The effect of changes in the distribution of employment

The official estimates of changes in hourly earnings are based on data collected in April and October each year. Owing to the seasonality of wage bargaining and employment, the two series are not compatible. The analysis in this paper relates to annual changes from October to October.²

In trying to assess the influence of employment conditions upon the rate of wage inflation, one is essentially concerned with the rate at which pay in particular employments is increased by wage-fixing decisions, that is, either centralized wage bargaining or decisions at the enterprise or plant level. One is not concerned with changes in average hourly earnings that arise independently of wage-fixing decisions, such as those that are due to shifts in the composition of employment. There are wide differences in labor productivity (i.e., in value added per worker-hour) in different industries and associated differences in average hourly pay in different industries. Thus, average hourly earnings may increase (or decline) with changes in the distribution of employment, reflecting changes in the composition of aggregate demand; but if hourly pay for the same work remains the same in each industry, one can hardly regard such changes as “inflationary” or “deflationary.” ³

The official estimates of changes in hourly earnings are derived from, current-weighted averages, reflecting the estimated level of employment by industry at the time of each biannual survey. The observed change in average hourly earnings (including the effect of changes in the distribution of employment) is, therefore, liable to overstate, or understate, the rate at which hourly pay is increasing because of wage-fixing decisions, whenever the industrial distribution of employment changes markedly. However, the effect of changes in the distribution of employment upon average earnings is much less important for men and women treated separately than for all workers taken together. This is so because industries typically employ predominantly men or women; there is a large differential between the standard rates for men and women in each industry, and standard rates for each sex are generally higher in industries employing predominantly men. Thus, important changes in the industrial distribution of employment lead to changes in the proportion of hours worked by men and women, and consequent changes in average hourly earnings owing to the difference between their rates of pay. The effect of changes in the industrial distribution of employment upon either men’s or women’s hourly earnings is limited by the relatively small share of their respective employment accounted for by low-pay (high-pay) industries.

The rise in hourly earnings is also liable to be influenced by changes in the proportion of young persons in employment, as youths and boys under 21 and girls under 18 are paid at lower rates than adults. Demographic changes in the age structure of the population and changes in the proportion of young persons receiving higher education or serving in the armed forces can therefore have a significant impact upon the rate of increase in average hourly earnings.

Tentative estimates suggest that the rise in average hourly earnings (including overtime) as published was moderated between 1959 and 1963 by a decline in the share of hours worked by men and a rising share of hours worked by juveniles, as the labor force began to be influenced by the sharp rise in the birth rate immediately after World War II. Between 1963 and 1968, on the other hand, the rise in average hourly earnings was strengthened by an increase in men’s share of hours covered; this was apparently associated first with a decline in the proportion of hours worked by women full-time workers up to 1966 and thereafter with a sharp fall in the employment of juveniles.⁴ Because the increasing share of hours worked by men was associated with a rising proportion of men’s hours paid at overtime rates, it had a disproportionately large effect on average hourly earnings.

An alternative index of changes in average hourly earnings

As the (estimated) employment in each industry at the time of the earnings survey is not published, it is not possible to distinguish the precise effect of changes in the industrial distribution of employment upon the rate of increase in hourly earnings. However, average hourly earnings and the hours worked by men, women, youths and boys, and girls are given, and so it is possible to construct an index of average earnings (including or excluding the effect of overtime) in which each of the four groups has a fixed weight. Such an index provides a measure of the change in hourly earnings of an “average worker,” which takes account of the fact that men work more than 70 per cent of all hours excluding overtime, women full-time workers under 20 per cent, and youths and boys most of the remainder. The changes in the average hourly earnings of each demographic group are influenced by the distribution of employment by industries, but by holding the weights of the groups constant the effects of demographic factors, and of shifts in the industrial composition of employment resulting in changes in the proportions of hours worked by men, women, and juveniles, are excluded. For reasons explained above, such an index in effect abstracts from a large part of the effect of changes in the industrial distribution of employment upon average hourly earnings;⁵ it therefore provides a considerably better measure of the increase in hourly earnings owing to wage-fixing decisions than the published figures.

Correction for overtime

Although an increase in overtime can be considered the result of a wage decision by management, it is customary to correct earnings for overtime because the consequent rise in earnings can be expected to be reversed when overtime is reduced and because it is desirable to exclude the effect of overtime when trying to measure the importance of “wage drift.” Unfortunately, the published figures of changes in average hourly earnings excluding the effect of overtime are distorted by the method of correction, which underestimates the proportion of all hours paid at overtime rates and does not adequately allow for the fact that almost all overtime is worked by men.⁶

A realistic view of wage developments in the United Kingdom is also hampered by the treatment of substantial overtime work by men as though it were a temporary and exceptional phenomenon, when in fact it has been regular practice in some industries ever since World War II or the immediate postwar period. It has to be recognized that in these circumstances overtime earnings will come to be regarded as part of regular average hourly earnings. Throughout the 1940s and 1950s weekly take-home pay of a large proportion of men was regularly augmented by several hours’ overtime. For many, the “standard” workweek had in effect been superseded by a “guaranteed” workweek including overtime; and from the individual’s point of view, average hourly pay had come to include a certain proportion of overtime premium. This situation tended to focus the unions’ attention on reductions in standard working hours as a means of maintaining the amount of overtime work and protecting average hourly earnings against the consequences of reductions in actual hours of work.

Actual hours of work by men were no closer, on average, to the normal workweek in the late 1960s than during the 1950s—in fact, the amount of overtime worked by men had actually increased (see Table 1 and Chart 1). It therefore seems appropriate to consider the customary workweek for men as including a certain number of overtime hours, over and above the standard workweek, and to consolidate average hourly earnings for their customary workweek on this basis. The customary workweek for men will be taken to include an average of four hours’ overtime, so that average hourly earnings excluding the effect of “overtime proper” will include four hours of work at the premium (one and a half times the standard) rate.⁷ No similar problem arises for the other groups, whose actual hours of work were usually below or in line with their normal workweek. Actual hours of work for youths and boys were close to the normal week during the 1950s, but overtime has averaged one to one and a half hours a week since the standard workweek was reduced, mainly in 1960 and 1965.

Table 1.

United Kingdom: Difference Between Average Actual Hours Worked and Average Normal Weekly Hours of Full-Time Workers, 1953–70¹

(In hours)

Sources: Figures in first column are derived from data on hours published in Statistics on Incomes, Prices, Employment & Production, and in the Ministry of Labour Gazette, for years prior to 1960. Figures in other columns are derived from data on actual hours worked published in the biannual earnings inquiries, Ministry of Labour Gazette (now entitled Department of Employment Gazette); from figures of average normal weekly hours calculated from the indices of normal weekly hours for men, women, and juveniles (January 1956 = 100) published in Statistics on Incomes, Prices, Employment & Production; and from data on normal weekly hours in January 1956, and October of earlier years, published in the Ministry of Labour Gazette.

¹October figures.

²The series on actual hours of work includes two part-time workers as equivalent to one full-time worker, causing overstatement of average hours worked by full-time workers.

Table 1.

United Kingdom: Difference Between Average Actual Hours Worked and Average Normal Weekly Hours of Full-Time Workers, 1953–70¹

(In hours)

Year	All Full-Time Workers	Men	Youths	Women	Girls
1953	1.7	4.1	0.2	–3.22	–2.0
1954	2.1	4.2	0.3	–3.32	–2.0
1955	2.3	4.5	0.2	–3.42	–2.0
1956	2.0	4.1	—	–3.9	–2.2
1957	1.8	3.9	—	–4.1	–2.5
1958	1.5	3.4	0.1	–4.0	–2.3
1959	2.3	4.3	0.5	–3.6	–2.1
1960	2.8	4.9	0.8	–3.7	–2.0
1961	2.9	5.0	1.0	–3.4	–2.2
1962	2.7	4.8	0.9	–3.5	–2.0
1963	3.2	5.4	1.2	–3.2	–2.0
1964	3.5	5.7	1.2	–2.8	–1.9
1965	3.9	6.1	1.7	–3.0	–1.5
1966	3.5	5.4	1.5	–2.7	–2.0
1967	3.8	5.9	1.4	–2.9	–1.8
1968	4.1	6.1	1.4	–2.8	–1.7
1969	4.2	6.3	1.3	–2.8	–1.9
1970	…	5.6	1.0	–2.9	–2.4

Sources: Figures in first column are derived from data on hours published in Statistics on Incomes, Prices, Employment & Production, and in the Ministry of Labour Gazette, for years prior to 1960. Figures in other columns are derived from data on actual hours worked published in the biannual earnings inquiries, Ministry of Labour Gazette (now entitled Department of Employment Gazette); from figures of average normal weekly hours calculated from the indices of normal weekly hours for men, women, and juveniles (January 1956 = 100) published in Statistics on Incomes, Prices, Employment & Production; and from data on normal weekly hours in January 1956, and October of earlier years, published in the Ministry of Labour Gazette.

¹October figures.

²The series on actual hours of work includes two part-time workers as equivalent to one full-time worker, causing overstatement of average hours worked by full-time workers.

View Table

Table 1.

United Kingdom: Difference Between Average Actual Hours Worked and Average Normal Weekly Hours of Full-Time Workers, 1953–70¹

(In hours)

Year	All Full-Time Workers	Men	Youths	Women	Girls
1953	1.7	4.1	0.2	–3.22	–2.0
1954	2.1	4.2	0.3	–3.32	–2.0
1955	2.3	4.5	0.2	–3.42	–2.0
1956	2.0	4.1	—	–3.9	–2.2
1957	1.8	3.9	—	–4.1	–2.5
1958	1.5	3.4	0.1	–4.0	–2.3
1959	2.3	4.3	0.5	–3.6	–2.1
1960	2.8	4.9	0.8	–3.7	–2.0
1961	2.9	5.0	1.0	–3.4	–2.2
1962	2.7	4.8	0.9	–3.5	–2.0
1963	3.2	5.4	1.2	–3.2	–2.0
1964	3.5	5.7	1.2	–2.8	–1.9
1965	3.9	6.1	1.7	–3.0	–1.5
1966	3.5	5.4	1.5	–2.7	–2.0
1967	3.8	5.9	1.4	–2.9	–1.8
1968	4.1	6.1	1.4	–2.8	–1.7
1969	4.2	6.3	1.3	–2.8	–1.9
1970	…	5.6	1.0	–2.9	–2.4

Sources: Figures in first column are derived from data on hours published in Statistics on Incomes, Prices, Employment & Production, and in the Ministry of Labour Gazette, for years prior to 1960. Figures in other columns are derived from data on actual hours worked published in the biannual earnings inquiries, Ministry of Labour Gazette (now entitled Department of Employment Gazette); from figures of average normal weekly hours calculated from the indices of normal weekly hours for men, women, and juveniles (January 1956 = 100) published in Statistics on Incomes, Prices, Employment & Production; and from data on normal weekly hours in January 1956, and October of earlier years, published in the Ministry of Labour Gazette.

¹October figures.

²The series on actual hours of work includes two part-time workers as equivalent to one full-time worker, causing overstatement of average hours worked by full-time workers.

Chart 1.

United Kingdom: Average Actual Hours Worked Weekly and Average Normal Weekly Hours of Men, Women, and Youths and Boys, 1952–70

Hours per week

View Expanded

Table 2 gives the revised series of increases in average hourly earnings—calculated by excluding the effect of all overtime or the effect of more narrowly defined overtime and assigning constant weights to the average earnings of men, women, youths, and girls—and illustrates how these figures differ from the published data.⁸

Table 2.

United Kingdom: Increases in Average Hourly Earnings, Excluding the Effect of Overtime, 1954–70 ¹

(In per cent)

Source: Published series, Department of Employment Gazette (February 1971), p. 169, October-to-October figures. Other estimates are described in text.

¹The published data on women’s earnings before 1956 do not distinguish full-time from part-time women workers.

²Average hourly earnings, excluding the effect of overtime, of men, women (full-time workers), youths, and girls are combined with weights of 71.0, 19.7, 7.4, and 1.9, respectively.

³See text for definition of “overtime proper.”

⁴Difference measures the combined effect of the more exact measure of the effect of overtime (i.e., for each demographic group separately); of abstracting from changes in groups’ shares in hours worked; and of adopting the narrower definition of overtime. (The effect of this change alone is shown by the difference between column 3 and column 2.)

Table 2.

United Kingdom: Increases in Average Hourly Earnings, Excluding the Effect of Overtime, 1954–70 ¹

(In per cent)

		Alternative Estimates
	Published Series	Excluding the effect of all overtime2	Excluding only the effect of “overtime proper”3	Difference 4 (col. 3–col. 1)
Year	(1)	(2)	(3)	(4)
1956	8.2	8.1	8.2	—
1957	6.6	6.5	6.6	—
1958	3.4	3.5	3.4	—
1959	2.9	3.2	3.2	0.3
1960	7.3	7.7	8.0	0.7
1961	6.9	6.9	6.7	–0.2
1962	4.4	4.2	4.1	–0.3
1963	3.6	3.8	3.9	0.3
1964	8.1	7.8	8.0	–0.1
1965	9.5	9.2	9.2	–0.3
1966	6.5	7.0	7.0	0.5
1967	5.0	4.1	4.1	–0.9
1968	7.0	7.0	6.9	–0.1
1969	8.0	7.6	7.6	–0.3
1970	16.2	15.8	15.9	–0.3

Source: Published series, Department of Employment Gazette (February 1971), p. 169, October-to-October figures. Other estimates are described in text.

¹The published data on women’s earnings before 1956 do not distinguish full-time from part-time women workers.

²Average hourly earnings, excluding the effect of overtime, of men, women (full-time workers), youths, and girls are combined with weights of 71.0, 19.7, 7.4, and 1.9, respectively.

³See text for definition of “overtime proper.”

⁴Difference measures the combined effect of the more exact measure of the effect of overtime (i.e., for each demographic group separately); of abstracting from changes in groups’ shares in hours worked; and of adopting the narrower definition of overtime. (The effect of this change alone is shown by the difference between column 3 and column 2.)

View Table

Table 2.

United Kingdom: Increases in Average Hourly Earnings, Excluding the Effect of Overtime, 1954–70 ¹

(In per cent)

		Alternative Estimates
	Published Series	Excluding the effect of all overtime2	Excluding only the effect of “overtime proper”3	Difference 4 (col. 3–col. 1)
Year	(1)	(2)	(3)	(4)
1956	8.2	8.1	8.2	—
1957	6.6	6.5	6.6	—
1958	3.4	3.5	3.4	—
1959	2.9	3.2	3.2	0.3
1960	7.3	7.7	8.0	0.7
1961	6.9	6.9	6.7	–0.2
1962	4.4	4.2	4.1	–0.3
1963	3.6	3.8	3.9	0.3
1964	8.1	7.8	8.0	–0.1
1965	9.5	9.2	9.2	–0.3
1966	6.5	7.0	7.0	0.5
1967	5.0	4.1	4.1	–0.9
1968	7.0	7.0	6.9	–0.1
1969	8.0	7.6	7.6	–0.3
1970	16.2	15.8	15.9	–0.3

Source: Published series, Department of Employment Gazette (February 1971), p. 169, October-to-October figures. Other estimates are described in text.

¹The published data on women’s earnings before 1956 do not distinguish full-time from part-time women workers.

²Average hourly earnings, excluding the effect of overtime, of men, women (full-time workers), youths, and girls are combined with weights of 71.0, 19.7, 7.4, and 1.9, respectively.

³See text for definition of “overtime proper.”

⁴Difference measures the combined effect of the more exact measure of the effect of overtime (i.e., for each demographic group separately); of abstracting from changes in groups’ shares in hours worked; and of adopting the narrower definition of overtime. (The effect of this change alone is shown by the difference between column 3 and column 2.)

An alternative measure of wage drift

The faster rise in earnings than in rates is usually taken to reflect wage drift, i.e., the willingness of individual employers to grant wage concessions at the plant, to upgrade workers, or to pay higher than prevailing rates in times of strong demand.⁹ The difference between the change in hourly earnings as calculated here and the change in the index of hourly wage rates (which is calculated with fixed weights) provides a more satisfactory measure of wage drift than the published figures. In the latter, the effect of wage drift is obscured by the influence of changes in the industrial and demographic composition of employment upon hourly earnings and also by the errors introduced by the method of correction of earnings for overtime. The revised series for wage drift, given in Table 3, shows the difference between the percentage change in average hourly earnings, excluding narrowly defined overtime, and the percentage change in the index of hourly wage rates, adjusted to take account of overtime regularly worked by men and to include the effect of agreed reductions in “normal” hours in raising average hourly pay during the customary workweek.¹⁰

Table 3.

United Kingdom: Estimated Changes in Hourly Earnings and in Hourly Wage Rates and Revised Estimates of Wage Drift, 1956–70

(In per cent)

Source: Published series from Department of Employment Gazette (February 1971). Estimated series described in text.

¹Assuming fixed proportions of hours worked by men, youths and boys, women (full-time workers), and girls. See text for definition of overtime proper.

²The adjustment for overtime regularly worked by men increases the weight given to changes in men’s wage rates and has the effect of reducing the average rise in wage rates when men’s rates are rising less than those for women and juveniles, and vice versa; marked reductions in normal hours slightly increase the proportion of customary working hours paid at overtime rates. The combined effects are shown in the difference between column 3 and column 2.

Table 3.

United Kingdom: Estimated Changes in Hourly Earnings and in Hourly Wage Rates and Revised Estimates of Wage Drift, 1956–70

(In per cent)

	Estimated Change in Average Hourly Earnings, Excluding Overtime Proper 1 (1)	Change in Average Hourly Wage Rates		Wage Drift, Revised (col. 1 - col. 3) (4)	Wage Drift, Published Series (5)
Year		Published series (2)	Adjusted series2 (3)			Difference (col. 4—col. 5) (6)
1956	8.2	7.6	7.7	0.5	0.6	–0.1
1957	6.6	5.6	5.7	0.9	1.0	–0.1
1958	3.4	3.7	3.8	–0.4	–0.3	–0.1
1959	3.2	1.4	1.4	1.8	1.5	0.3
1960	8.0	5.5	5.8	2.2	1.8	0.4
1961	6.7	6.4	6.4	0.3	0.5	–0.2
1962	4.1	4.2	4.3	–0.2	0.2	–0.4
1963	3.9	2.3	2.1	1.8	1.3	0.5
1964	8.0	5.7	5.8	2.2	2.4	–0.2
1965	9.2	7.3	7.5	1.7	2.2	–0.5
1966	7.0	5.6	5.8	1.2	0.9	0.3
1967	4.1	5.3	5.5	–1.4	–0.3	–1.1
1968	6.9	6.7	6.6	0.3	0.3	—
1969	7.6	5.4	5.3	2.3	2.5	–0.2
1970	15.9	12.2	11.8	4.1	4.0	0.1

Source: Published series from Department of Employment Gazette (February 1971). Estimated series described in text.

¹Assuming fixed proportions of hours worked by men, youths and boys, women (full-time workers), and girls. See text for definition of overtime proper.

²The adjustment for overtime regularly worked by men increases the weight given to changes in men’s wage rates and has the effect of reducing the average rise in wage rates when men’s rates are rising less than those for women and juveniles, and vice versa; marked reductions in normal hours slightly increase the proportion of customary working hours paid at overtime rates. The combined effects are shown in the difference between column 3 and column 2.

View Table

Table 3.

United Kingdom: Estimated Changes in Hourly Earnings and in Hourly Wage Rates and Revised Estimates of Wage Drift, 1956–70

(In per cent)

	Estimated Change in Average Hourly Earnings, Excluding Overtime Proper 1 (1)	Change in Average Hourly Wage Rates		Wage Drift, Revised (col. 1 - col. 3) (4)	Wage Drift, Published Series (5)
Year		Published series (2)	Adjusted series2 (3)			Difference (col. 4—col. 5) (6)
1956	8.2	7.6	7.7	0.5	0.6	–0.1
1957	6.6	5.6	5.7	0.9	1.0	–0.1
1958	3.4	3.7	3.8	–0.4	–0.3	–0.1
1959	3.2	1.4	1.4	1.8	1.5	0.3
1960	8.0	5.5	5.8	2.2	1.8	0.4
1961	6.7	6.4	6.4	0.3	0.5	–0.2
1962	4.1	4.2	4.3	–0.2	0.2	–0.4
1963	3.9	2.3	2.1	1.8	1.3	0.5
1964	8.0	5.7	5.8	2.2	2.4	–0.2
1965	9.2	7.3	7.5	1.7	2.2	–0.5
1966	7.0	5.6	5.8	1.2	0.9	0.3
1967	4.1	5.3	5.5	–1.4	–0.3	–1.1
1968	6.9	6.7	6.6	0.3	0.3	—
1969	7.6	5.4	5.3	2.3	2.5	–0.2
1970	15.9	12.2	11.8	4.1	4.0	0.1

Source: Published series from Department of Employment Gazette (February 1971). Estimated series described in text.

¹Assuming fixed proportions of hours worked by men, youths and boys, women (full-time workers), and girls. See text for definition of overtime proper.

²The adjustment for overtime regularly worked by men increases the weight given to changes in men’s wage rates and has the effect of reducing the average rise in wage rates when men’s rates are rising less than those for women and juveniles, and vice versa; marked reductions in normal hours slightly increase the proportion of customary working hours paid at overtime rates. The combined effects are shown in the difference between column 3 and column 2.

The results given in Tables 2 and 3 are discussed in the following section.

The evidence of a shift in the Phillips curve

In order to see how the relation between increases in hourly wages and the level of unemployment has changed since the mid-1950s, it is convenient to plot the published and the estimated series of changes in earnings against the percentages of unemployment as shown in Charts 2 and 3. This presentation does not, of course, imply that other factors (such as changes in import prices or changes in the balance of indirect taxes and subsidies) have no influence upon the rise in money wages. Nor should the statistical relation between unemployment and the rise in wages be taken to mean that it is the level of unemployment per se that influences the rate at which wages rise; this question is discussed in Section IV.

Chart 2.

United Kingdom: Unemployment and Increases in Hourly Earnings Excluding Overtime, as Published and as Estimated, 1954–70

(In per cent)

View Expanded

Chart 3.

United Kingdom: Unemployment and Increases in Hourly Earnings Excluding Overtime, as Published and as Estimated, 1954–69

(In per cent)

View Expanded

The official series of changes in average hourly earnings excluding the effect of overtime seems to indicate a marked upward shift in the Phillips curve between the mid-1950s and the 1960s. As can be seen in the upper panel of Chart 2, later observations are generally above earlier observations at the same level of unemployment, apparently indicating a gradual tendency for wages to rise faster at any given level of unemployment.¹¹ (The two most striking exceptions to this tendency, the observations for 1961 and 1966, are years of wage restraint.) The upward shift is less evident in the estimated series shown in the lower panel of the chart. The association between the level of unemployment and the rate of wage increase appears closer, and the hypothesis that there has been continuous upward movement over time of wage increases at a given level of unemployment appears less plausible than on the basis of the published figures. The increase in earnings at much the same level of unemployment is now lower in 1961 than in 1954, the same in 1964 as in 1960, and little higher in 1967 than in 1963.

While a scatter diagram provides a good illustration of the relation between two variables over a number of years, the timing and context of changes in the relationship are more apparent from a simple diagram comparing the curves of the two time series, as in Chart 3.¹² One fact that is immediately obvious in this presentation, but that can be easily overlooked in a scatter diagram, is that one is comparing series that oscillate sharply over short cycles of similar duration. There is a tendency for unemployment in the cyclical troughs to increase over time. The existence of relatively heavy unemployment for a third and fourth successive year in 1969 and 1970 has no postwar precedent.

The pattern in the lower panel of Chart 3 suggests that the rate of wage increase is related to the stage of the cycle rather than to the amount of unemployment prevailing at that point. When unemployment decreases in the cyclical upswing, wage increases are high (about 8 per cent); when unemployment increases sharply after a boom, wage increases are low (3½ to 4 per cent), and they remain low in the following year if unemployment does not fall. But the current level of unemployment during the boom does not indicate whether wage increases will further intensify, continue high, or moderate. At the top of the cycle the rate of wage inflation usually moderates before there is a marked increase in unemployment.

Changes in wage rates and wage drift

The relation between the level of unemployment and changes in earnings in the course of the cycle is more easily understood when a distinction is drawn between the increase in negotiated wage rates and “statistical” wage drift, i.e., the difference between the percentage increase in hourly earnings and that in hourly rates. It then becomes apparent that, while the increases in wage rates tend to vary with the fluctuations in the level of unemployment in the course of the cycle, the changes in wage drift precede the changes in unemployment at the top and bottom of the cycle and occur while the increase in rates is still accelerating or decelerating (Chart 4). This difference explains some of the variation in the rate of increase in hourly earnings at similar levels of unemployment shown in the lower panel of Chart 2.

Chart 4.

United Kingdom: Unemployment and Increases in Average Hourly Wage Rate and in Wage Drift, 1953–70

(In per cent)

View Expanded

^* No revised estimate of wage drift is available before 1956.

Chart 5 illustrates the relation between the level of unemployment and the adjusted changes in wage rates. The regression line indicates the relation estimated for the period 1953–65, before the disturbances introduced by wage restraint in 1966 and the devaluation of sterling in 1967; WR = 11.1—4.22u, where WR is the percentage change in wage rates from October to October and u is the percentage of unemployment in June.

Chart 5.

United Kingdom: Unemployment and Actual and Expected Increases in Wage Rates, 1953–70

(In per cent)

View Expanded

Curve shows locus of expected change in average hourly wage rates, based on relation estimated over the period 1953–65, WR = 11.1 – 4.22u.

It seems reasonable to suppose that the tendency for earnings to rise in relation to rates will be stronger when vacancies are increasing, and vice versa. The strength of wage drift is related to the change in the number of unfilled vacancies in Chart 6. The line indicates the approximate relation between the change in the number of unfilled vacancies from October to October and the percentage increase in hourly earnings owing to wage drift, estimated for the period 1956–65; WD = 1.11 + 0.0131v, where v is the change in unfilled vacancies in thousands. This indicates that wage drift is positive unless the number of vacancies falls by nearly 100,000; when there is no change in vacancies, wage drift raises earnings by 1.11 per cent, and by 2.42 per cent when vacancies increase by 100,000. An increased tendency to wage drift appears to have had the effect of augmenting wage increases during the 1960s. Substantial wage drift of from 1.0–2.5 per cent was to be expected in every one of the seven years since 1962 except 1967,¹³ but in only two of the preceding seven years (1959 and 1960).

Chart 6.

United Kingdom: Changes in Vacancies and Actual and Expected Wage Drift, 1956–70

Wage drift

View Expanded

Curve shows locus of expected drift based on relation estimated for over the period 1956-65, WD = 1.11 + .0131v, where v is the change in vacancies in thousands. The vertical scale of this chart is twice that used in Charts 2–5.

The series of expected increases in earnings derived from the two formulas, plotted in Chart 7, provides an index of demand conditions in the labor market. Discrepancies between the expected and actual increases in earnings and between the expected and actual increases in wage rates, also shown in the chart, indicate (1) that wage increases were unusually low in relation to conditions in the labor market in 1953-55 and higher than usual in 1956; (2) that there was a marked deceleration of the actual rate of increase relative to the expected increase in the late 1950s culminating in a much-smaller-than-expected change in 1959; and (3) that there was an acceleration of the actual increase relative to the expected increase in the second half of the 1960s, with very-much-higher-than-expected increases after 1967. These tendencies appear to be related to the markedly different movement of prices of imported industrial products at these times, indicated by the dotted curve in Chart 7.

Chart 7.

United Kingdom: Expected Increase in Earnings and Change in Price of Imports (Manufactures); Discrepancies Between Expected and Actual Increases in Hourly Earnings and in Hourly Wage Rate, 1953–71

(In per cent)

View Expanded

^* Assuming “normal” productivity growth of 3½ per cent per annum. † Discrepancy on wage rates only.

The influence of changes in the price of imported industrial goods

It is usually assumed that, under given conditions in the labor market, pressure for wage increases will be stronger and larger increases will be secured when the cost of living is rising rapidly, and that a rise in import prices influences the rise in wages indirectly via its effect upon the cost of living, or the general level of prices.¹⁴ But one must suppose that a rise (or fall) in the import prices of competing goods directly influences the scale of wage increases that industrialists are prepared to grant. This would imply that changes in the price of imported manufactures have a stronger influence upon wage increases than their share in expenditure would suggest, and probably a swifter impact than is usually supposed. It might also imply that changes in the price of manufactured imports would tend to have a more decisive impact upon wage increases than changes in the cost of imported food or raw materials.

Ceteris paribus, the resistance of industrialists to wage claims is likely to be strengthened by a decline (or lessened by an increase) in the prices of imported manufactures. One may, for instance, suppose that industrial employers generally are conscious that basic wage rates can rise only slightly faster than the prices of imported products if they are not to feel an increased pinch of foreign competition.¹⁵ Thus, one can visualize the employers’ side in many wage negotiations explicitly or implicitly using the current, or somewhat earlier, rate of increase in the price of foreign goods as a yardstick in deciding an upper limit for acceptable increases in wage rates.¹⁶ However, as Chart 7 shows, up to 1966 import prices for manufactures declined or scarcely increased in years of relatively high unemployment and increased in most years of low unemployment.¹⁷ This association may reflect both the fact that cyclical fluctuations in demand in the United Kingdom were in phase with those in other industrial countries and a direct influence of demand conditions in the United Kingdom upon the prices of its imports of manufactures. As a consequence of this association, some of the influence of changes in the prices of imported manufactures may already be included in the expected increases in earnings under given labor market conditions.

Chart 7 provides an indication of whether the movement of prices for imported manufactures could be expected to have an unusual impact upon wage increases in any year. The percentage increase, or decrease, in the price of imported manufactures is plotted against the expected increase in unit labor cost implied by the expected increase in earnings and a growth of productivity of 3½ per cent per annum. The position of the two curves suggests that the movement of import prices constituted an unusually strong constraint upon wage increases in 1953–55 and 1957–58, and had the opposite effect, exercising an exceptional upward pull on wage increases, in 1956, 1965, and 1967–70. These effects appear to be reflected in the lower-than-expected wage increases in 1953–55 and 1959 and the higher-than-expected increases in 1956, 1965, and 1967–70. The fact that the influence of declining import prices in 1957–58 upon wage increases does not show up until 1959 may be due to the impact of the Suez crisis in 1957–58.¹⁸

The effect of wage restraint in 1961 and 1966

Some idea of the effect of wage restraint in 1961 and 1966 can be gained from Chart 7. In 1961 the effect on negotiated rates was very slight, but wage drift was relatively lower than in 1960 or 1962. There was little, if any, subsequent reaction in the form of a faster-than-expected increase in wages in 1962. Although unemployment decreased between 1960 and 1961, some moderation in the rate of increase in hourly earnings was to be expected because the number of unfilled vacancies had declined.

Wage restraint in 1966 had a greater impact on negotiated settlements, but this was accompanied by much-heavier-than-usual wage drift and was followed by much-larger-than-expected increases in rates in 1967. (See Table 4.) It might appear that considerable wage drift in 1966 was due principally to the substitution of increases at the local level for increases in rates not obtained in centralized bargaining; this would seem to support the pessimistic view that restraining negotiated increases in wage rates by incomes policy will not succeed in moderating increases in earnings. Such a conclusion needs to be qualified, however. When unemployment is already very low, and the number of vacancies very high, as in 1965, neither statistic may fully reflect a further intensification of demand for labor.¹⁹ Thus, the slight decrease in unemployment and virtually unchanged number of vacancies from June 1965 to June 1966 (see Table 5) most probably indicate a quite marked increase in pressure in the labor market over this period. This would mean that, despite a decline in the number of vacancies from October to October, conditions conducive to heavy wage drift applied for most of the 12 months up to October 1966.

Table 4.

United Kingdom: Expected and Actual Wage Increases, 1966–70

(In per cent)

Source: Estimates described in text. Actual increases in earnings, rates, and drift are as given in columns 1, 3, and 4 of Table 3.

Table 4.

United Kingdom: Expected and Actual Wage Increases, 1966–70

(In per cent)

		1966	1967	1968	1969	1970
Expected increase in earnings		6.6	2.4	3.2	3.0	1.7
	Increase in rates	6.4	2.1	1.7	1.8	1.0
	Wage drift	0.2	0.3	1.5	1.2	0.7
Actual increase in earnings		7.0	4.1	6.9	7.6	15.9
	Increase in rates	5.8	5.5	6.6	5.3	11.8
	Wage drift	1.2	−1.4	0.3	2.3	4.1
Discrepancy on earnings		0.4	1.7	3.7	4.6	14.2
	Rates	−0.6	3.4	4.9	3.5	10.8
	Wage drift	1.0	−1.7	−1.2	1.1	3.4

Source: Estimates described in text. Actual increases in earnings, rates, and drift are as given in columns 1, 3, and 4 of Table 3.

View Table

Table 4.

United Kingdom: Expected and Actual Wage Increases, 1966–70

(In per cent)

		1966	1967	1968	1969	1970
Expected increase in earnings		6.6	2.4	3.2	3.0	1.7
	Increase in rates	6.4	2.1	1.7	1.8	1.0
	Wage drift	0.2	0.3	1.5	1.2	0.7
Actual increase in earnings		7.0	4.1	6.9	7.6	15.9
	Increase in rates	5.8	5.5	6.6	5.3	11.8
	Wage drift	1.2	−1.4	0.3	2.3	4.1
Discrepancy on earnings		0.4	1.7	3.7	4.6	14.2
	Rates	−0.6	3.4	4.9	3.5	10.8
	Wage drift	1.0	−1.7	−1.2	1.1	3.4

Source: Estimates described in text. Actual increases in earnings, rates, and drift are as given in columns 1, 3, and 4 of Table 3.

Table 5.

United Kingdom: Unemployment, Increase in Active Labor Force, ¹ and Vacancies, 1951–70

(In thousands)

Sources: United Kingdom, Central Statistical Office, Annual Abstract of Statistics; Department of Employment Gazette. Figures relate to Great Britain.

¹Covers employed wage and salary earners and persons seeking employment; excludes self-employed persons and members of the armed forces.

²Probably understates the true level of unfilled vacancies; see text.

³The level of unemployment was temporarily reduced as a consequence of the Suez crisis.

Table 5.

United Kingdom: Unemployment, Increase in Active Labor Force, ¹ and Vacancies, 1951–70

(In thousands)

Year		Unemployment in June (U)	Increase in Active Labor Force over Previous June (ΔL)	Unfilled Vacancies in June (V)
Low unemployment
	1951	191	120	487
	1954	240	310	381
	1955	211	270	443
	1956	223	240	407
	1961	266	300	383
	1965	276	210	449
	1966	261	130	4502
Moderate to high unemployment
	1957	(265) 3	120	309
	1960	305	380	361
	1962	397	310	252
	1964	322	150	368
Highest unemployment
	1952	440	90	305
	1953	298	80	296
	1958	429	–40	213
	1959	419	130	247
	1963	480	120	215
	1967	500	–260	280
	1968	517	–140	303
	1969	499	–80	322
	1970	547	…	296

Sources: United Kingdom, Central Statistical Office, Annual Abstract of Statistics; Department of Employment Gazette. Figures relate to Great Britain.

¹Covers employed wage and salary earners and persons seeking employment; excludes self-employed persons and members of the armed forces.

²Probably understates the true level of unfilled vacancies; see text.

³The level of unemployment was temporarily reduced as a consequence of the Suez crisis.

View Table

Table 5.

United Kingdom: Unemployment, Increase in Active Labor Force, ¹ and Vacancies, 1951–70

(In thousands)

Year		Unemployment in June (U)	Increase in Active Labor Force over Previous June (ΔL)	Unfilled Vacancies in June (V)
Low unemployment
	1951	191	120	487
	1954	240	310	381
	1955	211	270	443
	1956	223	240	407
	1961	266	300	383
	1965	276	210	449
	1966	261	130	4502
Moderate to high unemployment
	1957	(265) 3	120	309
	1960	305	380	361
	1962	397	310	252
	1964	322	150	368
Highest unemployment
	1952	440	90	305
	1953	298	80	296
	1958	429	–40	213
	1959	419	130	247
	1963	480	120	215
	1967	500	–260	280
	1968	517	–140	303
	1969	499	–80	322
	1970	547	…	296

Sources: United Kingdom, Central Statistical Office, Annual Abstract of Statistics; Department of Employment Gazette. Figures relate to Great Britain.

¹Covers employed wage and salary earners and persons seeking employment; excludes self-employed persons and members of the armed forces.

²Probably understates the true level of unfilled vacancies; see text.

³The level of unemployment was temporarily reduced as a consequence of the Suez crisis.

The different outcome of wage restraint in 1961 and 1966 was almost certainly influenced by a marked difference in the underlying labor market situation in these years, owing to the unusually large number of new entrants in 1961 and the reduction in the number of young persons entering employment, which led to a marked decline in the growth of the labor force, in 1966.

The influence of the growth of the labor force upon the rate of wage increase is discussed in the next section. Before turning to that topic, however, it may be useful to summarize wage developments in 1967–70.

Wage developments in 1967–70

The greater-than-expected increase in wage rates in 1967 can be ascribed partly to the sharp rise in import prices for manufactures during 1966, partly to the easing of the wage freeze, and partly to the extension of productivity agreements (providing for exceptional wage increases as a quid pro quo for the adoption of new work procedures and involving the overhaul of systems of payment with the consolidation of various premiums and other payments into the rates structure). The last factor probably explains the exceptional negative drift in 1967 and the smaller-than-expected drift in 1968, when the number of vacancies again increased. (See Table 4.) Making rough allowance for this special factor, it appears that wage rates rose about 1½ percentage points faster than “expected” in 1967, about 3½ points faster in 1968 and 1969, and nearly 11 per cent faster in 1970. In contrast to 1967 and 1968, 1969 was a year of substantial excess drift and 1970 was a year of quite exceptionally strong wage drift.

Although the discrepancies from the expected increases in 1968 and 1969 appear large in comparison with the experience of the late 1950s and 1960s, the actual increases in earnings seem rather moderate when seen in the context of the exceptional rise in import prices. The implied increases of 4–5 per cent in unit labor costs each year (assuming “normal” growth in productivity of 2–3 per cent) followed increases of 5 per cent and 12 per cent in the price of manufactured imports in 1967 and 1968. The competitive gain following devaluation was maintained in 1969.

It is not possible to give any comprehensive explanation here of the increase of almost 16 per cent in hourly earnings in 1970, revealed in the recently published earnings survey for October 1970. But it is clear that external factors contributed to the acceleration, in addition to the special circumstances connected with the election and change of government. The faster rise in import prices for foodstuffs than in 1969 or even 1968 ²⁰ had more effect in raising the cost of living and influencing the size of wage claims. These effects were reinforced by the increase in indirect taxes, introduction of charges for some social services, etc. Together with the reduction in direct taxes on higher incomes, the proposals for legislation on industrial relations, and the. general abandonment of a policy of incomes restraint, this led to greatly increased labor aggressiveness. More intense pressure for wage increases at the local and central level coincided with reduced resistance to wage increases that was due in part to the faster rise in the price of competing foreign manufactures: the 9 per cent rise in the price of imported manufactures during 1969 gave scope for a substantially faster rise in earnings in manufacturing than had actually occurred in 1968 or 1969.

In the event, the rise in hourly earnings during 1970 far outstripped the 1969 rise in import prices for manufactures and, as there was virtually no increase in such import prices during 1970, led to a substantial deterioration in the underlying competitive position by the end of the year.

The relation between the level of unemployment, the growth of the labor force, and unfilled vacancies

It has been suggested that the strength of wage drift is related to the change in vacancies. Since the number of unfilled vacancies usually varies inversely with unemployment, this broadly implies that wage drift, like wage rates, will vary inversely with the level of unemployment, except when the relation between the number of unemployed and the number of unfilled vacancies changes. It will be useful to consider briefly what determines this latter relationship and how it is influenced by the growth of the active labor force.

Unfortunately for statistical neatness, the size of the labor force is estimated in June each year; thus, it is necessary in this section to study the development of unemployment and vacancies from June to June.²¹ The term “active labor force” here covers wage and salary earners at work and persons seeking employment; it excludes self-employed persons and members of the armed forces.²²

Since 1951 the sum of the number of unemployed (U) and the number of unfilled vacancies (V) in June has seldom fallen below 600,000 or exceeded 725,000. As Chart 8 shows, the total was at the lower end of that range from 1953 to 1957 and was rising gradually during the 1960s. By 1968 the total exceeded 800,000.

Chart 8.

United Kingdom: Labor Market Developments, 1951–69 ¹

Thousands

View Expanded

¹ Data relate to Great Britain.

One reason for the relative stability of this total is the responsiveness of the active labor force to variations in the demand for labor. Except during the early 1960s—when the number of new entrants to the active labor force was unusually large owing to the postwar bulge in the birth rate, high immigration, and the reduction in the armed services—the increase in the labor force did not exceed 200,000 unless the number of unemployed was well below 300,000 in June; and if unemployment exceeded 400,000, the labor force rose by substantially less than 200,000, or actually declined.²³ (See Table 5.)

The “natural” growth of the labor force is determined by the growth in the population of working age. However, in a country like the United Kingdom, where the population of working age is growing very slowly, the actual year-to-year growth of the labor force is very sensitive to the level of economic activity, since this influences the number of persons retiring from the labor force. When demand for labor is strong, and vacancies are difficult to fill, enterprises encourage older employees to remain at work, and older workers can find employment if they become unemployed. Since enterprises prefer to effect temporary reductions in their work forces by releasing older workers, easier demand conditions result in the release of workers who are at or past the age of retirement—many of whom quit the labor force. Also at higher levels of unemployment, older workers among the unemployed find it more difficult to obtain work and are more likely to leave the labor force. The employment of married women is similarly affected by variations in the strength of demand for labor. Consequently, the strength of demand affects both the growth of the labor force and the level of unemployment.

One can see how this adjustment process accounts for the relative stability of the sum of vacancies and unemployment. Suppose that the demand for labor declines. At some point there will be a tendency for the sum of unemployment and vacancies (U + V) to increase, because V will be approaching a minimum (set by labor turnover and the minimum time taken to match up applicants and openings), while U is increasing. But as U increases relative to V, so the tendency for increased retirements from the labor force strengthens. Conversely, suppose that the demand for labor increases. At some point U + V will tend to increase because U will be approaching minimum frictional unemployment (set by labor turnover, as well as the number of “unemployables”), while V is tending to increase. As V increases relative to U, the tendency for increased participation in the active labor force strengthens, lessening the further rise of V.

There is, however, an important difference between this and the preceding case. Because of unemployment compensation, unemployed workers have an incentive to remain on the registers for at least as long as they qualify for unemployment benefits; but when the number of outstanding vacancies is very great, employers may not bother to notify vacancies. Thus, as mentioned earlier, when unemployment is very low and vacancies are already high, neither series may indicate a further intensification of pressure in the labor market. It also follows from this that, while the number of unemployed may increase relative to the number of vacancies in recession years, the reverse is less likely to occur in boom years. The sharpest increases in U + V in fact occur in years of sharply declining demand for labor and rapidly increasing unemployment—1952, 1958, 1963, and 1967 (Chart 8).

Now consider the relation between unemployment, growth of the labor force, and the number of unfilled vacancies. The number of persons available for employment depends on both the number of unemployed and the potential growth in the labor force. Consequently, the ease of filling vacancies at any level of unemployment will be greater (or less), the higher (or lower) is the potential growth in the labor force; and one can therefore expect the number of vacancies to be lower (or higher) at a given level of unemployment. Similarly, if the potential growth in the labor force declines or increases for any reason, one may expect the level of vacancies associated with a given level of unemployment to rise or decline.

The potential growth of the labor force at any time depends on (1) the natural growth of the population of working age; (2) the existing participation of the population of working age in the active labor force (i.e., on whether or not participation has been held down at the previous level of demand for labor); (3) exogenous factors that from time to time alter the participation of that population in the active labor force at any given level of demand for labor. Such things as changes in the school-leaving age, the extension of higher education, changes in conscription policy, improved benefits or increased eligibility for unemployment compensation, and greater eligibility for pensions may be important in this connection.

A marked reduction in the potential growth of the labor force appears to have caused the changed relation of vacancies to unemployment in the late 1960s. During the boom years of the mid-1950s and early 1960s when unemployment declined to a low level (i.e., well below 300,000 in June), the labor force increased by 270,000–300,000 or more. The number of vacancies did not reach 450,000 even when unemployment fell to 211,000 in 1955. In 1965 the number of vacancies rose to almost 450,000 with 276,000 unemployed, and in 1966 the true number of vacancies was probably considerably higher, with 261,000 unemployed. The number of vacancies in 1965 was thus at least 80,000 higher than could have been expected, at the same level of unemployment, in 1961. The much slower growth of the labor force in 1964–66 than in 1960–62 (averaging less than 165,000, against 330,000, a year) is not attributable to demand factors and is reflected in the exceptionally sharp rise in outstanding vacancies in 1964 and 1965.²⁴

In 1967 and 1968 the labor force declined considerably, whereas it had continued to expand in all the previous recession years except 1958. The decline was due largely to a marked reduction in the number of young persons entering the labor force. The argument stated above would indicate that the sharp reduction in the potential growth of the labor force was probably sufficient to account for the considerably higher number of vacancies than in the previous years of high unemployment. This fact in itself would tend to encourage more of the unemployed to remain in the labor force; that is to say, it would at least partly explain the higher absolute level of unemployment reached in spite of the decline in the labor force. But exogenous factors also worked in this direction. The introduction of the Selective Employment Tax made employers in service industries and construction more inclined to release workers and more selective in their hiring. This probably led to more dismissals than previously among workers not approaching retirement and in localities where alternative employment opportunities were limited. Unemployed workers may also have become rather less willing to accept the relatively less attractive jobs because of higher unemployment benefits and, sometimes, redundancy payments.²⁵

This description suggests that the higher level of unemployment in 1967 and 1968 than in 1958 or 1963 probably did not indicate an easier labor market than in those years, because of the severely restricted potential growth of the labor force and because of the special factors tending to raise the amount of unemployment without making vacancies correspondingly easier to fill. It would appear that the comparatively high level of outstanding vacancies gives a truer picture of labor market conditions in those years.²⁶

From an economist’s point of view, it is a great pity that the recent declines in the labor force coincided with the 1966 wage freeze and its aftereffects and with the devaluation. Otherwise, they would have provided a rare opportunity to study the effect of an increase in “pressure” on the labor market not associated with an increase in the strength of demand for output. It would have been interesting to see whether the decline in the labor force, and the rise in the number of vacancies, was reflected in unusually heavy wage drift, and whether this was subsequently followed by a larger rise in wage rates than would normally be expected at the prevailing level of unemployment. As it is, any such developments are concealed in the much-faster-than-expected rise in wage rates in 1967–69, following the postponement of increases in 1966 and partly reflecting cost of living adjustments in 1968 and 1969. It is not excluded, however, that the change in the underlying labor supply situation had its effect in inducing a rapid rise in actual earnings, and this in turn played a part in predisposing unions to seek—and employers to grant—larger-than-usual increases in rates.

Some general conclusions can be drawn from this analysis. It appears that one reason for the faster rise in wage earnings at any given level of unemployment since 1965 has been the decline in the potential expansion of the active labor force. This seems to have resulted in a gradual rise in the number of vacancies associated with any given level of unemployment, and so in a larger increase in vacancies and hence stronger wage drift than formerly for any given change in unemployment.

The analysis underlines the importance of supplementing aggregate demand policy by an active labor market policy aimed at securing the maximum improvement in the rate of filling vacancies following any easing of demand for labor (e.g., by increasing labor mobility, reducing barriers to new entrants, and eliminating fiscal disincentives that discourage work after marriage or retirement). The brief description of developments in the labor market between 1951 and 1968 also indicates that, when aggregate demand is sharply restrained, the losses of real output (i.e., potential utilization of manpower for production) involved in increasing the degree of unemployment are substantially greater than the unemployment figures alone would indicate.

the level of unemployment and the increase in wage rates

The relation between unemployment and the rate of increase in wage rates is best regarded as a case of codetermination rather than of cause and effect. What conditions the bargaining power of the unions in the short run is the strength of employers’ desire to maintain or expand production in the immediate future. When demand conditions are strong, employers are prepared to offer a larger number of jobs and to accede more readily to union wage claims than when demand conditions are less favorable. Unions, for their part, are inclined to press larger claims when final demand is strong, mainly because the chances of success are greater. The unions’ bargaining power is reduced when aggregate demand slackens because the strike weapon generally becomes less effective against enterprises that are working below capacity, or that are accumulating, or have accumulated, inventories. Hence, the level of unemployment, which reflects the changes in demand for labor, provides a good indicator of wage bargaining conditions. It is probably a better indicator of bargaining conditions than other series, such as profits, because it provides a better index of how short-term fluctuations in demand conditions are affecting employers.

Now this does not mean that the level of unemployment per se is the causal factor that influences the bargaining power of the unions. The cause of the strengthening (or weakening) in the bargaining position of the unions is not the changing pressure of demand for labor relative to the available manpower supplies but the strength of aggregate demand in the economy. Thus, the inverse relation between unemployment and the rise in wage rates should be seen to arise from two distinct effects of the same underlying cause, namely, the variation in aggregate demand.²⁷

The willingness of employers to give in to a wage claim, or to submit to interruption of production, will depend on their collective view of the profitable openings and opportunity costs of forgoing production and of the possibility of passing on increased wage costs in higher prices. Their collective view is, naturally, influenced by expectations of the effects of government financial policy on the level of aggregate demand at home. It is, therefore, not surprising that changes in wage rates in different industries are quite highly correlated,²⁸ suggesting that “wage negotiations are affected by demand conditions in the economy as a whole—not the ‘local’ demand conditions obtaining in each industry.” ²⁹ In any given demand situation, the views of industrialists concerning the possibility of passing on increased wage costs in higher prices will be conditioned by the current trend of prices for competing foreign products. Since the level of wages strongly influences wages in the more sheltered sectors of the economy, changes in the price of imported manufactures are liable to have a considerable impact on the movement of wages in general.

The introduction of the movement in import prices for competing foreign products as a determinant of the rate of increase in wages carries some important implications concerning the position and shape of the Phillips curve for various countries. It suggests that, in those countries that maintain the lowest levels of unemployment, the rate of wage and price increase at a given level of unemployment will be moderated by the price competition of other countries’ imports. Hence, the apparent trade-off between unemployment and wage (or price) increases may be relatively more favorable than in those countries that maintain higher levels of unemployment (e.g., the United States). It also suggests that an industrial country that, for any reason, experiences fluctuations in activity out of phase with the fluctuations in the other industrial countries will be likely to have a flatter Phillips curve than one whose fluctuations coincide with the cycle in the other countries.

What does this imply concerning the interpretation of the Phillips curve? It means that, so long—but only so long—as unemployment fluctuates around a similar mean, and the cycle remains consistently in (or out of) phase with the fluctuations in the other industrialized countries, and there are no marked changes in the exchange rates, the Phillips curve will be reasonably stable, i.e., the level of unemployment will provide a good indication of the rate of wage increase to be expected. This relation may be useful in forecasting the probable rate of inflation in wages, rise in money incomes, and effects on prices in the short run.³⁰ However, the provisos introduced above suggest that the relation cannot be regarded as a generally stable one.

Although widely interpreted as such, the relation does not indicate a trade-off between the level of unemployment maintained and the increase in wage rates. The observed tendency for slight differences in the level of unemployment to be associated with marked differences in increases in wage rates arises in large part from the postponement, or advancement, of wage claims in the course of short-term cyclical fluctuations. The observed relation provides little indication of how the increase in wage rates might vary with differences in the level of unemployment maintained in the longer run.

It seems probable that, within the range of politically tolerable unemployment levels, slight differences in the unemployment level over the longer term would have a comparatively slight effect on the rate of wage increase.³¹ Thus, one could not assume, as Phillips suggested, that “if the average level of unemployment were kept at a little less than 2½ per cent, the average rate of increase in wages [i.e., wage rates] over a period of years” would be the same as the increase observed at 2½ per cent unemployment in the recession years of some recent period.³² For example, if unemployment had fluctuated around 2½ per cent instead of the 1¾ per cent average of the years 1956–67, the average annual increase in wage rates might well have been little, if any, smaller than the 5 per cent actually experienced over that period.³³

The relation between changes in wage rates and changes in earnings

Changes in hourly rates do not indicate exactly how hourly earnings are moving, because the relation of earnings to rates varies. One may abstract for a moment from the problems raised by inconsistencies of coverage and calculation in the earnings and rates indices, and assume that the earnings series correctly measures the average increase in earnings, excluding the effect of overtime, for the same fixed sample of workers that is covered by the rates index.

Even in a single industry, one could not expect the rate of increase in earnings to be consistently in line with the increase in wage rates, because earnings are continually being modified, whereas changes in rates occur at infrequent intervals after protracted negotiations and because earnings are influenced by local factors as well as by nationally determined rates. Even when there is no constraint on centralized wage bargaining, earnings tend to rise independently of changes in rates for many reasons, such as the fixing of piece rates at the factory,³⁴ raising of actual rates relative to standard rates, or upgrading of workers in order to avoid labor turnover, to obtain labor more quickly, to secure the “pick” of the labor market, etc. Almost all such forms of wage drift are liable to be influenced by the demand situation facing the particular enterprise (or plant) and the tightness of the local labor market situation. Since wage drift occurs as a consequence of day-to-day decisions, and does not—like wage rate negotiations—involve lengthy negotiations, the increase in earnings is likely to accelerate (or decelerate) before a similar change occurs in the rate of increase of negotiated rates. One may suppose that the strength of local pressures tending to raise earnings relative to wage rates is related to the keenness of employers to fill vacancies and their difficulties in doing so; the number of unfilled vacancies may provide a reasonably good indication of this. Thus, it seems plausible that earnings in industry generally should tend to rise more relative to ruling rates when the number of vacancies is increasing and less when the number of vacancies is stationary or declining, while the increase in wage rates (which, like the level of unemployment, reflects changes in demand conditions with a lag) is correlated with the level of unemployment. This would imply that earnings will rise faster than rates during the upswing of the cycle but will rise little faster or less than in line with wage rates during the downturn.

From this description, it would appear that the usually closer relation between unemployment and changes in wage rates than that between unemployment and changes in earnings is partly a question of timing. A change in the level of aggregate demand, and in employers’ demands for labor, has a speedier impact on vacancies and earnings than on the level of unemployment and negotiated increases in wage rates, which take time to eventuate.

Considerable ingenuity has been expended in trying to prove from the timing of changes in the two series that changes in rates induce changes in earnings rather than vice versa.³⁵ In fact, the chain of causation is complex, and mere coincidence of timing cannot indicate whether the rise in earnings is a consequence of the change in rates or vice versa. A marked tendency for earnings to rise relative to rates may stimulate wage rate claims and strengthen the bargaining power of the union; and an increase in rates may set off increased drift in the earnings of other workers. Because of this latter effect, there is likely to be a tendency for any given increase in wage rates to have more (or less) effect upon earnings, the stronger (or weaker) is the pressure of demand for labor currently experienced by employers.

Although wage decisions at the national and local level are influenced by different factors, at different points in time, neither set of decisions is insulated from the other. Consequently, unless a marked change in the economic situation intervenes, a faster rise at either level is liable to induce an acceleration of increases at the other level.

To sum up, this discussion suggests that the rate of increase in wages is strongly influenced by the expected level of aggregate demand in the near future, which largely determines the current strength of employers’ demand for labor. The percentage of unemployment (which normally reflects changes in the strength of employers’ demand for labor) provides a good index of wage bargaining conditions and hence of the probable scale of increase in centrally negotiated wage rates. In any given demand situation, industrialists’ views concerning the possibility of passing on wage increases in higher prices are liable to be influenced by the current movement of prices for competing foreign products; changes in the price of imported industrial products may therefore have a direct influence on the rate of increase in wages. A third factor influencing the rate of increase is the pressure of demand for labor relative to supply, especially as this affects the ease or difficulty that employers experience in filling vacancies, and hence their inclination to raise wage earnings above the going rate in order to maintain or expand their respective labor forces. The strength of this factor can be gauged from changes in the number of outstanding vacancies.

Résumé

Les chiffres publiés sur les variations des salaires horaires moyens des manoeuvres sont influencés à la fois par la façon dont la répartition des emplois varie dans le secteur industriel et par des facteurs démographiques qui influent sur la proportion des heures travaillées par les hommes, femmes et adolescents; ils sont également déformés par la méthode d’ajustement utilisée pour exclure l’effet des heures supplémentaires. Si l’on soustrait l’effet des variations de la proportion des heures travaillées par ces trois groupes d’individus et si l’on n’exclut pas le salaire versé pour les heures supplémentaires habituellement ouvrées par les hommes, les estimations obtenues ont moins tendance à augmenter au cours des derniers cycles que les chiffres publiés et fournissent en outre des estimations corrigées du glissement salarial. Alors que les modifications des taux de salaires varient avec le pourcentage de chômage(u), approximativement selon la formule WR = 11,1 – 4,22u, l’importance du glissement salarial est, elle, liée aux variations du nombre des offres d’emplois non satisfaites (v), approximativement selon la formule: WD = 1,11 + 0,0131v.

Une analyse de l’évolution du marché du travail semble indiquer qu’une diminution de l’accroissement potentiel de la population active à la fin des années 1960 a entraîné une augmentation du nombre des offres d’emplois non satisfaites par rapport à un niveau donné de chômage, ce qui a intensifié le glissement salarial.

Les disparités qui existent entre les variations effectives des gains horaires et celles qui sont “prévues” en vertu des formules susmentionnées font apparaître vers la fin des années 1950 un ralentissement des hausses des salaires par rapport aux conditions de la demande régnant alors sur le marché du travail et, pendant la deuxième moitié des années 1960, d’une accélération qui s’est accentuée très nettement après la dévaluation. L’auteur donne provisoirement comme explication à ces tendances la baisse marquée des prix des biens manufacturés importés à la fin des années 1950 et leur hausse exceptionnellement rapide vers la fin des années 1960.

En conclusion, l’auteur estime qu’à court terme la relation qui existe entre le nombre de chômeurs et la variation des taux de salaires devrait être considérée comme un exemple de “codétermination” plutôt que comme une relation de cause et effet. Cette conclusion diffère ainsi de l’opinion généralement acceptée selon laquelle la courbe de Phillips définit un “trade-off” entre le chômage et la hausse des salaires.

Resumen

Las cifras publicadas sobre la variación en el promedio de los ingresos por hora de los trabajadores manuales vienen influidas por las variaciones en la distribución industrial del empleo y factores demográficos que afectan la parte de horas trabajadas correspondiente a hombres, mujeres y jóvenes, y vienen asimismo desfiguradas por el método de corrección aplicado al trabajo en horas extraordinarias. Las estimaciones que dejan de lado el efecto de la variación en la proporción de horas trabajadas por cada grupo, y sin excluir la compensación pagada por las horas extraordinarias que trabajan de forma regular los hombres, indican mayor número de movimientos semejantes durante los ciclos recientes que las cifras publicadas, y proporcionan estimaciones revisadas del deslizamiento de los salarios. Aunque las variaciones de la tasa salarial varían con el porcentaje de desempleo (u) aproximadamente de acuerdo con la fórmula de WR = 11,1—4,22 u, la fuerza del deslizamiento se relaciona con la variación del número de vacantes (v) aproximadamente de la forma siguiente: WD = 1,11 + 0,0131v.

Un análisis de la evolución del mercado laboral indica que, a finales del decenio de 1970, una disminución del crecimiento potencial de la mano de obra asalariada condujo a un aumento en el número de vacantes existentes a un nivel dado de desempleo, y se fortaleció el deslizamiento salarial.

Las discrepancias entre las variaciones efectivas de los ingresos y las “previstas”, sobre la base de las fórmulas anteriores, indican una desaceleración de las subidas de salarios en relación con las condiciones de la demanda en el mercado laboral a fines del decenio de 1950, y una aceleración en la segunda parte del decenio de 1960, que se hace marcadísima tras la devaluación. Esas tendencias se explican de momento por la enorme bajada, a finales del decenio de 1950, de los precios de las manufacturas importadas, y la rapidísima subida que experimentaron a finales del decenio de 1960.

Para concluir, se sugiere que la relación de corto plazo entre el desempleo y la variación de las tasas salariales debe considerarse como un caso de codeterminación, y no de causa y efecto. Esto difiere de la interpretación generalmente aceptada de la curva de Phillips como indicativa de una equivalencia entre el desempleo y las subidas de salarios.