Some Empirical Evidence on the Determinants of Wage and Price Movements in Japan, 1950–73: A Survey

International Monetary Fund. Research Dept.
Published Date:
January 1975
  • ShareShare
Show Summary Details


One of the most controversial economic issues at present in Japan is the problem of inflation. During the decade of the 1960s and the first three years of the 1970s, the annual rate of inflation fluctuated between – 2 and 3 per cent in terms of the wholesale price index and between 3 and 8 per cent in terms of the consumer price index. However, in 1973 the rate of price increases accelerated to an annual rate of 16 per cent in terms of wholesale prices and to 21 per cent in terms of consumer prices.

These events have occasioned lively discussions, particularly about the main factors causing inflation and about remedies for curtailing it. The possibility of introducing incomes policies has also been attracting the attention of public officials, academicians, and businessmen. Many have argued that the vicious circle created by increases in wages and prices must be stopped by incomes policies, but such arguments have not yet gained wide support from economists and government officials. In any attempt to combat inflation, however, the relationship between money wages, labor productivity, and prices must be closely scrutinized. As a first step, an understanding of the mechanism of price and wage determination is necessary, as well as an assessment of the effects of various factors on the rate of increase in prices and wages.

The purpose of this paper is twofold: (1) to review critically some of the econometric studies on money wages and prices in Japan after World War II and to summarize the major findings; (2) to construct, on the basis of these studies, a prototype model describing the relationships that determine money wages and prices in Japan. Some implications are drawn about possible incomes and demand management policies that may be necessary to bring the rate of inflation down to a more tolerable level.

Section I of this paper contains a critical review of existing studies;1Section II presents a prototype model and discusses its implications for controlling inflation. A summary of the major findings appears in Section III, and the Appendix contains a short note on econometric problems encountered in some of the studies surveyed.

I. Econometric Studies on Wages and Prices in Japan After World War II

Since the pioneering works by A. W. Phillips (1958) and R. G. Lipsey (1960), several empirical studies have been carried out, analyzing the relationship between money wages and inflation in postwar Japan. Most of these have followed the theoretical framework developed by Phillips and modified by Lipsey. Generally, changes in wages are regarded as being dependent on price movements and the level of demand pressure in the labor market, while changes in prices are regarded as dependent on changes in labor productivity, money wages, and some other factors.

The studies were reviewed basically to attempt to clarify four issues: (1) the possibility of simultaneous interdependence between changes in money wages and prices; (2) the best proxy variable representing the demand pressure in the labor market; (3) the role of firms’ ability to pay higher wages; and (4) the “dual structural relationship” in the price-wage determination mechanism. Although the findings of the studies are not necessarily consistent and by no means uniform, they have shed considerable light on factors influencing money wages and prices, in spite of difficulties encountered in quantifying the significance of these factors.

interrelationship between wages and prices

The possible two-way causation in the relationship between changes in money wages and prices was approached by estimating the simultaneous equation model first; a judgment was then made on this question, depending on the statistical significance of estimated coefficients of money wages and prices. The first such attempt was made by Watanabe (1966). His simple simultaneous equation system can be expressed in general form as follows:


P*t=percentage changes in prices
Pr*t=percentage changes in productivity
W*t=percentage changes in money wages
Ut=level of unemployment.

His estimation of the equation system (1) by the two-stage least-squares method, using quarterly data for the period 1956/IV to 1962/IV, resulted in the following statement in his study (p. 43):

This result suggests that the hypothesis of a simultaneous determination between price and money wage earnings cannot be supported empirically, since the sign condition of the price variable in the wage equation became negative, in addition to general declines of the coefficients of determination.

Spitäller (1972) performed a similar exercise using annual data for the period 1955–70, and his empirical results led him to state that the hypothesis of a simultaneous interrelation between concurrent movements in earnings and consumer prices could not be confirmed. As a result, both Watanabe and Spitäller concluded that the changes in prices could be explained primarily by an induced demand effect originating from the rise in money earnings through a successive determination system. In such a system, they also argued, changes in prices in the previous period, but not in the concurrent period, tended to have a strong influence on money wage changes (see Watanabe, 1966, p. 45, and Spitäller, 1972, pp. 16–20).

In spite of poor statistical results from estimating the coefficient of the price variable, the conclusions of Watanabe and Spitäller cannot be taken for granted for a number of reasons. The fact that the estimated coefficient for the price variable has a wrong sign, or has a very low t-ratio with the right sign, does not always imply that there is no simultaneous relationship between changes in wages and prices. The estimated equation may have been incorrectly specified. The data for the variables in the system may not reflect what the variables should be representing. The low t-ratio, or large standard error of estimates for the coefficients, could still suggest that the existence of a simultaneous relationship is highly probable.

In addition, the fact that P1* in the Watanabe model has a statistically significant coefficient in the wage equation suggests the existence of simultaneity; since P* is defined as a percentage change in the price index of the current quarter (t) compared with a year earlier, and W* is defined in a similar manner, there is an overlapping period of three quarters between P1* and W*. Toyoda’s (1972) results indicate a similar relationship between changes in nominal wages and prices. Therefore, the conclusions drawn by Watanabe and Spitäller seem premature.

In contrast to the difference of opinion regarding the impact of prices on wages, there is strong statistical evidence that increases in money wages in a given period tend to have a positive impact on price changes in that period. However, contributors to the literature disagree about the magnitude of that impact, depending on how the equation is specified and how percentage changes in prices and wages are defined. The elasticity of price changes with respect to wage changes is estimated to be about 0.7 by Watanabe (1966), 0.5 by Spitäller (1972), 0.3 by the Economic Planning Agency (1971), and somewhere between 0.4 and 0.2 by Minami and Ono (1973), while Toyoda assumed it to be unity by definition.

According to theoretical models developed in the past, there are two ways to measure the effects of wages on prices. One is to relate price changes directly to changes in nominal wages and labor productivity. In this formulation, it is implicitly assumed that the impact of changes in productivity on prices would be different from that of changes in wages. The above formulation was adopted by Watanabe, Spitäller (1972), and the Economic Planning Agency. Another method would be to relate changes in prices to those in unit labor costs, where it is assumed that changes in labor productivity and nominal wages would have the same impact on prices. Such an approach was adopted by Minami and Ono (1971, 1972, 1973) and Toyoda (1972).

There seems little support for the validity of the second approach. True, firms try to adjust prices of their output according to changes in money wages paid to their employees and to changes in their labor productivity. However, in adjusting prices, entrepreneurs tend to give more weight to wage changes than to productivity changes, since they are more likely to have an accurate picture of wage movements than of productivity changes. These a priori considerations seem to be supported by the empirical results of Watanabe, Spitäller, and the Economic Planning Agency.

Only a few empirical studies strongly support the hypothesis of a simultaneous two-way causation in the relationship between changes in wages and prices. The Economic Planning Agency (1971) has estimated the simultaneous equation system (as a subsystem of the 54 equation model) by the two-stage least-squares method, using quarterly data for the period 1954/1 to 1969/IV. The result indicates that both the estimated coefficient of the price variable (percentage changes in consumer price index) in the wage equation and that of the wage variable (percentage changes in the per capita earnings of regular workers) in the price equation have the right sign and are statistically significant.

Minami and Ono (1973) give partial support to the simultaneous interdependence between price and wage changes in their six-equation model. Their model shows that fluctuations in wages of the high productivity sector are influenced by changes in the wholesale prices of this sector, among other things, and that changes in wholesale prices are in turn dependent on wage fluctuations in this sector.

Up to this point, the discussion of the possibility of a simultaneous interrelationship between changes in money wages and prices has been conducted on the basis of the statistical evidence, which suggests that the existence of such a relationship is highly probable. However, it is now necessary to examine an institutional arrangement of wage adjustments in Japan and to investigate the validity of such statistical inference. What is the mechanism under which wage adjustments are reflected by changes in the general price level in the concurrent period or in the most recent past in Japan? How are wage adjustments reflected in price changes in the same period? A recent report by the Committee on Prices, Income, and Productivity (1972) offers illuminating answers to these questions.

One of the important characteristics of the wage adjustment process in Japan is that adjustments are made every year, instead of once every two or three years, as in the United States or some countries of Western Europe. Therefore, rising prices almost immediately increase the ability of business firms to pay higher wages. Second, the extent of the wage adjustments is usually related to aggregate demand pressure, particularly in the labor market, as well as to general price movements. During a boom period, tight labor market conditions and strong upward pressure on prices not only induce labor union members to demand higher wages but also encourage business firms to offer higher wages in order to attract new employees, including new graduates and workers from other industries. Reflecting these developments, firms in the rest of the economy tend to follow suit and adjust the wages of their workers. During a downswing, price movements in the current and immediate past periods tend to be the dominating factor in wage negotiations between labor unions and firms, given the level of labor productivity. The outcome of such negotiations tends to set an example for others to follow. Also, higher wages have an immediate impact on unit labor costs. Firms tend to respond quickly by raising prices in order to maintain their profit margins.2 These characteristics of the institutional arrangements in the adjustment process of wages and prices suggest that a two-way causal relationship between wages and prices does exist in some sectors of the economy.

The above discussion leads to the following conclusions:

(1) The institutional arrangements in the wage-setting machinery in Japan indicate that a simultaneous interaction between price changes and wage fluctuations exists in principle, at least in some sectors of the economy.

(2) When such variables are measured at annual rates, using either annual or quarterly data, the empirical evidence provides some support for the hypothesis of the existence of a simultaneous interaction between the two variables.

proxy variables for demand pressure in the labor market

It is generally accepted among economists that demand pressure in the labor market plays an important role in influencing changes in money wages. However, there seems to be little consensus on what variable most adequately represents demand pressure. Unemployment statistics have often been used as proxy variables, but every empirical researcher recognizes the deficiencies in this variable. Watanabe (1966, p. 40) points out two important defects in the Japanese labor force survey:

(i) the conceptual definition of unemployment is too broad …, e.g., the persons who worked only one hour during a specified week per month are counted as part of the active labor force; and (ii) the definition with respect to population is not uniform throughout the survey.

Minami (1972) also argues that when the rate of unemployment decreases to a certain level—say 1 per cent—the rate of unemployment begins to be unresponsive to changes in the excess demand for labor, as happened in Japan in the 1960s and early 1970s. His argument can be extended even further to include the prevailing Japanese system of “employment for a lifetime,” which generally does not satisfactorily reflect the extent of demand pressure in the labor market whatever the rate of unemployment is. Therefore, the official unemployment statistics do not reflect disguised or hidden unemployment, which is extraordinarily high compared with other countries (see Minami, 1973, and Toyoda, 1972). Minami (1968, 1970a, 1970b, 1972) also pointed out that statistics on the rate of unemployment and on unemployment insurance only reflect ex post differences between demand for and supply of labor.

For these reasons, many writers have either adjusted the official unemployment statistics by allowing for disguised unemployment or they have looked for other variables to reflect demand pressure in the labor market. Watanabe (1966) argued that the number of persons receiving unemployment insurance payments was a more suitable variable than the rate of unemployment. Spitäller (1972) observed that the job opening rate, defined as the ratio of vacancies to job applicants, has a highly significant explanatory power in a wage equation. At the same time, he also found that in an equation relating price changes to unemployment (which may be interpreted as a price equation in reduced form), the unemployment rate and its rate of change performed better as explanatory variables than the job opening rate did. Toyoda used the sum of the official unemployment rate and a potential rate, which is the percentage rate of working persons wanting and actively seeking additional or different work in the total labor force. Minami (1972, p. 45) proposed to present “a new WAF [wage adjustment function] which does not include the rate of unemployment nor any other substitutes for the excess demand for labor, but does include explicitly the demand and supply functions of labor.” He posits a three-equation model consisting of (1) a wage adjustment equation, (2) a labor supply function, and (3) a labor demand function. By substituting the demand and supply function of labor into the wage adjustment function, he obtains a reduced form of the wage formation equation which is used for estimating the structural parameters. His main conclusion (p. 53) is that “wage changes are well explained by the excess demand for labor.”

Despite these efforts to find suitable proxy variables or estimates of excess demand conditions in the labor markets, there are still a number of problems inherent in the existing approaches. For example, the proxy variable used by Watanabe for the employment rate does not reflect as unemployed those who do not receive unemployment insurance payments for one reason or another, even though they are seeking jobs.

Toyoda’s reason for adjusting the official unemployment with a potential rate is (1972, pp. 269–70):

In Japan, workers intensively dislike long duration of unemployment and feel compelled to find some kind of work, even of a distasteful nature. Young unmarried workers usually prefer getting even undesirable jobs to being unemployed, preferring and anticipating to rise in the future along the seniority ladder system. Married male workers usually must continue to work even under unsatisfactory conditions to support their families, the custom of the working wife being alien to Japanese culture.

However, the justifications he gives for incorporating the potential unemployment seem awkward. If the workers’ disaffection with their current employment encourages them to seek additional or different job opportunities, the ratio of those “potential unemployed” to the total labor force would tend to increase during a period of strong demand pressure in the labor market. Dissatisfied workers would seek additional or different employment, with the expectation that they would have a better chance of obtaining a more satisfactory job opportunity during the boom period. Thus, the rate of potential unemployment tends to be positively related to demand pressure.

The ratio of job openings to applicants seems a far better indicator of demand conditions in the labor market, since, unlike the unemployment statistics, these statistics refer to the ex ante conditions. Nonetheless, this ratio is also somewhat deficient, because these statistics cover job openings and applicants registered in the public employment offices. Many employers in Japan still seek laborers through relatives, friends, and school teachers, and many job hunters use the same channels.

The estimation process for finding the excess demand condition in the labor market proposed by Minami (1973) seems most suitable from a conceptual point of view. There is no need to rely on the survey statistics of employed or unemployed labor. Instead, the estimation is based on the reduced form of the wage equation, which is dependent on past wages, income, population, price levels, and related factors. Since the published data on these variables are much more accurate than employment and unemployment statistics, the estimation process may reduce the inaccuracies caused by the poor quality of the statistics. Unfortunately, the process could run into econometric difficulties; since the estimates of one or more structural parameters in the reduced form equation may have a large standard error, as in Minami’s 1973 study, it would be unwise to rely only on the point estimates of these parameters to estimate the potential demand for and supply of labor. It is highly probable that the true magnitude of the excess demand condition in the labor force would be different from the estimates shown in his paper.

From this discussion it can be concluded that the most appropriate proxy variable for the excess demand condition in the labor market in Japan has yet to be unequivocally identified.

the ability-to-pay argument

Many empirical studies of the wage determination model have been based on a highly oligopolistic market structure. In such a model, the “ability-to-pay” argument is closely linked to “bargaining power,” and its importance has been generally accepted. However, there is wide disagreement over what is the best proxy variable to reflect this ability to pay. Kaldor (1959) argues that the rise in money wages depends on the bargaining strength of labor, which is closely related to the prosperity of industry: the greater the prosperity, the higher the labor union’s demand for wage increases. It is further argued that, under these circumstances, employers have great ability to grant higher wages. Bhatia (1962, p. 255) examines Kaldor’s concept of profit in more detail and argues that “a priori, one might expect that employers would be more willing to grant, and labour more eager to extract, higher wages when profits are unusually high than when they are unusually low, and vice versa, irrespective of whether profits are rising or not.”

Perry (1966) suggests that the ability of employers to pay a given wage increase is closely related to profits, since they are most clearly linked with the aspect of the bargaining situation. However, he points out that the measure of profits does not reflect all the dimensions of a firm’s ability to grant a wage increase but rather that this ability is only partially reflected in profits.

In contrast to these contributors to the literature, Kuh (1967) puts forward a productivity theory of wage levels by which he demonstrates that profits are likely to be a proxy for productivity, which is a more fundamental determinant of wages than profit. His argument suggests that during the period of rising productivity when profits are rising in the short run, the demand for labor, and thus wage rates, will increase. Consequently, Kuh argues that a profit variable should be replaced by a productivity variable, which is a key variable in the demand function for labor, and attaches little validity to the ability-to-pay argument.

Minami and Ono recognize that the ability of firms to grant higher wages plays an important role. In their 1971 and 1972 articles, they use the labor value productivity in the industries to reflect the demand condition for labor. This argument is very similar to Kuh’s productivity theory of wages. Watanabe used the ratio of profit (including depreciation) to net worth of own capital in nonagricultural industries, while the Economic Planning Agency considered the average share of corporate income to national income adjusted for inventories.

The disagreement over the most approximate variable representing the ability to pay stems from the complexity of the mechanism under which the “bargaining power” itself is generated.3 Because of this complexity, any measurable variable seems to reflect only partially the extent of the bargaining power; thus, how well a given variable represents the ability to pay becomes an empirical question.

Minami and Ono’s 1973 study shows that changes in labor value productivity in the high productivity sector have a significant correlation with changes in money wage rate in this sector, and the elasticity of the latter variable to the former is estimated to be about 0.6. The Economic Planning Agency also found significant correlation between the proxy variable for the ability to pay and money wages; however, the definition of the variables adopted in the study makes it difficult to interpret the estimated elasticity. Statistical results in Watanabe show considerable correlation between changes in the profit ratio and those in money wages but with a wrong sign, which he does not explain. Perhaps such results were obtained by a specification error in terms of the variables that are included or excluded, as well as the lag structure of the profit ratio.4

Judging from the results obtained by these studies, empirical evidence supports the claim that the ability-to-pay argument is meaningful in explaining money wage increases, but it is still not possible to decide on the best proxy variable for it.

dual structure

The so-called dual structure in the context of the Japanese wage-price determination mechanism is well known to Japanese economists.5 The dual structure argument was based on differential rates of price and wage increases in different sectors. Since the early 1950s, the wholesale price index for all commodities had been relatively stable, while the consumer price index had increased at a considerably high rate. If one examines price movements of the components included for the construction of the wholesale price index, it will be observed that the prices of products of the high productivity sector tended to decline, while those of the low productivity sector increased considerably. The hypothesis was advanced, therefore, that price increases are uniquely influenced by the efficiency of various sectors. Minami and Ono (1972, 1973) extended the dual structure hypothesis to explain wage determination. As discussed in the preceding section, wage increases in the low productivity sector tended to follow those in the high productivity sector. The results obtained in Minami and Ono (1973) generally support the dual structure hypothesis of a price and wage determination mechanism. Statistical results indicate that the industry’s ability to pay (as represented by labor value productivity) and past price increases have a significant influence on money wages, in addition to the influence of unemployment rate. Changes in the money wage rate in the high productivity sector influence changes in the service sector, which in turn affect those in the low productivity sector. The significance of the statistical results for the price equations is not as clear, since these equations are in the log-linear form with some ad hoc specifications, and there is a high degree of autocorrelation in the residual terms. In spite of these defects, some evidence supports the dual structure argument in price determination. According to the price equations, prices in the high productivity sector are largely influenced by the unit labor cost in this sector and by import prices. Prices in the low productivity sector are affected by those in the high productivity sector, as well as by the unit labor cost and aggregate demand pressure as represented by the real gross national expenditure (see Minami and Ono, 1973).

II. A Prototype Model on Wage and Price Determination

The preceding section has reviewed some of the econometric studies on money wages and prices in postwar Japan. These studies produced some evidence to support the simultaneous relationship between prices and wages, but the question of the most suitable proxy variable for excess demand in the labor market remains unsettled. Nonetheless, these studies have shed considerable light on the structural characteristics of the mechanism by which wages and prices are determined in Japan. On the basis of these econometric studies, a prototype of the model will be developed,6 and some implications will be drawn for incomes policies that may be necessary for bringing inflation down to a more tolerable level.

prototype model

Minami and Ono’s 1973 model seems to be the best foundation on which a prototype model can be developed, since their model appears to capture developments in prices and wages that are unique to the low and the high productivity sectors. A close examination of their model reveals the following characteristics in the wage and price determination mechanism:

(1) During a period of tight labor market conditions, as in recent years, wage increases in the low productivity sector (including the service sectors) tend to be much larger than those in the high productivity sector, with other things remaining constant.

(2) Since the productivity increase in the low productivity sector tends to be lower than that in the high productivity sector, the rate of increase in unit labor costs tends to be higher in the low productivity sector than in the high productivity sector.

(3) Price increases in the high productivity sector will be significantly influenced by increases in imports, because the raw materials used for production make up the bulk of imports in this sector, while increases in wholesale prices of the low productivity sector and consumer prices are affected indirectly by import price increases through price increases in the high productivity sector.

Fluctuations in the consumer price index are mainly affected by changes in the wholesale price index of the low productivity sector and in wages in the service sector.

These characteristics seem to explain satisfactorily the much higher rates of price increases in the consumer prices relative to the rates of wholesale price increases in all the industries during the period 1960–72, and to a certain extent the reverse developments in 1973.

Taking into account the above mechanism and some slight modification in the wage and price determination, the following model can be constructed:

where fij represents the first partial derivative of the ith function with respect to the jth argument.

Endogenous variables

W*h,W*l=percentage changes in wages of the high and the low productivity sector, respectively
W*he=expected rate of changes in wages in the high productivity sector
P*h,P*l=percentage changes in wholesale prices of the high and low productivity sector, respectively
P*c=percentage changes in the consumer prices
P*ce=expected rate of changes in the consumer price index.

Exogenous variables

U=demand pressure in the labor market, measured by, say, unemployment
Q*h,Q*l=percentage changes in the labor productivity in the high and the low productivity sector
P*m=percentage changes in import prices.

Of the seven endogenous variables, W*l and W*h are percentage changes in nominal wages in the low and high productivity sectors, respectively; P*l and P*h are percentage changes in the wholesale price index of the low and high productivity sectors, respectively; and P*c is the percentage change in the consumer price index. Expected price and wage variables (P*ceandW*he) are generated by the current and lagged values of these variables.7

Determination of wage changes

One of the explanatory variables for W*l and W*h is the demand pressure represented by the unemployment rate (U) in the labor market. However, U is a variable reflecting the condition of the labor market in the economy as a whole and is not likely to reflect it satisfactorily on a sectoral basis. In order to supplement the deficiency of this variable, an additional variable was included. For the high productivity sector, the rate of changes in the labor value productivity added (P*h+Q*h) in that sector reflects the ability of the firms to pay wage bills:8 the higher this variable, the greater the wage increase in the high productivity sector. For the low productivity sector, the expected rate of changes in the nominal wages in the high productivity sector influences the rate of changes in the wages of this sector. This element seems particularly important during labor shortages when small and medium firms of the low productivity sectors are eager to attract workers from the high productivity sector as well as school graduates. In addition, the expected rate of inflation in terms of the consumer price index is included as an explanatory variable for both W*l and W*h, since there seems to be considerable empirical support for the hypothesis that the expected rate of inflation plays an important role in wage changes (see Toyoda, 1972).9

Determination of price changes

It is hypothesized that changes in wholesale prices are influenced by unit labor costs, as well as by the cost of inputs from other sectors. However, since the responses of price changes to wage increases are different in their magnitude from those to changes in labor productivity, the rate of changes in the unit labor cost is separated from that of wage and productivity changes. One of the reasons behind the hypothesis that the cost of inputs is a propelling force in wage fluctuation is that as far as wholesale prices are concerned, prices are determined by the mark-up principle. The validity of this reasoning concerning the high productivity sector is obvious, when the market structure has oligopolistic characteristics. A similar hypothesis applied to the price determination of the low productivity sector can be explained as follows. As can be seen from the way rice prices are determined, government policies toward agricultural products and public services suggest that a direct relationship exists between the prices of outputs and those of inputs in the low productivity sector. This pattern can be seen in the relationship between import prices and wholesale prices in the two sectors, as well as in the relationship between price changes in the low productivity sector and fluctuations of wholesale prices in the high productivity sector. The former relationship reflects the fact that more than 80 per cent of imports consist of raw materials, most of which are used in the high productivity sector. The latter relationship demonstrates that the low productivity sector draws on important inputs, such as steel and other products of the high productivity sector.

It is assumed that fluctuations in consumer prices are influenced by those in the wholesale prices of the low productivity sector, as well as by import prices in the same period. It is well known that many products in the low productivity sector are included for computation of the consumer price index.

implications of the model for controlling inflation

There are two major methods for investigating how the prototype model can be used to study inflation control. One is to look at the reduced form equation; the other is to treat wage variables as policy instruments and inflation variables as target variables. With this latter method, the implications of an incomes policy can be discussed.

Reduced form approach

Since the prototype model expressed above has not been estimated, it may be worthwhile to discuss the hypothetical coefficient matrix for the reduced form. This matrix will be based on some empirically estimated coefficients that have appeared in the studies reviewed. Before presenting the matrix, a few points must be made. First, the matrix will represent a steady-state level of changes in wage and price variables—that is, if any lagged variables are included in the estimated structural equation, it can be assumed that the current and lagged variables take the same value in the steady state (for example, W*he=W*h,P*ce=P*c,W*l=W*lt1, and so on). Secondly, the constant term will be included in each equation to show the effects of the trend. Thirdly, demand pressure is measured by the unemployment rate (ratio of job applicants to job openings). Under these conditions, the structural relationship of the wage-price determination mechanism and its reduced form can be shown as follows:

Structural equations

Reduced form equations

Typical values of the steady-state variables may take such values as 0.8 for U, 10 for Q*h, 8 for Q*l and 4 for P*m on an annual basis.10 Then the value for endogenous variables will be:

As this example shows, steady-state money wages of the high and low productivity sector increase at virtually the same rate of about 16 per cent, and wholesale and consumer prices increase at a moderate rate.11 Since the values of the exogenous variables are obtained from actual data, it is expected that values of the endogenous variables would be fairly close to the actual data for the period 1967–72, if the hypothetical coefficients of the equations (Γ) through (7’) are consistent with actual behavioral relationships. Indeed, the values of the estimated endogenous variables are very close to the actual data. Therefore, it can be said that the reduced form coefficient matrix (and, of course, the hypothetical parameters for the structural equations) would be reasonable. However, these results do not reflect price movements in 1973, when both wholesale prices and consumer prices increased at a substantially higher rate. But, if the unexpected events in 1973 (speculative hoarding, sharp price increases of imported materials, and accompanying mark-ups of unusually high proportions) were not of a recurring nature, then steady-state price increases would be about the same in the future as in the period 1967–72. As can be seen from the reduced form coefficient matrix, about 35 per cent of higher prices of import materials will be transmitted to money wages and wholesale prices of the two sectors, as well as consumer prices. The level of unemployment seems to influence considerably the rate of changes in money wages and prices, except for price changes in the high productivity sector. This implies that a tight fiscal and monetary policy will eventually slow down the rate of increases in the long run. Since tight fiscal and monetary policy has been in effect for about a year and a half since early 1973, the effects are expected to appear in prices and money wages in the future. If these policies were prematurely reversed, the rates of increases in prices and money wages would probably be adversely affected.

Another important element in the price-wage determination mechanism is the rate of increase in labor productivity in the low productivity sector, which tends to lower the rate of increases in money wages and prices in all the sectors. An increase in productivity is desirable, of course, but a significantly higher level would be difficult to achieve in the short run because of the required long-term improvements that must be made in such areas as technology and education. These analyses of the reduced form coefficient matrix indicate that a demand management policy would be effective in controlling inflation.

instrument-target approach

If an incomes policy includes government control of wages and if one assumes that the rates of inflation in terms of wholesale prices of the high and low productivity sectors and consumer prices are the target variables, then one can ask what kind of wage guidelines the government can impose, given the target rates of inflation and data on other uncontrollable exogenous variables.

Since there are three target variables (P*h,P*l,andP*c), at least three instrument variables are necessary to achieve such targets according to the Tinbergen tradition. Since there are only two wage variables available as instruments, it is necessary to add one more instrument variable. Some of the exogenous variables are not qualified. For example, because of the nature of the model, the demand management policy variable cannot be included in the set of instruments, since this variable only affects prices indirectly through wage variables. When wages are controlled, the potential influences of the demand management policy are cut off. The variables representing productivity variables are not strictly under government control, since they are largely determined by such factors as technological progress and education. The only variable that qualifies as the instrument is the price of imports, since they can be influenced partly by the exchange rate policy. Therefore, it is assumed that the set of instrument variables consists of W*h, W*l, and E*R, where E*R is defined as the percentage change in the exchange rate of the yen vis-à-vis the U.S. dollar.

The relevant equations under consideration are now equations (4’), (5’), and (7’). They can be written in the matrix form as follows:

where B^, Â, Γ^ is a 3 x 3 matrix of the estimated coefficient of the endogenous variables, the instrument variables, and uncontrollable exogenous variables. A 3 x 1 vector includes: Y=(P*h,P*l,andP*c),C=(W*h,W*l,andE*R), and X=(Q*h,Q*l,andP*f), where P*f represents the percentage changes in the import prices in the world market, expressed in U. S. dollars, and where all the other variables are as previously defined. The solution of the value of C, given the target values Ȳ and the data on X¯, is

A typical matrix for Â, B^, and Γ^ is shown below:

Therefore, (ii) can be rewritten as

In order to estimate the values of the instrument variables which are consistent with inflation targets, a simulation exercise was carried out. Throughout the exercise, the assumed values of the uncontrollable variables Q*h,Q*l,andP*f are uniformly applied.12 The results shown in Table 1 suggest that if the policymakers set the target rate of inflation at 8 for P*h and 10 for P*l and P*c, then the value of W*h, W*l, and E*R is found to be about 20, 24, and 16, respectively. That is to say, the rate of increase in the money wage rate must be set at 20 and 24 per cent a year in the high and the low productivity sector, respectively, while the exchange rate of the yen vis-à-vis the U.S. dollar may be depreciated by about 16 per cent a year.

Table 1.Simulation Results of Incomes Policies1
Values of Target VariablesValues of Instrument Variables

Assume Q*h = 10, Q*l = 8, P*f = 4. All the figures represent percentage changes per annum.

Assume Q*h = 10, Q*l = 8, P*f = 4. All the figures represent percentage changes per annum.

If the target rates of inflation are set at 0, 2, and 2 for P*h, P*l, and P*c, respectively (Case V in Table 1), the money wages can increase by only 4 per cent in the high productivity sector and 13 per cent in the low productivity sector. However, the exchange rate can remain unchanged.13

Even though the meaning of the numerical solutions obtained above is quite straightforward, a number of practical problems remain in implementing the wage controls and exchange rate policies. A detailed discussion of these problems is outside the scope of this paper,14 but a few points can be mentioned. First, the depreciation of the yen over time may be unrealistic in view of balance of payments considerations. Secondly, the effects of exchange rate changes on demand pressure through the balance of payments is neglected here. It is obvious that the depreciation of the yen would result in a reduction of the balance of payments deficits or an increase in the surplus.

As a result, the aggregate demand for goods and services and, therefore, the demand pressure in the labor market tends to increase. This increased demand, in turn, exerts pressure on higher wage rates. Nonetheless, from the incomes policy point of view (through the instrument-target approach) a lower rate of increase in the wage rate is called for. Therefore, the artificial control over wage rate tends to result in a very serious distortion in reflecting natural forces of market conditions, and such distortion is not likely to be maintained for long. In addition, the depreciation of the yen and the incomes policy may bring about instability in the system. For example, the depreciation calls for a slower rate of wage increases, which in turn raises the competitiveness of Japanese exports in the world market; this again leads to an improvement in the balance of payments position, and to the demand pressure in the labor market as well as in the goods market. In order to offset that pressure, a further reduction in the rate of increases in wages is necessary, and so on. For these reasons, such an incomes policy will break down before long.

Thirdly, public opinion is likely to oppose any wage control that suggests a much higher rate of increase in money wages in the low productivity sector than in the high productivity sector. Therefore, the policymakers may not be able to implement an incomes policy (including an exchange rate policy) because of either adverse public opinion or balance of payments considerations.

A further question can now be raised: what target rates of inflation would be consistent with a publicly accepted incomes policy and a stable exchange rate? Suppose that such a policy consists of 16 for W*h and W*l and 0 for E*R, and assume that Q*h, Q*l, and P*f are 10, 8, and 4, respectively.15 In this case, prices may increase by only about 2 to 4 per cent a year. Such an increase does not seem unrealistic in view of wage-price movements in the latter half of the 1960s, when wages increased at a rate of 15 to 20 per cent a year, wholesale prices were quite stable, and consumer prices increased 4 to 6 per cent a year.

This result is very close to the one obtained from the reduced form approach. In both approaches, the steady-state changes of productivity in the high and low productivity sector and the price of imported materials (in the world market) are assumed to be the same. An incomes policy which limits the rate of money wage increases to 16 per cent is shown to be consistent with a rate of inflation of about 2 to 4 per cent. A demand management policy (fiscal and monetary policy) which can maintain the rate of unemployment (measured by the ratio of job applicants to job offers) at 0.8 is shown to determine endogenously the rate of inflation and money wage increases;16 thus, monetary and fiscal policy can achieve a similar result in the rate of inflation as does the incomes policy indicated by the above example.

III. Concluding Remarks

This paper has reviewed critically some of the econometric studies on price and wage movements in Japan after World War II. These studies have examined the hypothesis that a simultaneous interrelationship exists between wages and prices, and they have tried to determine the most appropriate proxy variable for demand pressure in the labor market and for the industry’s ability to pay. They have also touched upon the dual structural relationships in the wage-price determination mechanism.

Some studies do not support the hypothesis of simultaneous interaction between the wage and price variables, but the Economic Planning Agency (1971) and Minami and Ono (1973) have produced some supporting evidence. A review of the institutional arrangements of wage determination in postwar Japan has also produced some support for the hypothesis. Consequently, it can be reasonably concluded that wages and prices tend to interact simultaneously within the concurrent period, when percentage changes in variables are measured on an annual basis.

The question of the most appropriate proxy variable for demand pressure in the labor market remains unsettled, because measures of demand pressure are deficient from the conceptual point of view. However, if one variable must be chosen among available measures, the ratio of job openings to job applicants may be the most suitable, since the statistical significance of this variable is fairly well established in many econometric studies and reflects the ex ante demand pressure. The only shortcoming of this variable is that the studies on which it is based have been derived from a limited sample. For example, data on job openings and job applicants refer only to records at the official employment exchange centers.

The ability-to-pay argument is found to be relevant in explaining wage increases. However, it is not clear by any means what is the best proxy variable to represent the industry’s ability to pay. Further empirical studies appear necessary to clarify this issue.

Constructing a wage-price determination model which incorporates the dual structural relationship in the Japanese economy was found very useful. Without this relationship, a model would be severely limited in its ability to explain the fundamental mechanism of wage-price determination.

On the basis of the studies reviewed here, a prototype model of wage and price determination was constructed. Policy measures for curbing inflation implied by the model include: (1) tight fiscal and monetary policy leading to the reduction of demand pressure in the labor market; (2) increases in the labor productivity in both the high and the low productivity sectors, if possible; (3) an incomes policy consisting of wage controls and exchange rate policies (but this would be rather difficult to implement, due to adverse public opinion opposing a much higher rate of money wage increases in the low productivity sector than the high productivity sector, to balance of payments considerations, or to inherent instability in the system when the control variables include changes in both the wages and the exchange rate). However, it was suggested that if money wages were allowed to increase by no more than 16 per cent a year and the exchange rate under the equilibrium condition of the balance of payments were kept constant, the rate of increases in wholesale prices or consumer prices could be kept at 2 to 4 per cent a year. Similar rates of inflation and money wage increases were obtained by demand management policies, which can set the rate of unemployment at 0.8 per cent. Thus, demand management policies seem superior to the incomes policy from an economic point of view. These observations, however, are very tentative and should not be regarded as conclusive, since an analysis of demand management policy and various aspects of the balance of payments was not made in detail in this paper.

APPENDIX: Notes on Econometric Problems

Below is a brief summary of econometric problems encountered in the studies under consideration.

Specifications of lags

In the models of Watanabe (1966), Toyoda (1972), the Economic Planning Agency (1971), and Minami and Ono (1973), the unemployment variable enters the wage equation with the wrong lag. This problem arises because the rate of increase in money wages is calculated on the basis of the level in the current quarter and in the preceding fourth quarter (at annual rate), while the unemployment rate is the average in the current quarter. Therefore, the unemployment rate enters the equation with a lead which is incompatible with the theoretical specification unless further assumptions are considered (see Black and Kelejian, 1972).

Simultaneous equation bias

The ordinary least-squares method that is adopted in Toyoda (1972), Minami and Ono (1973), and Eguchi (1970) is not appropriate in view of the simultaneous interaction specified in their model. Therefore, this method may have resulted in a biased estimation of coefficients, but it is difficult to determine the extent of any bias involved in the coefficients estimated by the ordinary least-squares method.

Autocorrelation in residual terms

There seems to be a strong positive autocorrelation in the wage equation of Toyoda (1972) and in all the price equations of Minami and Ono (1973). Therefore, the significance of the estimated coefficients in these equations may be overstated.


Toyoda (1972) and Minami and Ono (1973) argued that the rate of unemployment alone is not adequate to reflect the general demand pressure in the economy. In order to supplement the deficiency, they included a variable representing percentage changes in real expenditure. Obviously, some correlations between these variables exist. For example, one would expect that when the rate of unemployment is low, the rate of increase of the real expenditure would be high, and vice versa. However, it is difficult to determine the extent of the multicollinearity between the two variables from the studies surveyed.


Mr. Otani, an economist in the Asian Department, is a graduate of the University of California at Berkeley and the University of Minnesota. In addition to colleagues in the Fund, the author is grateful to Professor Ronald Findlay of Columbia University and to Professor Yusuke Onitsuka of Osaka University for their useful comments on the paper.

Econometric studies surveyed include the Economic Planning Agency (1971), Eguchi (1970), Minami (1973), Minami and Ono (1971, 1972, 1973), Saito (1972, 1973), Spitäller (1971, 1972), Toyoda (1972), and Watanabe (1966). Non-econometric studies reviewed include the Committee on Prices, Income, and Productivity (1972), Minami (1968, 1970a, 1970b, 1972), Ohkawa (1973), Shirai (1971), and Takasuka (1963).

An exception to this process is found in the public sector, where there would be some time lag between price adjustments and wage increases.

See John T. Dunlop (1966, pp. 74–94). He cites the factors influencing (determining) bargain power as follows:

“1. Tastes of workers and employers, with respect to wages and manhours bought and sold…. Institutional factors, such as property rights, and wage-hour legislation….

“2. Market conditions, especially the degree and type of competition in the labor market, the product market, the market for complementary factors of production, and the market for competitive factors of production.…

“3. ‘Pure’ bargaining power: ability to get favorable bargains, apart from market conditions: (a) the extent of knowledge of tastes and market conditions influencing the behavior of the other party to the contract; and (b) intrinsic ‘toughness’; the ability to get the desired result with a given amount of energy and unpleasantness.”

The regression result referred to above is as follows:

where ΔWt = WtWt–4 and ΔPt = Pt+1Pt–1. See Watanabe (1966, p. 41). There is no economic theory that justifies the lag structure in the profit ratio variable (P).

See for example, Minami and Ono (1972, 1973), Ohkawa (1973), Shirai (1971), and Takasuka (1963).

This prototype model should not be regarded as the best model of wage-price determination; it is a model that represents the main characteristics of the econometric models surveyed in the paper.

Since most of the equations in this model are overidentified, simultaneous estimation methods can be applied to estimate the structural parameters.

Two points must be made concerning this variable (P*h+Q*h): (1) The unemployment rate (U) in the labor market may have possible multicollinearity with (P*h+Q*h); when the unemployment rate decreases statistically, there is a tendency for measured price increases and productivity gains to rise, and vice versa; (2) In principle, there is little reason to expect that money wage changes will react in the same magnitude as those in price and productivity. However, in order to incorporate the ability-to-pay argument, the two variables (P*h and Q*h) are combined.

When the adjustment coefficient in the adaptive expectation hypothesis is very high, the expected rate of changes in prices and wages can be replaced by the actual rate of changes in these variables.

U is the ratio of job applicants to job offers. The assumed values, except the value of P*m, are estimated by taking averages of the actual data during the period 1967–72. The rate of increase in the import prices (P*m) is assumed to be 4 per cent a year, compared with the historically observed value of zero per cent for the period. This modification was made in order to reflect the current trend of world inflation more fully.

One of the implications of the negative gap (about –6 per cent) between the rate of inflation and the increase in the unit labor costs is that there must have been a profit squeeze in the period 1967–72. The actual data on ratios of profits to sales in corporate sectors suggest that the average ratio declined by about 4 per cent in 1967–72, compared with 1962–66.

The assumed values are estimated by the actual data during the period 1967–72. Therefore, it is implicitly assumed that the steady-state changes in Q*h, Q*l, and P*f in the future will be virtually the same as in 1967–72.

It is interesting to note that the traditional wage guideline for limiting money wages to productivity gains does not hold in the case of Japan. This result is attributed to the dual structural relationship in the wage price determination mechanism, as well as to different elasticities of inflation variables with respect to changes in wage rates.

Qualitative arguments for and against incomes policies are well summarized in Pohlman (1972).

Implicit in this assumption is that the balance of payments position is in equilibrium from a long-run point of view.

This rate of unemployment was the average of annual data for 1965–72. Therefore, it is assumed that the government can introduce monetary and fiscal policies that will achieve a rate of unemployment of 0.8.

Other Resources Citing This Publication