New Zealand Sociology November 1995, Vol 10, No 2, pp.182-214.
Note This version has yet to have the graphs added.
Keywords: Distributional Economics; Statistics;
After around two decades of modern poverty research in New Zealand which has focused on poverty at a point in time, it is now possible to provide estimates of the changing numbers of poor over time.
The overall conclusion may not seem surprising. If the standard poverty level of the Royal Commission on Social Security (the BDL) is used, the proportion of the poor in the population as a whole hardly changed between the March year 1981 and 1990. In the following three years to March 1993 year (the latest currently available) the number of the poor rose sharply. This conclusion is not particularly dependent on the level used to measure poverty, if it is set in constant price terms. However this paper shows that if the line is based on a population mean or median income the interpretation is much more problematic.
Because there is an evident shift in the level of poverty in the 1990s the paper concludes by tracing the factors which caused these changes. The conclusion that the benefit cuts seem to have been the main contributing factor will not be surprising, but the research also identifies a labour market deterioration before the benefit cuts.
Before providing the new results the paper provides a brief review of poverty research, to place those results into a context and development.
Poverty Research in New Zealand
Modern poverty research in New Zealand began in the early 1970s, with the survey of large families in Hamilton by Peter Cuttance (1974), the Department of Social Welfare’s Survey of the Aged (1975), and my work using a diverse set of sources and approaches (Easton 1973a, 1973b, 1973c, 1976). Exact precedence is unimportant, publication dates need not reflect when the work was done, nor its results disseminated to fellow researchers. What is relevant is the development of a number of research studies at about the same time.
That the research began then, does not mean there was no poverty before the 1970s, or that it increased markedly about that time. We simply do not know. There has never been the data bases to estimate the magnitude of poverty earlier – indeed the possibilities of a quantitative tracing poverty through time only really begin in the 1980s. However it seems likely that there was considerable poverty, in the modern sense, before the research began.
The research seems to have been precipitated by three major factors. First research overseas -especially in Australia, Britain, and the US – percolated to New Zealand. Second there was a sharp deterioration in New Zealand’s economic performance from late 1966, which lasted about a decade and added to the financial pressures of many families. Third, the Royal Commission on Social Security (RCSS), established in the wake of the economic stagnation and reporting in 1972, focused attention on income adequacy questions. Certainly the RCSS’s report has been a seminal document for all subsequent thinking on poverty issues.
A fourth factor which facilitated research was the establishment by the Department of Statistics (now known as Statistics New Zealand) of a regular household economic survey, called Household Economic Survey (HES) today, but previously known as the Household Expenditure and Income Survey (HEIS). Easton’s review paper (1976) used the first available survey of 1973/4, to provide national estimates of aggregate poverty. Initially the income side of the survey was not very reliable, but by the early 1980s there had been sufficient development to give confidence to reliance on the HES income statistics.
This critical data base was supplemented by various special surveys which provided information on which the HES did not (and often could not) cover. The statistical quality of the surveys varied from the rigor of the Department of Social Welfare’s survey of the Aged and the Christchurch Child Development Study’s longitudinal investigation of a birth cohort (1978, 1980a, 1980b), to ad hoc surveys of little professional competence which nevertheless sometimes gave useful insights. (Easton 1986)
By the late 1980s the foundations of the research program had been set down. Easton (1991) Subsequent work has been to build on the paradigm, and there have been few new innovations.
The Poverty Research Paradigm
Much poverty research, especially that oriented towards estimating the size and composition of the poor, may be thought of as identifying and measuring a particular characteristic of a population (usually income) distribution. We are familiar with the notion of a mean of a distribution as a measure of the overall population – so faced with a statement like “real disposable per capita income rose 1 percent this year”, most people accept without much reflection that this is a statement about the average. However the mean is only one measure of a statistical distribution, which while popular may not always be the most appropriate. In particular the middle of a distribution is not relevant to examine when considering poverty.
There are a number of other ways of describing the distribution of a population, including the lorentz curve, the gini coefficient, and the coefficient of variation (Easton 1983). Each has its strengths, and weaknesses including the difficulty of computation and the understandability of subtle statistical concepts by the general population. Instead poverty research in New Zealand has tended to focus on the number (or proportion) of the population below a particular income level called the “poverty line”. A variety of poverty lines have been proposed, but the basic idea of counting the number at the bottom of the distribution has been a central theme of almost all of New Zealand’s quantitative poverty studies.
The intention of the exercise is not to conclude that those who are a dollar below the poverty line (whatever that is) are terribly badly off, and those who are a dollar above are in a satisfactory life style, although this is an often unfortunate misinterpretation of the gee-whiz approach. Rather the single figure is intended to give some order of magnitude, and perhaps guidance to policy directions.
(Another measure, until recently difficult to calculate, is the “income gap” which is the amount of income necessary to bring those up to the income of the poverty line. Thus it gives some notion of the distribution of the hardship within the poor. We shall see that there are problems in the interpretation of the measure. It is likely to remain a much less important measure in the foreseeable future.)
The Population Distribution: Disposable Household Income Adjusted for Household Composition.
At an early stage the accepted population distribution was disposable household income adjusted for household composition. Disposable income is total household income from market and benefit sources (but rarely inter-household private transfers), less income taxation and other compulsory outgoings. The income is then adjusted for the household composition. At the simplest level this might be thought of as placing income on a per capita basis so that households with more people have their disposable income scaled down on pro rata basis. In practice a more sophisticated procedure based on household equivalence scales is used, which allows for economies of scale in household spending, and the lesser needs of children. More is said about equivalence scales below.
In the 1970s the calculation of the resulting distribution was very ad hoc, but the Statistics New Zealand ASSET model of the household, means today each sample household in the HES sample can be operated on separately.
A couple of alternative population measures have been household expenditure rather than income, and to deduct housing expenditure from disposable income. Deducting housing expenditure from income generates a mongrel concept, signalling the crudity of the method. The problem of housing in poverty studies has been long recognized – essentially the problem is that many people have an implicit addition to their income from a favoured housing situation, such as a mortgage free home – but the deduction method is embarrassingly coarse. 2
Determining the Poverty Line
Initially a number of possible poverty lines definitions were explored, but by the end of the 1970s the most commonly used one was the RCSS BDL, or Royal Commission on Social Security Benefit Datum Line. Originally described as the Pensioner Datum Line (PDL) it had been the level set by the 1972 Royal Commission for a couple on the “pension” (i.e. aged benefit) or other social security benefits, adjusted in line with changes in prices.
Behind this pure dollar rate was some subtle argument, which has not always been recognized in subsequent developments. The 1972 Royal Commission, which devoted much space to discussing various aspects of poverty, argued that the aims of the social security system was “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.” (1972:65 original’s italics) This objective was a major consideration when they set their benefit levels. Thus the BDL sets a poverty line which implies a definition that a typical person with that income is able to participate in and belong to their community.
How did the 1972 Royal Commission decide its recommended poverty line was adequate for the purpose? It heard a lot of (typically qualitative) evidence from submissions, and on this basis made an informed judgement. Now that judgement could be wrong – certainly, it was not precise. However the evidence of the systematic and ad hoc surveys is that the BDL appears to be near a material standard of living which enables the average couple to fulfil the Royal Commission objective. The true figure, assuming “true” is a meaningful notion in this context, might be 10 percent higher or lower, but the available evidence suggests that the Royal Commission made a reasonable stab at the level. (Easton 1986)
There have been three major proposals for an alternative poverty line. 3 The first argues that the notion of participating in and belonging to involves behaviour and material consumption relative to the population as a whole. As the average level of material welfare changes, so should the poverty line. The adjustment is easy enough if an economy is growing (or declining) uniformly, but suppose there is a sharp contraction which lowers average real incomes by 10 percent say. Does the poverty line go down by this amount too? Easton (1980) exhibits angst over this problem – perhaps not as much as he would have had he known what was going to happen to average incomes in the 1980s (see below). But as best as one can interpret him from hindsight, his position was that the poor should not be expected to carry the full burden of adjustment of short term swings in material prosperity (especially in a downward direction), but if there is a secular change (including declines) in average real incomes the poverty line would probably be expected to adjust in parallel over time (unless there was a change in social attitudes).
The second alternative, proposed by two Treasury official, is to relate the poverty line to a food entitlement. (Brashares & Aynsley 1990, Brashares 1993) A detailed critique is published elsewhere (Easton 1995), but for this paper’s purposes we note the exercise was almost entirely bereft of any empirical input, and certainly made no attempt to incorporate the behaviour of the poor in any assessment of the measurement of their deprivation. It chose a not entirely arbitrary bundle of food4, which it costed, and then multiplied up by a parameter (of 4, with alternatives of 3 and 5) which was arbitrarily chosen by the researchers. More fundamentally, it is far from clear what was the underlying notion of poverty.
The third alternative proposal involved using a particular proportion of a median income level. (Stephens 1994) We shall be examining this notion in greater detail below, and will find it gives paradoxical outcomes. Unfortunately the underlying conceptual framework is as enigmatic as that of the Treasury officials’.
The choice of the median appears to arise because it is used in international poverty comparisons. However the international academic studies do not chose the median because of any inherent merit, but because it is a practical solution to a problem of standardization of cross-national comparisons. It would be foolish to hang a nation’s social policy on some decisions of a few foreign academics struggling with a very different issue.
Stephens seems to acknowledge this, because rather than using the 50 percent proportion which the international studies do, he argues that the poverty line should be 60 percent of the median. The choice of ratio is no less arbitrary than that of Brashares and Aynsley. Indeed Treasury officials could be entitled to argue that if an international criteria is to be used, than so should the international parameter, and that the 50 percent is more appropriate. Either choice is subjective, containing little reference to the living circumstances of the poor. That Stephens uses expressions like “preferred”, without any criteria of preference, and in the passive voice to disguise who is preferring, adds to the general picture of capriciousness of the approach. 5
A project involving “focus groups” was introduced to give some empirical content to a poverty line. Unfortunately the material has still not been written up, so comment depends on fragmentary reportage. Apparently a small number of selected families are asked how much they think is necessary to maintain an adequate standard of living. We are told their conclusion is not very different from Stephen’s preferred poverty line.
Supposing this is an accurate account of the work, there are two points to be made. The first is that the procedure is dependent upon small samples. Unfortunately large sample studies from overseas show those in different material circumstances give different answers. Regrettably the focus group approach uses such small samples that the sort of evaluative procedures for their more efficient overseas counterparts are not available.
Second and even more fundamental, the focus groups are not asked to provide an estimate based on some proportion of a median defined income, the notion that Stephens is proposing. Instead they give estimates of absolute expenditure (and not income), which is then compared to the outcome of the median income notion. We are told that in 1990/1 their selection was similar to the 60 percent level. Now this level changes in later years, and it is not clear at all that the focus groups would change their assessments of needs by the same amount. For instance in 1991/2 the median fell about 8 percent compared to a year earlier (see below). It seems likely that the focus groups in 1991/2 would chose the previous real outlays, rather than adjust their assessments down too, so the relativity becomes 60*1.08 = 65 percent of the median. Thus the calibration applies at best for a single year. The procedure thus fails miserably to provide a regular updating of the poverty line.
The Data Base
As already explained, the usual data base by which poverty numbers are calculated comes from the HES, after transforming unit household records. Usually the transformation involves calculating the disposable income, and adjusting the total for household composition by using a household equivalence scale.
The most readily accessible data is that published by Mowbray (1993) which provides adjusted household disposable income for seven of the ten years from 1981/2 to 1990/1. 6 In addition Mary Mowbray has kindly made available her estimates for 1991/2 and 1992/3, based on the latest HES available. The Mowbray data is presented as the mean equivalent disposable income in each income decile. Figure 1 shows the income shares of the ten deciles. It is based on appendix Table I. 7
Unfortunately the middle deciles are rather crowded and it is difficult to identify patterns within them. However there are two obvious effects. Income in the highest decile increases relative to the rest after 1988, because of the income tax cuts on upper incomes in the late 1980s, while the Income in the lowest decile bumps around due to losses from the self employed. This phenomenon is returned to later.
Inspection of Appendix Table I enables a quantitative assessment of the changes. Over the 13 year period, the income share of the top decile increased around 4 percentage points. This was at the expense of the bottom seven deciles. The big losers were those in the 3rd to 8th decile (above the groups which are typically treated as the poor) who each lost at least .5 percentage points. However the 9th decile, while experiencing a cut in its income share of around .3 percentage points, experienced a proportional cut of around 5 percent. 8
These changes should not surprise us. The substantial tax increases given to high incomes in the late 1980s, meant that others’ disposable income relativity would suffer. However we need a finer method than decile shares to explore the actual impact on the incomes of the poor.
Before further using the data a couple of caveats should be noted. The first is that like most work in recent years the data uses a household equivalence scale due to Jensen (1988). Although fashionable, like some of the poverty lines we have examined the Jensen scale is based upon a priori theorization, with little empirical content. There are some five New Zealand household equivalence scales, three of which are based upon empirical evidence of sorts. (Easton 1994) It is noticeable that the Jensen scale 1988 is the one with the strongest economies of scale, that is the one which thinks household costs rise most slowly as the size of the household increases, and the second lowest weighting for children. These are a priori assumptions of the Jensen scale, sometimes justified by crude comparisons overseas (e.g. Brashares 1993). However when Easton (1973b) made a careful comparison between New Zealand and New York, and found quite different economies of scale effects.
Suppose the Jensen economies of scale assumption is significantly wrong (as the empirically based scales suggest). Results based on it will be distorted. The likely effect will be to underestimate the degree of poverty in large households. especially those with children. Thus we should be cautious when interpreting the poverty estimates – they are probably conservative. 9 The development of empirically based authoritative accepted equivalence scales, is one of the priorities if poverty studies are to progress.
The second issue is that the Mowbray figures, like many others, report households not people in households. There is a tendency for the poorest to be in large households. Thus in 1992/3 the 10 percent of poorest households contain 11.4 percent of all people, and the twenty percent of poorest households contain 21.3 percent of people. Where appropriate the figures reported here are for people rather than households. 10
Poverty Lines: 1981/2 to 1992/3
The easiest poverty line to understand is those which are constant throughout the period. The first is the 1972 Royal Commission’s BDL which was equivalent to $14050 p.a. for a couple in 1991/92 prices. 11 It should be noted that the BDL is not the same thing as the actual benefit level, although for many years they were closely linked. In addition we have provided estimates for the Treasury level, labelled as TDL, which comes to about $11050 p.a. In addition an upper level, labelled as UDL, of $17050 p.a., an equivalent amount above the BDL is also shown.
The second group of poverty lines arise from adjusting the BDL for changes in disposable income. On earlier occasions Easton adjusted by real per capita private disposable incomes (e.g. 1980, 1990). However between 1983/4 and 1992/3 this measure fell .7 percent, 12 while mean household disposable incomes fell by 4.8 percent in constant prices, according to the Mowbray figures. That represents an annual difference of .44 percent a year, which might be due to error and definitional differences, but may also reflect changing household composition. Perhaps it hardly needs worrying about. Even so it is incorporated in the calculations.
Being unable to find a suitable disposable income series before 1983/4 the calculations use a real private consumption expenditure as a proxy. It rose 4.5 percent in per capita terms between 1971/2 when the Royal Commission set the BDL and 1981/2. After adjusting for the .44 percent p.a. effect, the poverty line would have been about at the same level as the BDL in 1981/2. It has a different path after, so we call this level indexed to the mean equivalent (disposable) income MNDL. As Figure 2 shows it tended to deteriorate through the 1980s and early 1990s, a decrease of 6.0 percent in total, or an average of .6 percent a year. This reflects the falling levels of income over the period, especially after 1987/8.
The third poverty measure uses the median income as the reference point, taking Stephen’s proposed 60 percent as the relevant proportion. We call this MDDL. As shown in Figure 2 it has experienced an even greater decline that MNDL, a total of 19.2 percent or 1.9 percent a year over the eleven years. This reflects a changing income distribution, in which the income share of middle income households are falling relative to top income households in a general period of stagnation. The main cause of the shift was the increasingly favourable income tax treatment on upper incomes at the expense of those in the middle.
Figure 3 shows the estimates of the headcount poverty for the three central poverty measures. They show quite different patterns. The estimates based on the BDL – that is a constant real poverty level – conform most closely to the conventional wisdom, rising mildly through the 1980s, and leaping in the 1990s. We shall be considering them in further detail in the next section.
The MNDL – which allows for changes in average incomes – shows a general drift downward of poverty numbers in the early 1980s, as average incomes fell, but the real value of benefit levels were maintained and unemployment fell slightly. The reversal in the unemployment trend in the late 1980s (with some weakening of real benefit levels towards the end) meant poverty numbers on this measure increased mildly, and they increased again in the early 1990s, although not as dramatically as for the BDL.
The third measure, the MDDL gives an almost entirely counter intuitive pattern of headcount poverty. It shows the proportion declining from 15.0 percent in 1981/2 to 11.7 percent in 1992/3, with no obvious rise in poverty in the early 1990s. This conclusion is an artefact of the falling level of the MDDL over time. Basically the index is falling faster than poverty is rising.
To understand why this has happened, consider the following thought experiment, not unrelated to what actually occurred over the period. Take a income distribution and calculate the headcount poverty based on some proportion of the median as the measure. Now suppose the government taxes those on middle incomes, and transfers the revenue gains to the rich in income lower taxes. The mean income will not change, but the median (which is based solely on the distribution of middle incomes) will fall. As a result the MDDL will fall, and so the numbers in poverty will fall. Paradoxically, a transfer of income from the middle income to the rich (with no change in the incomes of the poor) reduces the degree of poverty in the community on this measure. This outcome is so illogical as to suggest that poverty levels based on median incomes as reference points are almost completely useless. 13 It is difficult to see any merit in the measure, other than for those who think reducing taxes on the rich is socially beneficial.
In regard to the MDDL, this writer confesses to being sympathetic to relating the poverty line to changes in average incomes, since the “participating in and belonging to” notion depends on some overall community standard. Indeed an important shift in thinking on poverty, officially initiated by the 1972 Royal Commission, was from an absolute notion of poverty to a relative one. However I remain perplexed on how to incorporate changes in average real income levels into the analysis. Here I want to suggest a tentative strategy. Recall that attempts to calibrate the BDL left a margin of error of about 10 percent. Mindful of this, I suggest that the poverty level be a constant real value one until the mean disposable incomes move 10 percent above (or below) the level at which they level was calibrated.
Thus we are still in a situation where the concept of a constant real value suffices as the best measure we have of poverty level, given the limited changes to average real income over the last two decades.
Poverty Numbers Since 1981/2
However I have never been committed to the precise BDL level: hence the searching for evidence to verify its appropriateness. It is now possible to offer poverty estimates based on different constant real price levels. The following analysis provides estimates for seven constant real value poverty levels for the period from 1981/2 to 1992/3. The middle level is the BDL at $14050 p.a. for a married couple, and with three levels above and three below, at $1000 p.a. steps. the upper level is the UDL and the bottom level is the Treasury TDL.
The resulting estimates are shown in Figure 4 and Table 1. The pattern is similar for the seven levels. Basically the proportion of the poor did not change much between 1981/2 and 1989/90. After that it rose sharply, typically by between a quarter and a half. The pattern is not simple. The lowest increase is for the BDL (26.2 percent) while the extremes of the TDL experience a 55.1 percent increase and the UDL 37.4 percent increase.
Table 1: Changes in Poverty Proportions
|% INCREASE 89/90 TO 92/3||55.1||39.3||29.7||26.2||42.5||50.3||37.4|
This means that there were complicated processes affecting the poor in the late 1980s and early 1990s. An evident effect was the benefit cuts which occurred in early 1991, typically on 1 April 1991 (and hence first impacting in the 1991/2 year).
Another complication is that careful examination of the graph shows that there were some deterioration in the incomes of those in the disposable equivalent income brackets between $14050 and $16050 p.a. in the two years before the cuts occurred. The most likely explanation is that the deterioration of the labour market in that period was impacting on labour earnings (including supplementary earning for beneficiaries).
|Table 2: Two Estimates of Poverty in 1989/90
|NUMBERS OF POOR||Easton (1993)||This Paper|
Before leaving these figures it is appropriate to revisit briefly an earlier estimate of the change in poverty over the cuts period. Using the equivalent disposable income distribution derived by Brashares and Aynsley (1990), and adjusting for subsequent distributional change as reported in the HES, I found that poverty defined by the BDL measure increased by around 43 percent, with a range of 35 to 52 percent. The new figures suggest the increase was 35.3 percent at the bottom of the earlier estimate. This seems to arise because I had slightly underestimated the numbers of poor. (Table 2) Readers are left to judge whether the earlier estimate was misleading. Here I note that the figures were not robust to a percentage point, but the ad hoc procedure was reasonably satisfactory at giving an order of magnitude. 15
Who are The Poor?
As interesting as gee-whiz figures may be, of greater interest can be what is the composition of the people who are in the poorest part of the distribution. When considering the issues of the composition of the poor there are two separate questions: which are the largest groups among the poor, and which groups are more likely to be poor. The difference is that the second question may be answered in terms of groups who are small in the community, even though they may not make up a large proportion of the total poor.
Mowbray has supplied figures by household type by equivalent disposable income decile for 1992/3. The answers to the two questions are as follows:
Table 3: Who are the Poor?
The largest groups in the bottom deciles are (in order):
Two adult households with three or more children (about 27% of total)
One adult households with one or more children (about 19%)
Two adult households with two children (about 16%)
(no other group above 10 percent)
The groups with the largest proportions in the bottom quintile are (in order)
One adult households with one or more children (about 43 percent of the group)
Two adult households with three or more children (about 28 percent)
(no other group over 25 percent).14
In summary the largest group of the poor remain children and their parents. Almost 29 percent of all children are in the bottom quintile (in contrast to a fifth of the population). The relative proportion in the lowest decile is even higher.
In contrast the rich are adults only household, who make up 76 percent of the topmost quintile, despite being only 41 percent of the population.
Allowing that these figures may be conservative, because of the equivalence scale used, we must conclude that the situation first identified two decades ago – of children being the greatest reason for poverty – remains true to this day. That is despite of various family assistance measures (which have not been large in actual expenditure outlays) (St John 1994). Probably the most effective means of dealing with family poverty is still the simple strategy of increasing family assistance.
The Income Gap
The “income gap” or “income deficit” is the amount that is required to bring the poor up to the poverty line. It measures the intensity of the poverty, because it is larger when the lower the average income of those below the poverty line. 17 Table 4 gives the income gap for the BDL for each year, and for the seven poverty levels in 1992/93. There is a number of ways of representing the measure. Here we report it as a proportion of the poverty level.
|Table 4: Income Gaps as Percentage of Poverty Line16|
|Income Gap for BDL (by years)|
|Income Gap for different poverty lines (1992/3)|
Disappointingly the data does not make a lot of sense, leaping about in unpredictable ways. Inspection shows this is related to those years in which households with income losses (typically from self employment) are larger than usual. The instability, which is partly sampling error but also may reflect the volatility of self employed incomes, could be eliminated by clipping unwanted observations from the bottom. Such an approach leaves considerable room for arbitrary decisions, and the statistical properties of the clipped series are far from clear.
The data seems to suggest the size of the income gap seems to have gone up in the early 1990s, relative to the 1980s. Thus there are not only more poor, but the hardship of the poor has been greater.
Readers are unlikely to be surprised that poverty and hardship rose, on any reasonable definition in the 1980s – the statistics confirm the anecdote and survey (e.g. Dalziel (1993), List et al (1992), NZCCSS (1992), People’s Select Committee (1992)). They may be surprised that the distributional structure was so stable through most of the 1980s. The most evident impact was the 1988 tax cuts, and there is a hint of the effect of labour market deterioration shortly after. But in summary, there is little evidence from the available data that factor market changes had a major influence on the distribution of household income over the period. Distributional policies involving taxation and benefit levels, and even demographic change seem to have had more effect.
The study has cleared away some misleading developments in thinking about poverty. Median based poverty levels seem useless for tracing poverty over time. hopefully too a little progress has been made on the thorny question of indexing the Royal Commission BDL for changes in real incomes: it proves broadly unnecessary over the last two decades, but at some stage the question of recalibration has to be addressed. Nor should some of the quantitative innovations developed in this paper be ignored.
Problems remain, even confining the concern to the measurement of the size of the poor, and leave definitions of poverty line definition, and the actual experiences of the poor. The equivalence scale is still not acceptable, and problems like housing outlays and the losses of the self employed have still to be addressed with more finesse than they have in the past. Probably the different price experiences for different parts of the income distribution and household types becomes more significant as social welfare moves to a more targeted approach for user charges.
In policy terms we know that benefits underpin the bottom of the income distribution, that our poor are children and their parents, and that the aged are not our worst off. We knew this all two decades ago. Hopefully over the next few years policy will develop in manner which will mean such insights need not remain research results with so little practical policy input.
Brashares, E. (1993) “Assessing Income Adequacy in New Zealand”, New Zealand Economic Papers, 27(2), p.185-207.
Brashares, E. & M. Aynsley (1990) Income Adequacy Standards for New Zealand, The Treasury, Wellington.
Cuttance, P. (1974) Poverty Among Large Families in New Zealand, MSS Thesis, University of Waikato, Hamilton.
Dalziel, P. (1993) Taxing the Poor: Key Assumptions Behind the April 1991 Benefit Cuts: What are the Alternatives? Department of Economics, Lincoln University, Canterbury.
Department of Social Welfare (1975) Survey of Persons Aged 65 Years and Over, Government Printer, Wellington.
Easton, B.H. (1973a) Family Incomes, Mimeograph, University of Canterbury.
Easton, B.H. (1973b) A Needs Index, Mimeograph, University of Canterbury.
Easton, B.H. (1973c) Poverty in New Zealand, Mimeograph, University of Canterbury.
Easton, B.H. (1976) “Poverty in New Zealand: Estimates and Reflections”, Political Science, Vol 28, No 2, December 1976. Reprinted in Easton, B.H. (1983) p.264-279.
Easton, B.H. (1980) “Poverty in New Zealand”: Five Years After, Paper for the Conference N.Z. Sociological Association, 1980.
Easton, B.H. (1983) Income Distribution in New Zealand, NZIER Research Paper No 28, Wellington.
Easton, B.H. (1986) Wages and the Poor, Allen & Unwin, Wellington.
Easton B.H. (1991) Updating the Economic Model of the Household, Paper to the Conference of the Social Policy Research Centre, University of NSW, Sydney, July 1991, Economic And Social Trust On New Zealand, Wellington.
Easton, B.H. (1993) Poverty and Families: Priority or Piety? Paper to “Issues for Families” workshop, October 1993, Working Paper 93/62 Economic And Social Trust On New Zealand.
Easton, B.H. (1995) “Properly Assessing Income Adequacy in New Zealand”, New Zealand Economic Papers, forthcoming.
Ferguson, D.M. & L.J. Horwood (1978) The Adequacy of the Domestic Purposes Benefit for Women with Young Infants, Christchurch Child Development Study, Christchurch Clinical School, Christchurch.
Ferguson, D.M., L.J. Horwood & A.L. Beautrais (1980a) The Measurement of Family Material Wellbeing, Christchurch Child Development Study, Christchurch Clinical School, Christchurch.
Ferguson, D.M., L.J. Horwood & A.L. Beautrais (1980b) Living Standards in a Birth Cohort: The Correlates of Family Living Standards, Christchurch Child Development Study, Christchurch Clinical School, Christchurch.
Jensen, J. (1988) Income Equivalences and the Estimation of Family Expenditures on Children Department of Social Welfare, Wellington.
List, P., A. Hubbard, & G. Dolan (1992) Budgeting for Cuts: A Study of Poverty in Palmerston North after the April 1st Benefit Cuts, Methodist Social Services Centre, Palmerston North.
Mowbray, M. (1993) Incomes Monitoring Report: 1981-1991, Social Policy Agency, Wellington.
New Zealand Council for Christian Social Services (1992) Windows on Poverty, NZCCSS, Wellington.
People’s Select Committee (1992) Neither Freedom or Choice, People’s Select Committee, Palmerston North.
Royal Commission on Social Security (1972) Social Security in New Zealand, Government Printer, Wellington.
St John, S. (1994) Financial Assistance for Families from the State, Mimeo, Department of Economics, University of Auckland.
Stephens, R. (1994) The Incidence and Severity of Poverty in New Zealand, 1990-1991, Working Paper 12/94. Graduate School of Management, Victoria University of Wellington.
1. I am grateful to comments from two referees and Sally Jackman
2. A related problem is that the household equivalence scales have to be adjusted if housing expenditure is removed.
3. One option which has not been explored is a poverty line based on the rate of national superannuation. it could be argued that an adequate minimum income for the elderly should apply to the entire community. (See Easton 1980.)
4. Athough less than that which would be fed to prisoners. While it is not obvious why in welfare terms beneficiaries should be less well fed than those incarcerated, it is to be noted the latter are more likely to riot.
5. There is an interesting parallel between the poverty lines in that Stephens omits housing expenditure and Brashares and Aynsley omit everything except food expenditure.
6. Because slightly different equivalence scales are used in the first three years of the data and the last three years it has been necessary to meld the series, using the 1987/8 year data which is calculated using both scales. The splice is small, and does not obviously alter the conclusions presented here.
7. Because the mean real adjusted disposable income changes little over the period the graph is also similar to that for average incomes.
8. The Gini coefficients – a measure of overall inequality – show the same story, increasing from an average of .291 in the 1981/2 to 1987/8 period, to .304 in 1990/91 to 1992/93. A shift of .013 may seem small, but Easton (1993) points out that under certain assumptions, a shift of that magnitude would represent roughly a 10 percent increase in the coefficient of variation of a log normal distribution.
9. At least in regard to those based on an absolute poverty line. Those dependent on the mean or median are further complicated by the reference point itself being partly based on the Jensen scale.
10. A final difficulty with the figures, which reflects more upon the method to be used here than the underlying data, is that the method assumes the distribution is somewhat smoother than it is. Statistical distributions of this nature are usually smooth, but the impact of benefits can mean that there are sharp changes in the distribution, as when a large number of people on the minimum benefit cut into the distribution.
11. All the following calculations use 1991/2 prices, those being the ones used in the Mowbray data base. Each poverty level refers to the required annual income of a couple. Other household compositions can be adjusted from this by using the (Jensen) equivalence scale.
12. There is no comparable official figures for 1981/2.
13. Indeed it suggests that their use for international comparisons may be totally misconceived too.
14. The population increased by 2.4 percent over the two year period.
15. Part of the problem may be that the two distributions I used were different (with the Treasury based one a little higher – to the right – of the Mowbray one).
16. There are two anomalies.
a) 17 percent of single adult households (who are not superannuitants) were in the lowest decile, but only 5 percent in the next. Thus there appear to be a group of single adults who are poverty stricken, and who are quite unlike the rest. One might conjecture these are unemployed and others in similar work situations who are not eligible to benefits, plus students
b) While only 13 percent of national superannuitants are in the bottom decile, the proportion who are is skewed towards single households, who tend to be in the top part of the second to bottom decile, and in the third decile.
17. It is not, however, a measure with direct relevance to policy. It does not tell us how much it would cost to bring everybody up to the poverty line, since the implicit marginal tax rate is an impractical 100 percent.
Appendix Table 1
Household Income Shares (in Percent) by Deciles 1981-82 to 1992-93
Equivalent Disposable Income
Notes: Data from Mowbray (1993:18), plus later years provided by M. Mowbray. The authour is grateful for assistance from Vasanatha Krishna and Mary Mowbray.
To calculate the actual in come levels (in $1991) multiply income share by mean year, and divide by 10.