Economic Inequality in New Zealand: a User’s Guide

Published in The New Zealand Journal of Sociology , Vol 28, Issue 3, 2013, pages 9-66. 

http://ndhadeliver.natlib.govt.nz/delivery/DeliveryManagerServlet?dps_pid=IE18625730&dps_custom_att_1=ilsdb

Keywords: Distributional Economics; Statistics;

“When inequality is the common law of a society, the greatest inequalities do not call attention to themselves.” Democracy in America, A. de Tocqueville.

 

B. Perry (2103) Household Incomes in New Zealand: Trends in Indicators of Inequality and Hardship 1982-2012 (Wellington, MSD) 242pp.

 

M. Rashbrooke (ed) (2103) Inequality: A New Zealand Crisis (Wellington, BWB) 279pp.

 

Contents

Tables

Prologue

Key Messages

1.         Why is Economic Inequality Important?

2.         Measuring Inequality

3.         The Census Income Distribution

4.         Household Income Inequality

5.         Household Market Incomes

6.         Explaining Changes in the Level of Inequality: Market Influences

Unemployment

Top Market Incomes

Prices and Wages

7 .        Explaining Changes in the Level of Inequality: Social Influences

8.         Explaining Changes in the Level of Inequality: Redistribution

9.         International Comparisons: 2009

10.       Distributional Measures in a National Account Framework

11.       Changes in Inequality Internationally 1985-2009

12.       What Happened Before 1985?

13.       What Happened After the Global Financial Crisis?

14.       Poverty Measurement

15.       The Distribution by Social Groups

16.       The Dynamics of Inequality

17.       Income and Health

18.       Wealth

19.       Housing

20.       Inequality and Growth

21.       Epilogue: Towards Policy Responses

References (preliminary)

 

Tables

1: Income Reported by Adult Deciles in New Zealand Censuses 1945-2013

2: Decile Mean of Equivalised Household Income

3: Summary Indicators of Household Inequality 1982-2012

4: Gini Coefficients for Different Measures of Household Equivalised Income

5: Gini Coefficient for Household Income Distribution: 2008-2010

6: Gini Coefficient for Household Income Distribution: 1983-1987

7: Equivalised Median Household Income by Ethnicity

8: Proportions in Quintiles Who Rate Themselves as ‘Fair’ or ‘Poor’ Health

 

Prologue [1]

 

This survey is a response to two recently published works on inequality in New Zealand. The first, a report by Bryan Perry, although published in a government department, is an extensive scholarly study bringing together much material. The second, a book edited by Max Rashbrooke, although with contributions by academics, is a popular exposition which offers a starting point for a discussion on inequality but both undersells the issue by not being up-to- date with the research, and oversells it – if we are in crisis we have been in it for over two decades.[2]

 

Reporting on either presents a reviewer with a considerable challenge. The Perry study is so marvellously detailed that one could well end up with a review as long as its 242 pages; tackling the Rashbrooke book at the same level would involve tediously correcting or elaborating a plethora of statements.

 

Instead, this has been written as a survey of what we know – and dont know – about economic (mainly income and wealth) inequality in New Zealand. The works are referred to where they are a part of the exposition.

 

A surprising conclusion of the survey is just how much New Zealand research on economic inequality there has been. The incomplete list of references at the end cites more than 70 empirical studies. It is not practical to mention them all in this survey, even if those writing on the topic should be.

 

Reading the Rashbrooke book though, one recalls the story of the drunk who uses a lamppost for support rather than illumination. So it is for too many public policy rhetoricians.[3] This survey is consciously explanatory about the underlying analytics. In each case the test is whether those who contributed to the Rashbrooke book would have benefited from having a better grasp of them.

 

Key Messages

 

Section 1: Why is Economic Inequality Important?

 

1. The section identifies four main issues as to why inequality may be important

– equity

– plutocracy

– efficiency

– inequality may contribute to economic instability

 

Section 2: Measuring Inequality

 

2. There is no single perfect measure of inequality. There are numerous imperfect measures which may be used for different purposes. Sometimes they contradict one another.

 

3. The exact variable which is being measured matters. Among the options for measuring income (or wealth) are that of persons (adults with or without children) or of households and, in the case of income, before or after tax and transfers. Always check definitions (and dont trust that inexpert writers have). Be aware that the measures may not be comprehensive.

 

Section 3: The Census Income Distribution (Table 1)

 

4. Total personal income of adults as measured by the census (which is – before tax – market income only to 1981 and includes social security from 1981) shows relatively stable inequality from 1926 to the 1950s and then falling to 1986, followed by rises.

 

Section 4: Household Income Inequality (Table 2,3)

 

5. There is an unequal distribution of (equivalised) household income distribution. In 2012 about two-thirds of New Zealanders were in households with an annual household income of between $29,000 and $48,400 per (equivalised) person. A sixth were in higher income households; a sixth were in poorer households.

 

6. Today New Zealand is a more unequal society than it was three decades ago. However most of the increase in inequality occurred in the period between the mid-1980s and the mid-1990s. The width of the distribution (as measured by the Coefficient of Variation) increased by over a half. The trend after the mid-1990s is more ambiguous. The best interpretation is that the income distribution has remained at roughly the same level of inequality over the last two decades.

 

Section 5: Household Market Incomes (Table 4)

 

7. It is difficult to interpret changes in the household market income distribution. It probably follows much the same pattern as household disposable income. It may not be a particularly relevant indicator.

 

Section 6: Explaining Changes in the Level of Inequality: Market Influences

 

8. The rise in structural unemployment seems to have increased income inequality. Cyclical unemployment is important.

 

9. A major influence was a rise in the income share of the top 1 percent of adults. Their share rose from about 6 percent in the early 1980s (that is 6 times the adult average) to 10 percent today (10 times the adult average). Most of the shift occurred in the period between 1998 and 2003. The two major influences seem to have been a change in the tax treatment of dividends and an increase in margins for management and professionals over average workers.

 

10. There is not much evidence on the impact of prices and wage relativities (other than the margins mentioned in paragraph 9).

 

Section 7: Explaining Changes in the Level of Inequality: Social Influences

 

11. Changes in the  family and household structure and in the socio-demographic attributes of households had some impact – raising the level of inequality.

 

Section 8: Explaining Changes in the Level of Inequality: Redistribution

 

12. A major reason for the rise in equivalised household disposable income inequality between 1985 and 1993 is partly explained by the dramatic changes in redistribution between income levels that New Zealand experienced. The income tax scale was flattened and the introduction of GST imposed more heavily upon those on lower incomes. There were savage cuts to benefit levels in 1991 and more user pays for health and education so that more costs were pushed onto households, which the poorer ones found harder to bear.

 

13. While there was relative stability in the household income distribution after 1993, the relative share of the bottom quintile fell from the late 1990s. The group’s real incomes rose but less than average, probably because benefits were increased in line with prices rather than wages, so beneficiaries did not share in the rising real wages, beneficiaries were not entitled to the child tax credit, which was later expanded and called the ‘in work tax credit’ under working-for-families, and because many beneficiaries were unable to take advantage of the booming labour market.

 

Sections 9, 10.11: International Comparisons (Tables 5, 6)

 

14. New Zealand was 9th out of 34 in the OECD ranking of inequality in about 2009, after adjustment for population and per capita GDP. It was about 20th in 1985, so it moved from being in the bottom half of the OECD to the top half. (The measure used here is internationally comparable Gini coefficients.)

 

15. Over the whole three decades between 1982 and 2012 the increase in New Zealand income inequality was not the greatest.(The Swedish change seems to have been much higher.) But it had the greatest increase in income inequality in the decade to 1995.

 

Section 12: What Happened Before 1985?

 

17. We do not know what happened to household incomes before the 1980s because there are no data.

 

18. However personal market income seems to have been stable in the interwar period and declined in the early post-war period to 1981.

 

Section 13:. What Happened After the Global Financial Crisis?

 

19. The preliminary indication is that the Global Financial Crisis impacted more on top disposable incomes than bottom ones, so that there was some reduction in income inequality. This was despite the post-GFC income tax changes being biased towards the rich and despite some tightening of benefit entitlements. The probable explanation is that returns on investment fell while unemployment in New Zealand has not been too heavily affected by the downturn. (The Canterbury Earthquakes adds to the difficulties of interpreting the short data series.)

 

20. While the 1972 Royal Commission on Social Security pointed towards a notion of relative poverty, official policy since 1991 seems to be concerned with absolute poverty only.

 

Section 14: Poverty Measurement

 

21. Poverty lines based on median incomes are flawed, because poverty can be reduced by transferring income from the middle to the top.

 

22. A higher proportion of the population experienced relative poverty after 2000 than in the 1980s. – perhaps 2 to 4 percentage points.

 

23. The majority of the poor are parents with jobs and their children (although they may have had only one or two), living in their own home albeit usually with a mortgage. The proportion in poverty is higher among solo parents, those without jobs, living in rental accommodation and with a brown ethnicity (but there are fewer of each category). More women than men are poor. While the incidence of poverty is higher among Maori and Pasifika, there are more poor who are not of their ethnicity.

 

24. The salient conclusion from the research is that over 80 percent of the poor are children and their parents (and others in their households) and that proportionally more children are in poverty than adults (especially those adults who are not parents).

Section 15: Distribution by Social Groups (Table 7)

 

25. Mean Maori equivalised household incomes were 90 percent of average in 2012 and Pasifika ones were 89 percent. European/Pakeha ones were 107.5 percent.

 

Section 16: The Dynamics of Inequality

 

26. Although there is considerable dynamism in the income distribution as some people and households shift their relative locations over time, there is also considerable inertia.

27. One study found about half of those below the study’s poverty threshold at a point in time were in chronic (‘permanent’) poverty – the figure for children was 60 percent. The proportions will be higher if a higher poverty threshold is used.

 

Section 17: Income and Health (Table 8)

 

28. In any given age group, those in the lower income quintiles are in poorer health than those in the higher income quintiles. Some infectious diseases seem to be associated with poverty.

 

Section 18: Wealth

 

29. Physical and financial wealth is much more concentrated than personal income. While there is a life cycle to wealth holdings (peaking at about the age of 60), within each age cohort, wealth is also very unequally distributed. The main form of this wealth holding is housing.

 

30. The vast majority of the adult population had little physical and financial wealth. In 2003/4 6.5 percent had negative net worth, although this may be dominated by those with student debt. Conversely 1 percent of adults had 16.4 percent of total wealth, about the same share as the bottom 70 percent of the population. There is not a lot of gender inequality, but each European/Pakeha owns about 2.6 times that of the other ethnic groups. Larger families (with more than two children) have less wealth.

 

Section 19: Housing

 

31. Very little is known systematically about the impact of housing on inequality but from what is known differences in housing outgoings tend to increase effective income inequality.

 

Section 20: Inequality and Growth

 

32. While per capita National Income in constant expenditure prices rose at a trend rate of 1.7 percent p.a. between 1982 and 2012, Sen’s real national income – which is more sensitive to distributional change – rose only 1.3 percent p.a. because of the rise in income inequality. In effect on Sen’s measure the additional inequality cost New Zealand economic growth almost a fifth.

 

Section 21: Epilogue: Towards Policy Responses

 

33. While the focus of the survey is on the facts and related analytics rather than policy – on getting things right – there is a preliminary discussion of predistribution and redistribution policies.

 

34. This survey of economic inequality concludes

 

… those who command policy – whether effectively or ineffectively – have to decide to what extent reducing (or increasing) economic inequality is a policy objective. Is New Zealand satisfied with shifting from a low inequality to a high inequality society? What would its founding nineteenth century migrants have thought about the fact that, after allowing for each country’s size and affluence, New Zealand is now more unequal than the countries they left? And what would those who invited them here have thought had they known their descendants would be firmly in the bottom end of the unequal distributions?

 

References

 

The bibliography includes over 70 items of New Zealand empirical research plus some international ones and some of a more philosophic nature. Readers are invited to submit further New Zealand work which has been inadvertently overlooked.

 

1. Why is Economic Inequality Important?

 

Public concern over economic inequality is rising both internationally and in New Zealand. It probably reflects four main themes which are inextricably mixed together.

 

1. The Equity Argument

 

Many people judge that inequality (or, more often, severe inequality) is morally wrong per se. Their arguments as to why this is so can be tortuous or simple. One of the most sophisticated is that of philosopher John Rawls whose ‘second principle of justice’ is that social and economic inequalities should be arranged to be of the greatest benefit to the least-advantaged members of society and that offices and positions should be open to everyone under conditions of fair equality of opportunity. (Rawls 1971/1999) [4]

 

Implicit in this formulation is a belief that some inequality benefits the poorest. A simple version of it is that a society with perfect equality would give the poorest a lower quality of life than one with some inequality, because there would not be the economic incentives to improve oneself, thereby benefiting others too, while the costs of enforcing such perfect equality are high.

 

However this does not tell us how much inequality is ideal. Such evidence, as there is, indicates that inequality tends to be lower in higher income societies, while high inequality societies do not seem to grow any faster. These are but correlations; there is not a lot of systematic evidence to explain what the underlying causal processes might be.

 

There are two main subsidiaries of the equity argument. The first arises from the need for ‘social coherence’ – an unequal society is a divided one. It has an efficiency dimension since a divided society is likely to require more public spending on justice, policing and corrections and on private security measures.

 

The second argument is that while there is nothing wrong with social inequality at a point in time, it has long term consequences in terms of loss of opportunity for future generations. Sometimes it is said that the purely ethical case against inequality does not carry much weight (in the mind of the exponent) but there is a concern that children (in particular) in deprived homes will be denied life chances. (Someone who holds such a view may also be concerned about the three non-equity arguments which follow.)  Note that it also has an efficiency dimension, because arguably an unequal society is wasting talents.

 

2. The Plutocracy Argument

 

There is a concern that if there is great inequality then those at the top of the (income, say) distribution have an unfair influence in the direction of the governing of society. While the rich always have more political leverage than the poor, the amount varies depending on institutional arrangements and the degree of inequality. It is possible that while a political system may have the formal appearance of a democracy, it may in practice be a plutocracy in which its effective rulers are its rich.

 

Plutocracy undermines the claim of a democratic society. Where the economic disparities outweigh the formal structures of democracy, the practical operation of the democratic principle suffers – and with it the rights and voice of the underprivileged. The unequal distribution of national resources is consequently affected and with it the denial of opportunities for the disadvantaged sectors of the population. Stiglitz’s point about the cumulative effects through unequal media and national dialogue is especially relevant.

 

Where a particular system settles on the democracy-plutocracy spectrum is partly a matter of institutional arrangements (such as the degree to which money can influence elections) but inequality is also relevant because it affects how much various groups can afford for influencing purposes. Indeed the successful ones may use that influence to bias the institutional arrangements in their favour (such as the laws which control contributions to elections). This influence is not confined only to elections and the policies of the elected governments; it may also affect the media and the national dialogue. (Stiglitz 2012, Wade 2013)

 

The plutocracy issue is sharper in America and Britain, where the top 1 percent of income recipients have double the share they have in New Zealand. In any case any such class analysis in New Zealand has to pay more attention to offshore influence.

 

3. The Efficiency Argument

 

Recently The Spirit Level has argued that unequal societies are more likely to have poorer social performance. (Wilkinson & Pickett, 2009) It finds a relationship between inequality, on the one hand, and, on the other, various variables: drug and alcohol infant mortality, educational performance, homicides, imprisonment rates, life expectancy, mental illness, obesity, social mobility, suicide, teenage births and trust.

 

Some of the relationships are empirically stronger than others. Thus far, the research provides but a correlation and every scientist knows that correlation is not causation – that two events are associated does not prove that one causes the other. There are various hypotheses about the mechanisms which might cause the social distress, but none are yet proven in a scientific court. [5] Any inefficiency would mean that a society with higher inequality has to spend more – whether publicly or privately does not matter here – on its health and justice systems to attain its desired outcomes.

 

A related issue is that the greater the economic inequality, the lower the intergenerational mobility. Even if economic inequality at a point in time were not to matter, it seems likely to affect life chances over time. A child who comes from a deprived background is likely to have poorer health throughout the rest of their life, to have less access to educational and vocational opportunities and to be more likely to experience social delinquency. The equity argument says this is wrong; for efficiency concerns it makes measures to correct the effect more expensive.[6]

 

4. The Macroeconomic Argument

 

One element of the Global Financial Crisis may have been that the increasing incomes among rich Americans were saved and lent to those below for housing purchase, which led to the sub-prime loans and the resulting financial instability. (Stiglitz 2013) This argument applies to the American economy only, and does not seem to apply to New Zealand. It is mentioned here as one part of the main issues in the world debate.

 

In summary there are a number of reasons why we might be concerned with economic equality. Aside from moral concerns, there is good reason to believe that inequality impacts on the way in which a society functions. If some of the impacts are unattractive, reducing inequality may result in some social gains.

 

2. Measuring Inequality

 

There is no unequivocal measure of inequality or changes in inequality. While what is happening to inequality is clear when only two individuals are involved – either their incomes are closer or further apart – the addition of just one extra person can lead to ambiguities. It is only possible to make unequivocal comparisons where the ‘Lorenz’ curves do not cross. [7] (Atkinson 1975: Ch 3; Easton 1983: Ch 2) [8]

 

The core of the problem is that economic distributions are typically complex and cannot be characterised by a few parameters. Students are introduced to simple statistical distributions; the commonest of which is the normal (or Gaussian) distribution which requires but two parameters (a mean and standard deviation) which may be compacted for many comparative purposes into a single parameter of the coefficient of variation – the ratio of the two.[9] Commonly there is not a sole parameter to compare two economic distributions. That is why there can be no authoritative index of inequality.

 

This is illustrated by the following simple example. Suppose there are three people who have incomes of 3, 10 and 17 respectively. If there is a transfer of one unit from the middle income person to the bottom and one unit to the top the distribution is now 4, 8, 18.

 

Observing the bottom person has had an increase in their income one might assume there has been a reduction in inequality. But, paradoxically, observing the top person has also had an increase, the conclusion seems to be there has been an increase in inequality.

 

A widely used measure is the ‘Gini’ coefficient which measures the average difference between the incomes of the individuals in the distribution (scaled by the overall average income). It ranges from 1 when there is total equality, to 0 when all the income is received by a single individual. (Sometimes it is reported as a percentage/out of 100.)

 

Because it is not well understood the Gini coefficient has a mystique. For the general user a measure of the ‘width’ of a distribution – such as the coefficient of variation – may be more intelligible.[10] A CV of 1 is equivalent to the standard deviation equalling the mean; of 2, twice the mean; and so on. A doubling of the CV is equivalent to a doubling of the width of a distribution.

 

 

Unfortunately it is rare for a research study to provide the CV. However there is a simple conversion of a Gini coefficient to a CV when the variable has a lognormal distribution – which many economic distributions approximately follow.[11] To assist understanding the magnitude of the Gini coefficient – and even more to assist understanding the meaning of a change in inequality between two distributions – whenever the Gini coefficient is reported the CV equivalent for the lognormal is also reported.

 

To add to the paradox, the Gini coefficients for the two distributions we considered at the beginning of the section are identical (GC = 0 .655; CV = 6.2), which might lead one to conclude there has been no change in inequality.

 

So the three measures tell different stories, illustrating the difficulty of judging changes of inequality (or comparing them) when the changes are small. Fortunately in practice where there have been substantial changes in the income distribution, the measures of inequality have been consistent. But when the changes in the distribution are small they can contradict one another.

The counsel of perfection is to give (income shares). Space means that is not always possible and, in any case, much of the public finds the resulting tables too complicated.

 

A recent development at the OECD has been the use of disparity indexes rather than the Gini Coefficient. Essentially a disparity index is the coefficient of variation scaled so that the mean of the group (say of countries) is unity. It is the main indicator in the OECD paper discussed at end of Section 7.

 

3. The Census Income Distribution

 

The Population Census asks respondents to report their incomes for the previous year. Until 1981 they asked only for market incomes, but from 1981 they asked for total income including social security benefits. (Conveniently in 1981 they asked both.) The request is for before tax income but it is likely some (especially social security beneficiaries ) report their after-tax income. In any case their recall is not always accurate.

 

The advantage of the series is its length, although it is less valuable since superior series have become available. Census data is also useful for tracing social groups such as Maori. The available series is shown in Table 1. (The Interwar data is discussed in section 11.)

 

Table 1: Income Reported by Adult Deciles in New Zealand Censuses 1945-2013

Percent of Reported Income

March Year

Bottom 3 Deciles

7th

6th

5th

4th

3rd

2nd

Top

Gini

Coeff

Market Incomes (Non-Maori)

1926

0.0

0.1

5.1

7.8

13.0

16.9

20.6

35.6

0.62

1936

0.0

2.2

2.9

7.6

9.5

15.1

22.4

40.3

0.59

1945

0.0

2.7

4.8

7.9

11.2

15.4

20.0

37.8

0.59

1951

0.0

.2.5

4.5

9.2

13.4

16.4

20.7

35.7

0.60

Market Incomes (Total Population)

1951

0.0

0.1

3.9

8.5

14.0

15.1

20.1

38.5

0.61

1956

0.0

0.2

3.9

8.9

13.1

15.8

20.2

38.3

0.61

1961

0.0

0.4

4.7

8.2

13.9

16.0

19.7

37.1

0.60

1966

0.0

1.2

5.2

9.4

12.6

15.4

18.7

37.5

0.58

1971

0.1

1.5

5.1

9.9

12.2

16.1

19.5

35.6

0.57

1976

0.3

1.6

5.8

9.6

12.7

15.7

19.8

34.6

0.56

1981

0.4

1.4

5.9

9.6

12.7

15.6

19.6

34.9

0.56

Total Income Including Social Security Benefits

1981

4.4

4.5

6.6

9.5

11.7

14.2

17.7

31.3

0.48

1986

6.9

5.5

7.2

9.1

11.3

13.8

17.6

28.7

0.43

1991

7.5

4.8

6.6

8.6

11.6

13.3

17.4

30.3

0.45

1996

5.9

5.1

5.9

8.2

10.7

13.7

17.1

33.4

0.48

2001

5.7

4.8

6.1

8.1

10.6

13.3

17.5

33.9

0.49

2006

5.8

4.8

6.5

8.7

10.9

13.1

17.7

32.4

0.47

2013

Sources: Statistics New Zealand Population Censuses.

Note: the 1945 Census was in September 1945. It did not include overseas paersonel. This data corrects and error in Easton (1996).

 

The inconsistencies of measurement (and inaccuracies of recall) over time mean that it is not always easy to identify exactly what is going on. However the pattern seems to have been that market income was becoming less unequal in the post war era up to 1986. (Easton 1983) An important factor was increased market income to those at the bottom of the distribution. The probable main reason is women (especially mothers) entering the paid labour force.

 

The post-1981 series is complicated by the addition of social security to market income, so some of the changes are the result of changes in in the levels of benefits and the numbers of recipients. Even so, it shows what is a characteristic feature of the post-war New Zealand income distribution. There is a very large increase in inequality between 1986 and 1996, equivalent to a 50 percent increase in the coefficient of variation.[12]

 

After 1996 the inequality as measured by the Gini Coefficient is reasonably constant and may not reflect structural change. We await the 2013 Census data (to be published in December 2013) before drawing any structural conclusions.

 

The income shares reported in this section are before taxes are levied. The next section discusses the impact of taxation but this has to be at a household level and the data is only from 1981.

 

4. Household Income Inequality

 

Inequality is a multidimensional phenomenon, including inequality of income, or wealth, or inequality of access to health care, justice and educational opportunity.

 

Moreover there are sub-dimensions within the categories. Income may refer to market income, or to disposable income, in which taxes are deducted and government transfers (generally described as ‘benefits’ in this review) are added, or it may refer to taxable income, which often adds public transfers to market income.[13] It may be on the basis of individuals, typically excluding children[14]; or it may be the income of households which include children. The population under consideration usually matters.[15] Most of the other dimensions have similar complications.

 

We shall focus here mainly on the ‘equivalised household income’ data base, which is at the centre of New Zealand analysis. The method was introduced in 1974 when the first Statistics New Zealand household survey became available. (Easton 1976) There have been steady improvements over the years, including the ability to access data at unit record level, and steady improvements in the quality and consistency of the records.

 

The basic procedure is to take the reported (annual) income – which may be adjusted for under-reporting – deducting income tax and adding government-provided benefits (if they are not included in the reported income). The income is then adjusted for the household’s composition.[16] The result is called ‘equivalised’ income. The purpose is to treat households of different size and composition equivalently. For most purposes the notion may be as if it is ‘per capita’. (A variation is to focus on spending rather than on disposable income. The income focus arises because dis-saving in one period is likely to lead to lower spending in future ones.)

 

The household equivalence scales which are used for the adjustment usually allow for the economies of scale for larger household, and often treat adults and children differently. There may be adjustments for housing circumstances (discussed later) and other differences.

 

The scales are fraught with complications not always appreciated by their users. (Easton 2004) Some will mentioned later when they become important for interpretation.

 

Since each household has different numbers of individuals and, especially, larger households tend to be poorer ones, each person (including any children) in a household is assigned the equivalised household income.[17] All individuals are then ranked by the equivalised incomes of the household they belong to into deciles.[18]

 

Table 2 shows the resulting tabulation for the 2012 year.[19] The second column shows each decile’s average equivalised income (it may best be treated as the income of a couple; in April 2012 their New Zealand Superannuation would have been at least $25,953 p.a.). The third column shows the decile share of total equivalised income.

 

Table 2: Decile Mean of Equivalised Household Income

DECILE  Average Income  $p.a.2012 prices

Share %

Top

97300

24.3

2

60200

15.1

3

49300

12.3

4

41700

10.4

5

35800

9.0

6

31400

7.8

7

28000

7.0

8

23800

6.0

9

19500

4.9

Bottom

12900

3.2

ALL

38700

100

Source: Perry (2103:238). (Negative incomes set at zero.)

 

The basic message is that there is considerable income inequality in New Zealand, with the top decile receiving about 2.4 times the average income and 7.5 times as much as the bottom decile. The Gini Coefficient is 0.306, equivalent to a Coefficient of Variation of 0.25 and a standard deviation of around $9,700. Roughly two-thirds of New Zealanders in 2012 were in households with an annual income of between $29,000 and $48,400 per (equivalised) person.

 

The income shares can also be constructed for many years since 1982. To simplify, Table 3 provides the share of the top 10 percent (decile) of households, the Gini coefficient (with CoV) and the bottom 20 percent (quintile). The third indicator reflects that estimates of poverty (discussed later) usually conclude about a fifth of the population are in poverty. Columns five and six are poverty-based estimates to be discussed later. The final column gives the average income for a couple measured in 2012 prices.

 

Table 3: Summary Indicators of Household Inequality 1982-2012

Year

Top Decile

 Share %

Gini Coefficient

(CoV)

Bottom Quintile

Share %

Absolute Poverty

%

Relative Poverty

%

Average Income

 (2012 Prices)

1982

19.9

.268 (.18)

8.9

12

12

30000

1984

20.2

.270 (.18)

9.0

13

13

29400

1986

20.3

.265 (.17)

9.2

14

14

28400

1988

20.1

.262 (.17)

9.1

12

12

28900

1990

23.1

.300 (.24)

9.0

14

14

30700

1992

23.3

.311 (.26)

8.1

24

24

27500

1994

23.8

.318 (.28)

8.1

26

26

27000

1996

25.0

.325 (.30)

8.1

20

20

29100

1998

25.0

.327 (.30)

8.0

16

15

31200

2001

25.4

.334 (.32)

7.5

16

17

32600

2004

24.5

.329 (.31)

7.7

13

20

34400

2007

24.6

.328 (.31)

7.8

11

18

36200

2009

25.8

.331 (.32)

7.9

7

18

39800

2010

24.6

.318 (.28)

7.9

8

18

39300

2011

26.7

.343 (.35)

7.6

8

16

39800

2012

24.3

.317 (.28)

8.1

6

15

40000

Source: Perry (2013) ibid. pp.70, 87, 110, 238.

 

Broadly all the indicators show increasing inequality over the three decades between 1982 and 2012. The income share of the top decile rises, that of the bottom quintile falls, while the Gini coefficient rises. The year-to-year fluctuations are generally small.[20]

 

In particular, on all indicators most of the change in inequality occurred in the period between the mid-1980s and the mid-1990s. The width of the distribution (as measured by the Coefficient of Variation) increased by over a half.

 

The trend after the mid-1990s is more ambiguous. The share of the bottom quintile tended to fall, the share of the top decile was roughly stable and the Gini coefficient was probably stable, possibly declining. In my view the best interpretation is that the income distribution has remained at roughly the same level of inequality over the last two decades, although different groups experienced gains and losses.

 

5. Household Market Incomes

 

In the course of examining the impact of fiscal measures on household material standards of living, Aziz et al calculated Gini Coefficients for different measures of income for the four years for which they had data (1988, 1998, 2007, 2010). Their work is reproduced as Table 4. A comparable series from Perry (2013) is added. The researcher’s primary interest is the final row of the table, which adds to household income the services provided by the government (such as education and health) and suggests the government has been less supportive to those with low incomes in recent decades.

 

Table 4: Gini Coefficients for Different Measures of

            Household Equivalised Income

1988

1998

2007

2010

Market Income

0.42

0.49

0.54

0.52

Disposable Income (Aziz)

0.3

0.35

0.38

0.36

Disposable Income (Perry)

0.26

0.33

0.33

0.32

Final Income.

0.27

0.3

0.35

0.35

Source: Aziz et al (2012) Table 3, Perry (2013) see Table 3 above.

 

Unfortunately the four years come from two data studies which may not be exactly aligned. Crawford & Johnston (2004) provide the first two observations; Aziz et al (2012) the last two. The consistency problem is illustrated by comparing the Aziz household equivalised disposable income Gini Coefficients with the Perry calculated ones which are constructed to be consistent over time.

 

The concern here is not the different levels – which is both puzzling and disappointing – but the different patterns. The Perry series shows a leap between 1988 and 1998 and then remaining at roughly the same level; the Aziz one has a continuing rise from 1998 to 2007 and then the flattening out. This could be reconciled with the Perry pattern if we assume that the Aziz and Crawford estimates are not on the same basis. If that it true for household equivalised disposable income it must also be true for household equivalised market income.

 

But even if the series are consistent we need to be careful interpreting household equivalised market income. Economists are interested in personal market income because it reflects the factors of production held by individuals and the (factor) prices that reward them. However equivalised household market income is not as interesting from this perspective, because household formation is also relevant.

 

If a couple of workers marry that will change the distribution of household equivalised market income but not the personal market. If they have a child that again changes the household market equivalised distribution even if they continue to work. As we report below Hyslop &Maré (2001) observe one of the causes of increase in inequality is changes in household structure.

 

Ultimately is in unclear why we should be interested household equivalised market income except as a step on the way from personal market income to household disposable income.

 

Is the same true for household disposable income? No, because the variable has a meaning as an indicator of the material standard of living of the inhabitants of a household. The challenge is to explain what are the determinants of that measure. We begin with the determinants of market income.

 

6. Explaining Changes in the Level of Inequality: Market Influences

 

While it is true that income inequality has been increasing over the last three decades, almost all the significant change occurred in the first decade. This is not a trivial issue. What serious policy needs is an analytic account of what determines the inequality, preferably with an indication of the magnitude of each effect; getting the right story to analyse is important.

 

Methodologically it is easier to explain changes in the level of inequality than to explain the level of inequality itself. The latter usually needs cross-country comparisons or heavy use of theory with some microeconomic analysis (which rarely gives an indication of magnitudes).

 

The available explanations for changes in economic inequality are centred on two general economic areas. The first involves changes in the macro-economic (or market) environment – especially unemployment, remuneration and investment income.; the second, redistribution policies discussed in the next section.

 

Unemployment

 

Changes in the level of unemployment level affect just about everyone but apparently those in the second-to-bottom quintile are affected most. (Easton 1995) Unemployment rose to unprecedented post-war heights in the early 1990s, reflecting the major industry restructuring going on at the time (plus some loss of aggregate demand). It may well be half of the labour force was unemployed over four years 1988/9 to 1992/3, albeit in most cases for a short period as they shifted between jobs, but usually sufficiently long for them to register with the employment service of the Department of Labour. (Easton 1996:110-112) This suggests that part of the rise in inequality in the 1985 to 1993 period could have been due to the weaker labour market.

 

Unemployment levels fell from the mid-1990s, especially after 2000. The effect on income inequality is evident in the falling Gini coefficient (but less so in the top decile and bottom quintile which are less affected by unemployment). (Perry 2013:56, 163) However the effect is small, at most equivalent to a 20 percent fall in the width of the distribution (the CoV) between 2001 and 2011. (Moreover, some of that fall must be attributed to the introduction of Working for Families). It appears the lift in unemployment between the early 1980s and early 1990s cannot have made a major change to income inequality in the period.

 

More generally, there is a case that the structural (i.e. cyclically-corrected) rate of unemployment rose about 2 to 3 percentage points between the 1960s and the 1980s. (Easton 2012) This would have added to inequality but the shift occurred outside the period of main focus.

 

Top Market Incomes

 

What happened to top of the income distribution measured before tax? Analysis is treacherous because it is based on an administrative data base which changes over time as law and administrative practices change. (Additionally it is based on income reported for tax purposes which is subject to tax avoidance and excludes capital gains.)

 

To reduce some of these complications the analysis is based on all adults (over 15) and not those just who are recorded in the tax statistics. The income in the denominator is the primary income receivable as reported in the System of National Accounts. Four measures from 1981 to 2011 are available: the share of the top 0.1 percent, the share of the top 1 percent the share of the top 10 percent and the Pareto coefficient. The discussion below is confined to the top 1 percent. The other indicators are of different magnitudes but show generally the same pattern. (Easton 2013T)

 

The top 1 percent’s share of total income was around 6 percent of total income until 1989, or 6 times that of the adult average.. It then rose rapidly to just over 10 percent by 1993, or 10 times the average. After this jump of 4 percentage points (or two thirds) the share remains broadly constant.[21]Arguably the share increased by, say, 0.3 percentage points in the early 2000s; if so it fell by roughly the same amount in the following decade. There is no evidence of the sharply rising income share at the top since 2000 we see in jurisdictions with more sophisticated financial sector.

 

Why did the share of the incomes at the top rise? Part, perhaps a third, is explained by the ending of the double taxation on dividends in 1989 led to more being recorded as taxable income.[22] (This would mean that unrecorded capital gains probably fell.)

 

The best explanation for the remainder seems to be that margins for management and professionals (i.e. over average workers) rose both in the public and private sectors. The big shift occurred in the early 1990s (probably as a consequence of legislative change and changed private sector attitudes following the market reforms); the structural explanation is probably the opening up of the global labour market for these skills.

 

Unfortunately we cannot simply map individual adult market income shares onto households. Those with top decile incomes are likely have more adults – some of who may, or may not, have incomes in the top decile – and they are more likely to be larger and have more dependants. Even so it seems safe that perhaps 2.5 or 4 percentage points of the top decile households probably came from higher market receipts in the early 1990s.)

 

Prices and Wages[23]

 

Little is know about the effect of other changes in relative market prices. Did market liberalisation period benefit the rich more than the poor? The short answer is that there is little evidence that they had a marked effect.[24]There may have been some, but it seems to have been minor in comparison to the impact of the effects we are about to describe. (Easton 1996)

 

There is some evidence that the wage distribution became more unequal. However its impact on the household distribution is unclear.[25] (Dixon 1998)

 

7: Explaining Changes in the Level of Inequality: Social Influences

 

Dean Hyslop and David Maré (2003) carried out a careful analysis of changes in the distribution of Gross Household Income (that is before income taxes are deducted).[26] They use five indicators of inequality for three periods: 1983-6, 1989-93, 1995-98.

 

They examined five effects whose change might be expected to impact on the level of inequality: household structure, National Superannuation, household attributes employment outcomes and economic returns to factors of production. Collectively they explain about half the total change.[27] Here is (part of) Hyslop and Marés conclusion.

 

Our analysis shows that the changes in the distribution of income involved a complex set of factors which are difficult to summarise using a single measure of inequality. Examining income inequality across all households, we find that the main factors which contributed to the change in inequality were changes in family and household structure (primarily a pronounced drop in the fraction of two parent households and a rise in the fraction of sole parent households), and changes in the socio-demographic attributes of households. These factors each explain one-sixth of the total increase in the Gini coefficient [of Gross income] over the period, and up to one-third and one-half (respectively) of other measures of inequality. … our results show that changes in the employment outcomes of households had a more modest impact on income inequality. However, within household types, we find that employment changes do have a large effect on the observed change in income inequality. Finally, we find little evidence of any systematic effects of changes in the economic returns to socio-demographic attributes on the distribution of household income and inequality.

 

(The failure to get purchase from the unemployment variable may be surprising, but may reflect timing.)

 

8. Explaining Changes in the Level of Inequality: Redistribution

 

As it happens, the rise in income inequality between 1985 and 1993 is associated with some of the most dramatic changes in redistribution between income levels that New Zealand has experienced.[28] Top income tax rates were cut, while the removal of tax exemptions and the introduction of GST imposed more heavily upon those on lower incomes. Additionally there were savage cuts in benefit levels in 1991 while union power to maintain and increase real wages was weakened. These largely explain the change in inequality in that period.

 

There have been subsequent changes in tax and benefit levels, but none apparently dramatic enough to disturb greatly the income distribution. There is one exception.

 

It is clear from Table 2 that the relative share of the bottom quintile fell from the late 1990s. Their real incomes rose but more slowly than average incomes. There seem to have been three reasons. First, benefits were increased in line with prices rather than wages, so beneficiaries did not share in the rising real wages. Second, they were not entitled to the working-for-families in work tax credit, which benefited those in immediately higher deciles.[29] Third, many beneficiaries were unable to take advantage of the booming labour market of the late 1990s and early 2000 to find jobs. It is true that the level of top income taxes fell further, but apparently not enough to markedly change the after-tax share of the rich.

 

In summary, we know that the change in redistributional policies had the major impact on income inequality, but so did changes in the market income of the top 1 percent. Probably they contributed about equally to the rise in the share of the top 10 percent of households. Cyclical changes in unemployment levels had some smaller but perceptible effect (unfortunately they are most important at the very time the biggest changes in redistribution and top income rising shares happened, and they may obscure the length of the transition). We dont have a lot of evidence about the effects of market prices including wages.

 

9. International Comparisons: 2009

 

In recent years the OECD has compiled comparable Gini coefficients for all its members, although not for every year.[30] Table 5 shows the ranking for the average of the 2008 to 2010 years (recall the coefficient is not an ideal measure of inequality – none exists).[31] (The adjusting is explained and interpreted below.)

 

There is a major caveat to the tabulations for they use exactly the same equivalence scales for all countries. However true equivalence scales will reflect differences in market prices. The differences can be dramatic as I found when I valued the same basket of goods and services in New York and New Zealand prices. It generate quite different equivalence scales. The most obvious reasons for this was the cost of housing with additional rooms (for children) in New York being far more expensive than New Zealand, so that the New York scale showed weaker economies of scale for household expenditures.[32]  (Easton 1973N) We do not know what the effect of using country specific equivalence scales would be on the rankings in Table 5.

 

Table 5: Gini Coefficient for Household Income Distribution: 2008-2010

Rank by

Unadjusted

Adjusted

Inequality Country

Gini Score

Country

Gini Score

Most Chile

0.508

Chile

0.459

2 Mexico

0.471

Luxembourg

0.392

3 Turkey

0.411

Mexico

0.389

4 United States

0.379

Israel

0.369

5 Israel

0.373

United States

0.362

6 Portugal

0.345

Australia

0.348

7 United Kingdom

0.343

Ireland

0.345

8 Japan

0.336

Iceland

0.336

9 Australia

0.335

New Zealand

0.336

10 Greece

0.332

Switzerland

0.333

11  Spain

0.329

Portugal

0.333

12 New Zealand

0.317

Turkey

0.333

13 Canada

0.320

Greece

0.331

14 Estonia

0.316

United Kingdom

0.329

15 Italy

0.315

Canada

0.322

16-17 Average

0.313

Estonia

0.315

Korea

0.313

Spain

0.315

Ireland

0.312

Average

0.313

18 Poland

0.305

Japan

0.306

19 Switzerland

0.298

Norway

0.303

20 France

0.296

Netherlands

0.303

21 Germany

0.287

Italy

0.297

22 Netherlands

0.286

Austria

0.289

23 Luxembourg

0.278

Sweden

0.284

24 Hungary

0.272

Korea

0.284

25 Iceland

0.270

France

0.281

26 Sweden

0.266

Finland

0.281

27 Austria

0.265

Belgium

0.276

28 Belgium

0.261

Denmark

0.272

29 Slovak Republic

0.260

 Germany

0.272

30 Finland

0.258

Slovenia

0.260

31 Czech Republic

0.255

Poland

0.254

32 Norway

0.248

Slovak Republic

0.250

33 Denmark

0.244

Hungary

0.246

Least Slovenia

0.243

Czech Republic

0.245

Source: OECD Income Distribution and Poverty http://stats.oecd.org/Index.aspx?DataSetCode=IDD (1 August 2013)

 

The third column shows that in about 2009 New Zealand was 14th to most unequal among the 34 countries on the OECD list. Thus it is in the top half of the list. While the country may appear to be only a little above the average Gini coefficient score (0.317 versus 0.313), that amounts to almost an 11 percent larger coefficient of variation.

 

The ranking in the list is higgledy-piggledy, not least because the level of inequality is affected by population size (a population of one has no inequality) and affluence (high income economies tend to be more equal than low income ones). The fifth column adjusts for this – in effect assuming that all countries have the same sized population and GDP per capita.[33]

 

We can now see a pattern in the column. At the bottom are the five East-Central economies which had been in the Soviet sphere of influence before 1991.[34] Above them are the remaining continental Europeans with the ‘Club Meds’ more unequal. Scattered among them are Japan and Korea. Higher (more unequal) are the Anglos – Canada, United Kingdom, New Zealand, Ireland, Australia and United States mainly above the average and at the top in inequality terms are the Latinos and Middle East.[35] In summary:

 

MORE UNEQUAL

Latinos & Middle East

Anglos

Club Meds

Northern Continental Europeans & North East Asia

East-Central Europeans.

LESS UNEQUAL

 

New Zealand is the middle of the Anglos, more unequal than Canada and the United Kingdom. In 2009 it was certainly in the top half of the OECD (as it was in the unadjusted ranking).

 

9. Distributional Measures in a National Account Framework

 

A new, and potentially fruitful, development has been the OECD’s Working Party on National Accounts developing ‘Distributional Measures Across Household Groups in A National Accounts Framework’ published in September 2013.[36]

 

A major effort has been to ensure international comparability in income and expenditure measure by anchoring them into national accounts concept. This involves realigning the numbers reported in household surveys. The surveys are known to under-report some items (notoriously alcohol consumption).[37] National Accounts estimate usually estimate the items independently (say from production, imports and sales in the case of alcoholic beverages). They do so based on agreed international standards. In principle then the aggregates are more comparable internationally and so would be the household distributions following realignment.[38] However the adjustments probably may not have a lot of impact on domestic comparisons over time, although the may affect levels (shares).[39]

 

The preliminary work involves only eight countries: France, Italy, Korea, Mexico, Netherlands, New Zealand and Slovenia (for various years between 2003 to 2010). This is really too few to make much of an assessment but the basic impression is that New Zealand’s ranking is similar to that in Table 5, a little above the middle.[40] Interestingly in the small sample New Zealand’s poor seem to have relatively high expenditures on health and housing. [41]

 

10. Changes in Inequality Internationally 1985-2009

 

Unfortunately there are Gini coefficients for only 17 countries in about 1985. (Table 6)

 

New Zealand proves to be below average among these 17 countries in about 1985.[42] Making the reasonable assumptions that those for which we have not got data behave similarly to those for which we have, New Zealand income inequality was in the bottom half of the OECD.

 

Table 6: Gini Coefficient for Household Income Distribution: 1983-1987

Rank by

Unadjusted

Adjusted

Inequality Country

Gini Score

Country

Gini Score

Most Mexico

0.452

Mexico

0.528

2 Turkey

0.434

Turkey

0.454

3 Greece

0.345

United States

0.387

4 United States

0.338

Israel

0.382

5 Israel

0.326

Greece

0.362

6 United Kingdom

0.309

Luxembourg

0.325

7 Japan

0.304

United Kingdom

0.325

8 Canada

0.294

Canada

0.302

Average

0.293

Average

0.301

9 Italy

0.287

Japan

0.297

10 Netherlands

0.272

New Zealand

0.294

11 New Zealand

0.271

Italy

0.271

12 Germany

0.251

Netherlands

0.262

13 Luxembourg

0.247

Germany

0.209

14 Denmark

0.223

Norway

0.212

15 Norway

0.222

Denmark

0.182

16 Finland

0.209

Finland

0.177

Least Sweden

0.198

Sweden

0.169

Source: OECD Income Distribution and Poverty http://stats.oecd.org/Index.aspx?DataSetCode=IDD (1 August 2013)

 

Perhaps we can be more precise. The likelihood is that New Zealand was about 20th out of the current 34 in the mid-1980s. Twenty-five years later it was 9th.[43] Its Gini-coefficient moved from an adjusted 0.294 to 0.336 equivalent to a 47 percent increase in the coefficient of variation. The other 16 countries moved an average of 31 percent.

 

It is true that on the available data, New Zealand had the greatest increase in income inequality in the decade to 1995 (among rich countries for which we have data), but it would be foolish to assume that was true over the next two decades. Sweden’s coefficient of variation more than doubled over the period, but this was from an exceptionally low level in 1983 to a low-to-moderate one in 2009 (when Sweden was ranked 23rd on the list, which now included the East-Central European economies).[44]

 

Indeed as already reported in regard to the 17 countries and confirmed by a more detailed OECD study, the general level of income inequality in the OECD has been trending (roughly linearly) upward over the period. (Perry 2013:171) The New Zealand income inequality (measured by the Gini coefficient) has been broadly steady since the mid-1990s. As a result its level has been converging towards the OECD mean (from above) but has yet to reach it.

 

Why the trend increase in OECD mean inequality? Three reasons come readily to mind. One is that the increasing political power of the plutocracy has resulted in blunting the application of redistributional policies. A second is that globalisation has increased the international mobility of (especially) higher paid labour and countries have felt compelled to lower top tax rates to discourage their movement. Given the middle class hunger for public goods, such as public health care, any government spending cuts consequent on the lower tax-take result in lower benefits for the poor. Additionally the financial sector boom has increased inequality in those countries with a significant international financial sector.

 

12. What Happened Before 1985?

 

We do not know what happened to the household income distribution before 1982 because there is no suitable data base.[45]

 

There is more data on the personal income distribution. The census data reported in Table 3 suggests there was a tendency for inequality in personal market incomes of Non-Maori to be stable in the interwar period (at least in the handful of years which are available for the story during the Great Depression may be different).

 

After 1951 market (and subsequently market plus social security) incomes fell through until 1981. The census data is consistent with Easton (1983) which explored a wide range of data from various sources, that concluded that the evidence points to the personal income distribution narrowing in the post-war era to 1981.

 

However, the personal income distribution does not map simply into the household distribution. Among the important complications in the postwar era are the diminishing size of households as the fertility rate fell, the greater variation in household size and composition (including the rising share of single adult with children families) and the increasing proportion of mothers in the paid workforce.[46]

 

An attempt has been made to estimate an indicator of inequality going as far back as 1921 based on the shares of income among taxpayers, especially of the top 1 percent. (Atkinson & Leigh 2007a, 2007b) Unfortunately the series is not consistent over time.[47] Not all personal income was taxed (especially before 1938), while the rising numbers of earning women in the postwar era changes the proportion of adults who are taxpayers.

 

Historical narratives based on the series can be very unsatisfactory.[48] However, as Robert Solow observed, addicted gamblers will play a roulette wheel they know to be biased because it is the only one available.

 

13. What Happened After the Global Financial Crisis?

 

Suppose that the data distortions from the Canterbury earthquakes occur only in the 2011 year.[49] The preliminary indication is that the Global Financial Crisis impacted more on top disposable incomes than bottom ones, so that there was some reduction in household income inequality.

 

This was despite the post GFC income tax changes being biased towards the rich and despite some tightening of benefit entitlements. The probable explanation is that returns on investment fell (or alternately that they were over-elevated before the crisis struck) while unemployment has not been too heavily affected by the downturn (in New Zealand, but not everywhere).

 

14. Poverty Measurement

 

Systematic studies in poverty in New Zealand began in the 1970s with Peter Cuttance’s pioneering survey of large families. (Cuttance 1974) A comprehensive survey of the state of poverty research would be bigger than this survey on economic inequality. This section focuses on the insights the measurement of poverty sheds onto economic inequality generally.

The proportion of the population in poverty below some (equivalised) income level can be thought of as a measure of inequality. It is, of course, no more definitive than any other indicator and, like each of the others, it reflects the purpose for which it is being used. If the concern is the power of the plutocracy, a measure such as the share of the top one percent is more useful; if the concern is the deprivation of opportunities for future generations, the proportion of the young in poverty is more relevant.

 

Defining the poverty threshold has not been easy, in part because of the New Zealand tendency to ignore what has gone on before rather than engage in a scholarly dialogue of research and analysis.[50] A useful foundation for discussing poverty is the report of the 1972 Royal Commission on Social Security. While it did not directly address a suitable level, its deliberations while setting the social security benefit level apply.

 

The Commission argued that the aims of the social security system should be:

(i) First, to enable everyone to sustain life and health;

(ii) Second, to ensure, within limitations which may be imposed by physical or other disabilities, that everyone is able to enjoy a standard of living much like that of the rest of the community, and thus is able to feel a sense of participation in and belonging to the community. (Royal Commission 1972:65, original’s italics.) [51]

 

The first aim is a statement that people should not be in ‘absolute’ poverty. The second is that they should not be in ‘relative’ poverty, a version of the moral argument that the need for social coherence requires that those at the bottom should share in the rising prosperity of the community as a whole.

 

A poverty threshold can be derived from the empathetic but frugal Commission’s deliberations, since it would have set the benefit level close to, if not on, the threshold. In practice early research used the benefit level for a married couple as the benchmark.

 

How to update a relative poverty threshold? Obviously it needs to be increased with inflation but what to do about changing living standards? The first known attempt to do this argued for increasing the poverty threshold as mean incomes increased. (Easton 1980) But for some reason, the median income became more popular because – it was claimed – it was more reliable when the upper tail of the distribution was poorly measured.[52] Unfortunately the resulting poverty threshold is sensitive to redistributional policy, sometimes in paradoxical ways.

 

Consider the following income distribution for three individuals: 4,10, 16. Suppose the poverty threshold (the proportion in poverty is another measure of inequality) is defined as 50 percent of the median. The median income is 10 in this case, so the poverty threshold is 5 and one of the three individuals is in poverty.

 

Now suppose that 2 units are transferred from the middle individual to the top individual; the distribution is now 4, 8, 18. The median income falls to 8, the poverty threshold is now 4 and nobody is below it. Paradoxically an increase in inequality by transferring income from the middle to the rich eliminates poverty according to a definition based on median incomes.

 

This is not just a theoretical possibility. The changes in tax and benefits in the late 1980s and early 1990s had exactly this impact so, according to the median-based measure, poverty fell – much to the public delight of the inegalitarians – despite objective evidence that New Zealand’s poor were suffering from greater economic stress as their incomes fell. (Easton 2002)

 

There have been attempts to provide other measures of poverty thresholds. One involved asking focus groups (selections of people with chosen characteristics) from lower incomes what they thought the thresholds should be.[53] One would like to report that the conclusion was a level similar to that recommended by the Royal Commission. That was true in the case of two adult-three child households but not for the one adult-two child group. In fact the proposed thresholds for the two groups were not rationally coherent. (Stephens, Waldegrave & Frater 1995; Easton 1997) This may explain why subsequently the research group seems to have abandoned the focus group approach and used a proportion of the median instead. (From the frying pan into the fire.)

 

One other significant attempt to estimate a poverty threshold, albeit implicitly, was by a visiting American scholar who used American-based methods that had been abandoned by the Royal Commission almost two decades earlier. The proposed ‘minimum adequate’ income for benefit levels, funded by Treasury, became the foundation for the savage budget cuts of 1991. (Brashares 1993; Easton 1995)

 

The cuts heralded a new way of thinking about benefit levels and hence of poverty (or perhaps a reversion to a very old way). In effect the Royal Commission’s principle of relative poverty was abandoned to be replaced by absolute poverty, sufficient to enable the sustaining of life and health (some would argue that even that is not met by current benefit levels) but not sufficient to enable people to feel a sense of participation in and belonging to the community. Since 1991 the benefit level has been increased with consumer prices but not with real incomes.[54] Inevitably, as we saw from Table 1, the income share of the bottom quintile fell.

 

To analyse this a little more closely we use the poverty threshold in the MSD study – 60% of the 1998 median which is constant in real terms over time. (Many would think this threshold is too low.[55]) The estimate is shown in column 5 of Table 2. (Perry 2013) In the 1980s the percentage below the threshold was in the low teens. The 1991 benefit cuts almost doubled the percentage to the mid-20s, but from the mid-1990s it fell back to below the 1980s level; in 2012 (the last available year) it was about 6 percent.

 

However this is conceptually an absolute level of poverty; a threshold that is fixed in real terms does not allow for the poor sharing in prosperity. Column 6 of Table 2 illustrates what happens with a relative poverty threshold which parallels changes in average standards of living. In the period from 1982 to 1997 the mean of equivalised household income was broadly flat albeit with some decline to 1994. It is assumed that in such depressed circumstances the threshold is not changed, that is, the poor do not share the pressure of a short downturn.[56] (Easton 1980) After 1997 living standards, as measured by equivalised household income, rose steadily through to 2009 (by about 2.4% p.a.) A relative poverty level assumes that the poverty threshold rose in parallel. Average incomes have been broadly stagnant since 2009 so the threshold is maintained at its 2009 level.

 

The story of proportions in poverty, told in column 6 of Table 2, remains the same until 1997 because the threshold is the same.[57] After that the relative poverty proportions do not fall as much as in the absolute poverty case, and remain in the high (rather than low) teens. One might conclude that on this measure there was higher (relative) poverty after 2000 than there was before 1990. On that measure income inequality also increased.

 

The next step is to ask whether it matters to be in poverty. In what way do those on low incomes have life experiences different from those with higher incomes? Any answer requires an analysis of the whole of the distribution. Data bases which assist answering such questions have only recently come available and are yet to be fully exploited. (Perry 2012: Sections K, L) They are likely to show that disposable income is a useful indicator of life experiences and hardship, but there are other important factors (some of which can be influenced by policy) including education, health, household assets, housing and urban/rural location,

 

15. Distribution by Social Groups

 

Social groups are not randomly scattered through income distribution but tend to cluster in particular parts of it. The New Zealand research has focussed on who the poor are and on the ethnic distributions.[58]

 

In addition to drawing attention to the poverty issue and to tracing changes in the income distribution over time, a significant analytic use of poverty numbers is to identify who are among the poor, The conclusion, first identified in the mid-1970s and elaborated since, still has not been entirely absorbed by the conventional wisdom. It tends to think of the typical poor as a brown solo mother, with many children, living on the benefit in rental accommodation.

 

In fact the majority of the poor are couples with jobs, with some – but not a lot of – children living in their own home albeit with a mortgage. (Because the New Zealand Superannuation benefit level has tended to be above the poverty threshold, the stereotype that the New Zealand poor contain a lot of elderly has almost disappeared; Overseas the minimum public provision for the elderly is often less generous.[59]) The proportion in poverty is higher among solo parents, those without jobs, living in rental accommodation; the proportion of the poor with a brown ethnicity is higher but there are fewer of them. (More women than men are poor.)

 

What the conventional wisdom has confused is that while a higher proportion of a particular category, say solo mothers, may be in poverty, the numbers of poor are this proportion times the total numbers in the category. Thus there may be more poor in a group with a lower prevalence but a larger total number.

 

The salient conclusion from the research is that over 80 percent of the poor are children and their parents (and others in their households) and that proportionately more children are in poverty than adults (especially excluding parents).[60]

 

It is well established that the brown ethnicities tend to be over represented among the poor.[61]  Mean Maori equivalised household incomes were 90 percent of average incomes in 2012 and Pasifika ones were 89 percent. (European/Pakeha ones were 107.5 percent.) However there may well be factors other than ethnicity which explain whole or part of the differences. For instance, a high proportion of the incarcerated are Maori, but this is in part of due to their younger age distribution. Allow for that and the proportion comes down (but is still higher than the Pakeha incarceration rate). The same applies to unemployment, which like incarceration is more concentrated among young adults.[62] Part, but not all, of the lower incomes of those with brown ethnicity may be due to the population being younger and having more children. This is rarely investigated.

 

There is a curious feature of income trends over time, summarised in Table 5.

 

Table 7: Equivalised Median Household Income by Ethnicity

$2012 prices

Ethnicity

1988

2012

Change

European/Pakeha

28000

35800

27.9%

Maori

23200

30000

29.3%

Pasifika

22700

29800

31.3%

ALL

26700

33300

24.7%

Source: Perry op cit, p.80.

 

Apparently over the 24 years the median income of each group has risen faster that the median for all ethnic groups. This may be a peculiarity of medians. But it also may reflect the changing shares of the categories if, say, the brown share of the population had risen relative to the Pakeha one.[63]

 

Discussions on ethnic inequality are disappointing.[64] They tend to draw attention to some aspects of the inter-ethnic inequality but do little deeper analysis.[65]

 

A particular problem for analysts arises from changes in pricing. To illustrate suppose free schooling for children was withdrawn but families were given exactly the same allowance as an income grant. They would not be markedly better off, but their disposable income would seem to have risen. Their equivalised income should not.[66]

 

Over time, the household equivalence scales should be regularly updated. They are not. (Easton 2006) An even trickier problem arises when the government subsidises the service for some groups rather than others (e.g. free visits to general practitioners for the young). Ad hoc attempts to assess their overall impact suggest the effect is not great although they matter a lot to those in need.[67] (Easton 1996, 1999)

 

16. The Dynamics of Inequality

 

Income inequality would not matter so much if for each period one was randomly assigned to an income level so that, for instance, those in the bottom decile had as good a chance as anyone else of being in the top one next time. In reality a major determinant of income in one period is the past record of income. But to what extent is this the case?

 

A study by John Creedy (1997b) based on personal income tax data came to the conclusion that ‘the present value, over the age 20 to age 65, displays less inequality than in any single year’. One would be surprised if this was not true, although it is well to be reminded. However the result arises from simulation – the data base is for only three years – whereas it is possible – if unable to be tested – that the auto-correlations of incomes involve much longer lags. So the specific quantitative results are not decisive.

 

Also using income tax data, Harry Smith and Robert Templeton (1990) found that 25 percent of those in the bottom quintile were in a higher quintile a year later, while seven years later some 45 percent were. Note, however, this includes shifts from being a low-paid student to full-time well-paid employment. It also includes those who moved from the very top of the bottom quintile to the very bottom of the second bottom quintile. The relative income gains may be trivial.

 

The Survey of Family, Income and Employment (SoFIE) followed a sample of households over seven ‘waves’ (years), so it is possible to track the change in relative income over the period.

 

Kirstie Carter and Fiona Imlach Gunasekara (2012, Table 5) measure the shifts between quintiles. Some 45 percent of those in the bottom quintile remain there six years later. Over 80 percent were below the median income. In contrast 54 percent of those in the top were still there, and 84 percent of them were above the median income six years later. Whether this is a high or low level of immobility is a value judgement but, given all the potential life events that can occur in six years which might have the potential to disturb an (equivalised) household income, one is inclined to think there is considerable inertia.

 

Bryan Perry (2013:194-197) points out the New Zealand experience is not greatly different from that of other rich countries. His calculations suggest that about two-thirds of the change occurs in the first year, and shifts over the following years are much smaller. This may suggest that while some households temporarily change quintiles to revert the following year, longer term change is not great, and often the result of a life event.

 

Perry distinguishes those in ‘current’ poverty from those in ‘chronic’ poverty with average annual income over the seven waves being below the poverty line. He concludes there is considerable persistence:

– in any wave, around half are in both chronic poverty and current poverty, the other half being only in current poverty (i.e. more temporary or transient poverty)

– the people in this more transient group change a lot over seven waves, which is why it turns out that the number in low income at least once in seven waves is more than double the number in low income at any one time

– in addition to those identified as being in current poverty in a wave there is another one in five (i.e. 3% of the whole population (20% of 15%)) who are in chronic but not current poverty

– for children, 60% of those in current poverty are also in chronic poverty, and there are another one in five in chronic but not current poverty at each wave

– very similar findings have been produced for the UK and Australia. (Perry 2013:199-201)

 

If a higher poverty threshold than the 50 percent of median equivalised income is used here (60 percent is preferred by some commentators), the proportions quoted here will be on the low side.

 

The issue of intergenerational dynamics is hardly explored. We know that children from homes with lower incomes are likely to be sicker and get inferior education and training than those from homes with higher incomes. The quantitative extent to which this translates into inferior life opportunities is not known.

 

17. Income and Health

 

Suzie Ballantyne (Carson) and the author found an association between health status and income, illustrated by Table 6.

 

Table 8: Proportions in Quintiles Who Rate Themselves as ‘Fair’ or ‘Poor’ Health

Household Quintiles

 

ALL

Age Group

Bottom

2

Middle

4

Top

15 Under Female

7.4

3.6

2.9

3.1

1.2

4.5

Under 15 Male

5.8

5.8

3.4

3.0

3.1

4.7

15-64 Female

10.7

11.0

8.7

7.2

3.9

7.9

15-64 Male

10.2

8.3

7.4

4.2

3.8

6.2

Over 65 Female

42.5

29.4

27.5

22.3

21.2

26.5

Over 65 Male

40.2

29.8

25.2

21.3

19.4

25.5

ALL

9.3

11.0

11.0

7.0

4.8

8.6

Source: Easton & Ballantyne (2002) Chapter 9.

 

The general pattern is that in any given age group, those in the lower income quintiles are in poorer health than those in the higher income quintiles.

 

Note that as age increases health status deteriorates, and that after the age of 15 women tend to report being in poorer health than do men. This generates an apparent anomaly in the aggregate pattern, with those in a household in the second to bottom and middle quartiles prone to poorer health than those at the bottom. This is because the age groups are not equally distributed through the quintiles. Because the elderly are concentrated more in the second and middle quintiles, and because they are more likely to be sick, those quintiles report the highest incidence of sickness for the population as a whole.

 

That there is an association does not prove causation. Indeed it is possible that the causation is in either direction. A history of poor health may reduce current income. On the other hand low income may cause poor health via inadequate spending on food and other health-promoting but standard consumer products (especially inadequate housing) or on health care. This data does not tell us which mechanism is working – probably both are.

 

Research at the University of Otago’s Wellington school of Medicine has observed that the incidence of tuberculosis, other infection diseases and acute rheumatic fever and TB is higher in areas (Census Area Unit) with lower incomes. (Baker et al 2008, Baker et al 2012, Jaine et al 2004)

 

18. Wealth

 

Wealth is the stock which generates the flow of income. The focus tends to be on physical assets (such as housing) and financial assets (such as shares and bank deposits). Additionally there is human capital which generates labour market income; there may also be state entitlements such as eligibility for social security and New Zealand superannuation, which also generate a flow of income for some.[68]

 

Measuring the distribution of wealth is not easy; once more it depends on the data sources. An early attempt to measure physical and financial wealth used returns of estates reported for death duties, treating the wealth of those who have died as a random sample of the living. Since wealth varies over the life cycle it is necessary to treat each age group separately; ideally there needs also to be an adjustment for those with wealth having a lower mortality than those without, although that has not been practical. In any case the growing practice of exempting the joint family home wrecked the method, since it meant the reported dead person was not representative of the living whenever there was a surviving spouse.

 

The estimates which may be least criticised give the following general conclusions for 1956 and 1966:

– physical and financial wealth is much more concentrated than personal income;

– there is a life cycle to wealth holdings, but even within each age cohort, wealth is very unequally distributed;

– the main form of this wealth holding is housing.

None of these conclusions are particularly surprising. (Easton 1983: Ch 7) [69]

 

A more recent and refined estimate came from SoFIE which in 2003/04 asked individuals the value of their assets, liabilities and net worth.[70] (Cheung 2007)

 

Again it found that the vast majority of the adult population had little physical and financial wealth – about 6.5 percent of them had negative net worth, although this may be dominated by student debt. Conversely 1 percent of adults had 16.4 percent of total wealth, the same as about 70 percent of the population.[71] Again the study shows a marked life cycle in net worth, peaking at about the age of 60.

 

There is not a lot of gender inequality, but each Pakeha owns about 2.6 times that of the other ethnic groups. One of the report’s curious findings is that Auckland wealth levels are markedly below those of the rest of the country; we do not know the extent to which age and ethnic factors drag them down. The report illustrates that a lot of material is available at the unit record level, but alas it has not been exploited, so we cannot be sure the extent to which other variables mask the true correlations.[72]

 

A subsequent publication adds to the earlier report. (Statistics New Zealand 2008) While a couple with two dependent children has higher median net worth than a family with one dependent child, those with three or more children have less than those with two. The same applies for one parent families. Because of the way the data is presented it is not possible to make an exact comparison but a family with two parents certainly averages more than twice as much net worth as a single adult family – probably four times as much. (SNZ 2008: Tables 2 & 3, pp.7 & 8) The pattern for the mean (average) is not quite as regular. However there may be intervening variables.

 

19. Housing

 

While owner-occupied housing is a part of total wealth, it is unusual because it does not generate cash income for the owner (and hence itds income effect is not directly measured). Instead it saves expenditure. Thus two otherwise identical households with the same cash income may have very different standards of living if one is servicing a crushingly heavy mortgage while the other is mortgage free. How to allow for this?

 

A common adjustment is to deduct housing costs from the income. Since one element constitutes expenditure and the other constitutes income, this must be conceptually confused. Moreover the same household equivalent scale is applied whether the measure is before or after housing costs. That is surely nonsensical since the main reason for economies of scale that the household equivalence scale is adjusting for is housing; a larger family does not need proportionally as much housing – double its size and one does not double the number of kitchens, laundries and so on.

 

Suzie Ballantyne and the author suggested an alternative conceptually robust procedure in which an income was imputed to a household reflecting the difference between average household outlays (based on household characteristics) and actual outlays. We concluded

 

… adjusting for housing circumstance reinforces the inequalities in household type, benefiting the groups who are generally better off. This is not surprising. Because of the taxation advantages – the return on capital invested in owner-occupied housing is not taxed to the extent of other investments, including rental housing – the wealthy are likely to invest in their own accommodation. Additionally, there is a life cycle effect. Households with children owning a house are likely to have higher outlays than if they were renting because they are still paying off a mortgage. The households benefit later in life on the lower outlays of freehold owner-occupier housing (by which time the children will have probably left). (Easton & Ballantyne 2002: Chs 3,7)

 

While this approach has not been generally taken up, neither have the conceptual inadequacies of ignoring differences in housing circumstances or of deducting housing expenditure for income been addressed.

 

Housing well illustrates the lamppost problem. There is considerable anxiety about house ownership, the proportion of which has been falling. Despite it being obvious that there is a home ownership life cycle, until recently there was no data. Too often researchers and commentators have been left with the lamppost of the aggregate proportion across the life cycle.

 

The historical experience has been that since the Great Depression, proportionally more homes have been owned (with and without mortgage) by their occupiers. (Schrader 2013) However the proportion has been falling off since the 1986 census. It is too easy to attribute this to the rising income inequality. However there has been demographic change and there may well have been a shift to people owning a home in one location but living elsewhere while renting. There has also been social change. The impression is that the younger generation is generally settling down later than their parents and grandparents, more often in their thirties.[73] That probably means they prefer renting to ownership in their twenties. We do not know how much this (and the changing accommodation choices of the elderly[74]) explain the falling proportion of ownership.[75]

 

20. Inequality and Growth

 

It is conventional to interpret per capita constant price GDP (or better still national income) in welfare terms, ignoring the distributional impact of any gains (or losses) of output. Thus there is exactly the same outcome if the all gains go to the very rich as if the gains are shared equally among everybody. Sen suggested an alternative measure which he called ‘real national income’, which takes into account the distributional impact of changes. (Sen 1976; 1979) Ultimately it deducts from national income a share, measured by the Gini coefficient, to reflect the degree of inequality.

 

Thus while per capita National Income in constant prices (adjusted for the additional spending power arising from improving terms of trade) rose at a trend rate of 1.66 percent p.a. between 1982 and 2012, Sen’s real national income rose only 1.34 percent p.a. because of the rise in income inequality. In effect the additional inequality cost New Zealand almost a fifth on Sen’s measure. (Easton 2013C)

 

More generally, there is little evidence that inequality generates faster growth (on the conventional measure of output or GDP). High per capita economies tend to have lower inequality although the causal processes and directions are not well understood.

 

21. Epilogue: Towards Policy Responses

 

Those who were not there may find the extremism of the New Zealand government’s economic and social policy in the 1980s almost unbelievable. It had many aspects to it but relevant to this paper was a seminal shift away from the egalitarianism of the post-war era as the basis for policy to downgrading fairness and hence accepting considerably higher levels of inequality. It is unclear how conscious this abandonment was.

 

The New Zealand way tends to be policy without analysis. As André Siegfried said a century ago, New Zealanders’

outlook, not too carefully reasoned, and no doubtful scornful of scientific thought, makes them incapable of self-distrust. Like almost all men of action they have a contempt for theories: yet they are often captured by the first theory that turns up, if it is demonstrated to them with an appearance of logic sufficient to impose upon them. In most cases they do not seem to see difficulties, and they propose simple solutions for the most complex problems with astonishing audacity. (Siegfried, 1992:54)

 

Too often the simple solution fails, but by that time the proponent has retired, been promoted or moved on to another complex problem so that no one is ever held to account for the failure (especially given that retrospective evaluation of policy is rare, so policy makers cannot learn from failure).

 

Rather than propose some simple solutions to poorly formulated problems, this epilogue draws attention to two classes of policy responses if excessive economic inequality is judged to be requiring attention.

 

The classical policy response is redistribution, that is, the use of taxation, transfers and subsidies to move resources from those with more to those with less. Since the rise in inequality a quarter of a century ago can be mainly attributed to regressive changes in distributional policy, the simple policy is to reverse those changes. It is, of course more complex than that but …

 

The second policy response is predistribution, which is the notion that the state should try to prevent inequalities in market incomes occurring in the first place rather than ameliorate inequalities through redistribution. (Hacker 2011) Although the term has only come into prominence recently, the idea is a long held one, insofar as numerous traditional interventions in education, health care, housing and training, can be seen as precursors. It can be extended to wages, as British Labour leader Ed Miliband has recently proposed; it could be applied to the current government’s proposals to shift beneficiaries into the labour force. Indeed, the term is so loose it may be used for just about any policy with distributional implications, effective or otherwise (and therefore is an ideal notion for New Zealand policy pragmatists).

 

However, we have seen that there is not a simple relationship between personal market incomes and household market incomes. The relationship is so complex that it is not impossible that reducing inequality in personal incomes could increase inequality in household incomes partly because children may not be a beneficiary and partly because higher incomes for supplementary earners which would add to already well of households.

 

Nevertheless, predistribution focuses on the options when distributional policy has been fine tuned as much as practicalities allow. Over the last quarter century we have become more alert to the disincentive and perverse incentive effects of redistribution (which seem to have been magnified by globalisation – the greater international mobility of capital and skilled labour). Other measures which shift social inequality in the desired direction should not be ignored.

 

But predistributional policies need to work with effective distributional polices not replace them, especially as they often have an investment element which makes them fiscally expensive in the short run.

 

Prior to that, those who command policy – whether effectively or ineffectively – have to decide to what extent reducing (or increasing) economic inequality is a policy objective. Is New Zealand satisfied with shifting from a low inequality to a high inequality society? What would its founding nineteenth century migrants have thought about the fact that, after allowing for each country’s size and affluence, New Zealand is now more unequal than the countries they left? And what would those who invited them here have thought had they known their descendants would be firmly at the bottom end of the unequal distributions?

 

References

 

This list of references includes those mentioned in the survey, plus other significant New Zealand publications in the area.[76]

 

Atkinson, A. (1975) The Theory of Inequality (Oxford OUP)

Atkinson, A. & A. Leigh (2007a) ‘Top Incomes in New Zealand 1921-2005: Understanding the Effects of Marginal Tax Rates, Migration Threat and the Macroeconomy’; Review of Income and Wealth, 54(2):149-165.

Atkinson, A. & A. Leigh(2007b) ‘The Distribution of Top Incomes in New Zealand’ in A. B. Atkinson & T. Piketty (eds) Top Incomes over the Twentieth Century. A Contrast Between Continental European and English-Speaking Countries (Oxford OUP) Ch 8. Updated in World Top Income Data Base

http://topincomes.g-mond.parisschoolofeconomics.eu/

Aziz, O., M. Gibbons, C. Ball & E. Gorman (2012) ‘The Effect on Household Income of Government Taxation and Expenditure in 1988, 1998, 2007 and 2010′, Policy Quarterly 8.1, February 2012, 29-38

Baker, M.  D. Das, K,  Venugopal & P. Howden-Chapman (2008) ‘Tuberculosis Associated with Household Crowding in a Developed Country.’ J Epidemiol Community Health; 62: 715–721.

Baker, M., L. Barnard, A. Kvalsvig, A. Verrall, J. Zhang, M. Keall, N. Wilson, T. Wall &  P. Howden-Chapman (2012) ‘Increasing Incidence of Serious Infectious Diseases and Inequalities in New Zealand: a National Epidemiological Study,’ www.thelancet.com  Vol 379 March 24, 2012, 1112-9.

Bakker, A. & J. Creedy (1999). ‘Macroeconomic Variables and Income Inequality in New Zealand: An Exploration using Conditional Mixture Distribution.’ NZ Economic Papers 33(2):59-79.

Barker, G. (1996) Income Distribution in New Zealand (Wellington, IPS)

Brashares, E. (1993) ‘Assessing Income Adequacy in New Zealand’, New Zealand Economic Papers, 27(2):185-207.

Brosnan, P. (1986). ‘The Uses and Limitations of Census Income Data’ in Crothers C (ed) The Uses and Limitations of Census Data: Recent Censuses to 1981, Occasional Paper No. 6. December 1986. NZ Demographic Society.

Carter, K. & F. Imlach Gunasekara (2012). Dynamics of Income and Deprivation in New Zealand, 2002-2009. A Descriptive analysis of the Survey of Family, Income and Employment (SoFIE) Public Health Monograph Series No. 24, Department of Public Health, University of Otago, Wellington.

Cheung, J. (2007) Wealth and Disparities in New Zealand (Wellington, SNZ)

Coleman, A. (Sept 1999). The Life-Cycle Model, Demographic Change, Savings and Growth. Unpublished Treasury Working Paper.

Crampton, P, & P. Howden-Chapman (1997) Socioeconomic Inequalities and Health. Proceedings of the Socioeconomic Inequalities and Health Conference (Wellington, IPS)

Crawford, R & . G. Johnston (2004) Household Incomes in New Zealand: the Impact of the Market, Taxes and Government Spending, 1987/88–1997/98, Treasury Working Paper 04/20

Creedy, J. (1997a). ‘Earnings Dynamics over the life cycle: new evidence for New Zealand’. New Zealand Economic Papers. 30, 131-153.

Creedy. J. (1997b). Statics and Dynamics of Income Distribution in New Zealand. (Wellington IPS)

Cuttance, P. (1974) Poverty Among Large Families in New Zealand, (University of Waikato, MSS Thesis)

Dixon, S. (1995) ‘The Inter-Industry Wage Structure’, Labour Market Bulletin, 1995:1;41-71.

Dixon, S. (1996) ‘The Distribution of Earnings in New Zealand: 1984-94′, Labour Market Bulletin:45-100.

Dixon , S. (1998). ‘Growth in the Dispersion of Earnings: 1984-97’. Labour Market Bulletin:71-107.

Easton, B. (1973F) Family Incomes, Mimeograph, University of Canterbury.

Easton, B. (1973N) A Needs Index, Mimeograph, University of Canterbury.

Easton, B. (1973P) Poverty in New Zealand, Mimeograph, University of Canterbury.

Easton, B. (1976) ‘Poverty in New Zealand: Estimates and Reflections’ Political Science, Vol 25, No 2, December 1976; reproduced in (Easton 1983)

Easton, B. ‘The Economic Life Cycle of the Modern New Zealand Family’, Australian and New Zealand Journal of Sociology, February 1977:85-89.

Easton, B. (1980) Poverty in New Zealand; Five Years After, Paper to the 1980 Conference of the New Zealand Sociological Association (Christchurch, University of Canterbury)

Easton, B. (1983) Income Distribution in New Zealand, NZIER Research Paper 28.

Easton, B. (1995) ‘Properly Assessing Income Adequacy in New Zealand, New Zealand Economic Papers, 29(1), 1995:89-102

Easton, B. (1989) ‘Notes Towards the Distributional Consequences of Policy Changes’, Social Policy and Inequality in Australia and New Zealand, Social Welfare Research Centre Reports and Proceedings, No 78, Sep 1989: 171-194.

Easton, B. (1995) ‘Poverty in New Zealand: 1981-1993′, New Zealand Sociology, Vol 10, No 2, November 1995.

Easton, B. (1996). ‘Income Distribution’ in B. Silverstone, A. Bollard & R. Lattimore, A Study of Economic Reform – the Case for New Zealand (Amsterdam North Holland) p.101-138.

Easton, B. (1996) In Stormy Seas (Dunedin UOP) p.92-95.

Easton, B. (1997). ‘Measuring Poverty: Some Problems’. Social Policy Journal of New Zealand 9, November:

Easton B. (1999) ‘What has Happened in New Zealand to Income Distribution and Poverty Levels,’ S. Shaver & P. Saunders (ed) Social Policy for the 21st Century: Justice and Responsibility (SPRC, Sydney, 1999) Vol 2:55-66.

Easton, B (2000) How Accurate Are the Incomes Reported in the Household Economic Survey?

http://www.eastonbh.ac.nz/1997/02/how_accurate_are_the_incomes_reported_in_the_household_economic_survey/

Easton, B. (2002) ‘Beware the Median’ SPRC Newsletter No 82, November 2002: 6-7.

Easton, B. Econometrics of Household Equivalence Scales, Wellington Statistics Group, 11 February, 2004:

http://www.eastonbh.ac.nz/2004/02/the_econometrics_of_household_equivalence_scales/

Easton, B. (2010) ‘Income and Wealth Distribution’ Te Ara: the Encyclopedia of New Zealand

http://www.teara.govt.nz/en/income-and-wealth-distribution

Easton B. (2012) ‘The Return of Unemployment’, Chapter 39, Not in Narrow Seas: The History of New Zealand from an Economic Perspective, forthcoming, draft available from the author.

Easton, B. (2013C) Calculating Sen’s Real National Income for New Zealand

http://www.eastonbh.ac.nz/2013/09/calculating-sens-real-national-income-for-new-zealand/

Easton, B. (2013E) Ethnicity, Gender, Socioeconomic Status and Educational Achievement: An Exploration (Wellington, PPTA)

Easton (2103T) Top Taxpayers: 1981-2011 http://www.eastonbh.ac.nz/2013/09/top-taxpayers-1981-2011//

Easton, B. & S. Ballantyne (2002) The Economic and Health Status of Households (Wellington School of Medicine)

Expert Advisory Group on Solutions to Child Poverty (2012) Solutions to Child Poverty in New Zealand (Wellington, Office of the Commission for Children)

Hacker, J. (2011) the Institutional Foundations of Middle-class Democracy (London, Policy Network)

Hansen, P. (1997). ‘Inference on ‘Earnings Dynamics over the Life Cycle: New evidence for New Zealand’ New Zealand Economic Papers. 31(2):221-7.

Howden-Chapman P, N. Wilson & T. Blakely T. (eds.) (2000). Socioeconomic Inequalities and Health. Report to the Ministry of Health, Wellington.

Hyslop, D (1999). A Dynamic Analysis of Individual’s Market and Disposable Income. Unpublished research paper, NZ Treasury.

Hyslop D & D. Maré 2001). Understanding changes in the distribution of household incomes between 1983-86 and 1995-98. Treasury Working Paper 01/21

Jaine, R., M. Baker & K. Venugopal (2011) ‘Acute Rheumatic Fever Associated With Household Crowding in a Developed Country,’ The Pediatric Infectious Disease Journal Vol 30, Number 4, April 2011 1-5

Jensen, M. & W. Meckling (1983) Democracy in Crisis (St. Leonards, N.S.W, Centre for Independent Studies)

Maloney T. & G. Barker (1999). The Low-Income Dynamics of Families in the Christchurch Health and Development Study, Law and Economics Consulting Group. Report to Treasury.

Martin, B (1997) Away from Equality: Change in Personal Incomes, 1951 to 1991 Discussion paper No. 20. Population Studies Centre. University of Waikato

Martin , B. (1997) Income Trends among Individuals and Families, 1976 to 1996 Briefing Paper 2, Population Studies Centre, University of Waikato.

Martin, B. (1998) Incomes of Individuals and families in New Zealand, 1951 to 1996 (PhD thesis, University of Waikato)

Martin, B. (1998) Men, Women, Employment and Earnings: Change in the Distribution of Income among Individuals, 1976 – 1996 Population Studies Centre, University of Waikato.

Martin, B. (1999) Sub-National Income Differentials, 1986 – 1996 Population Studies Centre, University of Waikato.

Martin, B. (2000) ‘Tracking Down Inequality: Assessing the Impact of Socio-Demographic and Employment Changes upon Inequality of Family Incomes 1976-1996’ New Zealand Population Review.

Milliband, E (2012) Speech to the Stock Exchange, 6 September 2012http://www.politics.co.uk/comment-analysis/2012/09/06/ed-miliband-s-redistribution-speech-in-full

Mowbray, M. (1993) Incomes Monitoring Report 1981-1991 (Wellington Social Policy Agency)

Mowbray, M (2000) Incomes Monitoring Into the 1990s (Wellington , Ministry of Social Policy)

New Zealand Planning Council (1988), For Richer or Poorer – Income and Wealth in New Zealand (Wellington)

New Zealand Planning Council (1990) Who Gets What? – The Distribution of Income and Wealth in New Zealand (Wellington)

O’Dea (2000) The Changes in New Zealand’s Income Distribution New Zealand Treasury Working Paper 00/13.

OECD Working Party on National Accounts (2013) Distributional Measures Across Household Groups in A National Accounts Framework STD/CSTAT/WPNA(2013)10/RD

Perry, B. (2002), ‘The Mismatch Between Income Measures and Direct Outcome Measures of Poverty’ Social Policy Journal of New Zealand 19:101-127.

Perry, B. (2005) the Choice of Income Sharing Unit for Poverty Measurement: Household or Economic Family Unit Unpublished discussion paper, Ministry of Social Development, Wellington.

Perry, B. (2009) ‘Non-income measures of material wellbeing and hardship: first results from the 2008 New Zealand Living Standards Survey, with international comparisons.’ Working Paper 01/09, Ministry of Social Development, Wellington.

Perry, B. (2014, forthcoming) ‘Non-income measures of material wellbeing and hardship in New Zealand: full findings from the 2008 Living Standards Survey and the Household Economic Survey, 2006-07 to 2011-12.’

Perry, B. (2103)There are earlier editions.

Podder, N. & S. Chatterjee (1998) Sharing the National Cake in Post Reform New Zealand: Income Inequality Trends in Terms of Income Sources. Paper to 1998 Conference of NZ Association of Economists.

Rashbrooke. M. (ed) (2103) Inequality: A New Zealand Crisis (Wellington, BWB)

Rawls, J. (1999) A Theory of Justice: Revised Edition, Cambridge: Harvard University Press. (Original edition 1971)

Royal Commission on Social Security (1972) Social Security in New Zealand (Wellington, Government Printer)

Schrader, B. (2010) ‘Housing – Tenure’, Te Ara – the Encyclopedia of New Zealand

http://www.TeAra.govt.nz/housing-tenure

Sen, A. (1976). ‘Real national income’, The Review of Economic Studies, 43 (1), 19-39.

Sen, A. (1979). ‘The welfare basis of real income comparisons’, Journal of Economic Literature, 17 (1), 1-45.

Sen, A. (2009) The Idea of Justice (London, Penguin)

Skeptic’s Play (2013) The Gini Coefficient of Log Normal Wealth,

http://skepticsplay.blogspot.co.nz/2013/03/the-Gini-coefficient-of-log-normal.html

Smith, H. & R. Templeton (1990) A Longitudinal Study of Incomes Statistics New Zealand

Siegfried, A. (1904, 1992)) Democracy in New Zealand

Statistics New Zealand (2001) Household Savings Survey

Statistics New Zealand (2008) Family Net Worth in New Zealand

Stephens, R., P. Frater & C. Waldegrave(1995) ‘Measuring Poverty: Some Rebuttals of Easton.’ Social Policy Journal of New Zealand 9 December:

Stephens, R., P. Frater & C. Waldegrave (1997) ‘Measuring Poverty in New Zealand.’ Social Policy Journal of New Zealand 5, December:

Stephens, R., C. Waldegrave & P. Frater (1995) ‘Measuring Poverty in New Zealand’ Social Policy Journal of New Zealand 5 December:88-112.

Stiglitz, J. E. (2012) The Price of Inequality (NY, Norton)

Stiglitz, J. E. (2013) ‘Inequality is Holding Back the Recovery’ New York Times 19 January, 2013.

Stroombergen, A. D. Rose & J. Miller J (Aug. 1995) Wealth Accumulation and Distribution: Analysis with a Dynamic Microsimulation Model (Wellington, BERL)

Wade, R. (2103) Inequality and the West (Wellington, BWB)

Waldegrave, C., S. Stuart & R. Stephens (1996) ‘Participation in Poverty Research: Drawing on the knowledge of low-income householders to establish an appropriate measure for monitoring social policy impacts.’ Social Policy Journal of New Zealand 7

Wilkinson, R. & K. Pickett (2009)The Spirit Level: Why More Equal Societies Almost Always Do Better (London, Allen Lane)

 

Endnotes

 

[1] I am grateful for helpful comments from Warwick Armstrong, Micahel Baker, Geoff Bertram, Elizabeth Caffin, Charles Crothers, Jeff Cope, Margaret Galt, Diane Owenga, Susan St John, Sandra Watson and some people who asked not to be named.

[2] Rhetoricians with policy agendas will claim there is a crisis and propose their policy as a solution or partial solution to, say, the change in the level of inequality even though there is no apparent connection between the two. The classic political analysis of crisis politics is Democracy in Crisis (Jensen & Meckling, 1983) which explains how crises are manufactured to pursue policy ends. Ironically, as their title indicates, the writers use exactly the same political strategy themselves.

[3] The comparable story about some researchers is of the drunk who looked for the car keys under a lamppost light, despite having lost them elsewhere, because the light was better there.

[4] There is a caveat. The ‘just savings principle’ requires that justice been maintained through time. The first principle is that each person has an equal right to the most extensive basic liberty compatible with a similar liberty for others. Sen (2009) presents a critique of the Rawlsian Theory of Justice which does not affect this exposition.

[5] The causal hypothesis Wilkinson and Pickett offer involves psychological phenomenon like anxiety and lower self-esteem in higher inequality societies.

[6] Even so, honesty requires acknowledging that a lot of the social expenditure on responding to inequality is not very effective.

[7] A Lorenz curve plots the cumulative share of people from lowest to highest income on the horizontal axis against the cumulative share of income earned on the vertical axis.

[8] There are many other measures than the ones used in this survey – e.g. percentile ratios, such as 80 to 20. None resolve the basic problem that no single measure can capture all the change in a distribution, while a plethora of partial measures adds to the confusion.

[9] The same applies to the log-normal distribution. A useful first approximation to many income distributions is to treat the logarithm of the income as normal. However the approximation is usually crude for actual distributions

[10] The coefficient of variation of the distribution is its standard deviation (a measure of the dispersion or width) divided by the mean (so that it is scaled).

[11] A reasonable approximation for the lognormal distribution for GC in the 0.1 to 0.8 range is CV = 0.15*exp(9.2*GC) (Easton 1983, ibid, p.22-230, Skeptic’s Play 2013) However few distributions are exactly lognormal, certainly not the ones discussed in this survey.

[12] The insert is actually greater because in 1987 most social security benefits were grossed up (NZ superannuation and its predecessors always had been). Insofar as beneficiaries report their pre-tax benefit, some of the increase in low incomes between 1986 and 1991 was spurious and the inequality increase was even greater.

[13] While taxable income has included most transfer incomes since 1987, inclusion was not as comprehensive before then.

[14] Sometimes the data base forces the treatment of trusts as individuals.

[15] In the 1970s it mattered whether the measurement of income inequality covered all adults or just taxpayers. The former showed a fall in inequality over the period; the latter a rise. This was because women who had not been earning entered the paid labour force. Because they generally worked part-time and their pay rates were lower, they were at the bottom of the taxpayer distributions which appeared more unequal, yet their additional earnings boosted lower adult incomes so that the audlt income distribution was less unequal. (Easton 1983: Ch 4),

[16] Among the complications is how to handle households with negative income (typically they are those with self -employed occupants) and what to do with households whose expenditures far exceed their income. A common practice is to eliminate households with negative incomes or to set their incomes at zero.

[17] This assumes that households share their income equally. It may not be true even in families (mothers may deny spending on themselves to protect their children). Furthermore all households are not families (e.g. a student flat) and have no reason to share income equally.

[18] Smaller quantiles (e.g. percentiles) are not generally practical because of the Household Survey’s sample size.

[19] The 2012 year refers to the year ending June 2012. From 1998 and earlier the year ends in March.

[20] There is a major fluctuation in 2011, due to the impact of the Christchurch earthquakes because insurance receipts are treated as income. This limits the ability to assess the effect of the macroeconomic turmoil which followed the Global Financial Crisis.

[21] After allowing for a blip up due to some tax avoidance in the 2000 tax year and a recovery down phase there after.

[22] Once dividends were taxed as a part of the corporation tax regime and then taxed in the hands of the dividend recipient. Tax changes in the 1980s had (roughly) the corporation tax being treated as a prepayment of income tax.

[23] See also Hyslop and Maré (2003)

[24] It might, however, be worth exploring whether there has been an increase in margins for skill and management, which have widened the income distribution.

[25] It is true that the employee share in domestic income has fallen but this is in part due to shifts between employee status and self-employment and an increasing share going to foreign owners of capital. (Easton 2010) Earlier in the 1970s, real wages had risen faster than productivity and then stagnated from about 1985. (Easton 1996)

[26] This caveat is especially important given the social security benefits were ‘grossed up’ in 1987 – that is increased so they were taxed back to the same net level. This almost certainly means that the study did not capture the reductions in the effective levels of benefits as well as the reductions in income tax on upper incomes.

[27] Measured by changes in the Gini Coefficient. The levels for the three periods were 1983-6 = 0.347; 1989-92 = 0.386, 1996-98 = 0.398. They represent a change in the CV of 43percent and 12 percent between the two periods.

[28] The other candidates for the claim are a series of incremental changes over a longish period (the Great Inflation from 1968 to 1990 might be an example) or possibly the introduction of social security in 1939. However the latter involved horizontal redistribution more than vertical redistribution.

[29] In any case it was an extension of earlier tax credits and partly compensated for their loss of value from inflation.

[30] While Israel is not a member of the OECD, it is included in the data base (and here).

[31] Rankings can be sensitive to the imprecise third decimal place of the Gini coefficient..

[32] Using the New York priced equivalence scale is the reason that some of my very early estimates of poverty in New Zealand were much higher than the refined ones.

[33] The adjustment equation is based on the equilibrium equation (standard errors are shown below) that the Gini coefficient =    0 .032 Log(Population) –          0.164 Log (GDP per cap) -0.170                                             (0.013)                                    (0.055)

[34] Estonia, which had been a part of the USSR, is in the middle of the rankings. The other two Baltics are not members of the OECD. The OECD also provides an estimate for the Russian Federation of 0.428 unadjusted, placing it third most unequal between Mexico and Turkey. Adjusting it for being large and poor its Gini would have been 0.360 or seventh, between the United States and Australia.

[35] Not quite fitting into any pattern at the top are Iceland, Switzerland and Luxembourg. Ad hoc reasons may be advanced in each case. (Luxembourg GDP per cap is always problematic because a substantial part of its workforce resides outsides its boundaries.) Note also that the Latinos, the Middle East and North East Asia are only two countries each.

[36] STD/CSTAT/WPNA(2013)10/RD

[37] While it is easier to illustrate the point with alcohol consumption, the issue also applies to income aggregates which may be more important for distributional purposes. Even so the distribution of expenditure aggregates are of interest.

[38] Not perfectly though, because it is possible that the ratio of under-reporting varies by household characteristics.

[39] For an early example of the measurement problem see Easton (2000). SNZ promptly addressed it when their attention was drawn to it.

[40] A caution about primary (i.e. market) income. New Zealand has largely a non-contributory system so its social security payments are not included in primary income with the result that its primary income seems particularly widespread. (See Op. Cit 35).

[41] Op. Cit. 43-4.

[42] The adjusted uses the same equation and variables as for the 1989 period. The population and GDP per cap will definitely have changed and the parameters may have (but probably not by much). However, given the limited number of data points it was not worthwhile doing the re-estimation.

[43] As well as New Zealand inequality overtaking that of Canada, Greece, Japan and the United Kingdom (which are in the 17), it may also have overtaken Estonia, Portugal, Spain and Switzerland.

[44] Sweden’s Gini coefficient is so exceptionally low in 1983 that Moser’s Law suggests we should be cautious when using it.

[45] The handful of spot surveys is not sufficiently consistent to enable comparisons.

[46] Much popular commentary remains, implicitly or explicitly, based on the notion of a standard household of a couple with two children. In 2006, however, only 30.8 percent of households consisted of a couple and some children.

[47] A clear indication of the unreliability of the series is the dramatic fall in the share of the top decile in the late 1930s, when the tax base was extended as a result of the introduction of social security tax. For a cleaner series – albeit it over a shorter (post-war) period, seee Easton (1983) Chapter 10.

[48] E.g. Rashbrooke (2013:25-27).

[49] As already footnoted the Canterbury earthquakes distorted measurement of household income inequality because insurance payments are treated as income receipts rather than offsets for capital losses.

[50] The Expert Advisory Group on Solutions to Child Poverty (2012) is a recent example of using a superficial definition with little attention to past scholarship.

[51] There was a third aim: ‘where income maintenance alone is insufficient (for example, for a physically disabled person), to improve by other means, and as far as possible, the quality of life available.’

[52] The standard error of the estimate of a median is larger than of a mean by about 25 percent (for a big sample from a normally-distributed population).

[53] Groups on higher incomes are likely to recommend an even higher level.

[54] This assessment is based upon work being carried out by the author, but not yet publicly available for confidentiality reasons.

[55] Because the numbers near the threshold are high, a small change in its level can change dramatically the proportions in poverty and the pattern over time. For instance, setting the poverty threshold just above rather, than below, the basic New Zealand Superannuation benefit level has the numbers leap.

[56] Of course, if overall living standards faced a prolonged decline the relative poverty threshold should decline too (to some extent). This does not change markedly the paragraph’s overall story.

[57] The proportion are estimated by the author using a linear interpolation procedure based on the numbers in Table F3 of Perry (2103:110). The method is similar to that used in Easton (1994).

[58} Perry (2012) has numerous tables which can be used to place a variety of groups in the context of the whole distribution.

[59] Because New Zealand superannuation is indexed to average wages and more adults are working, superannuitants may not share in prosperity arising from increasing labour force participation. As a result their benefit level may decline below the poverty threshold because it is set on average household incomes not average worker incomes.

[60] The effect of the working for families tax credit has yet to be assessed.

[61] This survey does not cover the ‘other’ (none of European/Pakeha, Maori or Pasifika) group which is very heterogeneous. Often the sample size is small and the estimates subject to a large standard error.

[62] Educational attainment (say PISA scores) also illustrates the point. It is affected by socioeconomic status. That Maori have a lower SES than Pakeha explains about one third of the difference between their educational attainment levels. (Easton 2013E)

[63] Ethnic identity is self-categorization, which may change over time (and is also affected by the way the statistician aggregates the various responses). If some households were to categorize themselves as Pakeha in 1988 and Maori in 2012 the medians of both groups could go up.

[64] I have looked at income inequality within Maori, finding that it was less (say measured by the coefficient of variation) than for the population as a whole. The reason seems to be that social security benefit levels for Maori are higher relative to the mean than for the non-Maori (because benefits levels do not ethnically discriminate), thereby pushing up the bottom of the distribution.

[65] Recent examples are Chapters 6 (by K. Mila) and 10 (by E. T. Poata-Smith) in M. Rashbrooke (2103) op cit. Incidentally Figure 6.1 on page 97 is mislabeled. The Pasifika unemployment rate is certainly not a third of the overall rate.

[66] The impact of the student tertiary funding in the early 1990s probably undermines comparisons of their standards of living over time using the methods described here (unless there is a specific adjustment). Rising school fees have a similar effect, but they have not been as big.

[67] Housing is particularly tricky. Housing grants are included in income but not cheap housing for state tenants. There is a separate section on housing below.

[68] None of the estimates reported below include private (including GSF) pension entitlements.

[69] It was not possible to provide estimates for later dates because of the increasing importance of joint family homes.

[70] Another source of data is the Statistics New Zealand 2001 Household Savings Survey. There is a publicly available synthetic unit-record file of 300 unit records which is not about real people, but was generated using statistical techniques to have similar characteristics as respondents to the survey.

[71] Inferred from Cheung (2007: Figure 2). The comparable figures for the 1966 adult distribution were 19 percent and about 75 percent; this might suggest the wealth distribution had become more equal over the 38 years, but the difference may reflect definitional differences and measurement errors.

[72] For instance a simple correlation between income categories and their average wealth suggests that an extra $1000 of the former (before tax) is associated with an extra $3700 of the latter (the correlation is high but not linear). However, the retired may be high in wealth and low in income, so there is a need to control for age.

[73] Between 1986 and 2006 the (media and mean) age of first nuptial birth rose 4 years. There is not comparable data for ex-nuptial births.

[74] In 2006 home ownership peaked at near 80 percent in the 60 to 74 age group (the proportion for the entire population is 53 percent) and then fell off to 59 percent for those over 85. The percentages do not cover the institutionalised.

[75] Instead the aggregate ownership and far-from-rigorous measures of housing affordability are used by policy rhetoricians to draw conclusions about accessibility to home ownership.

[76] This list is preliminary. Additions of NZ publications will be gratefully receive. Despite the popularity of the practice, it is not my intention to omit any paper because I dont like the author or cant understand it. That is why the list is on the generous side of inclusion. Those which have been overlooked will be added to a list on the author’s website when his attention is drawn to them; possibly with a note explaining how they modify the survey conclusions.