Validation and the Health and Household Economy Project

Paper to the Wellington Health Economists Group, Thursday 29 November, 2002.(1)

Keywords: Distributional Economics; Health; History of Ideas, Methodology & Philosophy; Statistics;


This is a brief summary of a 100 plus page report, The Economic and Health Status of Households,(2) prepared by Suzie Ballantyne and myself. The data base was the Household Economic Survey (HES). For the three year period covering 1994/5-1996/7 the HES included questions on the respondents’ recent utilisation of health services together with as a subjective assessment of each’s health status, as well as socioeconomic variables such as income and expenditure and personal characteristics.

The Health Research Council provided the funding which gave the authors, the opportunity to research this data base. Other contributors to the project were the Prince Albert Trust, the Department of Public Health of the Wellington School of Medicine, and Statistics New Zealand, together with Rob Bowie, Paul Brown, Denise Brown, Len Cook, Dean Hyslop, Sandra McDonald, Diane Macaskill, Claudio Michelini, Clare Salmon, John Scott, Helen Stott, and Alistair Woodward.

None of these are responsible for the results or interpretation in the report. A special mention needs to be made that the access to the data used in this study was provided by Statistics New Zealand in a secure environment of the Statistics New Zealand Data Laboratory (SNZDL) giving the researchers direct access to HES data but also giving effect to the confidentiality provisions of the Statistics Act 1975. It should be emphasized that any results and the inferences drawn from the work are the judgements of the authors, not Statistics New Zealand.

The Background to the Study

The use of the Household Economic Survey to investigate the household distribution began with its first publication in 1975, and the theoretical model which underpinned the investigation was largely completed by the end on the 1970s, although there have been a few analytic developments in the following two decades, most notably the direct access to unit records and the resulting ability of trace distributional change over time. Statistics New Zealand has also improved the quality of the survey data. However some fundamental weaknesses in the method identified a quarter of a century ago, have still not been addressed, although this report makes some contribution towards resolving them.

A quarter of a century ago there was no health data in the survey. So this study explores a previously uninvestigated dimension. However, because the data is cross-sectional it is not – on the whole – possible to draw causal conclusions other than those already imbedded in the hypotheses being tested. In econometric terms, simultaneity is a problem. Even so it was possible to explore questions such as:
1. To what extent do poorer households have poorer health, and those with poorer health live in poorer households?
2. Does private health related spending differ between different income groups when controlled for health status?
3. What is the relationship between housing status and health status, when controlled for income?
4. Are there any specific issues relating to children’s health and income?
5. What is the effect of non-household health funding (such as medical insurance or the community service card) on health spending?
6. How effective is the community services card? How successfully is it targeted?
7. What is the impact of medical insurance on private health service spending and on health service utilisation?

The aim of this presentation is to give a brief overview of what we did and some of the more revealing conclusions. Before doing so, something has to be said about the underlying methodological problem, which is far from peculiar to this study.

The Validation Problem

Whenever we use or construct an index we need to ask how do we know the measure is valid. To take a simple case, at one stage there was a fashion for constructing indexes of wellbeing by country. This involved collecting all the available raw data together, selecting those indicators which seemed to be related to wellbeing, and combining them into a single measure. This involved a number of judgements. One problem was what to do with data on divorce rate. Some said that high divorce rates were indicative of poor wellbeing, others said they were an indicator of high wellbeing, and some said they were irrelevant. Whatever the details of the argument, at issue here is that what the index constructors did with the divorce rate indicator was a matter of judgement, with the reasonable question as to what extent their judgement merely reflected a matter of opinion and to what extent it reflected a valid assessment of some underlying real phenomenon.

Social research is riddled through with this problem, with little attempt is made to validate the index. Apparently the constructors are so sure of their judgement they think it unnecessary. Recent very public examples are the New Zealand Deprivation Index, and the just published MSP Economic Living Standards Index. (3,4) In each case there has been an enormous amount of effort in the construction of the index, and no attempt to validate it. In each case the user is faced with the possibility that the index merely summarises the opinions and prejudices of those who constructed it, and has no connection with reality, even though we proceed as if NZDep tells us something about the location of deprivation in New Zealand and the ELSI tells us something about the living standards of households. (The same problem applies to many economics indexes. Virtually every one provided by the private sector, including such measures as the world competitiveness indexes, are based on the opinions of the compilers, with no attempt to show the index is connected with the phenomenon it purports to measure, even though it is used as if it was.(5))

The validation issue is riddled through the household income distribution studies. How do we know equivalent income reflects anything about the real world? Leaving aside the accuracy of the HES survey data – and it is worth noting that the studies proceeded for almost two decades before it was observed that there was a difference of as much as twenty percent between aggregate household income as estimated directly and the implicit aggregate income reported in the HES (6) – what about the choice variables and the transformations done to them? There is external validation of the income variable, since we know that it is relevant to predicting economic behaviour such as expenditure.

Most problematic has been the adjustment for household composition. This usually involves a ‘Household Equivalence Scale’. Very often the scale chosen is based on introspection by its proponents, without any external validation or justification. This study confirms the intuition that choice of scale matters, and that the ones used in official studies are biased against large households, typically households with children, relative to most other scales. That means that if they are wrong – if the opinions and prejudices of those who constructed them do not reflect reality – we have been underestimating the degree of poverty among children and their parents. Unfortunately, while it was possible to make some progress, the issue is still unresolved. In summary – as I shall report – the reality of household composition is much more complicated than our opinions and prejudices.

The Health Status Index

An illustration of the differences in approach between this study and some earlier ones, is in the construction of the health status index. The HES survey included questions about health service utilisation, and also asked for a self assessment of the respondent’s health status. Structurally then, the data base is similar to the one the MSP used when it constructed its economic living standards index. The MSP chose some of the question responses and combined them into their Economic Living Standards Index. The choice of indicators and the method of combination entirely reflected their judgement and there has been no attempt thus far to show the index reflects any underlying reality, other than the choice decisions of the constructors.

Had Suzie and I used this approach we would have faced problems not dissimilar to the divorce indicator one. For instance, how is one to use the dentist utilisation indicator? Does a high usage typically indicate poor health or does it indicate good health. Instead reduced the degree of opinion, by using a quite different approach which used the respondents’ subjective answer to the question of their health status (unlike the MSP study which largely ignored the parallel question).

Respondents were asked ‘In general, how would you rate your health?’ with response options of ‘Excellent’; ‘Good’; ‘Not so good’; ‘Poor’.(7) These responses are, of course, subjective, especially as the respondent may not have a good comparator. Nevertheless they convey useful information.

We used the health service utilisation responses to predict the probability the respondent would give the subjective health status repose, given their record on the utilisation health services. Because the equations are non-linear the econometrics is a bit tricky – we used multivariate analysis to compact the variables, and probit analysis to give the functional relationships.

We then assumed that the subjective health response was a measure of objective health status together with an error (which we assumed was independent of the errors in the utilisation variables). So the equation we had constructed to relate health service utilisation to subjective health status can be used to predict the objective health status. (Note that the constructed health utilisation based index of health status, being a probability, is a continuous variable ranging between zero and one, in contrast to the subjective response which is a categorical variable.)

Mathematically the index is a linear combination of the various health service utilisation responses, weighted on the basis of how well each predicted the subjective health status variable. It is an index, just like the Economic Living Standards Index and like NZDep. The difference is that we did not choose the weights of the index components, and to that extent the index involves making fewer judgements and choices than those constructions.

Who chose our weights of the components? Directly, the weights are selected by the regression equation. Indirectly the weights are selected by the respondents because the regression weight for each component is based on the way the respondents’ utilisation of health services connects with their health status. The MSP could have done this when it constructed the ESLI, but instead of giving people a say in the judgement, they gave the weighting decisions to the public servants involved in the research. In effect our utilised health status index is validated against the actual subjective responses of people. There is a second validation I will report shortly.

Adjusting for Housing Circumstances

The second validation also involved an adjustment of income for housing circumstances. The problem has bedevilled the definition of income. There are two major aspects to it.

First, the standard measure of income makes no allowance for housing status. Thus a household which owns its dwelling freehold, may have quite a different standard of living to a household with the same recorded income but which is paying of a mortgage, which again may be different from a household also in receipt of the same recorded income, but which is renting. Second, there will be considerable variation of payments for similar houses under the same tenure. This reflects the imperfect working of the housing market, which will only generate price convergence in the long run – at best. (In contrast, we expect all purchasers of vegetables, say, to pay broadly similar prices.) The consequence of these two effects is that the measured income of the household does not correspond to the theoretical income which underlies the model which economists use.

The problem is well known, but usually disregarded, simply because there is no obvious procedure to deal with these deficiencies. One ad hoc approach has been to deduct the household’s housing expenditure from the household’s income, but the underlying theory – if any – is far from clear, since it involves deducting an expenditure from an income, and then uses household equivalence scales which including housing, despite the numerator omitting it. (8) Instructively, there is no attempt to validate the approach, but rather to state the procedure is justified so the reader has to rely on the authority of the researcher. (9)

The task, therefore, is to construct a theoretically rigorous way of incorporating the vagaries of the housing market into a better measure of income, and then show it is a superior approach to ignoring the problem. To give an intuition of what we did. Suppose a household would normally spend $100 a week if it were renting the accommodation it was living in, but its current outlays were only $60 a week – for whatever reason, such as the actual rent has not caught up to the market rent (or is subsidised), or it owns the accommodation (without mortgage). The effect is the household is $40 a week better off, because it is paying only $60 for $100 accommodation. In cash terms it can afford to purchase $40 a week other goods and services which a household paying the full rental on its housing cannot. The difference can be imputed to the households reported income, increasing it by $40 a week. Similarly, a household which pays $140 a week for $100 a week accommodation, is $40 worse off compared to a household paying the normal rental, and so the overpaying household should have $40 a week deducted from their income to give a fairer indication of its effective spending power. So we adjusted for erratic accommodation costs by adding (or subtracting) to household income the surplus (or deficit) between what the household should be paying for the accommodation, and what it actually pays.

But how to calculate what the household should be paying? In the ideal world we would know the true market price for each household’s accommodation. ‘True market price’ means the amount the household would be outlaying on rent, if the landlord only had only a commercial relationship with the household, and if the market was in a fully adjusted equilibrium. In practice we cannot know that amount, but let us assume that on average the accommodation in the rental sector reflects the equilibrium rent.

Aside from a detailed survey, a means of estimating the true market rent of a household would be to identify the average outlay for a household for a given set of circumstances. For rental households, this is in principle possible, because we know the actual rental payments, and can use them as an proxy for the equilibrium market rentals, assuming the difference is a random error. So we ask how do different characteristics of the households affect the rental payment. For instance, bigger households are likely to spend more than smaller ones, as would richer households compared to poorer ones. There will be differences between households located in different cities, and so on.

The various effects were combined together in a (regression) equation, so that we can predict the equilibrium rental payments of any household with a particular set of characteristics. This also applies to households which are other than renting. In effect we can work out what they would be paying for rent (on average). Their actual outlays will be different.

We then add the difference – the amount they should be paying on housing compared to the amount they are actually paying – to their reported disposable income to get income adjusted for housing circumstances. This new data is, of course, an index, But while this is an elegant way of dealing with the erratic disparities in accommodation costs and, moreover, consistent with economic theory (and adjusts income with an income variable rather than an expenditure one), a theory does not in itself validate the usefulness of the adjustment. That is done in the next stage of the research.

Validating the Indexes

The next step in the study was to estimate behavioural relations on the utilisation of, and private expenditure on, health care services. Again we used regression. Among the independent variable we used to predict the utilisation and expenditure were income and health status. In each case we have a choice of variables

reported disposable income or disposable income adjusted for housing circumstances
reported (subjective) health status or the health status index based on utilised health care.

In each case it was the constructed, rather than the reported, index which was the more powerful explanatory variable in the regression equation.

We might treat this as some validation of the constructed indexes insofar as they provide better predictions of actual behaviour than the raw reported indexes. Of course that is not a total validation of the indexes – that requires many such exercises. But what Suzie and I can claim is that we have paid far more attention to the problem of validation than is usual in New Zealand.

The Determinants of Health Care Utilisation and Expenditure

Because time is short, I only list our findings as far as the determinants of health care utilisation and outlay. As far as utilisation is concerned, the recorded events were:
– Accident and emergency at a hospital;
– Other hospital services such as outpatient clinic, hospital pharmacies, laboratories or day wards;
– An ambulance;
– Nights spent as a patient in a hospital;
– Length of time since visiting a GP (more recent coded highest);
– Number of visits to a GP or family doctor;
– Nights spent as a patient or in a nursing home or similar;
– Number of visits
– to a medical specialist or consultant;
– to a chemist or pharmacist;
– to a dentist;
– to an optician or optometrist;
– to an other medical personnel;
– Medical support services such as laboratories, x-ray clinics or health caravans.

The econometrics found that if all other personal characteristics are the same:
– income is not a major determinant of health care utilisation.
– holders of community services cards tend to use most services more than had they not one.
– holders of high use cards utilise all services more intensely.
– holders of medical insurance tend to use most services more intensely.
– those who describe themselves as not being in excellent health use health services more intensely. The exception is for dental care whose use is associated with excellent health.
– the age and gender effect is much as expected. Young children are lower users of services, but boys more so than girls. In early adulthood women tend to use the services more than men, although this may partly reflect their child bearing and rearing duties. However young men do use the accident and emergency and ambulance services more than women. In later adulthood there tends to be a diminution of use of health services (with some obvious exceptions such as for opticians), but with perhaps some rise for some services after retirement.
– ethnic minorities tend to use the health services less than the Pakeha majority (even after adjusting for other socioeconomic characteristics).

Aggregate private expenditure on health care includes spending on minor items – such as toothpaste – which the healthy will do too. If all other household characteristics are the same:
– Income is the single most important variable accounting for 39 percent of the total explained variation in spending. The income elasticity was .24 which means that while those on higher incomes spend more on health care, their private health spending decreases as a part of their total spending.
– Possession of medical insurance was second to income in explanatory power, accounting for 31 percent of the total explained variation in spending. The study also explored the characteristics of households which had medical insurance. Income is not an important determinant, while Maori households are less likely to be holders compared to others with similar socioeconomic characteristics. Household composition and age – but not gender – are factors, but the largest reason for health insurance may be the health status of the members of the household, especially households with particularly sick members.
– The third largest determinant (12 percent) of explained variation is household composition and gender.
– Another 8 percent of variation is explained by health status. However the effect is strong one for those who are in poor health. A household which is has a zero probability of excellent health spends over four times as much on health care as one all of whose members are in perfect health.
– Additionally, those with community service cards spend 25 percent less than those without (despite, them being higher utilisers of the services), while those with High Use Cards spend on average 15 percent more (but they spend less on prescriptions).

Equivalence Scales

It was not possible to validate the household equivalence scales to the same degree as for the other indexes.

Unfortunately the choice of equivalence scales matters. For instance, consider the impact of the scales on poverty. This paper avoids the issue of what is the appropriate poverty line by looking at the results for seven different ones, centred on the most commonly used poverty line, the Royal Commission’s Benefit Datum Line (BDL), set in real terms in 1972, as the minimum practical benefit level for a couple for which it was necessary to enable them to participate in and belong to their community, but three others at are provided on either side.

In the prices at the time of the surveys on which this analysis is based, the BDL amounted to $15,200 p.a. for the couple. In order to get the poverty lines for other household types this amount is adjusted according by an equivalence scale. That means that the poverty levels for different household types will vary according to the choice of scale. Thus the numbers of poor will vary according to the choice of equivalence scale. This is shown in the following table in which the proportions below the poverty line are given for six household equivalence scales. The definitions of these scales need not detain us here, except to note that J88 (the Jensen 88 scale) and SR (the square root) scales are the ones used in the official statistics. (10)

We see that the outcomes are sensitive to choice of equivalence scale. The Jensen 1988 give the lowest levels of poverty numbers for a given poverty line., with the Square Root not far behind. A comparison of the Jensen 1978 to the Jensen 1988 equivalence scale is instructive. The difference between them at the Royal Commission BDL amounts to over 130,000 people, so a change in the opinion of the scale constructor in 1988 took 130,000 New Zealanders out of poverty without spending a cent.

Percentage of People Below Given Equivalised Incomes By Equivalence Scale

Income of Two Adults Scale
$p.a. PC SR J78 E80 J88 CM*
12200 22.0 7.9 9.9 11.0 7.9 8.6
13200 26.2 10.4 13.0 14.0 9.7 10.6
14200 30.2 13.5 16.4 17.7 12.5 13.5
15200 (RCSS BDL) 34.0 17.8 20.3 21.7 16.6 17.3
16200 37.9 22.8 25.4 25.8 21.3 22.2
17200 41.8 27.1 30.9 29.8 26.7 27.2
18200 44.9 31.3 35.7 34.6 31.5 32.4

A major reason for the difference is each scale assumes strong economies of scale. As the next table shows the Jensen 1988 and the Square Root scales tend to favour small households, and thereby underestimate the numbers (relative to the other scales) of children in two adult families, and children and their parents generally compared to other scales.

Percentage of Household Type Below RCSS BDL By Equivalence Scale

Household Type PC SR J78 E80 J88 CM*
One adult 4.6 32.5 8.3 5.1 12.2 7.0
One adult & 1 child 48.1 48.1 33.8 17.9 33.5 17.0
One adult & 2+ children 93.9 70.2 69.6 60.3 60.5 47.9
Two adults 8.3 8.3 8.3 7.2 8.3 8.3
Two adults & one child 29.6 15.9 18.7 19.1 15.6 15.9
Two adults & 2 children 43.2 16.0 21.2 22.6 16.0 17.2
Two adults & 3 children 65.9 21.5 31.1 33.9 21.8 24.7
Two adults & 4+ children 83.5 23.7 46.9 55.8 26.8 34.6
Three adults 14.1 6.7 9.5 11.3 7.9 10.1
Three adults & children 53.2 20.1 29.6 36.0 23.9 27.7
Other 31.2 8.4 15.3 23.2 11.9 16.9
All children 46.4 20.6 26.2 22.8 20.1 20.8
All parents 39.3 17.6 22.5 24.7 17.9 19.2
Other adults 13.2 15.1 10.9 10.9 11.0 10.8

Validating the equivalence scales proved difficult. Most have no empirical derivation. (It turns out that the test for the other indexes does not apply, since there may be a difference equivalence scale for each expenditure component.)

The paper discusses a strategy when household composition has to be allowed for, mainly aimed at avoiding using Household Equivalence Scales. When it is unavoidable, it recommends that a particular one that comes out of Claudio Michelini’s work be used (CM* above). This have the merit of being derived econometrically from household expenditure functions, and thereby having some internal validation. It is also more in the middle of the range of available scales in terms of distributional outturn. Unfortunately Claudio died during the project. His work was not complete, so his scale is not the best potentially possible. It is the best we have, and it is the one used in the following tabulations.

Estimating and Locating Poverty

The full report provides a lot of detail about the shape of the household distribution, the characteristics of people who are located in each part of the distribution, and the sensitivity of outcomes to various assumptions such as choice of equivalence scale and definition of income. Space limitations means only a very limited selection of these results can be reported. The following are based on the Michelini Scale, household income adjusted for housing circumstances and the Royal Commission BDL, although the qualitative conclusions are robust to the choice of any reasonable poverty line even if the quantitative ones are affected in the obvious direction.

The next table illustrates the issue with regard to household type. Notice how that the single adult household with the children has the highest incidence of poverty – a matter frequently commented upon – but they make up only a sixth of the poor. In contrast households with two adults and children have less than half of the poverty incidence of the single parent households, but make up more than twice the the numbers. There is a danger by focussing on solo parents, we ignore the vast majority of the poor, who are children and their parents, most of whom live in two adult homes. Households with children contain over four fifths of the poor.

Michelini Equivalence Scale; Income Adjusted for Housing Circumstances; RCSS BDL Poverty Line

Household type Proportion in Poverty Proportion of the Poor
Adult not in the labour force 4.6 1.6
Adult in the labour force 2.6 0.4
Two adults, neither in labour force 6.0 3.5
Two Adults, 1+ in labour force 4.0 4.4
Adult & 1+ children 38.7 16.3
Two adults & 1 child 13.9 7.6
Two adults & 2 children 18.2 16.7
Two adults & 3+ children 21.5 18.0
All 2 Adult & children 18.4 42.3
Three adults, no ithout children 7.2 8.0
Three adults & children 22.8 23.5
In households with children 21.9 82.1
In households without children 5.5 17.9
ALL 14.2 100

The same pattern applies to ethnicity. Non-Pakeha poverty is higher than Pakeha poverty. Yet almost three fifths of the poor are Pakeha.

Michelini Equivalence Scale; Income Adjusted for Housing Circumstances; RCSS BDL Poverty Line

Ethnicity Proportion in Poverty Proportion of the Poor
Pakeha 10.7 58.5
Maori 23.7 19.9
Pacific Island 33.4 11.8
Asian & other 26.3 9.8
ALL 14.2 100

When we look at poverty by housing tenure (after adjusting for housing circumstances) we find that while the highest incidence of poverty are those that rent their homes, over half of the poor live in their own homes.

Such tabulations is they warn there is a danger of focusing on the poor groups with high incidence and ignoring the bulk of those in poverty.

Michelini Equivalence Scale; Income Adjusted for Housing Circumstances; RCSS BDL Poverty Line

Ethnicity Proportion in Poverty Proportion of the Poor
Rent 27.3 47.4
Owned with Mortgage 13.5 37.5
Owned without Mortgage 6.0 15.0
Rent Free 1.2 0.1
ALL 14.2 100

The Location of the Sick in the Income Distribution

The final chapter locates the sick in the household distribution by self-rated health. Again there are numerous tables and only a few can be reported here. There is an apparent paradox in the results, for the sick are more likely to be in the middle quintile than the bottom quintile of the overall household distribution.

Michelini Equivalence Scale; Income Adjusted for Housing Circumstances

Income Quntiles Bottom 2 Middle 4 Top All
Households 21.7 25.3 25.5 16.3 11.2 100

However an examination of the characteristics by age and gender that the poorest in each category have the highest incidence of poverty. The old are the sickest, but they are not as poor as the young. As far as poor health is concerned, age is a more important determinant than income, although within each age group the lower the income the poorer the health.

Michelini Equivalence Scale; Income Adjusted for Housing Circumstances

Income Quntiles Bottom 2 Middle 4 Top All
Under 15 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 19.7 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


Unfortunately the HRC funds have run out, and they were unable to extend them. The research team is now broken up with Suzie in Dunedin and Claudio where the good econometricians go. Nevertheless, the report The Economic and Health Status of Households both progresses our analysis of household income and sickness, and enriches our quantitative and qualitative understanding of economic and health status. It also demonstrates that while there are limitations in Statistics New Zealand’s Household Survey, it remains an extraordinary data base which has been barely exploited.

Many of the results reported here will confirm or extend informed observers beliefs about the income distribution and health. however I hope that the report has conveyed a couple of important important lessons.

First, looking at prevalences may mislead as to the importance of the overall group. Those more prone to poverty (or whatever) are not necessarily the largest group of the poor (or whatever).

Second without some external validation, an index must be distrusted as only the opinions of those who constructed it. The methodological issue is wider than this. The parallel is in the mid-1980s a small band of economists began to impose their theories on the economic management of New Zealand, even though the theories had no external validation. It is a matter of record that their theories did not work, and the New Zealand economy suffered, as did the health and wellbeing of New Zealanders.

1. While the original report was written with Suzie Ballantyne, the views expressed in this paper are entirely my own insofar as they deviate form the original report.
2. B.H. Easton & S. Ballantyne (2002) The Economic and Health Status of Households, report for the HRC, available from the authors.
3. P. Crampton, C. Salmond & R. Kirkpatrick with R. Scarborough & C. Skelly (2000) Degrees of Deprivation in New Zealand: An Atlas of Socioeconomic Difference, Auckland.
4. Department of Social Welfare (2002) New Zealand Living Standards 2000, Wellington.
5. An exception is the NZIER Quarterly Survey of Business Opinion. For instance, R. Bowie & B.H. Easton (1987) The Quarterly Survey of Business Opinion, NZIER Research Paper 36.
6. B.H. Easton (1997) How Accurate are the Incomes Reported in the Household Economic Survey?, Wellington. (2001) J. Archibald Household Economics Survey Integrated Weighting,
7. In practice there were so few ‘poor’ responses, we combined them with the ‘not so good’ and called the category ‘fair or poor’ health.
8. B.H. Easton (1997) ‘Measuring Poverty: Some Problems’, Social Policy Journal of New Zealand, No 9, November 1997, p.171-180.
9. R. Stephens, C. Waldegrave & P. Frater (1995) ‘Measuring Poverty in New Zealand’, Social Policy Journal of New Zealand, Issue 5, December 1995, p.88-112; R. Stephens, C. Waldegrave & P. Frater (1997) ‘Measuring Poverty: Some Rebuttals of Easton’, Social Policy Journal of New Zealand, Issue 9, November 1995, p.181-185; J. Jensen & V. Krishna (2001) ‘Tracking Living Standards: Is it Done Better by EDY or HEDY?’ Social Policy Journal of New Zealand, Issue 16, July 2001.
(10) PC – Per Capita; SR = Square Root; J78 = the scale Jensen proposed in 1978; E80 = Easton’s econometrically estimated scale in 1980; J88 = the scale Jensen proposed in 1988; CM* = generalised Claudio Michelini Scale. E80 has an age of head of household, which gives it a slightly different calibration for two adult households.