect of wage indexation per se on the rate of change of money wages. Unfortunately, despite a voluminous econometric literature on the subject (see the review of the empirical evidence in Nerlove, 1967), there seems to be no agreement on whether this elasticity is greater than, less than, or equal to unity.³⁴ Early direct estimates of the elasticity of substitution by Arrow and others (1961) suggested a value less than unity for most of the 24 countries in their sample, but these estimates have subsequently been challenged. Further, it is extremely doubtful that the effect of past wage indexation policies on the change in labor’s share could be identified by comparing changes in labor’s share between countries with and without indexation because indexation is sure to be only one of the many factors responsible for such changes (and probably a relatively unimportant factor). Thus, while it is by now well documented (for example, see Kravis, 1968, and Heidensohn, 1969) that labor’s share in national income has been steadily rising in most countries for at least the past three to four decades, there is as yet no evidence that wage indexation, where applied, has contributed to such a change.³⁵

In sum if, as sometimes assumed, wage indexation leads to a larger increase in money wages than would occur in the absence of indexation, and if the elasticity of substitution is less than unity, labor’s share in national income will increase as a result of the implementation of indexation. At present, however, neither of these two assumptions can be accepted with a great deal of confidence. What is more predictable is that if wage indexation induces a faster growth in real wages without affecting the productivity of labor, it will also induce, ceteris paribus, some decrease in employment.

other labor market effects of indexation

The analysis presented in this section suggests the following tentative conclusions about the effects of wage indexation on the labor market:

(1) Full wage indexation is likely to lead to greater stability of real wage movements over time; as such, wage indexation should also lead, ceteris paribus, to greater stability of employment over time and, hence, to a reduction in the fixed hiring, training, and lay-off costs associated with changes in employment.

(2) To the extent that wage indexation eliminates the need to forecast future inflation rates and to the extent that these forecasts in the absence of indexation are often incorrect, wage indexation may well lead to a more efficient allocation of labor resources in the economy by reducing the number of erroneous labor force and employment decisions based on incorrect perceptions of relative prices.

(3) Full wage indexation is quite likely to lead to a lengthening of the duration of labor contracts and in turn to some cost savings for both employers and workers, since the costs associated with frequent wage renegotiations will be diminished.

(4) Wage indexation may well lead to a decrease in aggregate strike activity in the economy if it leads either to an increase in real wages or to a lengthening of the duration of labor contracts, or to a reduction in those strikes caused by divergent views of management and labor on the expected rate of inflation.

(5) If wage indexation leads to an increase in money wages, it will lead to a decrease in employment, ceteris paribus; however, the probable effect of wage indexation on labor’s share in national income is not clear, since there is at present considerable uncertainty about the size of the elasticity of substitution between labor and capital in most countries.

wage indexation and devaluation

As is well known, one of the traditional arguments for devaluation is that it is a viable method of reducing the real wage (thereby improving the economy’s international competitive position) in countries where money wages are inflexible downward. In other words, it is argued that workers will accept a reduction in their real standard of living that is induced by devaluation which they would not accept in the form of a forced reduction of domestic money wages. This argument, however, rests on the implicit assumption that the money wage response to price changes (the coefficient a₁ is less than unity (see Goldstein, 1974). Therefore, it is sometimes concluded that full wage indexation will prevent devaluation from having a positive effect on the trade balance, since under such a policy a₁ will be equal to unity.

In evaluating the impact of wage indexation on the efficacy of devaluation policy, it is necessary to consider the following factors. First, to the extent that the a₁ coefficient is close to unity in the absence of wage indexation, full wage indexation will have very little effect on the efficacy of a devaluation—that is, devaluation will not bring about much, if any, reduction in the real wage in either case. Second, for the cases of partial or threshold wage indexation—which are the most popular in practice—the a₁ coefficient with indexation might be less than that in the absence of indexation; in such cases, devaluation might be more effective with indexation than without it. For example, with threshold indexation it will be possible to obtain a reduction in the real wage from devaluation as long as the devaluation does not push price changes up to or beyond the threshold point.³⁷ Third, it may be possible to reduce the wage response to a devaluation with indexation if the price index used in the indexation formula does not contain imported (finished) goods—for example, by linking wages to the gross national product deflator rather than to the consumer or retail price index. However, if imported goods represent a significant proportion of consumption and investment expenditures, workers are apt to become dissatisfied with a wage determination formula that permits the international purchasing power of their income to fall over time. Finally, it may be recalled that in the absence of indexation, wage changes will be based, inter alia, on the expected rate of inflation. If the expected rate of inflation, in turn, is a positive function not only of present and past actual rates of inflation but also of the devaluation itself, then indexation may reduce, ceteris paribus, money wage changes by substituting the actual rate of inflation for the higher expected rate of inflation. Once again, however, it is unfortunately not really possible to evaluate this effect because very little is known about the impact of government policy decisions on the formation of price expectations.

In sum, it does not necessarily follow that wage indexation will make devaluation a completely ineffective or even a less effective policy instrument. Further, even if indexation did have an adverse effect on the efficacy of devaluation policy, it could always be temporarily suspended so that devaluation could have its intended relative price effects. For example, cost of living escalator clauses in wage contracts were suspended for a one-year period in Denmark after the 1967 devaluation so as to increase the effectiveness of that devaluation (see Vega, 1973).