Are Incomes Becoming More Unequal? Are Real Incomes Increasing?


Current Issues Brief 15 1999-2000

Are Incomes Becoming More Unequal?
Are Real Incomes Increasing?

Geoff Winter
Statistics Group
20 June 2000

Contents

Major Issues
Introduction
The Overall Picture

The Gini Co-efficient
Real Incomes
Quantile Analysis

Selected Income Unit Groups

Income Shares
Real Incomes of Selected Income Units
Other Measures of Dispersion
Other Comparisons

Conclusion
Endnotes
Appendix

Table 1. All Income Units: Type of Income Unit and 'Gini Co-efficient'
Table 2. All Income Units: Type of Income Unit and Total Real Mean Income
Table 3. All Income Units: Income Shares by Quintile
Table 4. Couples with Dependent Children: Income Shares by Quintile
Table 5. One-Parent Families: Income Shares by Quintile
Table 6. Lowest (First) Quintile: Mean Incomes
Table 7. Couples with Dependent Children: Measures of Income Dispersion
Table 8. One-Parent Families: Measures of Income Dispersion

Major Issues

It has been the popular perception, and maybe the reality, for some time now that there is a widening gap between the rich and the poor in Australia. Whilst this is almost certainly true when considering earnings it seems by no means to be the case automatically for incomes.

The widely used Gini Co-efficients have tended to indicate increasing inequalities in total incomes from 1981-82 to 1997-98 for all income units combined in Australia and for each broad group of income units separately, except for one-parent families. But movements from the mid 1990s are far less conclusive, when there may not have been much change at all in inequality, except amongst one-parent families where it may have increased slightly.

Movements in real incomes however seem to be much more definite-real incomes have increased overall and for each main group of income units, both over the period 1981-82 to 1997-98 and from the mid-1990s.

The shares of total incomes going to various quintile groups overall tend to confirm what the Gini Co-efficients have shown: the share going to the lowest income quintile appears to have decreased significantly while the share going to the highest income quintile has increased substantially; this indicates a significant increase in income inequality. However from the mid-1990s there appears to have been little change. If the future of a nation is in its children then perhaps the most critical groups in society are income units with dependent children. These movements have been reflected amongst couples with dependent children, but not amongst one-parent families which have shown far more volatile changes.

Examination of the lowest quintile for these critical groups indicates that real incomes have increased generally, but by 1997-98 may have started to stabilise or even decrease; for all income units combined there appears to have been much smaller if any gains overall.

Other measures of dispersion have tended to provide conflicting evidence, but in comparisons with the average weekly earnings figures, a traditional benchmark in Australia, income units with dependent children may have improved their situation despite apparent increases in inequality, although even this 'evidence' is by no means conclusive given the changing nature and composition of Australian families.

This type of analysis inevitably leads to questions about the levels of poverty in the community. Using different measures, some international studies seem to suggest that the extent of poverty has been decreasing in Australia, although some studies undertaken in Australia do not necessarily agree with them. But there is no doubt that the social security system has had a dramatic impact on reducing the extent of 'market income' poverty. Despite apparent increasing inequalities, real incomes of those on the lowest incomes appear to have been increasing, not only because there has been a system of social security transfers but also because the actual social security payment rates were generally linked to changes in community earnings and prices, policies which it is imperative to maintain if poverty reduction is to continue.

Introduction

One of the continuing concerns of governments, and others, in western democracies, and one of the driving forces of social policy is the distribution of incomes amongst the population. With 'globalisation' and increasing deregulation and privatisation taking place in Australia, the question of income distribution has re-emerged as a prominent issue. Coupled with this is the question of whether real incomes are increasing, especially for those on the lowest incomes, as increasing real incomes can mollify or 'offset' the effects of growing inequality for these groups. Generally, increasing income inequality is seen as undesirable as it may impair social cohesion; increasing real incomes are considered to be desirable. Increasing real incomes for any particular group results in more purchasing power, and hence a higher (material) standard of living, and less poverty if it is occurring amongst those with the lowest incomes. Thus achieving increasing real incomes, especially for the lowest income groups, could perhaps be a higher priority than being concerned about increasing income inequality, especially if they are not directly related. However, because increasing income inequality leads to an increasing proportion of the total wealth accruing to those with the highest incomes it remains a matter of concern.

This paper looks at changes in income distributions from 1981-82 to 1997-98, the latest period for which data is available. Data which will include quantification of the outcomes of recent changes under the jurisdiction of federal industrial relations have not yet become available, as the full impact of those changes has yet to work its way through the economy; similarly for changes in the social security payments system. It should be noted therefore that the analysis below is somewhat incomplete; a much more representative picture will emerge in the next two to three years. The data analysed in this paper are total gross incomes from all sources. They therefore include earnings and income from investments and property, all these combined being known as 'market income', and government transfer payments. These incomes have not been adjusted for any taxes paid, nor for any concessions or rebates received. They have also not been adjusted for the numbers of members within the individual income units which have been grouped for this analysis, nor for the changing proportions of the total numbers on income units comprising each group, which may affect income distributions and comparisons. It should also be clearly noted that this paper does not analyse wealth or changes in wealth in any way.

The paper concentrates on analysing published data from the frequent surveys of income distribution conducted by the Australian Bureau of Statistics (ABS) as these are the longest-running consistent source of data for Australia. Data from these surveys which have been published for periods shorter than the span from 1981-82 to 1997-98, and data from other sources,(1) have not been considered in this analysis. Thus the conclusions may differ from those which could have been reached had unpublished and additional data been considered. It should also be noted that these data are from sample surveys, which means that they are subject to sampling error, i.e. the estimates may be different from the figures which would have resulted if the whole population had been surveyed. From 1994-95 the sample sizes have been much smaller than in the earlier surveys, which may have affected sampling variability and therefore the accuracy of the comparisons.

The paper looks at indicators of overall inequality and compares changes in inequality with changes in real incomes, first using the Gini Co-efficient and then quantile analysis. The discussion moves on to examine particular groups within society, again first using quantile analysis then looking at real incomes. The paper then considers other prominent measures of dispersion for critical groups and concludes with a mention of further measures which are not analysed.

Because the scope of the 1986 and 1990 surveys was different from that of the other surveys, in that income units with zero and negative incomes were excluded from those two surveys, estimates from those two surveys are not strictly comparable with those from the other surveys. It should also be noted that estimates from the 1981-82 survey are based on total annual incomes from all sources while estimates from all of the other surveys are based on current (weekly) incomes which may include some items such as income from property, e.g. interest and dividends, averaged out from the annual income of the previous year from these sources.

The Overall Picture

Analyses of income distributions are complex and technical. Measuring income inequality based on these analyses is very difficult to encapsulate and understand readily. Substantial increases in the incomes of a small group of people do not automatically cause an increase in income inequality overall, even if they then receive an even greater multiple of, say, the 'average' than previously. The apparent huge increases in executive salaries may well have led to increasing inequalities in earnings, but the resultant widening gap may well have been tempered by government transfer payments, especially where they are tied to measures which include those salaries, such as average weekly earnings (AWE). The 'average', or mean, and the median are measures of 'central tendency' within a distribution, which of themselves are not indicators of inequality (but do show levels and movements in levels). However there are measures of overall dispersion in income distributions which of themselves give an immediate indication of inequality.

The Gini Co-efficient

The Gini Co-efficient (G) is the most widely used and accepted measure of overall income dispersion internationally. (A detailed explanation of how G is derived is given in the Appendix.) It should be noted however that G never indicates the level of nominal or real incomes. It is possible therefore that members of a group or population could have very equal but low incomes, which do not give them much purchasing power, or could have unequal but much higher incomes, where most if not all members have a lot more purchasing power. Thus comparison of Gs between different groups does not indicate relative incomes between them, only the degree of inequality in incomes within each group.

Table 1 below shows the Gs for various types of income units in Australia over recent years. The Australian Bureau of Statistics (ABS) defines an 'income unit' as 'one person or a group of related persons within a household whose command over income is assumed to be shared'.(2) G for all income units combined appears to have increased during the 1980s and stabilised somewhat during the 1990s, meaning that income inequality was greater overall in the 1990s than earlier, but had probably not changed very much from the mid 1990s. The movements for each type of income unit appear to be more volatile, although for all types except the one-parent group the level of inequality appears to be lower at the start of the series in 1981-82 than it is at the end, 1997-98. Thus income inequality has increased for these groups.

Table 1. All Income Units: Type of Income Unit and 'Gini Co-efficient'

Type of income unit

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Couple -

 

 

 

 

 

 

 

With dependent children

0.283

0.31

0.31

0.34

0.33

0.32

0.337

Without dependent children

0.365

0.38

0.34

0.38

0.41

0.41

0.402

Total

(a)0.335

0.35

(a)0.33

0.37

0.38

0.37

0.377

One parent

0.350

0.32

0.30

0.29

0.32

0.28

0.326

One person

0.376

0.37

0.38

0.41

0.41

0.41

0.424

All income units

0.400

0.41

0.42

0.443

0.437

0.444

0.446

(a) Estimate.

Source: Income Distribution, Australia, various (ABS 6523.0).

Apparently small movements in G can indicate significant changes in income inequality. For example, the change from 0.365 in 1981-82 to 0.402 in 1997-98 for couples without dependent children is a movement of 10.1 per cent [(0.402-0.365)/0.365] further 'away' from 0 and 'closer' towards 1 and indicates therefore a significant increase in the inequality of incomes within this group. Incomes within the couples with dependent children group have become even more unequal-

(0.337-0.283)/0.283 = 19.1 per cent further 'away' from 0.

From these data it can be concluded that generally incomes had become more unequal by 1997-98 than they were in 1981-82.

Real Incomes

Movements in real incomes indicate changes in purchasing power, because the movements in nominal (actual) incomes are discounted for inflation. It is possible for those with the lowest incomes in a particular group to have an improving standard of living as their real incomes increase even though the inequality of incomes within the whole group is increasing. This usually occurs when the overall size of the 'cake', the economy, is growing but only a small proportion of the increase goes to those with the lowest incomes.

Inequalities of incomes indicate the relative differences within groups. Increases in inequalities of incomes do not automatically mean increases in the numbers of people in poverty (but that would certainly occur if these increases were accompanied by or were the result of falling real incomes for those at the lower end of the income scale). Increasing inequality does not mean that those at the lower end of the distribution are getting poorer in absolute or real terms, although it often does mean that. An example, which may assist in explaining and understanding this, is shown in the box below-Increasing Inequality and Increasing Real Incomes.

Table 2 below shows the movements in real mean incomes, i.e. nominal mean incomes adjusted for inflation as measured by the Consumer Price Index (CPI), for various types of income units in Australia over recent years. The table shows that real mean incomes overall and for each type of income unit had increased generally over the whole period, and were certainly higher in 1997-98 than they were in 1981-82, i.e. purchasing power had increased for all groups generally over the period.

Table 2. All Income Units: Type of Income Unit and Total Real Mean Income(a)
($ per week)

Type of income unit

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Couple-

 

 

 

 

 

 

 

With dependent children

454

464

480

503

496

504

538

Without dependent children

376

371

398

386

390

392

399

All couples

419

419

440

439

441

446

467

One parent

181

190

199

205

220

217

232

One person

186

199

196

188

188

196

206

All income units

303

310

320

307

309

313

330

(a) Mean income in December quarter 1982 prices adjusted by the CPI, eight capital cities, all groups, base: 1989-90 = 100.0.

Source Income Distribution, Australia, various (ABS 6523.0),

Consumer Price Index, Australia, various (ABS 6401.0).

Tables 1 and 2 show similar trends. Mean incomes had increased in value but incomes had become more unequal for all income units overall and within all groups except one-parent families, which seemed to have achieved both of the more desirable outcomes of increasing real (mean) incomes and decreasing inequality. However, this does not automatically mean that those with the lowest incomes in any or all groups had experienced increasing real incomes: see Real incomes of selected income units below.

Increasing Inequality and Increasing Real Incomes

In Year 1, family A, a couple with two dependent children where the husband/father works as an executive and the wife/mother is a full-time home-maker, has a gross annual income of $100 000. Family B, also a couple with two dependent children has no earner in the family and has a gross annual income of $20 000.

In Year 2, family A has had an increase in income. The husband/father has had a substantial increase in salary and the wife/mother no longer is a full-time home-maker but has gone into the paid labour force and also earns a full-time income. The family now has a gross income of $300 000. They employ a member of family B to do all the home-making tasks performed by the wife/mother in year 1 and other duties. Family B now has an earner and its gross income, after the reduction in social security payments, is now $40 000.

In year 1, family A had 5 times the income of family B; in year 2 family A had 7.5 times the income of family B. Both families have had a substantial increase in nominal incomes and, clearly, unless the inflation rate is 100 per cent or more, both families have had an increase in real incomes as well. Thus there has been a marked increase in income inequality while real incomes have also increased (substantially if usual rates of inflation occur) for both families.

Note that this is an example only for illustrative purposes-it is not intended to indicate that a situation such as this is common. However, this kind of development does result in increasing labour force participation rates which if sustained over long periods, as has happened in Australia over the last 30 years, is generally considered to be an indicator of a more robust, more highly developed, expanding and therefore richer economy. Also note that all of this deals only with personal/private individual/family money incomes and says nothing about other changes in the economy which may have a causal relationship with changes in real incomes, such as changes in taxation and movements in the 'social wage'.

Quantile Analysis

As for Gini Co-efficient analysis, populations or groups can be divided up into equal proportionate parts, ranking from lowest to highest in order, and calculating the share of the total income received by each. These parts are called quantiles, and the most common forms used are: percentiles-each one per cent of a population from 0 to 100 per cent; deciles-each 10 per cent; quintiles-each 20 per cent; and quartiles-each 25 per cent. (The Gini Co-efficient is a derivative of quantile analysis, summarising information about income shares for very small quantiles, usually percentiles or smaller quantiles.) Perhaps the most commonly used quantile analysis nowadays examines the income shares of quintiles-each 20 per cent of a population or group, particularly comparing the 20 per cent with the lowest incomes to the 20 per cent with the highest incomes. Table 3 below shows income shares by quintile group for all income units combined.

Table 3. All Income Units: Income Shares by Quintile
(per cent)

Quintile

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Lowest (first)

4.6

4.8

4.8

3.6

3.8

3.8

3.8

Second

9.8

9.7

9.7

9.3

9.1

9.4

9.0

Third

16.6

15.9

15.5

15.2

15.0

15.2

15.0

Fourth

24.8

24.4

23.9

24.0

23.7

24.0

23.9

Highest (fifth)

44.2

45.3

46.2

47.9

48.3

47.5

48.3

(Gini Co-efficient

0.400

0.41

0.42

0.443

0.437

0.444

0.446)

Source: Income Distribution, Australia, various (ABS 6523.0).

The figures in Table 3 clearly indicate that generally there has been a redistribution of incomes upwards, i.e. to the highest income group from all of the other groups, over the whole period from 1981-82 to 1997-98. This confirms what the Gini Co-efficient shows. Because the numbers in the distribution are so large-over 6.6 million income units in 1981-82, increasing to over 9.1 million in 1997-98, as these surveys represent all persons aged 15 years and over resident in Australia-apparently small movements in the income shares are significant. Thus the decrease in the share of the lowest income group, from 4.6 per cent of total income in 1981-82 to 3.8 per cent of total income in 1997-98, would mean increased economic hardship for many people, unless their real incomes had increased. If this had happened it would mean that the increase in inequality of incomes was occurring because real incomes for the other groups had increased even more. Exactly which people are in this situation can be determined by further analysis of income shares, as discussed in the next section of this paper.

Selected Income Unit Groups

The changes in the distribution of income shares shown in Table 3 may not be common to all types of income units. If the future of the nation is in its children then perhaps the most important or critical groups in society are income units with dependent children. If the income shares in these groups of income units have not declined and/or their real incomes have increased, then it could be said that they have been 'protected' from the undesirable effects of a redistribution of incomes upwards.

Income Shares

The income shares for couples with dependent children (Table 4) clearly show that over the period being analysed there has been an erratic but marked decline in the share of total income going to the lowest quintile in this group, from 8.1 per cent in 1981-82 to 6.7 per cent in 1997-98. This decline has been accompanied by an erratic but significant increase going to the highest quintile, from 36.5 per cent to 40.3 per cent. These movements are generally reflected in the Gini Co-efficients over the period-falls in income shares going to the lowest quintile with comcomitant rises in income shares going to the highest quintile show up as increases in the Gini Co-efficient.

Table 4. Couples with Dependent Children: Income Shares by Quintile
(per cent)

Quintile

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Lowest (first)

8.1

7.3

7.4

6.3

6.5

7.0

6.7

Second

14.2

13.3

13.1

12.8

12.6

13.0

12.6

Third

18.2

17.9

17.9

17.5

17.5

17.6

17.4

Fourth

23.1

23.2

23.5

23.1

23.6

23.6

23.0

Highest (fifth)

36.5

38.4

38.1

40.3

39.9

38.8

40.3

(Gini Co-efficient

0.283

0.31

0.31

0.34

0.33

0.32

0.337)

Source: Income Distribution, Australia, various (ABS 6523.0).

On these measures the increase in income inequality for all income units is clearly replicated in this group of income units.

The income shares for one-parent families (Table 5) show a rather more volatile trend than for couples with dependent children (Table 4).

Table 5. One-Parent Families: Income Shares by Quintile
(per cent)

Quintile

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Lowest (first)

7.4

8.5

9.3

8.5

8.4

9.8

8.3

Second

11.9

13.1

13.2

13.2

12.6

13.3

12.4

Third

14.8

15.1

15.8

17.5

16.3

16.4

15.8

Fourth

23.0

22.1

22.8

23.3

23.2

22.1

22.4

Highest (fifth)

42.8

41.5

38.9

37.5

39.6

38.4

41.2

(Gini Co-efficient

0.350

0.32

0.30

0.29

0.32

0.28

0.326)

Source: Income Distribution, Australia, various (ABS 6523.0).

It is apparent that there was a significant decrease in income inequality during the 1980s, which may have extended to 1994-95, although this is not reflected in the income share of the lowest quintile, which decreased from 9.3 per cent in 1990 to 8.5 per cent in 1994-95. Thereafter it seems that there have been significant swings in income inequality: the income share of the lowest quintile moved from 8.4 per cent in 1995-96 to 9.8 per cent in 1996-97 and down to 8.3 per cent in 1997-98 while the income share of the highest quintile moved from 39.6 per cent to 38.4 per cent and up again, to 41.2 per cent, over the same years. These movements may reflect the entries/re-entries into and exits from the paid work force which are often more spasmodic for single parents than for other people-moving on to or off the single parenting payment (and its pension predecessors) can make a substantial difference to incomes.

Despite the erratic trends in this series, on these measures it does seem that income inequality has decreased generally since 1981-82 for these income units, but the trend may be reversing if the 1997-98 figures are any guide.

Real Incomes of Selected Income Units

As noted earlier, a higher Gini Co-efficient, or a low-income share for the lowest quintile, does not automatically mean greater levels of poverty (especially absolute rather than relative poverty), although it often does. A situation where a small minority of people have relatively low (real) incomes while the remaining substantial majority have relatively higher comfortable (real) incomes (which gives a higher Gini Co-efficient) is arguably better than a situation where a substantially greater proportion have the same relatively low (real) incomes and only a few have higher comfortable (real) incomes (which gives a lower Gini Co-efficient). It is far easier to do something to alleviate the disadvantage of a small group than it would be to do something similar for a much larger group, political will notwithstanding. If real incomes are increasing for those in the lowest quintile then any absolute poverty they may have been experiencing is decreasing, even if their disadvantage relative to higher income groups is increasing. Table 6 below shows mean incomes for the lowest quintiles amongst couples with dependent children and one-parent families.

Table 6. Lowest (First) Quintile: Mean Incomes
($ per week)

Type of income unit

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Couples with dependent children

 

 

 

 

 

 

 

-nominal mean income

(a)156

225

310

299

316

354

357

-real mean income (b)

(a)156

170

176

158

161

177

179

One-parent families

 

 

 

 

 

 

 

-nominal mean income

(a)68

109

161

168

182

212

192

-real mean income (b)

(a)68

82

92

89

92

106

96

All income units

 

 

 

 

 

 

 

-nominal mean income

(a)57

99

134

96

117

121

124

-real mean income (b)

(a)57

75

76

51

59

61

62

(a) Estimate. (b) Nominal mean income in December quarter 1982 prices adjusted by the CPI, eight capital cities, all groups, base: 1989-90 = 100.0.

Source: Income Distribution, Australia, various (ABS 6523.0),
Consumer Price Index, Australia, various (ABS 6401.0).

It can be seen in Table 6 that real mean incomes rose significantly for the two types of income units with dependent children during the 1980s. One-parent families generally retained this increase throughout the 1990s but couples with dependent children regained their position only towards the end of the 1990s. Thus it would appear that, in general, the increasing disparity in incomes for all couples with dependent children (shown in Table 4) did not result in an increase in poverty amongst them. The decreasing disparity in incomes for one-parent income units (shown in Table 5) was accompanied by an increase in real incomes, probably a far more important change for a group with relatively low incomes than the inequality within it, so it is likely that the experience of poverty amongst this group also did not increase.

Other Measures of Dispersion

Two other measures of dispersion which may be considered to be indicative measures of income levels are: (i) comparisons with AWE, which is commonly used as a benchmark, and (ii) comparisons with the mean (average) and median within each income distribution itself. AWE is a measure of the total earnings of all individual employees (combined), i.e. those who are employed by either other individuals or by corporate bodies both private and public. Thus it is not a measure of family or group earnings or total incomes, i.e. incomes which include such items as rent, interest and dividends and government transfer payments. As such it can be described as a measure 'independent' of the income distributions of any income unit being analysed in this paper. Lots of small movements in the earnings of many people generally have a greater impact on AWE, but if say 100 people have an increase of $1 million each over a year, not an entirely unusual event in this era of individual bargaining and deregulation, the mean (average) income for the whole working population of Australia goes up by $12-13 in that year. Although this sort of amount may seem small, there have been many similar movements each year, which also add up to significant amounts. When so many transfer payments and concessions are tied to benchmarks such as AWE these movements can have a significant impact on costs and expenditures such as government outlays. Table 7 below compares means and medians for couples with dependent children and AWE.

This table provides several measures of income inequality. Looking first at the median and the mean (the average), the closer the median-the figure which divides the population into halves so that half of the population has incomes below the figure and half have incomes above it-is to the mean the more equal the distribution. The trend in Table 7 shows that the median is in fact becoming a generally smaller proportion of the mean, moving from 89.4 per cent in 1981-82 to 86.4 per cent in 1997-98, indicating that the distribution is becoming more unequal. This trend is 'confirmed' when comparing incomes to various proportions of the mean. As the table shows:

  • the proportion with incomes less than 100 per cent of the mean rose from 58.8 per cent in 1981-82 to 60.1 per cent in 1997-98
  • the proportion with incomes less than 75 per cent of the mean increased from 37.0 per cent to 40.4 per cent, and
  • the proportion with incomes less than 50 per cent of the mean moved from 16.3 per cent to 18.8 per cent.

Table 7. Couples with Dependent Children: Measures of Income Dispersion

Measure

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Nominal median ($ per week)

(a)406

546

755

829

849

882

928

Real median ($ per week)(b)

406

413

429

439

431

442

465

Nominal mean ($ per week)

(a)454

614

844

950

976

1006

1074

Real mean ($ per week)(b)

454

464

480

503

496

504

538

Median as proportion of mean (%)

89.4

88.9

89.5

87.3

87.0

87.7

86.4

Median as proportion of AWE (c)(%)

120.5

122.4

130.6

128.5

127.9

129.3

131.0

Mean as proportion of AWE (c)(%)

134.7

137.7

146.0

147.3

147.0

147.5

151.6

Couples with income-

 

 

 

 

 

 

 

Less than 100 % of AWE (c)(%)

36.1

29.1

32.2

33.1

34.9

32.1

31.8

Less than 75 % of AWE (c)(%)

19.9

19.0

17.8

19.0

19.2

18.1

18.4

Less than 50 % of AWE (c)(%)

10.1

8.4

6.7

7.9

7.8

6.8

7.2

Less than 100 % of mean (%)

58.8

60.0

58.4

60.0

59.4

58.7

60.1

Less than 75 % of mean (%)

37.0

36.7

37.7

39.4

39.9

38.7

40.4

Less than 50 % of mean (%)

16.3

15.9

16.7

18.4

18.4

17.5

18.8

(a) Estimate. (b) December quarter 1982 prices, adjusted using the CPI, all groups, eight capital cities, base: 1989-90 = 100.0. (c) All Males Total Average Weekly Earnings.

Source: Income Distribution, Australia, various (ABS 6523.0),
Average Weekly Earnings, various (ABS 6302.0),
Consumer Price Index, Australia, various (ABS 6401.0).

Thus, clearly all of these measures increased over the period shown. When an increasing proportion within a distribution falls below the mean, the distribution is becoming more 'skewed'. This happens when a very small number within a group gains very large increases. Australia has been a classic case of this phenomenon, as senior executive salaries have increased out of all proportion with other employees' wages and salaries over the last 10 years or so.(3) Even though this group may be relatively small in number these increases do have an impact on both the mean of any distribution which includes them and on the AWE figure which also includes them as they are still employees.

Comparisons with AWE, much lower than the mean for this distribution, however, apparently indicate a different trend. Table 7 shows:

  • the proportion with incomes less than 100 per cent of AWE declined from 36.1 per cent in 1981-82 to 31.8 per cent in 1997-98,
  • the proportion with incomes less than 75 per cent decreased from 19.9 per cent to 18.4 per cent, and
  • the proportion with incomes less than 50 per cent dropped from 10.1 per cent to 7.2 per cent.

Thus these proportions of AWE have all generally declined,(4) indicating less inequality. The different movements in the two series would indicate that an increasing proportion of these couples have incomes between AWE, an individual earnings figure independent of the distribution, and the mean of the distribution itself. This most likely reflects the fact that an increasing number of these income units contain more than one earner. If the AWE figure is a benchmark of income needed to live in some degree of at least moderate comfort, then the decreasing proportions with incomes below various proportions of AWE could be considered to be more critical than the increasing proportions with incomes below various proportions of the mean.(5) If also the families with the lowest incomes are those with less dependent children, because the parents are younger and have not completed their families-not an unreasonable expectation in many cases given the structure of government family payments and taxation-then the widening of income inequalities may not necessarily be leading to increasing poverty. Countering this argument however is the likelihood that young families generally have lower incomes because their earners generally earn less from their jobs as they are relatively inexperienced and perhaps have lower qualifications. Coming at a time when they most need financial support when raising young children may in fact indicate increasing poverty if incomes are becoming more unequal.

The main conclusion to draw for this group therefore is that incomes are probably growing more unequal in the technical statistical sense but movements in the actual standard of living of people at the bottom of the income scale are more uncertain, especially when financial needs are taken into account.

Table 8 below compares the same measures for one-parent families. By definition these families include dependent children-families including non-dependent children only are not included in this analysis.(6) It can be seen in this table that the median income as a proportion of the mean rose substantially up to 1994-95 and, although declining since then, remained significantly above the level in 1981-82. Despite the erratic movements in this measure the trend indicates decreasing inequality within this group. However, looking at the proportions with incomes less than 100 per cent, 75 per cent and 50 per cent of the mean, the picture is not so clear. As the mean for this group is low compared to other families with dependent children, and substantially lower than the 'independent' AWE figure, decreasing inequality would generally require that these proportions should fall over time. Again, up to 1994-95 it seems this was the case, but the figures since then are by no means conclusive. Perhaps the most that can be said is that the proportions with incomes less than 100 per cent and less than 75 per cent of the mean were apparently still lower in 1997-98 than they were in 1981-82, indicating less inequality, but for the lowest income sub-group-those with incomes less than 50 per cent of the mean-the movement from 1996-97 to 1997-98 indicates a marked deterioration in their position and a marked increase in inequality, both recently and over the whole period. The divergent trend for this group is probably a reflection of the employment/single parenting payment behaviour noted under Income shares on pages 7 and 8 above.

Table 8. One-Parent Families: Measures of Income Dispersion

 

1981-82

1986

1990

1994-95

1995-96

1996-97

1997-98

Nominal median ($ per week)

133

190

278

340

352

354

362

Real median ($ per week)(a)

133

144

158

180

179

177

181

Nominal mean ($ per week)

181

251

350

388

433

432

463

Real mean ($ per week)(a)

181

190

199

205

220

217

232

Median as proportion of mean (%)

73.5

75.7

79.4

87.9

81.3

81.9

78.2

Median as proportion of AWE (b)(%)

39.5

42.6

48.1

52.7

53.0

51.9

51.1

Mean as proportion of AWE (b)(%)

53.7

56.1

60.6

60.2

65.2

63.3

65.4

One-parent families with income-

 

 

 

 

 

 

 

Less than 100 % of AWE (b)(%)

87.4

90.3

87.8

95.1

85.6

86.9

84.4

Less than 75 % of AWE (b)(%)

76.9

78.1

73.3

74.3

72.2

73.7

70.8

Less than 50 % of AWE (b)(%)

62.7

59.4

51.2

46.2

44.6

45.0

45.5

Less than 100 % of mean (%)

65.6

66.8

62.3

58.8

62.5

63.9

63.8

Less than 75 % of mean (%)

52.3

45.1

41.5

39.2

43.3

40.8

43.9

Less than 50 % of mean (%)

13.0

15.8

14.6

10.4

13.8

9.8

14.8

(a) December quarter 1982 prices, adjusted using the CPI, all groups, eight capital cities, base: 1989-90 = 100.0. (b) All Males Total Average Weekly Earnings.

Source: Income Distribution, Australia, various (ABS 6523.0),
Average Weekly Earnings, various (ABS 6302.0),
Consumer Price Index, Australia, various (ABS 6401.0).

Thus, overall on comparisons with the mean, up to 1994-95 there were significant movements generally indicating improvements for the lower income groups, and thus less inequality, followed by an erosion of that position to some extent afterwards.

Turning to comparisons with the AWE figure, much higher than the mean for this group:

  • the proportion with incomes less than 100 per cent of AWE declined from 87.4 per cent in 1981-82 to 84.4 per cent in 1997-98
  • the proportion with incomes less than 75 per cent went down from 76.9 per cent to 70.8 per cent, and
  • the proportion with incomes less than 50 per cent decreased from 62.7 per cent to 45.5 per cent.

Thus proportions overall declined from 1981-82 to 1997-98, significantly for the proportions with less than 75 per cent and less than 50 per cent of AWE, and the situation has not deteriorated since 1994-95 to anywhere near the degree it has on other measures. This indicates that mean incomes for one-parent income units have risen faster than AWE. On AWE measures therefore it could be claimed that inequality amongst one-parent families has decreased over the period.

In summary, it could be concluded therefore for the group of one-parent income units that income inequality has probably decreased generally. But for those on the very lowest incomes, less than 50 per cent of the mean of the whole group, the trend is indeterminate. However, the mean real incomes for the lowest quintile (see Table 6) have increased as a proportion of the real mean income for the whole group (as indicated in Table 8), from 37.6 per cent in 1981-82 to 41.4 per cent in 1997-98. This would tend to confirm that any lessening of inequality has been accompanied by increasing real income for this group.

Other Comparisons

Among other measures of dispersion which could be used to examine changes in income inequalities are movements in disposable incomes and equivalence scales.(7) The Organisation for Economic Development and Co-operation (OECD) compiles equivalence scales and in Australia the 'Henderson' equivalence scales developed for the Australian Government Commission of Inquiry into Poverty in 1975 have been published from 1994-95. In a separate study conducted by the OECD,(8) for the period 1975-76 to 1993-94 there appears to be conclusive evidence that inequalities in incomes had increased in Australia. The Gini Index (Co-efficient) had increased overall, and the proportion of disposable income going to the three lowest deciles had declined, but the proportion of the population in poverty had fallen over the period, particularly from 1984.(9) It should be noted that this OECD study uses 'equivalent disposable income per household member' in its analysis, not total gross incomes discussed in this paper. The OECD approach therefore does adjust for any taxes paid, and does include government transfer payments to the various income units. It also adjusts for the numbers of members of the individual income units within the different groups of income units, which this paper does not.

Another international study(10) shows how social security transfers have increased real incomes for those on the lowest incomes and thereby reduced poverty. It estimates that in 1981 the poverty rate in Australia was reduced by 56 per cent, from 21.2 per cent of the population on a 'market income' basis, i.e. income from earnings and investments, to 9.3 per cent, because of social security transfers; in 1994 the poverty rate was reduced by 63 per cent, from 21.5 per cent of the population on a market income basis to 8.0 per cent similarly. These figures would tend to confirm that real incomes for the lowest quintile of the total population are increasing. However there have been some studies undertaken in Australia(11) which claim that poverty has not been reduced over similar periods.

Conclusion

Changes in income inequality are not easy to quantify. Popular perceptions that inequality may be increasing, because for example there have been massive increases in senior executive salaries, are not automatically accurate. However analyses of the measures of income dispersion available from 1981-82 and used in this paper appear to indicate that income inequality has increased overall and for most groups in Australia since then. This situation may have been 'offset' by targeting of government social security transfer payments, which appear to have ensured that real incomes, especially of those with the lowest incomes, have generally been at least maintained. Thus on the data analysed the apparent increases in income inequalities have been mollified to a sufficient degree that it is likely that the extent of poverty has not increased.

As Australia's social security payments are substantially based on movements in prices and AWE, and there is no doubt that social security transfers have a significant effect on real incomes for those on the lowest incomes, it therefore appears to be critical that a social security 'safety net' based on movements in general community earnings/incomes and prices be maintained if real incomes for those with the lowest incomes are to be maintained and increase.

Endnotes

  1. Such as the various household income and expenditure surveys also conducted by the ABS.

  2. ABS, Income Distribution, Australia, Catalogue No. 6523.0, various.

  3. See Tony Kryger, 'Private Executive Salaries', Research Note No. 24, 1998-99, Department of the Parliamentary Library, 1999.

  4. Although not in 1997-98 for the latter two proportions, i.e. less than 75 per cent and less than 50 per cent of AWE.

  5. 50 per cent of AWE is a significantly lower figure than 50 per cent of the mean for this distribution.

  6. ABS, op. cit.

  7. Equivalence scales adjust disposable income of income units according to differences in the characteristics of the income units, such as size and composition. Different scales have been used by different analysts of income distributions, such as the OECD and Professor Henderson in Australia.

  8. OECD, 'Income Distribution and Poverty in 13 OECD Countries', OECD Economic Studies, No. 29, 1997/11.

  9. ibid., p. 62.

  10. See Tiina Makinen, 'Structural pressures, social policy and poverty', International Social Security Review, Vol. 52, No. 4, October-December 1999, International Social Security Association, Geneva, pp. 3-24.

  11. See, for example, Anthony King, 'Income poverty since the early 1970s', in R. Fincher and J. Nieuwenhuysen (eds), Australian Poverty: Then and Now, Melbourne University Press, 1998.

Appendix

The Gini Co-efficient (G) is derived as a summary measure from the cumulative proportion of income received by a cumulative proportion of a population, ranked according to income from lowest to highest. These accumulations can be shown in graph form, known as the Lorenz curve, which shows the income share of any selected cumulative proportion of the population. The relationship between the Lorenz curve and the line of (absolute or perfect) equality, i.e. where any specified proportion of the population has exactly the same proportion of the total income of that population, gives G, as shown in the diagrams below.

Figure 1 & 2

All values of G for any population range from 0 (zero) to 1. Where the line of equality and the Lorenz curve coincide, G is 0 and there is absolute equality of incomes, i.e. every member of the group has exactly the same income. Where the area between the line of equality and the Lorenz curve is the same as the total area below the line of equality, G for that group is 1, meaning that one member of the group has all of the income. Thus the smaller the area between the line of equality and the Lorenz curve, the closer G is to 0 and the more equal the income distribution is for that particular population or group.

G is not the best available measure of inequality as it tends to concentrate more on inequalities between middle or modal incomes rather than inequalities between the extremes-'generalised entropy measures' better interpret inequality changes across entire distributions-but G is the most frequently available measure in Australia, and perhaps the most readily understood.

It should be noted here that any G is an estimate like any other from a sample survey and is thus subject to sampling variability. Hence the smaller the movement between two Gs the more likely it is due to sampling variability and is therefore more likely to be not significant, i.e. not a movement which has actually occurred within the population being analysed.

It is possible for the G to be stable overall while it is volatile within each or any constituent group, as the differences in the distributions within any group can 'smooth' themselves out when they are all combined into one overall distribution. It is also possible that large changes within a small group, e.g. one-parent income units in this analysis, can offset small changes within the other groups when they are all combined into the one overall distribution. It should be noted however that Gs derived from grouped data, as in the analyses in this paper, are less sensitive to changes in incomes than they would have been if ungrouped unit records had been available and used.

 

Facebook LinkedIn Twitter Add | Email Print