Chapter 7: Assessing a Poverty Line

A chapter of Globalisation and Welfare State

Keywords: Distributional Economics; Social Policy;

As Chapter 3 reported, poverty has long been a phenomenon in New Zealand life. Yet in the 1970s poverty research expanded. A poverty income line was established, numbers of people below the line were calculated, some behavioral investigations were undertaken, ethnographic studies increased, and some policy measures were undertaken to reduce the incidence of poverty.

Why at this time did research boom? When answering this question, we need to keep in mind that research on a phenomenon is not the same as the phenomenon itself. The flourishing of poverty research in the 1970s does not mean that poverty increased then. There were those who said that poverty was increasing (and/or income inequality was getting worse). The argument seemed to be based upon a pessimism that things are always deteriorating, together with a confusion between the existence of poverty (which they were observing in the 1970s), with increases in poverty (for which no evidence was provided).

I was once at a seminar in the 1970s, in which the pessimists were relating anecdotes about the poverty they saw about them, and arguing it was increasing. The sympathetic audience was persuaded. One dissenter rose to describe the poverty he could recall about him in his childhood in the 1950s. The meeting was transformed, because the pessimists were in a quandary. How could they deny these authentic experiences, which demonstrated there had been some poverty in the past? So they too rose, and told increasingly lurid stories of the poverty they recalled from their childhood. The thesis that there was a golden age of a poverty free society collapsed, although one suspected the pessimists left the seminar still believing that poverty was increasing.

The story cautions against the belief there was a time when there was no social stress and no severe social inequality. However we cannot be sure that inequality increased in the first decades of the postwar era. I started a study based on that premise, but the weight of the evidence is that economic inequality was decreasing at least through to the mid 1970s, although there are a few caveats. (1)

First, income data is available by individuals, and not by households. Decreasing inequality by individuals need not mean there was decreasing inequality by households because there are complicated interactions depending on household composition (whether there are more or less children, together with different ages), and the rise of the multiple income household (especially with the mothers moving into the workforce), plus the complication of the erratic incidence of housing expenditure. Second, the incomes of fathers was falling relative to average incomes. (2) If, as we shall see, households with children are the main group of poverty, their poverty may have been rising while other’s were falling. Third, while income inequality may have been falling steadily (but certainly not dramatically), there may have been other social changes which overwhelmed the effect. The study notes the migration of the Maori into the cities appeared to result in a deterioration in their housing circumstances, even though it may have increased their incomes. (3)

But supposing there had not been a general or substantial increase in poverty in the early post-war era. Why then the outbreak of poverty research? A number of factors seemed to be converging.

First, the movement of the Maori into the cities, was indicative of a wider phenomenon of social differentiation occurring by space and by perception. Urban Maori may or may not be better off than their rural cousins, but unquestionably they were more visible, and undoubtedly they were on average poorer than the urban Pakeha (just as they had been poorer than the rural Pakeha). However to see the urban poor as an ethnic phenomenon would be to miss the point. There are many more Pakeha than other ethnic groups, there are many more poor Pakeha. What seems to have been happening is that spatial differentiation was occurring within cities, and positional goods and services were becoming more evident, although their flaunting (conspicuous consumption) does not become widespread until the 1980s. (4)

A second factor was the growing recognition that the welfare state had not eliminated poverty in the way its earlier advocates had expected. A new generation observed the still existing poverty and highlighted it. This was not a peculiarly New Zealand perception, and it is not difficult to trace similar trends overseas. The British concerns about poverty and the US War on Poverty were both influential in shaping New Zealand research, as was Ronald Henderson’s poverty studies in Melbourne.

Third was the slower growth of the economy after 1966, and the evident rise in labour market stress. This not only led to worries about the state of household welfare, but also to the Royal Commission on Social Security which sat between 1969 and 1972. (5) It was deluged with a lot of anecdotal and ethnographic evidence about poverty and inequality, which it tried to shape into a consistent, and where possible quantitative, story necessary because its greatest test was to recommend the dollar level for the social security benefit.

The 1972 Royal Commission on Social Security

The report of the (1972) Royal Commission of Social Security is one of the landmarks in New Zealand social history. In a wise and detailed manner it codified the existing social security system, providing a foundation for subsequent social policy discussion. The Royal Commission was charged only with reviewing the social security system, but its vision was somewhat wider. Two and more decades after, the informed are till likely back to refer to it. So should the reader. Here are its `essential principles on which we consider our social welfare system and its administration should be based.’

“(1) The community is responsible for giving dependent people a standard of living consistent with human dignity and approaching that enjoyed by the majority, irrespective of the cause of dependency. We believe, further, that the community responsibility should be discharged in a way which does not stifle personal initiative, nor unduly hinder anyone trying to preserve or even enhance living standards on retirement or during times of temporary disability.
(2)The aims of the 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;
… (iii) Third, 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.
(3) Social security cash benefits are only one aspect of the total problem of maintaining incomes and raising living standards. Taxation, wages, employment, economic development, education, health, housing, social services, and cultural policies are all also of great importance. There is a manifest need therefore to co-ordinate those areas which impinge on, one another.
(4) Need, and the degree of need, should be the primary test and criterion of the help to be given by the community irrespective of what contributions arc made.
(5) Coverage should be comprehensive irrespective of cause wherever need exists, or may be assumed to exist.
(6) Identification and measurement of need is essential if the primary test is to be observed. We believe that this is best done by establishing categories of people who are most likely to bc unable to derive adequate incomes from the market system, or who are most likely to face unusual expense in maintaining an acceptable standard of living. It is still necessary either to:
… (i) Discriminate between those falling within a category (for example, the aged, the widowed, the sick) who need or do not need help, or to find out how much help is needed (the selective approach); or to
… (ii) Assume that need exists, and therefore dispense with further discrimination where the expectation of need within a category is high enough, and other considerations (such as the effect of taxation) are favourable. (This is the universal approach.)” (6)

It is not proposed to provide a detailed interpretation of these points, but the first two deserve some elaboration. The first emphasizes the role of the individual in the community, and therefore rejects the New Right view of rational economic man, which was central to the reforms of the 1980s and 1990s, which plays down the role of community. Thus the reforms had to be in direct conflict with the traditions of the New Zealand welfare state.

The second principle offers three standards, which correspond to three notions of poverty.
* The first aim is equivalent to eliminating absolute poverty, where people have insufficient to life and health.
* The second aim corresponds to the notion of the elimination of relative policy, where a person’s standard of living is judged relative to the rest of the community. The Royal Commission suggested that it should be judged on the ability to participate in and belong to the community, tying the notion into the first principle of the centrality of community in the welfare state.
* The third aim does not quite correspond to the elimination of what is sometimes called tertiary poverty, where an individual may have sufficient material goods, but lack a quality of life. This also challenges the New Right view, which does not – as John Stuart Mill put it – distinguish between the happiness of a pig and the happiness of a philosopher. Mill argued that the happiness of a philosopher was superior to that of a pig, that enlightenment was superior to ignorance. James Fitzgerald, Canterbury’s first provincial superintendent (and a university graduate), put it nicely in his inaugural address in 1852:

“’There is something to my mind awful in the prospect of the great mass of the community rapidly increasing in wealth and power without that moral refinement which fits them to enjoy the one or that intellectual cultivation which enables them to use the other.” (7)

Yet the report was the end of an era, rather than a beginning. The Royal Commission had arisen out of the 1968 National Development Council’s concerns about the affordability of the welfare state. The NDC itself was, in part, a response to the collapse in the price of wool in 1966, which we now know heralded the end of the first stage in the post-war era, a stage which made possible the welfare state codified by the Royal Commission. But even as it wrote the stage had come to an end. How distant sounds the following judgement of the Royal Commission:

“Even if ‘full employment’ is an imprecise concept, it is unlikely that any New Zealand Government will be able to escape from the public insistence that it must so manage the economy that there is a market for the services of all those able and willing to work.” (8)

That begged the question as to the degree that the government was able to so mange the economy.

The Royal Commission on Social Security Benefit Datum Line (RCSS BDL)

In doing so the RCSS provided a platform for the poverty research industry which was to come. It moved on past the notion that poverty was an absolute concept in a wealthy country and recognized the existence of relative poverty, which was also being adopted elsewhere. While it allowed that individuals may have enough to sustain life and health, and thus be out of absolute poverty, they may have insufficient to be able to participate in and belong to their community, and hence be in relative poverty.

How to operationalize this notion, preferably into a quantifiable and practical concept? The latter was more important. The next section describes some of the alternative approaches, which were tried in the 1970s, but the eventual answer was obvious enough with hindsight.

The favoured poverty line was the income recommended by the RCSS as the basic benefit level, now known as the BDL or benefit datum level, noting that the government was not obliged to follow the Royal Commission’s recommendation (and in 1991 indeed slashed benefits far below its recommendations). The RCSS would not have recommended a benefit level below that which they judged was required to enable the participation in and belonging to objective. But given the Commission’s fiscal prudence, they were unlikely to set a benefit much above their assessment of the poverty line. Thus anyone’s whose income was below the poverty line could be thought of as poor in the Commission’s relative poverty sense.

The poverty line was originally called the PDL, or pension benefit level, since the benefit for the elderly was set at the same level as other social security benefits. There were two reasons for this. First the `P’ echoed `poverty’. Second the implication was that the rest of the community were entitled to a standard of living similar to the poorest elderly who have been more looked after by the existing system. The social security benefit and pension levels diverged after the introduction of National – now New Zealand – Superannuation (NZS) in 1976, so the term BDL was adopted. Note that there could be a case for increasing the poverty line to the NZS level (which would mean the numbers below the poverty line would be higher). This was not done because the NZS level had become increasingly political. The logic of the BDL is that it was chosen by an expert panel, and it was important to avoid politicians or officials could manipulate the poverty level for public policy purposes. A further complication occurred in 1990 when the actual social security benefit level was markedly reduced relative to the BDL (the benefit having been broadly aligned with it until then). As a result the BDL is sometimes called the RCSS BDL, although habit – followed in this book – means the shorter expression is still generally used.

Had the RCSS chosen the right poverty level? It had been inundated with evidence about community living standards. Its judgement was likely to be as good, if not better, as any panel of four wise men and one women.

In subsequent years ethnographic surveys were used to test the validity of the conclusion, although they involve the converting of qualitative judgements into quantitative ones. There were only a few useful quantitative studies, most notably those by David Ferguson on the elderly in 1974, and his longitudinal Christchurch Child Development Study from 1976 (next section). In the end the BDL seemed to be about right. (9)

How to update the BDL for economic and social change after 1972? There has been little dispute that adjustments for inflation should use the Consumer Price Index. Influential here was Frazer Jackson’s finding that different weightings of goods and services in the basket which made up the CPI did not overly change its track. (10)

The theoretical conclusion in regard to real (i.e. inflation adjusted) changes in standards of living was that the BDL should be adjusted for long term changes in real incomes, but not for the stage in the business cycle. However the practical issue of how to do this proved irrelevant. There has been hardly any change in average real household incomes since the RCSS deliberates in the early 1970s, an issue to which we return.

A minor fracas was whether those on the actual BDL was in poverty, or only those below it. David Bedggood for instance argued the former. (11) The issue is not insignificant because a lot of people use to be in receipt of a benefit equivalent to the BDL, so the numbers of poor were markedly affected by this minuscule change. But it also follows that were those beneficiaries to receive one further cent they would edge out of that group. Making too much of this distinction would be grandstanding, but it points to the general issue that the poverty count is sensitive to the choice of poverty line.

Alternative Poverty Lines from the 1970s

During the 1970s other poverty lines were considered. British researchers, for instance, used a level of their commonly used poverty level based on their social security supplementary assistance levels plus 40 percent. This had the merit of reflecting the practical experiences of people living on low incomes, derived from the reports of social workers interacting with the poor. But it had the disadvantage that the supplementary assistance regime was administratively determined, so it can be changed by political fiat.

In the case of the US procedure based on the cost of food basket, had been discussed and put aside by the RCSS as primitive. It was revived by a Treasury official in the late 1980s, and is reported in that context below.

In New Zealand, before the BDL line was proposed, Peter Cuttance had used a measure in his study of large families in the Waikato which was based on the mean income less one standard deviation. (12)This is a perfectly viable notion, but it requires knowledge of the population income distribution, and so is not very practical because collection of a suitable data base is expensive.

A similar disadvantage applies to the otherwise attractive procedure used by David Ferguson in the DSW Survey of the Elderly in 1974, and in his subsequent work with the longitudinal Christchurch Child Development Study. (13)This involves asking a group of people are asked series of questions about the living experiences and possessions. Their answers are statistically processed into a scale which reflects their standard of living. By inspection one may be able to identify a point along this scale which will function as a poverty level. Thus an ethnographic approach melds into a quantitative one.

There is a notion here, which Peter Townsend in Britain has argued at length, that the is a reasonably abrupt change in behaviour and circumstance between those who are poor and those who are not. (14) For instance it appears there was a group in the community who could not afford to take their children to the doctor (a research finding which has resulted in a policy of a General Medical Services benefit high enough to have low or zero private charges for children), Suppose we could identify a cluster of such instances which tended to occur in some families but not others, with few cases where there was a mix of the two options. Then it would be easy to discriminate between those who are poor and those who are not, via a poverty criterion. It would not be an income level, but hopefully it would map reasonably accurately on to one, and hence provide a poverty level.

The trouble is that such a procedure is extremely data intensive, and therefore expensive. Studies which can obtain the data are very rare. In his Christchurch Child Development Study Ferguson found that his subjects objected to direct questions about their standard of living, and so only did the analysis once (and even then it only on a limited sample of the entire population of families).

Frazer Jackson suggested that investigation of the number of different goods and services that a household has may generate a similar scale. (15) For instance the poor family may have mutton stew every day, upgrading to sausages for a treat, while the rich family will have a different protein source for each meal. So variety is a useful indicator of relative poverty than overall quantities. Thus far this insight has been too data intensive to be followed up, although one would urge ethnographic studies to ask questions about variety as well as quantity.

Is there a abrupt change, which allows us to distinguish easily between the poor and the not so poor? That is an empirical question hardly tested in New Zealand. The few studies I have seen do not suggest the issue is that simple. As the great English economist, Alfred Marshall recalled `natura no facit saltum’ (Nature does not make a jump). (16) I would be surprised to see a jump in behaviour and circumstance of the population of all families.

A variation of the Ferguson approach was a Heylen Research Centre study commissioned by TVNZ in 1984. It gave a sample of the population a score of situations and asked whether it was an indication of a decent minimum standard of living, and whether the household was in that personal situation. (17) Thus 89 percent of the sample said that sometimes missing a meal in order to make ends meet was indicative of below the minimum decent level, while 2 percent of the sample said that was their experience. This is a promising approach, although one worries about the representativeness of the sample in this case. However it was never followed up.

Generally though, the 1980s were a period of quiet in poverty research, so there was little progress in defining or refining the poverty level. The BDL became a sort of quasi-official line, so much so that the by the end of the 1970s Department of Social Welfare reports were tiptoeing around the notion. It is never in the interests of a government to have an `official’ poverty line, since it will inevitably used to lambast them, because there will be people living below it. Better to leave it ambiguous so the politicians can slide away. For the same reasons, the poor and their advocates, press for such an official line. That they have been unsuccessful to establish an official poverty level or definition of full employment (whereas, for instance, the financial sector has been successful in getting the government to adopt an inflation target) reflects the relative power of the two sides, plus the exercise of setting an official line would long arduous and costly.

Adjusting for Household Circumstances: The Household Equivalence Scale

Thus far it would appear that there is a single poverty line, but it is going to differ for different household circumstances. In order to participate in and belong to their community, a couple living together are going to need quite a different income from a family of two adults and four children, Other circumstances may also affect this need: the ages of the children (as may to a lesser extent, the ages of the adults), whether one of the family is suffering from a disability, a Maori family may have tangihanga needs not relevant to a pakeha equivalent, the employed have costs which the unemployed do not, and so on. One could argue that each family’s circumstances so were so special that each should have its own poverty line. That is impracticable, but it is common to adjust for household size and composition.

This is done using a Household Equivalence Scale (HES), which converts the poverty line for a reference household into an equivalent line for other households. In New Zealand the reference household has been a married couple, mainly because that was the way that the reference unit used for social security purposes. It could have been for a single person living alone. Other potential reference households re much less attractive. for instance, the two parent, two child household requires agreement as to the ages of the children. So the BDL is usually defined as the amount required for a couple living by themselves to be able to participate and belong in their community. The RCSS recommendation was $33.00 a week in September 1971. Allowing for increases in consumer prices since them that would be about $305 a week in March 1997.

Suppose the HES scale was to say that a single person required 60 percent of the income of a couple to attain the same standard of living (in this case to be able to participate in and belong to their community in a similar way). Then the BDL for a single person would be 60% x $305, or $183 a week. Given the scale, the datum line can be calculated for any other household it covers. That the ratio is higher than half reflects the economies of scale of living in larger households. It is not true that two can live cheaper than one, but it is true that two can live more cheaply together than they can live separately.

How to construct a scale, Basically three methods have been used in New Zealand.

Valuing Necessary Goods and Services. A group of people decide what goods and services each household type needs to attain the same standard of living. The goods and services are then valued, and the ratios between the total valuations determine the HES. This is a complicated process and involves judgements which may be disputed. For instance the only one available in New Zealand (which comes from a New York study) provides for a woman in paid work having more nightwear than one who works at home. Even so, there is an important lesson to be learned from the exercise. The HES is quite different when New Zealand prices are applied to the New York scales, so different relative prices are important. (18) Thus it would be wrong to use an overseas scale to for New Zealand purposes, although sometimes this happens.

Econometric Studies (ES). Econometricians (who specialize in the statistical problems met in economics) analyze household behaviour based (usually) on expenditure and other characteristics, to construct demand relationships for each household type. The income necessary to attain the same demand for each household type can be calculated, and the ratio between these incomes gives the HES. This by far the most methodologically sound approach, for it depends upon actual household behaviour and minimizes the use of judgements. (19) There is a vast literature, which while technically sophisticated, heavily depends upon substantial data and major computing requirements, so that often one has a display of technical dexterity, rather than insights into household behaviour. Nevertheless there have been at least two attempts in New Zealand to construct reasonably comprehensive HES using the econometric approach,

A Priori Methods. in this context, a priori is little more than the constructor of the HES makes a series of judgements, which are based upon instinct or anecdote, but have little to do with actual household behaviour. Extraordinarily the most widely used HES in New Zealand – the Jensen scale – is an a priori scale. As a couple of Treasury officials remarked, the scale is `ad hoc’, a surprising criticism from them for we shall see that much of their work was a priori too.(20)

Because the Jensen scale has been so widely used more needs to be said about it. The pseudo-scientific nature of its approach is evident by going back to the original papers. Jensen writes down a mathematical function which requires two parameters.(21) The method sets the single adult household at .6 relative to 1.0 for a couple, and the two adult-four child household at 2, and solves to get the required parameters: one measures the economies of scale from living in a larger household, the other measures the relative weight in expenditure of a child compared to an adult. It could be said that the .6 comes from some empirical work, but even here there is a circularity, since the .6 was the ratio set by the social security benefits, which is what the Jensen scale is meant to be assessing. The 2 comes entirely from introspection as far as one can assess.

Ten years later Jensen altered the scale. Other than he now thought his previous judgement was wrong, it is not evident why he did so for there is, again, little reference to local evidence. The paper does refer to overseas household equivalence scales, but as already mentioned different relative prices in different countries lead to different economies of scale effects.

To appreciate the absurdity of the approach, consider what adjustments are necessary as user charges are increased. For simplicity consider if education was fully charged, so families with children now need more income because they pay school fees? There is no automatic mechanism in the determination of the Jensen HES which would adjust for the change. the other two methods would require the group adding to its lists of necessary goods and services education fees, while the econometricians would statistically observe the change in behaviour, and their ratio would automatically change. Jensen could make a series of ad hoc adjustments to their parameters, but what principles would he use? A prior approaches involve unstated and obscure assumptions, many of which are certainly not value free.

It is claimed that different equivalence scales give broadly equivalent income distributions.(22) However overseas studies demonstrate that different scales lead to significant differences in poverty profiles across different types so the choice of scale matters both for poverty measurement purposes, and for setting minimum income levels.(23)

How then do the different HESs available in New Zealand differ. The following table reports the scale levels for a single person and family of six (the two households Jensen focuses on, recalling the level for a couple is set at unity), and also the two parameters which Jensen derives.

Household Equivalence Scale
(Couple Household = 1)

Scale One Adult
Familyof Six Scale
Valuing Goods & Services
Easton (1973)* 0.64 1.83 .644 .778
Econometric Methods
Easton (1980)** .53 2.07 .916 .606
Smith (1989)*** .61 2.16 .737 .781
A Prior Methods
Jensen (1978) .60 2.00 .737 .781
Jensen (1988) .65 1.75 .621 .730

* Easton A Needs Index (1973).
** Easton ‘Three Household Equivalence Scales’ (1980).
*** Only reported here are the Canadian Method Scales (with adult clothes included in necessities), since the utility method gives problematic results.

The effect of using the ad hoc Jensen scale rather than one of the more empirically based scales is to reduce the income adequacy level for large households relative to small ones. This has important fiscal implications since the social security benefit structure in New Zealand is based on a two adult household. Thus the Jensen scale justifies a lower child support than the empirical evidence suggests, keeping down spending on social security. (24) Compared to the available empirically based ones it has very strong household economies of scale, making the unlikely assumption that it is very much cheaper (almost 40 percent) for four people to live together than separately. The advantage to fiscal policy of this assumption is that it downplays the cost of children in a household. Thus there is probably more children below the true poverty level than suggested by the estimates using this distorted scale, with the consequence that there has been even less attention to addressing the needs of children. The government monopoly control on poverty research has saved the government billions of dollars in family assistance – at the expense of children and their parents.

Nevertheless we need a HES, for poverty research, and to set benefit and minimum income levels. It is just that we need a better, empirically based one than the ad hoc Jensen scale.

The Treasury Food Based Line

In 1990 a couple of Treasury officials published a paper Assessing Adequacy Standards for New Zealand, a version of which was subsequently published by one of them in the 1993 New Zealand Economic Papers, as `Assessing Income Adequacy in New Zealand’. (25) We will refer mainly to the second version, which presumably responds to some of the widespread public criticisms of the first, even though the first was probably more influential on policy.

The paper is interesting because while Treasury officials may have fiddled around with poverty definitions and measurement in the past, this study seems to reflect a more concerted effort. However its definition of a poverty line was regressive, perhaps because the lead official was a FIFO, (26) and so never really got her head around the New Zealand debate. Instead she was overly dependent upon the US experience which has neither a representative nor the most advanced poverty research. She dismisses – and misrepresents – other research in a single sentence `the approach … that adopts the benefit level as the standard is not included.’ (27) To repeat the obvious – the RCSS BDL is not necessarily the actual benefit level.

The paper uses three general types of standards to determine income adequacy, but here we need look at only the food standard criteria defined as `The food share standard is based on the cost of food where these costs are converted by a multiplier, the inverse of the Engel coefficient, to total income levels.’ (28) The chosen bundle of food is a `minimum’. The particular total comes from the FOCAS Information Service of the University of Otago, who regularly value a food plan of the Department of Health at local (Dunedin) prices. The paper then adds an adjustment for `household consumption economies’ by deducting 20 percent for the amount for two adults and two ten year old children. The deduction quantum is derived from an American study.

The paper then says that the level was checked against the provision of food for the prison population, but it is unclear how this was done. (29) There appears to be no adjustment for food grown within the prison, for gifts of fruit to prisoners by visitors, and for prisoner purchases of food from their pocket money. In fact the basic food plan for males was $32.05 for an adult man and $29.47 for an adult woman, an average of $31.93 weighted by the gender balance in prison. Ignoring these extras the prisoners had, they were still getting 6 percent more food than the basic food plan. (This led to the advice to the poor to steal food. If they get away with it they will be better fed, if they get caught and jailed you will also be better fed.) The situation was anecdotally supported by prisoners giving their 1993 Christmas desserts to beneficiaries.

Three multipliers were used – three, four, and five. The diophantine emphasis on integers is intriguing. It is clearly unnecessary, since the RCSS recommended a $20.00 per week benefit level when the average weekly food costs for an adult was $3.93, a non-integer ratio of 5.09. (30) If there be any doubt about the idiosyncrasy of this approach, observe that men generally eat more than women, and so their food costs are higher. Using the same multiplier that would mean the poverty line would be lower for a woman than a man, and that for instance it is implicitly assumed she should have less to spend on clothing. This bizarre assumption can be avoided by assuming different multipliers by gender, which underlines just how arbitrary the whole exercise is. (31) There is hardly any justification for the choice of integer in the paper. It appears to favour `four’, saying that `three’ may be too stringent. (32) But so might four, or whatever.

The method comes from Mollie Orshansky. (33) Her work is widely misunderstood. Decades later Nancy Ruggles was to summarise her contribution as `[a]lthough Orshansky’s thresholds bore only a very approximate relationship to a `scientifically’ determined minimum level of subsistence, the fact they incorporate at least some adjustment for family size and composition made them an advance over much previous work’.(34)

Moreover Orshansky is not primarily using the measure to calculate the number of poor, but to identify the characteristics of the poor, the task to which the bulk of her 1965 paper is devoted. Rereading her paper one is struck by the sensitivity, understanding, and intelligence of the author to the problems inherent in her method, attributes not always evident in her successor users of her method. In 1965 Orshansky’s work represented progress on what had gone before. That is not true a quarter of a century later.

One would like to have said that the Treasury paper represents a good try, and would have been invaluable to the 1972 Royal Commission on Social Security, had it been written two decades earlier when it was near the state of the art. However something must be said about the underlying methodology.

First note that insofar as there is any empirical content it comes from overseas, in which points to a deferential mentality by those who do the work, compounded by their using approaches which, while progressive some decades ago, have become outmoded. Meanwhile they ignore superior work in New Zealand. But behind this approach, be it done here or offshore, there is a deeper problem.

Suppose an audience of low income New Zealanders were told that a measure of income adequacy (a poverty level) had been developed which was dependent upon a priori theorisation and the use of ad hoc assumptions, but which had no reference at all to the conditions in which they lived. Their least strong reaction would be astonishment. (35) Yet this is the approach of the paper. It constructs a deprivation index based without any reference to the living conditions of the deprived. If there be any doubt about the gap between they who were preparing the reality of those involved with this work (which acknowledges contributions by six senior Treasury officials), and the reality of the poor which it is addressing, note its reference to porterhouse steak, whereas more representative of the poor’s lot is stewing steak. (36)

The original work was carried out in a department of state which, shortly after the work was completed, recommended substantial cuts to benefit levels. It matters little whether one of the new benefit levels was almost precisely one of those suggested by the research. What is important is that the cuts were made in a policy context in which research could be carried out oblivious to the actual living circumstances of those who were to be affected. Familiarity with the empirical research in New Zealand would have led to the correct prediction there would be a sharp rise in severe hardship, as the new benefit levels were often well below the adequate income to attain the Royal Commission of Social Security’s aims of the system.

The Treasury may have dropped this approach. We report later on their discussion in the 1996 Briefing to the Incoming Minister. At this point we note that they move towards using benefit levels as measures of the poverty line.

The Poverty Measurement Research Group

The Poverty Measurement Research Group (PMRG) starts with the simple idea of asking people what they think would be a suitable poverty line, (37) or as the PMRG puts it to them the `minimum adequate weekly expenditure’. (38) Regrettably the PMRG has not discussed how this relates to the objective of the `participating in and belonging to the community’, of the RCSS, or even to absolute and relative poverty distinction, an issue complicated by it (and Treasury) talking about an `absolute poverty level’, when they seem to be referring to a relative one.

The question is posed in carefully selected focus groups, which appear very expensive to administer. It is evident that there is considerable variation between the focus group’s judgements, so that there is currently a high margin of error in the estimates.

The procedure is an ethnographic one which leads to quantitative estimates, although the amount of actual behavioral information which it provides is minimal. It also has parallels with the `valuing necessary goods and services’ approach for household expenditure scales, with the advantage that it is done by people whose experiences are apparently not dissimilar than those they are talking about. (39) However the procedure is subject to the usual ethnographic concerns, of a lack of scientific reproducibility (for instance, how outcomes are affected by different focus group leaders). A lack of rigorous statistical evaluation adds to these concerns.

There are also problems with the focus groups responses. They are asked to provide estimates for both a single adult and two child household, and a two adult and three child household. The chosen latter figure is not much higher than the former one, implying either outrageously strong economies of scale (very much stronger than even the Jensen HES), or that the focus groups are using different criteria for the two household types. (40)

The PMRG then uses the Jensen HES to calculate its standard poverty line for a couple with two adults and one child, rather than the standard two adult household. This is merely a rebasing exercise, and no explanation is provided as why they did this. But it adds to the confusion, especially as the age of the child is unspecified. (41)

A further complication arises because the PMRG then relates their poverty line to an estimate of median household income (adjusted for household composition). (42) In particular they set their poverty line as 60 percent of this median. Why they did this is a mystery, especially as far as we know at no stage were the focus groups consulted about whether they used some median notion when they made their deliberations.

Any statistician knows that a statistical median (the middle income) is inelegant, clumsy to calculate, and its estimators are inferior to the statistical mean (the average). Aesthetics aside (although they are not a bad test of the integrity of a theory), using a proportion of the median has a very grave defect. Like the British benefit measure rejected in the 1970s, it can be affected by policy and administrative change. For instance suppose the government were to take income from those in the middle of the income distribution and give it to the rich. Then the median income would fall. Thus any poverty level which was a proportion of that median would fall, and so the numbers below the level would fall. Using a median based measure the government could appear to reduce poverty by making the rich better off and doing nothing for the poor.

This possibility is not a figment of a theoretical imagination. In fact the median has been declining relative to the mean, since the early 1980s – as far back as we can go. The fall in the median income through time (especially relative to the mean) has been so great that the numbers below 60 percent of the median has been falling (despite poverty, by any sensible definition, rising). Thus a poverty level based on the median can be used to pretend poverty has been falling over the last 15 years. Not surprisingly, both Roger Kerr of the Business Roundtable and the Treasury seized on this result, making the absurd claim that poverty has been falling during the reforms of the 1980s and 1990s. The Treasury even presented the median based poverty line in their 1996 Post Election Briefing, (43) as though it were an intellectually coherent approach.

The PMRG have denied their median calibrated poverty line is intended for such uses. not only have they not offered an alternative approach, but they published a table which uses a median based poverty line over time. (44) In any case why bother to use a ratio of the median, and get in such a muddle, if the intention was not to use it as some mechanism to allow for changes overtime? Most astonishingly in all this confusion, the PMRG failed to note that their proposed poverty line is very close to the RCSS BDL, despite the latter being set two decades earlier. The PMRG could have made the modest but justified claim that their exercise has been another ethnographic contribution to the improved calibration of the BDL and supports the view it is a reasonable estimate of a poverty line (and our respect for the wisdom and competence of the RCSS). Instead, when they (rarely) refer to earlier work they misrepresented it. (45)


If two decades ago, when the modern poverty paradigm was being developed, it had been said that Royal Commission’s assessment of the poverty line would still be the best available today, there would have been widespread astonishment. There are probably two main reasons for this robustness. First average household incomes have changed very little in real terms over the period, so there has not been steadily rising incomes which would inevitably increase a relative poverty line. Second subsequent research on a poverty line has been desultory, and often misconceived, so there has been no serious challenge to it.

Perhaps there were two further reasons. We might admire the wisdom of the Royal Commission. But, second, it is also possible that their setting of the BDL, and its implementation as a basic social security level, set public expectations as that being a reasonable level to assess poverty.

Whatever the explanation, there is a need to improve – recalibrate – the poverty line, using a variety of means. There are two higher priorities. One is to get a scientifically based and broadly acceptable household equivalence scale, instead of relying upon one that is evidently wrong. Second we need to know more about the behavioral consequences of being below the poverty line.

This is not merely a matter of refining our understanding the extent and implications of poverty, as discussed in the next chapter. A poverty line has major policy implications, both in terms of setting the targets for intervention, and for assessing its success.

Next Chapter Ch 8: Labour Market Segmentation

1. Easton, Income Distribution in New Zealand, (1983).
2. ibid p.239.
3. ibid p.211
4. Pearson & Thorns (1983).
5. RCSS (1972).
6. RCSS (1972:65-6). The italics are as in the original. In order to capture the flavour wider social welfare principles rather than the narrow social security principles, their last two principles have been shifted to (2) and (3) in the ordering.
7. Webb (1957:212).
8. RCSS (1972:291).
9. Easton, Wages and the Poor (1986).
10. Jackson (1977). Housing has to be excepted from this generalisation, because of major divergences between mortgage free, and mortgaged housing, and between owner occupier and rental housing.
11. Bedggood (1980).
12. Cuttance (1974, 1976).
13. DSW (1975), Ferguson & Horwood (1978), Ferguson, Horwood, & Beautris (1980a,b).
14. Townsend (1980).
15. Jackson (1986).
16. on the title page of Marshall (1920).
17. Heylen Research Centre (1984), reported in B.H. Easton Wages and the Poor (1986). The data files are with the Social Survey Research Archive at Massey University.
18. B.H. Easton, A Needs Index, (1973).
19. McLements (1978).
20. Brashares & Aynsley (1990:24).
21. There is also a trivial scaling factor.
22. E.g. Rutherford et al (1990), and Brashares (1993).
23. Buhmann et al (1988).
24. and also the replacement ratio (the benefit to earnings relativity) for families.
25. Brashares & Aynsley (1990), Brashares (1993).
26. Fly in fly out, that is an overseas expert who visits the country for a short time.
27. (1983:186).
28. p.189.
29. But see Brashares & Aynsley (1990, Annex 4:2).
30. RCSS (1972:122).
31. In any case food prices do not move precisely with overall consumer prices, so either the income adequacy line would have to be updated by the food price index (which would be odd), or if another index (e.g. the consumer price index) were used an integer in one year will be a non-integer in the next.
32. Brashares (1993:190).
33. Orshansky (1963,1965).
34. Ruggles (1990:4).
35. Even that would be stronger than the reaction of scholars, for it seems possible to publish an article on the topic in a learned journal with little reference to the work that had gone before.
36. Brashares (1993:190).
37. Waldegrave et al (1995, 1996).
38. Although the question is about `expenditure”, responses include a savings element for life insurance and superannuation.
39. `Apparently’ because little information, other than ethnicity, is reported about the characteristics of the focus groups.
40. For instance one focus group thought a three person household required $70 a week, but a five person household required $100 a week. Ignoring economies of scale, simple algebra indicates that the allowance per child was $40 a week, and the allowance per adult was minus $10 a week. The focus group also assumed the housing requirements for the larger family were exactly the same as for the smaller one. (Waldegrave et al, 1995:95).
41. There are other complications, as they change the base year, and select one of the two focus group household averages as preferable. (B.H. Easton, Measuring Poverty: Some Problems (1997)).
42. And as it happens, for a different year (1991) from when the focus groups gave their judgements (1993 on).
43. Treasury (1996).
44. Waldegrave et al (1995:107).
45. For further debate on this topic see B.H. Easton Measuring Poverty: Some Problems (1997), and Waldegrave et al (1997)