Jump to ContentJump to Main Navigation
North-South Trade, Employment and InequalityChanging Fortunes in a Skill-Driven World$

Adrian Wood

Print publication date: 1995

Print ISBN-13: 9780198290155

Published to Oxford Scholarship Online: November 2003

DOI: 10.1093/0198290152.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017

(p.458) Appendix A5 Wage Dispersion and Income Inequality

(p.458) Appendix A5 Wage Dispersion and Income Inequality

Source:
North-South Trade, Employment and Inequality
Publisher:
Oxford University Press

This appendix completes the review of recent trends in skill differentials in the North (initiated in Appendix A3 and continued in Appendix A4). It examines movements in overall wage dispersion (Section A5.1), and possible causes of these movements (Section A5.2). It concludes by looking also at trends in the distribution of income among households (Section A5.3).

The two preceding appendices examined movements in the relative wages of skilled and less‐skilled workers, using education, age, and occupation as measures of skill. This appendix uses a rather different measure of skill, namely the wage itself. It attempts to make inferences about movements in skill differentials from changes in the dispersion (or inequality) of the size distribution of wages among workers. The underlying assumption is that high‐ and low‐earning groups of workers correspond roughly with more‐ and less‐skilled groups (cf. Williamson and Lindert 1980: 327). If so, a rise (say) in the wages of skilled workers relative to less‐skilled workers would clearly tend to increase wage dispersion. Conversely, other things being equal, greater wage dispersion would imply wider skill differentials.

This measure of skill has some advantages over the education, age, and occupation measures used earlier, which fail to capture the large and obvious diversity of skill levels among people with a given number of years of schooling and experience, or in particular occupational categories. By contrast, differences in wages are in principle capable of reflecting the differing economic value of all sorts and gradations of work‐related abilities, whether innate, or deliberately or fortuitously acquired. In practice, however, changes in wage dispersion are at best an imperfect indicator of changes in skill differentials, and are more reliable in this regard when (as here) studied in conjunction with other cruder indicators.

The most obvious limitation is that not all wage differences among workers are caused by differences in skill. Some of the other causes can be controlled for: for example, using data on the earnings only of full‐time workers eliminates most of the effect of differences in hours worked. But many non‐skill‐related causes of wage differences cannot be controlled for in size distribution data. These include differences in the nonpecuniary attractiveness of particular jobs, obstacles to labour mobility, unions, and other direct institutional determinants of wage rates. Observed movements in wage dispersion could thus arise from changes in these other influences.

Even if skill were supposed to be the sole influence on wages, movements in wage dispersion would not necessarily be a good indicator of changes in the relative wages of workers with specific degrees of skill. Such movements could also arise from changes in the skill distribution of the population of workers concerned. Even with no change in skill differentials, for example, wage dispersion would increase if the gap between the amounts of skill possessed by higher‐paid and lower‐paid workers widened over time. Wage dispersion would likewise tend to diminish if unskilled workers were given relatively more training, or if they ceased to work, and hence vanished from the wage statistics. Moreover, such (p.459) changes in the skill composition of the work‐force might be causally associated with changes in skill differentials, and might thus amplify or muffle the effects of changes in skill differentials on wage dispersion.1

A5.1 Trends in Wage Dispersion

Fig. A5.1 and Table A5.1 summarise information on changes in wage dispersion in eleven Northern countries during the 1970s and 1980s. In almost all cases the data refer only to full‐time workers. Inclusion of part‐timers would not only make the distribution appear more dispersed at any given time, but would also tend to

Appendix A5 Wage Dispersion and Income Inequality

Fig. A5.1 Wage Dispersion In Four Large Northern Countries, 1970–1990 (Q90/Q10 For Full‐Time Males)

Note: Because the coverage of the wage and of the work‐force differ somewhat across countries, levels of dispersion are not strictly comparable.

Sources: Canada (annual earnings): from Table A5.1, with interpolation. France (annual earnings): from Table A5.1, with interpolation. UK (weekly earnings): New Earnings Survey summary volumes; refers to adults unaffected by absence, and includes overtime pay. USA (weekly earnings): Juhn, Murphy, and Pierce (1989); forward and backward extrapolation of 1970 (1968–72) Q90/Q10 ratio derived from table 1, using indices in fig. 3.

(p.460) increase its dispersion over time, because of the rising share of part‐time employment. Where possible, the sample is also limited to males, in order to eliminate the effect of the rising share of (generally lower‐paid) female workers. The measure of dispersion is in most cases dictated by the sources used, and varies among countries, adding to the difficulty of these international comparisons.

The figure contains information for four large countries—Canada, France, the UK, and the USA—on the ratio of the ninth to the first decile of full‐time male wages. Increased dispersion during the 1980s is apparent in all four countries. This tendency was strongest in the USA, where it represented the continuation of a milder and less steady trend in the same direction during the 1970s. In Canada, the pattern was similar to the USA (though with a 1986–8 drop, explained in Section A3.1 above). In the UK, the increase in dispersion seems to have started in 1977, prior to which there had been a modest downward trend. In France, the increase in wage dispersion started only in 1984, and was much slighter than in the other countries. Before then, dispersion had been diminishing from the early 1970s (the pre‐1980 numbers are not strictly comparable, but the trend during the mid‐1970s is corroborated by other data in Marsden 1989b, table 2.2).

Table A5.1 covers seven other countries. The coverage of the data, the time‐periods, and the measures of dispersion vary considerably. But the impression that consistently emerges is of reduced dispersion in the 1970s and increased dispersion in the 1980s. This U‐shaped pattern is most evident in Sweden, where the data include females, and the trough appears to have been in 1983. The data for Denmark hint at a similar pattern, though the coverage and the dispersion measure are unsatisfactory. There were increases during the 1980s also in The Netherlands and New Zealand (the data for both these countries include females). In both Germany and Italy, there was declining dispersion in the 1970s, although it was only slight in the former country. The limited data on full‐time workers for Austria show increasing dispersion in the 1980s. (The longer Austrian series, which includes part‐timers, exhibits an entirely different pattern, with rising dispersion in the 1970s and no change in the 1980s.)

Davis (1992, figs. 1A and 1B) compares trends in wage dispersion among full‐time males in eight Northern countries from the early 1960s to the late 1980s. For the USA, Canada, the UK, and Sweden, the Davis series show much the same movements as in Fig. A5.1 and Table A5.1. For France, the timing and size of the upturn in the 1980s is similar to that in Fig. A5.1, but the prior downtrend is much milder in the Davis series. In The Netherlands, the Davis series show little change between 1983 and 1987 (compared with the 1980–6 increase in Table A5.1, which is suggested by another source to have continued to 1988—Marsden 1990, n. 11).2 Two extra pieces of information in the Davis study are that wage dispersion rose steeply in Germany between 1981 and 1984, and in Australia between 1981 and 1987.

Davis also makes some illuminating cross‐country comparisons of the upper and the lower halves of the wage distribution. He notes that the levels of dispersion in the top half tend to be more similar across countries than in the bottom half. In particular, there is less lower‐half dispersion in the UK and France than (p.461)

Table A5.1. Changes in Wage Dispersion in the North, 1970–1990

Country

Indicator (definitions below) and coverage

Austria

1970

1975

1980

1983

1987

1989

Q75/Q25: full‐time males

1.47

1.52

Q75/Q25: all males

1.70

1.74

1.81

1.79

1.81

Canada

1967

1973

1981

1986

1988

Q90/Q10: full‐time males

4.08

3.55

3.71

4.29

4.03

Denmark

1975

1980

1983

1986

1989

Mean/median: full‐time male salary‐earners

1.10

1.08

1.08

1.08

1.10

Q75/Q25: all males

1.60

1.59

France

1970

1972!

1980

1984

1986

1990

Q90/Q10: full‐time males

3.67

3.85!

3.19

3.10

3.16

3.20

Germany

1972

1978

Q90/Q10: full‐time manual males

1.59

1.57

Q90/Q10: full‐time non‐manual males

2.13

2.10

Italy

1972

1978

Q90/Q10: full‐time manual males

1.88

1.58

Q90/Q10: full‐time non‐manual males

2.99

1.51

Netherlands

1978

1980!

1985

1986

Q75/Q25: all full‐time employees

1.47

1.46!

1.52

1.54

New Zealand

1980

1983

1986

1990

Top/bottom quintile groups mean income indices: all full‐time employees

100

101

102

103

Sweden

1975

1980

1983

1984!

1988

Top/bottom decile groups mean earnings: all full‐time employees aged 20–64

3.53

2.98

2.87

3.01!

3.65

Note: Q75/Q25 = ratio of upper to lower quartile; Q90/Q10 = ratio of 90th to 10th percentile. ! = discontinuity in source of data.

Sources: Austria: full‐time (standard income) figures from Statistiches Handbuch 1990 (table 9.24), 1984 (table 9.18); all‐males figures from 1990 Handbuch, table 9.05 (1989 number spliced over 1987 break).

Canada: Survey of Consumer Finances, special tabulation from Statistics Canada.

Denmark: Statistisk Årbog 1990 (tables 193, 203) and corresponding tables of earlier issues.

France: Sawyer (1982, table 7.12), Marsden (1989b, table 2.3), Girard and Lhéritier (1991).

Germany and Italy: Marsden (1989b, table 2.2).

Netherlands: Statistical Yearbook (1980: 352; 1981: 358; 1986: 347; 1988: 350).

New Zealand: Pocket Digest of Statistics (1988: 80); Key Statistics (Dec. 1990, table 4.02). Data in table generally refer to September, but in 1990 to June.

Sweden: Marsden (1989b, table 2.4); for 1988 calculated from Income Distribution Survey (1988, table 28).

in the USA and Canada, which corroborates other evidence that institutional pressures tend to prop up low wages much more in Europe than in America. Moreover, Davis (1992, figs. 3A and 3B) shows that in France the two halves of the wage distribution moved in different ways, with increasing dispersion in the upper half from about 1973, but declining dispersion in the bottom half until 1984. In Canada, the two halves also moved in opposite directions in the 1970s, but with declining dispersion in the top half and increasing dispersion in the bottom half. In the USA and the UK, the two halves of the distribution behaved more symmetrically.

There are many other studies of trends in wage dispersion in the USA. Their findings vary somewhat, as do their methods and choice of data, but it is apparently agreed by all that wage dispersion among both male and female full‐time (p.462) workers increased in the 1980s. The study by Juhn, Murphy, and Pierce (1989) from which the US data in Fig. A5.1 are drawn focuses on male wage‐ and salary‐earners with strong labour force attachment, and adjusts for a change in census procedures that may have reduced dispersion between 1974 and 1975. It concludes that there was little change during the 1960s, but an accelerating trend increase in dispersion after 1968 or 1969. Other authors interpret the unadjusted data as indicating a U‐turn in dispersion, with its trough in 1975 for full‐time males and females (Bluestone 1990, and Harrison and Bluestone 1990).

Blackburn and Bloom (1987) stress the differences among earlier studies of the USA. Their own calculations for full‐time year‐round workers during 1967–85 (ibid., table 4) indicate a U‐shaped pattern for females with a trough in the mid‐1970s (which the authors interpret as no trend), and for males little change until the late 1970s, after which there was a marked increase in dispersion. Ryscavage and Henle (1990, table 3) also examine the earnings, including self‐employment income, of year‐round full‐time workers. Between 1968 and 1978, they discover no change in dispersion for males, and a decrease in dispersion for females. Between 1978 and 1988, earnings dispersion increased substantially among both males and females.

In Canada, Picot, Myles, and Wannell (1990) discover that the earnings distribution of full‐time full‐year employees became increasingly polarised between 1967 and 1986. Up to 1981, much of this increase was due to the rising labour force participation of females (which is consistent with the lack of a clear trend for males during the 1970s in the analysis of Dooley 1987, table 5). But in the 1980s changes in the sex and age composition of employment contributed little to the marked increase in polarisation. The source of the Canadian data for males in Table A5.1 confirms that earnings inequality increased also among full‐time females during the 1980s, and suggests that the increase for females started earlier, in about 1973.

In the UK, the J‐shaped movement of wage dispersion among full‐time males during the 1970s and 1980s shown in Fig. A5.1 has been widely noted, and occurred also among full‐time females, although for them the trough appears to have been in 1979 rather than 1977 (Adams, Maybury, and Smith 1988, table 3).3 If the ratio of the upper to the lower quartile, rather than the top to the bottom decile, is considered, the trough appears earlier and wider: 1974–7 for males and 1977–9 for females (Moll 1991, table 2). There are no comprehensive data on the distribution of earnings for the UK prior to the 1970s, but Saunders and Marsden (1981: 159–66) infer that its dispersion had generally been declining in the 1950s and 1960s.

Movements in wage dispersion in other countries have received much less attention than in North America and the UK. Anticipating Davis (1992), Marsden (1989b) and Harrison and Bluestone (1990) suggest, each on the basis of data for a couple of other countries, that this trend may have been ubiquitous. Marsden notes, however, that the increase in dispersion appears to have been later and milder in France and Sweden than in the USA and the UK, although the more (p.463) recent data for Sweden in Table A5.1 and in Davis (1992) show a steep rise. In the two other countries with separate data for each sex (Austria and France), the sources for Table A5.1 show that the increase in dispersion among males in the 1980s was paralleled among females.

A5.2 Causal Interpretations

In summary, wage dispersion among full‐time workers, both male and female, appears to have increased during at least part of the 1980s in all eleven of the Northern countries for which relevant data exist. What happened earlier, and when, is more varied by country and by sex. However, as explained above, causal interpretation of these trends is not entirely straightforward. The increase in dispersion during the 1980s is consistent with widening wage differentials between skilled and less‐skilled workers, but it could also reflect changes in other determinants of wage differences, or in the shape of the skill distribution of employment.

The hypothesis that the increase in dispersion was caused solely by these other forces can be rejected with considerable confidence. For there is strong evidence, presented in Appendices A3 and A4, of a simultaneous widening of wage differentials among skill‐related education, age, and occupational categories. It is also clear, though, that the enlargement of these inter‐group wage differentials was responsible for only part of the increase in overall wage dispersion, to which increased dispersion within these groups also made a substantial contribution.

As usual, the increase in intra‐group wage dispersion is best documented for the USA. Juhn, Murphy, and Pierce (1989, fig. 11) show that rising returns to education and experience explain only about half of the increase during the 1980s in the overall dispersion of wages among full‐time males. The other half reflected rising inequality within education and age groups, a trend which had been strongly in evidence since the late 1960s, and had accounted for virtually all the pre‐1980 increase in overall inequality. Ryscavage and Henle (1990, tables 7 and 8) also discover that the Gini coefficients of the earnings distribution of full‐time year‐round workers increased between 1982 and 1988 in all six broad occupational categories, both for males and for females, though only in one category was the increase statistically significant. The same authors had earlier noted significant trend increases in earnings inequality among full‐time year‐round males in eight out of ten occupational groups during 1958–77 (Henle and Ryscavage 1980, table 3).

The evidence for other countries is more limited, but most of it points in the same direction. Davis (1992) examines movements in wage dispersion among full‐time males within education, age, and occupational groupings in the USA and seven other Northern countries. During at least part of the 1980s, he finds increases in intra‐group dispersion in Australia, Canada, Germany, The Netherlands, Sweden (from 1984), and the UK, but not in France. During the 1970s, too, there was an increase in Canada (as in the USA), but in France, Sweden, and the UK (until 1977) intra‐group dispersion declined (and there are no earlier data for the other three countries).

Picot, Myles, and Wannell (1990: 15) likewise conclude that almost all the increase in earnings polarisation among full‐time workers in Canada during (p.464) 1967–86 ‘took place within industries and occupations’.4 In the UK, Adams, Maybury, and Smith (1988, table 3) record that earnings dispersion increased during the 1980s among both manual and non‐manual full‐time workers of both sexes. Moll (1991) notes increased wage dispersion in this period also within the ten largest manual and non‐manual occupations for men and women. In France between 1982 and 1986 earnings dispersion increased among males in all three broad non‐manual occupations, but decreased among male manual workers and among females in three of the four occupations (Marsden 1990, table 3B). In Sweden between 1980 and 1984 wage dispersion increased among both skilled and unskilled manual workers, and in the two lower non‐manual occupational groups, but decreased among both higher and intermediate non‐manual workers (ibid., table 4).

The increases in wage dispersion within education and occupation groups are of course subject to the same problem of interpretation as the increase in overall dispersion, since they too could have been caused partly by changes in other determinants of wage differences or in the distribution of skills. Juhn, Murphy, and Pierce (1989) argue that increased intra‐group dispersion largely reflects rising returns to aspects of skill that cannot be measured with conventional census data. They also argue that this has been caused by increased relative demand for these unobservable skills, on the grounds that observed changes in industry and occupation mix suggest a simultaneous rise in the skill intensity of employment. This interpretation is highly plausible, but obviously is not amenable to direct verification. For one of the limitations of the wage as a measure of skill (unlike, say, years of schooling) is that it is not possible to juxtapose changes in the relative wages and relative numbers of people in different skill categories.

To some extent the observed movements in wage dispersion, both overall and intra‐group, must have reflected pressures on wage differentials unrelated to skill. In particular, as noted in Section A4.4, incomes policies, minimum wage legislation, and union wage policies appear to have compressed all wage differentials in the UK, France, Italy, and Sweden in the 1970s (and in the latter three countries also at the beginning of the 1980s). It is therefore likely that the increase in wage dispersion during the 1980s partly reflected the restoration of differentials to their previous levels (Marsden 1990). But wage dispersion in the UK had by the late 1980s risen well above its level of the early 1970s. More generally, the pervasiveness across countries of the increase in dispersion, and its persistence over time, make it impossible to believe that relaxation of egalitarian constraints was the main cause. And there is no other obvious non‐skill‐related explanation for this general increase in wage dispersion.5

It remains to consider the possibility that the increase in dispersion was partly, though clearly not entirely, caused by a change in the shape of the skill distribution, rather than in the relative wages of skilled and less‐skilled workers.6 This (p.465) hypothesis is often connected with the debate over ‘hollowing out’ or the ‘disappearing middle’ referred to in Section A4.4, but it needs some disentangling. It also demands closer attention to the concepts and measurement of dispersion, inequality, and polarisation. The term ‘dispersion’, which has been heavily used in this appendix, is general but rather vague. ‘Inequality’ is a more precise concept, although it is well known that different formal measures of inequality may give different results. The term ‘polarisation’ is formalised by Wolfson (1989) as the proportion of people in the tails of the distribution, with wages beyond some specified range around the median.7 He shows that standard measures of inequality and polarisation need not move in the same direction.

The distribution whose middle is said to be disappearing is often simply the size distribution of wages among individual workers. However, changes in the shape of this wage distribution, which also reflect changes in the relative wages of skilled and less‐skilled workers, are a poor guide to changes in the shape of the skill distribution. Moreover, a tendency for the middle of the wage distribution to disappear could be entirely due to wider skill differentials in wages, which would make the distribution more polarised. Whether the distribution would also become more unequal is less clear, and would depend on the measure of inequality. But inter‐quantile wage ratios, which are used to measure dispersion for most of the countries in Fig. A5.1 and Table A5.1, would usually move in the same direction as measures of polarisation.8 So there need be no conflict between the trends described here and those emphasised in other studies which have focused on the disappearing middle of the wage distribution.

Where there is a potential conflict is between the present view that the middle has shrunk mainly because skill differentials in wages have widened (as argued also, for example, by Juhn, Murphy, and Pierce1989) and the view that this has happened mainly because the proportion of jobs or workers with middling levels of skill has declined. But there is unfortunately little direct evidence on changes in the shape of the skill distribution. The best available study is that of Myles (1987), who uses four measures of occupational skill requirements, and data on the occupational composition of employment, to assess changes in Canada between 1961 and 1981. His results (ibid., tables 1 and 2) give no support to the (p.466) view that middling‐skill jobs have expanded more slowly than low‐ and high‐skill jobs.9 On the contrary, with very few exceptions, his calculations indicate a general upgrading of the skill distribution: progressively fewer jobs with low skill requirements and more jobs with higher skill requirements. This upgrading, moreover, proceeded faster in the 1970s than in the 1960s.

All other studies of this issue appear to have been forced to rely on changes in the distribution of jobs ranked by their wage levels. In Section A4.4, it was noted that the distribution of employment among coarse occupational categories suggests some shrinkage of middle‐wage jobs relative to lower‐wage jobs, both in the USA during 1984–90 and in the UK during 1973–86, and that this shrinkage was concentrated on occupations associated with manufacturing. Juhn, Murphy, and Pierce (1989, fig. 13) calculate the effects of changes in the industrial and occupational composition of employment on the proportions of workers at different relative wage levels in the USA during 1967–87. Over the whole period these structural changes would have increased the relative demand for highly paid workers and reduced the demand for workers with average or below‐average wages. This shift began in the early 1970s. In 1979–87 demand for the lowest‐paid workers fell less than for lower‐middle‐wage workers, but only slightly. The authors conclude that increased wage dispersion has been caused by relative shrinkage, not expansion, of the number of low‐wage jobs.

Picot, Myles, and Wannell (1990) likewise conclude from an analysis of Canadian data that people who attribute increased wage dispersion in the 1980s to the growth of low‐wage jobs in consumer services are mistaking the symptom for the cause. The basic change, as they see it, has been a decline in the relative wages of young people, which ‘is simply more visible in consumer services because of the concentration of young workers there’ (ibid. 25). It should be emphasised that not all studies of the changing shape of the distribution of jobs have interpreted the wage as a proxy for skill. Moreover, since there are other influences on wages, it would in principle be possible to have a decline in the share of middle‐wage jobs without a decline in the share of middle‐skill jobs. In any event, there is little or no empirical support for the proposition that increased wage dispersion has been caused mainly by polarisation of the skill distribution of workers or of employment opportunities. Widening skill differentials in wages appear to have been the main or sole cause.

A5.3 Income Inequality

The principal reason for concern about greater wage dispersion is the undesirable social effects of increased income inequality. But there are several intervening linkages between the distribution of wages among full‐time workers of a particular sex (as in Fig. A5.1 and Table A5.1), and the distribution of income among households, which is the usual way of looking at inequality. So although wider skill differentials tend to increase income inequality, the connection between movements in the two sorts of distribution is not necessarily close.

(p.467) The intervening linkages include the gender balance of employment, the female–male wage ratio, and the extent of part‐time work. Also important are variations among households in the number of wage‐earners, and for the distribution of per capita income, variations also in the number of non‐earners (which depends partly on the level and pattern of unemployment). Moreover, there are other sources of household income, both from capital and from the state in the form of pensions and other social security benefits. Finally, some data on house‐hold income distribution are also net of the effects of taxation.

Not all these linkages are independent of changes in skill differentials. For example, a reduction in the relative wages of unskilled males may push their wives out to work. Increased unemployment among unskilled workers may augment the effect of wider wage differentials on income inequality. Greater wage dispersion adds cumulatively, via savings, to inequality in the ownership of (and income from) capital. But most of the intervening linkages are governed by other forces, changes in which may thus offset or amplify the influence of wider skill differentials on income inequality.

Table A5.2 summarises the available information on household income distribution in eleven Northern countries between 1960 and 1990.10 It draws on many different sources, which use a wide variety of concepts and measures of income inequality. Comparing the level of inequality across countries is thus in some cases impossible, and in all cases risky. However, comparisons of the direction of trends in inequality can be made with more confidence, although in some countries this requires looking at two or more overlapping series.

The data strongly suggest that trends in the 1980s were different from those in the two earlier decades, and in a direction consistent with the influence of wider skill differentials. In none of the nine countries for which information is available did household income inequality increase significantly during the 1960s or early 1970s, and in seven it decreased. In the 1980s, by contrast, inequality increased in six out of eleven countries, and in only three did it decrease. There are eight countries whose data span the periods before and after 1980. In four of them there was a change in trend towards greater inequality (in the 1970s in the UK and the USA, and around 1980 in Japan and The Netherlands). In three (France, Germany, and Italy) the former trend of decreasing inequality continued during the 1980s, and in Canada there was no trend in either sub‐period.

The detailed data that would be necessary to prove the connection between wider skill differentials and greater household income inequality—and to ascertain why in some cases there is no such connection—are not readily available for any of these countries. However, other studies cast some light on this issue. Wolleb (1989: 12–18) analyses trends in household income inequality in France, Germany, and Italy—the three countries in which there was a continuing decrease in the 1980s—as well as in the UK. In France, he notes that wage differentials continued to narrow until the mid‐1980s, and argues that this was an important (p.468)

Table A5.2. Changes in Household Income Inequality in the North, 1960–1990

Country

Indicator and coverage

Australia

1980

1985

Gini: gross income among households

0.40

0.42

Canada

1967

1973

1981

1986

1988

Gini: gross income among households

0.40

0.41

0.40

0.40

0.40

France

1962

1970

1975

1979

1985

Gini: gross income among households

0.51

0.44

0.42

0.40

Gini: a‐e expenditure among households

0.33

0.29

Germany

1960

1964

1970

1975

1980

1985

1988

Gini: net income among households

0.38

0.38

0.39

0.37

0.37

0.34

0.33

Italy

1967

1972

1975

1980

1985

1987

Q5/Q1 shares: net income among households

9.58

8.48

7.08

Gini: expenditure among households

0.34

0.35

0.35

0.35

Gini: gross income among households

0.32

0.31

0.30

Japan

1963

1968

1974

1980

1982

1986

1989

Gini: gross income among employee households

0.22

0.20

0.20

0.20

0.21

0.21

Gini: gross income among households

0.36

0.35

0.34

0.33

0.34

Netherlands

1959

1967

1975

1979

1981

1985

Q5/Q1 shares: net income among households

7.96

6.60

Q5/Q1 shares: net income among individuals

4.38

4.02

3.98

Q5/(Q1 + Q2) shares: net income among individuals

2.29

2.98

New Zealand

1980

1985

Gini: gross income among households

0.35

0.35

Sweden

1975

1980

1984

1988

Maxeq%: factor income among economically active households

21.6

22.2

24.4

24.8

Maxeq%: net income among all households

22.9

21.8

23.4

23.7

UK

1959

1965

1970

1977

1979

1985

1988

Gini: gross income among households

0.40

0.39

0.39

0.37

0.37

0.41

Gini: a‐e gross income among households

0.29

0.30

0.32

0.37

USA

1960

1965

1970

1973

1979

1985

1988

Q5/Q1 shares: gross income among households

8.60

7.87

7.57

7.47

8.02

9.57

Gini: gross income among households/individuals

0.41

0.40

0.41

0.43

Definitions: Q5/Q1 shares = income share (or mean income) of top quintile group as ratio of bottom quintile group.

Gini = Gini coefficient of inequality.

Maxeq% = maximum equalisation percentage (larger value indicates more inequality).

Factor income = income from all sources except government transfers.

Gross income = income from all sources (including transfers) before tax.

Net income = income from all sources, after direct taxes.

Expenditure = expenditure out of net income.

Among households = distribution among household or family units.

Among individuals = distribution among individual income recipients.

a‐e = household income or expenditure adjusted to an adult‐equivalent basis.

Sources: Australia, New Zealand: Saunders, Hobbes, and Stott (1989, table 10).

Canada: Wolfson (1989, table 1), updated to 1988 in a personal communication.

France: 1962–70 Sawyer (1976: 27); 1970–9 Wolleb (1989: 41); 1979–85 Eurostat (1990, table 5.6).

Germany: net income series 1960–80 Wolleb (1989: 67); 1985–8 calculated from DIW Wochenbericht (1986 No. 51–2, and 1990 No. 22).

Italy: gross income and expenditure 1975–87 Wolleb (1989: 110, updated in a personal communication); net income 1967–77 Sawyer (1982, table 7.14, figure in 1975 col. refers to 1977).

Japan: employee household series 1962–74 Mizoguchi and Takayama (1984, table 1–4); 1982–9 calculated from table 5 of the Annual Report on the Family Income and Expenditure Survey (current income row); all household series Mizoguchi (1985b, table 8, overall household column).

Netherlands: 1959–67 Sawyer (1982, table 7.14); 1975–85 net income among individuals Statistical Yearbook of The Netherlands (major coverage change 1981–2; Q5/(Q1 + Q2) figure in 1981 col. refers to 1982).

Sweden: Statistika Meddelanden, Income Distribution Survey (1988, tables 19, 20).

United Kingdom: a‐e series Economic Trends (March 1991: 118, table O); gross income series Economic Trends May 1978, July 1984, Nov. 1987 (data in 1970 col. refer to 1970–71, those in 1977 col. to 1975–6, those in 1978 col. to 1978–9, and those in 1985 col. to 1984–5).

United States: households (families) Mishel and Simon (1988, table 14), Bureau of the Census P‐60 Reports (No. 97 table 22 and No. 166 table 5); households/individuals (families plus unrelated individuals) Blackburn and Bloom (1987, table 1).

(p.469) (p.470) contributor to reduced income inequality, as were increases in social security benefits and tax progressivity, while demographic forces had mixed effects.

In Germany, Wolleb recognises that rising wage dispersion after 1978 tended to increase household income inequality. But this was more than offset by a rise in the proportion of households with more than one wage‐earner (a group within which there is a relatively low degree of income inequality), so that there was on balance a further reduction of inequality. In Italy, too, income inequality did not increase despite widening wage differentials after the early 1980s, but it is less clear what changes in the structure and behaviour of households broke the connection. In the UK, the marked widening of skill differentials surely contributed to the sharp increase in household income inequality during the 1980s, which was exacerbated by regressive tax and social security changes.

Widening wage differentials must also have contributed to rising household income inequality in the USA, though once again the connection is not well documented (Freeman 1987: 43; Mishel and Simon 1988). Blackburn and Bloom (1987: 37) find that increased wage dispersion among male principal earners was one cause of greater household income inequality, but put more emphasis on other influences. However, Danziger and Gottschalk (1989) note the insufficiency of several possible explanations of increased inequality, including cyclical forces, cohort size, and the extent of female headship of households. In Canada, where wage dispersion among full‐time workers increased after 1970, the stability of income inequality is the net outcome of various conflicting economic and demographic pressures (Wolfson 1986a, 1989). The level of household income inequality in Japan is disputed, but all sources agree that it increased after the late 1970s, and there is some evidence that wider wage differentials contributed to this shift (Ishizaki 1985–6; Ozawa 1985–6).

Notes:

(1) e.g. training or exit of unskilled workers might be induced by widening of skill differentials, and might thus offset a tendency towards greater wage dispersion. Alternatively, such training or exit might be autonomous, and hence cause skill differentials to narrow, which would further diminish the dispersion of wages.

(2) The Davis series for The Netherlands have the advantage of being limited to males, but exclude the top 5% of earners (Davis 1992, table 1).

(3) Katz and Loveman (1990) note an important difference between UK and US experience. In the USA, the increase in wage dispersion has been associated with a marked absolute decline in the real wages of the lowest decile, while in the UK the real wages even of this decile rose in the 1980s.

(4) However, Dooley (1987, table 6) finds only a slight tendency for wage dispersion within education and age groups to increase during the 1970s among full‐time Canadian males.

(5) Increased labour‐market mobility due to the abolition or weakening of union or legislative restrictions on entry and exit may have been another institutional feature of the 1980s, but this should presumably have tended to narrow wage differentials rather than to widen them.

(6) This distinction is particularly well explained by Adams, Maybury, and Smith (1988), who note that the disappearance of workers from the middle of the wage distribution would be sufficient to increase inter‐quantile wage ratios even if there were no change in the wages of those who remained in the distribution. This is because the absolute number of workers in the (say) upper quarter of the distribution would decline, and hence the wage threshold of this quarter (the quartile) would rise.

(7) Subsequently, Wolfson and Foster (1991) have devised a different measure of polarisation, which can be conceptually related to the Lorenz curve.

(8) Inter‐quantile ratios, as Wolfson (1989) notes, are poor measures of inequality in a strict sense, since they omit a lot of information and are not necessarily consistent with the usual minimum criterion of ranking by Lorenz curves. For this reason, these ratios may not move in the same direction as more formal measures of inequality. Fig. 1 of Wolfson (1989) also happens to provide an exception to the proposition that inter‐quantile ratios will usually move in the same direction as measures of polarisation. This hypothetical counter‐example, in which polarisation increases but the inter‐quartile ratio does not alter, depends on rather unusual assumptions, namely bimodality and movement of some people towards as well as away from the mean. But such contradictions could arise even with more usual distributions, e.g. if the quantiles were extreme while the income range used to define the tails was narrow.

(9) However, data from another source suggest that the skill distributions of blue‐collar and lower white‐collar jobs may be bimodal (Myles 1987, table 11).

(10) The table does not include all the available data. Other series which simply confirmed the trends shown in the table were omitted. Additional information on trends in relative poverty in the early 1980s in eleven countries is provided in Eurostat (1990, tables 4.4 and B7): for France, Germany, The Netherlands, and the UK, it confirms the trends shown in Table A5.2; but for Italy it suggests that inequality may have increased. In Belgium, relative poverty decreased, and in Denmark it remained unchanged.