Prices and Unemployment in Selected Industrial Countries
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

INDUSTRIAL COUNTRIES have recently experienced considerable increases in their general price levels. These increases have revived interest in the study of inflation and of policies to stabilize prices. Investigations into the causes of inflation have dealt frequently with the effects on prices of the pressure of aggregate demand against existing capacity. The degree of demand pressure is not observable directly, and, for this reason, it must be measured by some stand-in or proxy variable, such as the ratio of unfilled orders to sales, the ratio of inventory stocks to sales, and indices of capacity utilization.1 The main purpose of this study is to examine, for selected industrial countries, the relationship between inflation and one such measure of demand pressure, namely, the rate of unemployment. The description of the model employed and the discussion of empirical results are preceded by a brief review of some theoretical issues that arise in the context of the analysis of movements in wages and prices in relation to demand pressure.

Abstract

INDUSTRIAL COUNTRIES have recently experienced considerable increases in their general price levels. These increases have revived interest in the study of inflation and of policies to stabilize prices. Investigations into the causes of inflation have dealt frequently with the effects on prices of the pressure of aggregate demand against existing capacity. The degree of demand pressure is not observable directly, and, for this reason, it must be measured by some stand-in or proxy variable, such as the ratio of unfilled orders to sales, the ratio of inventory stocks to sales, and indices of capacity utilization.1 The main purpose of this study is to examine, for selected industrial countries, the relationship between inflation and one such measure of demand pressure, namely, the rate of unemployment. The description of the model employed and the discussion of empirical results are preceded by a brief review of some theoretical issues that arise in the context of the analysis of movements in wages and prices in relation to demand pressure.

INDUSTRIAL COUNTRIES have recently experienced considerable increases in their general price levels. These increases have revived interest in the study of inflation and of policies to stabilize prices. Investigations into the causes of inflation have dealt frequently with the effects on prices of the pressure of aggregate demand against existing capacity. The degree of demand pressure is not observable directly, and, for this reason, it must be measured by some stand-in or proxy variable, such as the ratio of unfilled orders to sales, the ratio of inventory stocks to sales, and indices of capacity utilization.1 The main purpose of this study is to examine, for selected industrial countries, the relationship between inflation and one such measure of demand pressure, namely, the rate of unemployment. The description of the model employed and the discussion of empirical results are preceded by a brief review of some theoretical issues that arise in the context of the analysis of movements in wages and prices in relation to demand pressure.

I. Some Theoretical Issues in the Analysis of Wage and Price Movements

Causes of inflation

Observed price increases cannot always be explained fully by demand pressure; for instance, inflation may rise in the face of unchanged levels of spare capacity in the economy. In the United States, price increases accelerated between 1955 and 1957, even though the pressure of aggregate demand did not increase over these two years. Apparently, the inflation had been caused by labor and management through the abuse of discretionary powers.2 Insofar as price developments reflect movements in prices of agricultural products and imports, these developments cannot be captured by a variable measuring demand pressure either, since the prices of the products vary mostly with supply conditions and the prices of the imports can be considered autonomous. Similarly, demand pressure alone cannot explain price increases occurring under conditions of inflationary expectations. Indeed, when inflationary expectations prevail, the rate of price increases may even rise despite a decline in economic activity.

The effect on prices of a given level of demand pressure varies with its distribution by sector or region and with the structural characteristics of the market. Concentration of demand pressure in a particular sector or region may induce sectoral price increases, which lead to more rapid general inflation than would be experienced if the pressure were more evenly distributed.3 Upward pressure on prices at any given level of demand can be dampened by reducing structural rigidities. However, structural inflation cannot easily be distinguished from demand inflation. There are indications that rising demand pressure increases structural rigidities.4

Observed price increases result, therefore, from interacting forces, and the isolation of individual determinants of inflation may not be feasible.

Demand pressure and unemployment

Measures of demand pressure, such as ratios of unfilled orders or inventory stocks to sales and indices of capacity utilization, relate chiefly to the industrial or manufacturing sectors of the economy. Since it is unlikely that demand in all sectors will vary exactly in proportion, such indicators may not be as good a measure of the pressure of aggregate demand as the rate of unemployment in the economy as a whole. However, while superior to indicators that relate to individual sectors only, the rate of unemployment is by no means a perfect measure of demand pressure. Its use involves at least three difficulties. First, the link between demand pressure and unemployment is neither direct nor uniform. Variations in the demand for goods and services affect the demand for labor and lead, in the first place, to changes in the number of vacancies. The change in vacancies ordinarily leads to a change in the rate of unemployment, but with a lag that depends on the structural characteristics of the market and the speed at which demand changes. Second, some unemployment is involved in the process of switching jobs. Instances of frictional unemployment of this type tend to rise with the number of vacancies because the inducement to change jobs is greater when job offers are abundant than when they are scarce, but the duration of frictional unemployment is assumed to be shorter during a cyclical upswing in the economy than during a downturn.5 Third, cyclical variation in economic activity may also influence the size of the labor force.6 It appears that, at least in the United States, such an influence has an asymmetric effect on the reported unemployment rate: while the rise in participation in the labor force following a rise in vacancies is reflected in reported unemployment data, the decline in participation after a fall in vacancies is not. As a consequence, the degree of actual demand pressure is overstated when the number of vacancies is small and unemployment is high.

Several studies of wage and price behavior in the United States have attempted to make the unemployment rate a better measure of demand pressure. Simler and Telia [40] and Vroman [45] adjusted the official unemployment data for the large reserves of marginal participants in the labor force in the early and mid-1960s.7 Vroman took account of trends in participation rates in the labor market. Vroman, Liebling and Cluff [25], and the authors of the Wharton model [12] restricted the measurement of unemployment to males of a certain age group whose influence on wage movements was considered dominant. Taylor [42] adjusted the rate of unemployment to include both hidden unemployment and hoarded labor, both of which measure the under-utilization of the labor supply. The fact that in all these instances the adjusted rate of unemployment8 appears to be superior to the unadjusted rate in explaining changes in prices and wages has tended—despite the difficulties involved—to confirm the basic hypothesis that unemployment is a valid measure of demand pressure.

Unemployment and wage movements

Any observed relationship between unemployment and inflation is presumed to reflect the effect of demand pressure on prices in general. The response of wages to unemployment should be more direct, since unemployment is more closely related to demand pressure in the labor market than to aggregate demand pressure.

The relationship between wages and unemployment was first expounded by Phillips [35], who found the rate of unemployment to be a significant determinant of the rate of change of money wage rates in the United Kingdom. The evidence suggested that the relationship, which Phillips regarded as stable over time, was nonlinear in the sense that the change in the rate of wage increase induced by a given fall in unemployment was progressively higher, the lower the level of unemployment. The property of nonlinearity can be attributed to two factors. The first relates to the employer’s competitive bidding for labor when unemployment is low. As unemployment declines and the labor market tightens, employers would bid up wages both to attract increasingly scarce unemployed labor and to hire employed labor away from existing jobs. A given reduction in employment would thus be associated with higher wage increases at lower levels of unemployment than at higher levels. The second factor relates to the reluctance of employees to accept, even when unemployment is high, wage increases below a certain amount or—although this may not occur in practice—wage reductions above a certain amount, On account of this rigidity of wages in the downward direction, successive increases in the rate of unemployment by a given amount would therefore involve a falling rate of change in wages.

Phillips and many later analysts assumed that the supply schedule for the labor market was given and that the structural characteristics of the labor market were constant, which implies a constant relationship between vacancies and the level of employment.9 Subsequent analysis of the microconditions of the labor market has stressed that market imperfections, such as labor immobility, incomplete information, and monopolistic power, determine the degree to which the position of the Phillips curve differs from what it would be under perfect market conditions where wage rates would not rise until full employment is reached. These imperfections constitute a “boundary” below which the Phillips curve cannot lie.10 Any level of unemployment therefore implies a certain change in wage rates that cannot be reduced, or made more negative, unless structural supply rigidities are reduced also.

Phillips argued that changes in wage rates were determined not only by the level of demand pressure but also by its rate of change. He inferred the importance of changes in demand pressure from scatter diagrams relating changes in wage rates, measured along the vertical axis, to the level of unemployment, measured along the horizontal axis. By inspection of the time path that connects in chronological order the actual combinations of unemployment and wage changes recorded in the scatter diagrams, he observed that wage increases tended to be higher than predicted when unemployment was falling and lower when it was rising, which led to “counterclockwise loops” of this time path around the fitted curve.11 Lipsey [26] presented quantitative evidence confirming the importance of changes in the level of unemployment and argued that the loops were a result of differences in the distribution of unemployment during periods of recovery and recession. These differences arise because recession is assumed to affect demand simultaneously in all markets, at least initially, while recovery is assumed to have less even effects. At the beginning of recovery, unemployment in individual sectors will decline and the rate of wage increases in these sectors will rise. If these wage increases are matched by similar increases in sectors where unemployment has not yet declined, the general rise in wages will be high relative to the reduction in total unemployment, and the time path of actual combinations of unemployment and wage changes will lie above the fitted curve. In a recession, by contrast, employment will fall off simultaneously in all sectors, and the time path will be below the fitted curve. Hansen [16] offers an alternative explanation of the observed counterclockwise loops, which is more persuasive because it can more easily accommodate changes in the sign of the coefficient of the change in unemployment. The loops are attributed to the lag with which unemployment reacts to a change in vacancies. A cyclical upturn in demand raises vacancies and wage rates, but the structural characteristics of the labor market may not allow an immediate reduction of unemployment. In fact, unemployment may still be rising as a result of the preceding recession. Similarly, unemployment might still be declining at the upper turning point of the cycle, even though a drop in vacancies may already have reduced the rate of increase in wage rates. At least at the turning points of the cycle, the relationship between changes in unemployment and changes in wage rates might thus be positive. However, as long as the duration of the cycle exceeds the length of the lag with which unemployment reacts to vacancies, the relationship between changes in unemployment and changes in wage rates over the cycle would be negative.

Phillips, Lipsey, and most subsequent analysts have allowed for the possible influence on wages of conditions in the product market as well as conditions in the labor market. The standard measure of these product market conditions has been the change in the consumer price index; their influence on wages has been referred to as cost-push inflation. Phillips assumed that wages would be affected by an increase in consumer prices only to the extent that such an increase would in fact reduce real wages.12 Lipsey contested this assumption and suggested that price increases would affect wage bargains even if there were no threat of a reduction in real wages.

The emphasis on real wages recurs in studies that stress the importance of expectations. In these studies, unemployment is dismissed as a factor affecting wage movements; the Phillips curve is considered a highly unstable relationship, the position of which is determined by the expected rate of price increases.13 Inasmuch as expectations depend on past wage movements, current wage increases should be comparatively low in cyclical upswings and high in downswings. The expectations hypothesis therefore implies clockwise rotations of the time path of actual combinations of unemployment and wage changes around the fitted curve—the exact opposite of the counterclockwise loops attributed to the lagged response of unemployment to vacancies. The expectations hypothesis altogether denies the existence of a long-run trade-off between unemployment and wage stability.14

Other critics accept the Phillips relationship between wages and unemployment but claim that it is too weak to explain wage movements adequately because it ignores other important wage determinants, such as profits, productivity, union activity, and wage leadership. The relationship between wages and profits is not immediately evident. While a rise in profits would enable employers to pay higher wages, they might not be willing to pay them if the rise in profits were thought to be of only short duration.15 Empirical evidence on the importance of profits as a determinant of wage movements has been inconclusive. Zaidi [46] suggested that profits were an important determinant of wage movements in Canada, while Reuber [37] arrived at the exactly opposite conclusion. Bhatia [4 and 5], Eckstein and Wilson [9], Eckstein [10], and Perry [31 and 32] agree that profits affect wage movements in the United States but disagree on their explanatory power. Bhatia discards unemployment and prices altogether and regards the level of, and change in, profits as the main determinants of changes in wage rates. Perry considers the importance of these latter variables as marginal. Vroman [45] suggests that their importance declines even further when an adjusted rate of unemployment is substituted for the unadjusted rate. Insofar as rising profits reflect a rise in productivity, the relationship between wages and productivity would also be uncertain.16 In any event, Kuh’s [24] findings from an analysis of wage behavior by sector that wage movements are explained better by productivity than by unemployment are not supported by statistical evidence when adjusted (Vroman [45]) rather than unadjusted unemployment data are used.

Attempts to capture the relationship between union activity and wage movements can be found in Pierson [36] and Hines [17], and the influence of wage leadership is examined in Eckstein and Wilson [9]. While Pierson considers union behavior an important determinant of wages in addition to the rate of unemployment, Hines denies the existence of a trade-off between unemployment and changes in wage rates and explains wage movements in terms of union behavior alone. Eckstein and Wilson find that wage patterns set by key industries are, in addition to the rate of unemployment, significant determinants of changes in the wages of non-key industries.

The problem of assessing the precise relationship between wages, on the one hand, and profits, productivity, and union activity, on the other hand, is aggravated by multicollinearity.17 In models of wage determination that include profits or growth in productivity as explanatory variables in addition to the rate of unemployment, intercorrelation among the explanatory variables has been found.18 Similarly, if unions are likely to raise their wage demands when profits, productivity, and consumer prices are rising, collinearity would exist in any equation that contained union activity in addition to any of these three variables. Since intercorrelation among explanatory variables reduces the reliability of their coefficients, various authors have suggested dropping such variables as profits and productivity from the estimating equation and returning to the original Phillips-Lipsey wage model.19 If union activity is as closely related to increases in the cost of living as has been asserted,20 growing union strength would be reflected at least in part in a rise in the coefficient of the percentage change in the consumer price index.

While it appears that the hypothesis relating changes in wages to the unemployment rate has not been invalidated by alternative explanations of wage behavior, it has been qualified in some respects. Evidence suggests that the basic relationship between wage changes and unemployment is not very stable over time. This lack of stability may be due to the variable impact on wages of the change in the rate of unemployment, inflationary expectations, changes in the distribution of unemployment, the discretionary behavior of labor and management, and other supply rigidities. Since reduction in market imperfections and dispersion of demand pressure are assumed to improve the trade-off between a change in wage rates and unemployment, governments pursue policies toward these ends. The observed relationship between unemployment and wages may therefore also be modified by government action.

Unemployment and price movements

Recent empirical analysis has extended the hypothesis relating changes in wages to the unemployment rate beyond the labor market and has examined the relationship between unemployment and inflation in general.21 This extension must hinge on either of two assumptions: (1) that variations in the demand for labor result from variations in aggregate demand in the economy; or (2) that changes in the general price level are determined by changes in wages. While changes in wage costs and demand pressure are not likely to be independent of one another, the fact that either assumption, if confirmed, would be a sufficient condition for a trade-off between inflation and unemployment can be seen when the following is considered. If, on the one hand, unemployment were an efficient measure of aggregate demand pressure and if changes in prices were, at least in the short run, determined mainly by factors other than changes in wages, an estimated relationship between unemployment and inflation would measure the effect on the general price level of demand pressure alone. If, on the other hand, prices were determined mainly by wage costs and if unemployment were a poor measure of aggregate demand pressure, the observed unemployment/inflation schedule would reflect the dependence of changes in prices on changes in wages. Either wage costs or demand pressure could therefore predominate in the determination of price movements.

Which of the two in fact predominates cannot be inferred from a curve fitted to unemployment and price data. However, a separate indication of the response of unemployment to aggregate demand pressure could be obtained, for example, by comparing movements in the unemployment rate to deviations of actual gross national product (GNP) from its trend value. With respect to the price/wage relationship, it has generally been assumed that price changes reflect some constant markup over changes in unit labor costs.22 The assumption that prices are determined by wage costs, which implies that nonwage components of prices move in proportion to wage components, is contested by Brechling [7], who argues that the short-run relationship between aggregate excess demand and the nonwage markup is negative. Since the relationship between aggregate excess demand and the wage markup is assumed to be positive, Brechling’s argument implies that wage and nonwage components of prices may move in opposite directions, at least in the short run. The transition from changes in wages, which may be inferred from the Phillips curve, to changes in prices may therefore involve some difficulty.

This study does not try to ascertain whether wage costs or demand pressure predominates in the determination of price changes. It merely attempts to examine for a number of industrial countries the extent to which changes in the general price level can be explained in terms of factors of the type that enter into the determination of wage movements.

II. Empirical Analysis

The model of price determination

The empirical investigation focuses first on the basic relationship between the rate of change in the general price level and the level of unemployment. Subsequently, an attempt is made to trace any impact on price movements that might be attributed to the speed at which demand pressure changes by specifying an estimating equation that contains both the level and the percentage change of unemployment. In two additional steps, past movements in the general price level and import price developments are included in the equation. This final form applies Phillips’s hypothesis, as modified by Lipsey, to the determination of price changes in an open economy. However, in two important respects the price equation used in this study does not correspond to that hypothesis. First, any impact that past price movements may have on current price movements cannot be attributed solely to the adjustment of wages, and thus prices, to an increase in the cost of living. Inasmuch as past price changes reflect the influence of past levels of, and changes in, demand pressure, the dependence of current on past price movements implies a distributed lag in the response of prices to demand pressure. Alternatively, the dependence of current on past price movements may reflect the influence of inflationary expectations. Thus, while the fit of the price equation may improve when the lagged dependent variable is included as an additional explanatory variable, this relation is subject to various interpretations.

The second respect relates to import prices. Phillips and Lipsey could not have used both import and consumer prices in their wage equations, since prices of imported consumer goods are contained in the consumer price index. The specification of their model precluded the possibility of isolating the influences on wage changes of changes in consumer prices and import prices; the inclusion of both would, in a sense, have measured the same effect twice. This problem has been, at least in part, avoided in the present model in which percentage changes in both the GNP deflator and import prices are included as explanatory variables, since, by definition, the GNP deflator does not contain import prices of final goods. Similarly, it is statistically unobjectionable to regress current movements in the GNP deflator on current movements in import prices: unlike regressing changes in the GNP deflator on changes in prices of agricultural products, this would not amount to explaining domestic price movements by themselves. Rather, an increase in import prices may lead to rising domestic prices owing to diminishing competition from abroad, substitution of domestic for imported products, rising cost of imported raw materials and intermediate products, and rising wage demands.

The present model of price determination tests the reaction of prices to demand pressure under the assumption of various discrete lags and on alternative hypotheses that the price/unemployment relationship is either linear or nonlinear. The model is applied to price developments in the following 12 industrial countries: Austria, Belgium, Canada, Denmark, France, Germany, Italy, Japan, the Netherlands, Sweden, the United Kingdom, and the United States. In general, the lack of comparable unemployment data among countries that ordinarily follow different procedures in the compilation of these data limits the usefulness of intercountry comparisons of the relationship between prices and unemployment. However, for 7 of the 12 countries—Canada, France, Germany, Japan, Sweden, the United Kingdom, and the United States—comparable unemployment data could be obtained for the period 1955-68.23 The present analysis was, mainly for this reason, limited to that period (for Germany and the United States the period of fit includes 1969). The early 1950s were excluded also because of autonomous influences in those years both on price movements and in the labor market for a number of European countries. In the wake of World War II, a number of European countries recorded substantial reserves of unemployed labor, which were absorbed only gradually as the economies of these countries recovered. Similarly, wartime and postwar controls of wages and prices in many instances were not removed until after the early 1950s.

The analysis is based on annual observations because reliable quarterly data are not available throughout. This leaves a maximum of 14 observations, and hence little opportunity, to establish whether the price/unemployment relationship in the various countries is stable when tested over different subperiods. Yet, in addition to the period 1955-68, the performance of the model was tested for the period 1959-68(69) to establish whether and how the explanatory power of factors affecting price movements had changed in the 1960s. Some further conclusions regarding the properties of stability can be drawn from the effect that the inclusion of additional explanatory variables has on the coefficient of the unemployment variable.

For the 12 countries and the two time periods, the year-to-year percentage change in the GNP deflator (GP) 24 was regressed on the concurrent level of unemployment (UN). The level of unemployment is centered at midperiod, and the year-to-year price change is centered at the end of period. Thus, in testing the lagless relation, the price change from one year to another is most appropriately associated with the average level of unemployment over the same two years (½ (UNt + UNt-1)).25 Alternatively, it was assumed that the level of unemployment would affect price movements with a lag of one-half year, so that GṖ was regressed on UNt-1. These equations represent the basic relationship between prices and the level of demand pressure. The percentage change in the rate of unemployment (U˙) was introduced to allow for the effect on prices of the speed at which aggregate demand expands and contracts; it was included in the equations in the alternative forms U˙tandU˙t1. Since theory suggests a nonlinear Phillips curve, GṖ was also regressed on the reciprocal values of UN, which is one possible way of capturing the nonlinearity. The equations were then expanded to allow for the dependence of current on past price movements, assuming a lag of one year. Past price movements were therefore included as an additional explanatory variable in the form GṖt-1. Subsequently, the proposition was tested that in an open economy changes in import prices might explain some of the residual variance in GṖ not explained by unemployment and past changes in prices. The percentage change in import prices (MṖ) 26 was assumed to affect GṖ concurrently (MṖt) or, alternatively, with a lag of one year (MṖt-1).

Statistical results

Empirical performance of the model was tested by multiple regression analysis. The results recorded in Tables 2, 3, and 4 show that the rate of unemployment, whose coefficient has the hypothesized negative sign, is a statistically significant determinant of changes in the general price level in 11 of 12 countries. Austria is the only country for which unemployment could not explain any of the variance in price movements. For France and Japan, the relationship between prices and the rate of unemployment is significant only when tested over the periods including the mid-1950s, while for Germany and Sweden a significant relationship could be found to exist only during the 1960s.27 The rate of unemployment is generally significant in both linear and nonlinear form, but the part of the variance of price movements that is explained by the equation tends to be greater when the nonlinear form is used. The significant nonlinear relationship indicates that the unemployment/ inflation schedule has the characteristic shape of a Phillips curve, but the fact that the linear relationship is also significant suggests that the curvature of the schedule is unlikely to be very strong. The charted relationship between the rate of inflation and the concurrent level of unemployment supports such an assumption—see Chart 3. This chart facilitates some comparison of slope and position of the unemployment/inflation schedules among those countries for which consistent unemployment data are available. It appears that during the 1960s Canada and the United States recorded a more unfavorable trade-off between inflation and unemployment than the five other countries with consistent unemployment data—France, Germany, Japan, Sweden, and the United Kingdom. The concurrent level of unemployment (½ (UNt + UNt-1)) was chosen for the diagrams because, in the single-variable equations, it performed in general somewhat better than UNt-1. However, when the single-variable model was expanded to include the percentage change in the unemployment rate, lagged unemployment performed better.

Table 1.

Selected Industrial Countries: Steady-State Relation Between Inflation and Unemployment1

(In per cent)

article image

For countries marked with an asterisk, unemployment data are comparable. For the remaining countries, such comparability is not assured.

For each country, the steady-state relation between inflation and unemployment corresponds, approximately, to the range of unemployment observed during the periods over which the regression equations, recorded in Chart 1, were fitted and from which the steady-state relations were computed.

For Germany, the equation was used that was estimated over the period 1959-69 (Table 3) and contains as explanatory variables UNt11, and

article image
t rather than UNt-1 and
article image
t as does the equation in Chart 1.

For Japan and the United Kingdom, the equations used are those without the import price variable. However, with MṖ = 0, results from the equations with the import price variable would be very close to those recorded in Table 1. Results for the United Kingdom were derived from the equation relating to the period 1955-67 (Table 3).

For the United States, the equation covering the shorter period, 1960-69, was selected. While derivation of the steady-state relation from the equation for the period 1956-69 would, in comparison with the results shown in Table 1, yield lower values, the ranking of the countries with comparable unemployment data would not change, irrespective of which equation is used. On the basis of the equation for the period 1956-69, a steady unemployment level of 4 per cent implies a rate of inflation of 3.8 per cent. This estimate is in line with corresponding results contained in three econometric models that relate to a similar timespan. See Hymans [20]. (The three models are the model of the Office of Business Economics, the FRB-MIT-Penn model, and the DHL-III model of the University of Michigan.)

Table 2.

Selected Industrial Countries: Relationship Between the Percentage Change in the GNP Deflator and the Concurrent Level of Unemployment (Regression Results) 1

article image

Symbols used: UN = average annual rate of unemployment; R¯2 = coefficient of determination adjusted for degrees of freedom; D-W = Durbin-Watson statistic; SE¯ = standard error of the estimate adjusted for degrees of freedom; figures in parentheses are t-ratios. Regression coefficients that are marked with one and two asterisks are statistically significant at the 5 per cent and 10 per cent levels, respectively.

For the United Kingdom, the single-variable equation was also estimated over periods excluding 1968, since price movements in that year may have been influenced considerably by the preceding devaluation of sterling. In part, this influence is captured by the inclusion of import prices in the multivariable model (Table 4).

Table 3.

Selected Industrial Countries: Relationship Between the Percentage Change in the GNP Deflator and the Reciprocal of the Concurrent Level of Unemployment (Regression Results) 1

article image

Symbols used: UN = average annual rate of unemployment; R¯2 = coefficient of determination adjusted for degrees of freedom; D-W = Durbin-Watson statistic; SE¯ = standard error of the estimate adjusted for degrees of freedom; figures in parentheses are t-ratios. Regression coefficients that are marked with one and two asterisks are statistically significant at the 5 per cent and 10 per cent levels, respectively.

In the equation over the longer period, statistical significance of the coefficient of the unemployment variable requires inclusion of 1969 in the timespan covered.

For the United Kingdom, the single-variable equation was also estimated over periods excluding 1968, since price movements in that year may have been influenced considerably by the preceding devaluation of sterling. In part, this influence is captured by the inclusion of import prices in the multivariable model (Table 4).

Table 4.

Selected Industrial Countries: Expansion of the Model to Include the Percentage Change in the Rate of Unemployment, the Lagged Dependent Variable, and the Percentage Change in Import Prices (Selected Regression Results) 1

article image
article image
article image
article image
article image
article image
article image

Symbols used: UN = average annual rate of unemployment;

article image
= percentage change in the rate of unemployment divided by 100; GṖ = percentage change in the GNP deflator; MṖ = percentage change in import prices; R¯2 = coefficient of determination adjusted for degrees of freedom; D-W = Durbin-Watson statistic; SE¯ = standard error of the estimate adjusted for degrees of freedom; figures in parentheses are t-ratios. Regression coefficients that are marked with one and two asterisks are statistically significant at the 5 per cent and 10 per cent levels, respectively.

For Denmark, France, and Sweden, none of the additional independent variables contributed to the explanation of price movements, and the expanded model did in no instance perform better than the single-variable model (Tables 2 and 3). Equations that include the percentage change in import prices as an explanatory variable are recorded only for countries where that variable appeared to be a statistically significant determinant of movements in the general price level, namely, Austria, Belgium, Japan, the Netherlands, and the United Kingdom.

In the equation for the longer period, statistical significance of the coefficient of UNt11 requires inclusion of 1969 in the timespan covered.

In the equations for the shorter period, 1959-68, the contribution of unemployment, import price, and lagged dependent variables proved consistently insignificant; results from those equations are therefore not recorded.