Developing Deprivation Questions University of Bristol

Working Paper Number 13  

Developing Deprivation Questions for the Family Resources Survey

Stephen McKay Sharon Collard

Personal Finance Research Centre University of Bristol

December 2003

Main contents

Acknowledgements....................................................................................................................7 Authors.......................................................................................................................................7 Recommendations on question selection...................................................................................8 1 Introduction......................................................................................................................14 2 Approaches to poverty measurement...............................................................................16 3 Effects of having shorter sets of questions ......................................................................21 4 Reducing sets of questions – naïve approaches and latent class analysis........................31 5 Checking validity of individual questions.......................................................................35 6 Exploring dimensions of deprivation using factor analysis.............................................45 7 Longitudinal analysis.......................................................................................................58 8 Validating low incomes ...................................................................................................65 9 Topic areas.......................................................................................................................69 10 References....................................................................................................................77 Annex A: Data sets used in this study .....................................................................................80 Annex B: Main tabulations by country....................................................................................83 Annex C: Agreed question list for Family Resources Survey..............................................85

Detailed contents

Acknowledgements....................................................................................................................7 Authors.......................................................................................................................................7 Recommendations on question selection...................................................................................8 Overview................................................................................................................................8 Proposed question list..........................................................................................................10 Revisiting the list of proposed questions.............................................................................13

1 Introduction......................................................................................................................14

1. Objectives ................................................................................................................14
2. Methods....................................................................................................................14

2 Approaches to poverty measurement...............................................................................16

1. Introduction..............................................................................................................16
2. Using deprivation indicators to measure poverty ....................................................16
3. The 1999 PSE ..........................................................................................................17

2.3.1  Measuring Child Poverty using the PSE..........................................................18 3 Effects of having shorter sets of questions ......................................................................21

1. Introduction..............................................................................................................21
2. Adults.......................................................................................................................21
3. Children....................................................................................................................25

3.3.1  FACS................................................................................................................26

4. Connecting parental and child deprivation..............................................................27 Annex: detailed comparison and adult and child poverty measures....................................30 4 Reducing sets of questions – naïve approaches and latent class analysis........................31

1. Adult poverty...........................................................................................................31
2. Child poverty ...........................................................................................................33

5 Checking validity of individual questions.......................................................................35

1. Introduction..............................................................................................................35
2. Perceptions of necessities ........................................................................................35

1. Families compared with the population as a whole.........................................35
2. Comparison across income groups ..................................................................37
3. Regional differences ........................................................................................39
4. Differences between ethnic groups..................................................................40

3. Correlations with income.........................................................................................43 6 Exploring dimensions of deprivation using factor analysis.............................................45

1. Introduction..............................................................................................................45
2. Methodology............................................................................................................46
3. Findings....................................................................................................................47
4. Implications..............................................................................................................48 Annex: Results from the factor analyses .............................................................................50 Annex: Technical details of the factor analyses conducted.................................................57

7 Longitudinal analysis.......................................................................................................58

1. Introduction..............................................................................................................58
2. Aggregate trends over time......................................................................................59
3. Individual-level analysis..........................................................................................62
4. A case for chain-linking'? ......................................................................................64

8 Validating low incomes ...................................................................................................65

1. Introduction..............................................................................................................65
2. FACS panel analysis................................................................................................66

1. Money management and Savings ....................................................................66
2. Low income – transitory or long term?............................................................67

9 Topic areas.......................................................................................................................69

1. Introduction..............................................................................................................69
2. Debt..........................................................................................................................69
3. Subjective well-being...............................................................................................73
4. Issues of question wording – consensual deprivation indicators.............................75

10 References....................................................................................................................77 Annex A: Data sets used in this study .....................................................................................80 PSE.......................................................................................................................................80 FACS....................................................................................................................................80 BHPS....................................................................................................................................81 FRS ......................................................................................................................................81 Annex B: Main tabulations by country....................................................................................83 Debt and deprivation............................................................................................................83 Subjective well-being (families with children)....................................................................84 Annex C: Agreed question list for Family Resources Survey..............................................85

List of Tables

Table 3.1 Validity of reduced scale in PSE: lacking two items...........................................24 Table 3.2 Composition of deprivation (lacking two items) using shorter series of questions

24

Table 3.3 Proportions of families unable to afford child-related items...............................27 Table 3.4 Holidays for adults and children, in families with children.................................29 Table 4.1 Unconditional analysis of adult poverty questions..............................................31 Table 4.2 Latent Class Analysis – 2 class solution..............................................................33 Table 4.3 Unconditional analysis of child poverty questions..............................................34 Table 5.1 Items and activities considered essential by people with families compared with

the general population (differences of 3% or more)........................................................36 Table 5.2 Items and activities by gross household income (excluding refused/DK)...........38 Table 5.3 Items and activities by region..............................................................................40 Table 5.4 Items and activities by ethnic group ....................................................................42 Table 5.5 Correlation between questions and various income concepts..............................44 Table 6.1 Unconstrained confirmatory factor analysis, oblique three-factor solutions (Layte et al, 2000) .......................................................................................................................46 Table 6.2 PSE survey, all items and activities.....................................................................50 Table 6.3 PSE survey, items and activities considered as necessities .................................51 Table 6.4 PSE survey, additional questions relating to deprivation ....................................52 Table 6.5 PSE survey, children's items and activities.........................................................53 Table 6.6 FACS survey........................................................................................................54 Table 6.7 BHPS survey........................................................................................................56 Table 7.1 Trends over time in deprivation questions - FACS .............................................60 Table 7.2 Trends over time in deprivation questions – BHPS.............................................62 Table 7.3 IS dynamics and living standards measures.........................................................63 Table 7.4 Chain-linking' different index series..................................................................64 Table 8.1 Saving, managing and deprivation status: bottom FACS quintile.......................67 Table 9.1 Deprivation and Debt (FACS 2002)....................................................................70 Table 9.2 Deprivation, debt and poverty (FACS 2002).......................................................71 Table 9.3 Deprivation, debt (and informal borrowing) and poverty (PSE 1999)................71 Table 9.4 Deprivation, subjective well-being and average equivalised incomes (PSE 1999)

73

Table 9.5 Subjective well-being by age group, in the bottom third of equivalised incomes

(PSE 1999).......................................................................................................................74 Table 9.6 Subjective well-being by family status, in the bottom third of equivalised

incomes (PSE 1999) – those aged 16-64 .........................................................................74 Table 9.7 Subjective well-being by ethnic status, by halves of equivalised income (BHPS

wave 11)...........................................................................................................................75 Table 9.8 Deprivation indicators: different approaches.......................................................76

List of Figures

Figure 2.1 Child-related items that families were unable to afford...................................19 Figure 2.2 The number of child items that families could not afford ...............................20 Figure 3.1 Overlap between full deprivation scale and reduced subsets of 2-10 questions

22

Figure 3.2 Main and remaining PSE deprivation questions..............................................23 Figure 3.3 Overlap between full child deprivation scale and reduced subsets of 2-10

questions 26

Figure 3.4 Distribution of parental and child poverty in families, using deprivation

indicators (mosaicplot' for banded groups)....................................................................28 Figure 3.5 Adult and child deprivation scales (scatterplot with small jittering') ............30 Figure 7.1 Model of product diffusion..............................................................................58 Figure 8.1 Incomes histories of those in the bottom quintile in 2001...............................68 Figure 9.1 Debt and deprivation by income level (AHC).................................................72 Figure 9.2 Debt and deprivation by income level (BHC) .................................................72

Acknowledgements

This analysis was funded by the Department for Work and Pensions. Liaison was managed by Ashley Kershaw and Anne Mason. Helpful advice was provided by Elaine Squires, Daphne White, Steve Ellerd-Elliott and colleagues. Advice and comments on the first draft were provided by the DWP technical experts group on measuring child poverty.

Data from the Family Resources Survey (FRS) and the Families and Child Study (FACS) were supplied by the Department for Work and Pensions.

The data used in this analysis from the Millennium Survey of Poverty and Social Exclusion (PSE) and the British Household Panel Study (BHPS) were supplied through the ESRC Data Archive at Essex University. We thank them for the supply of data, and thank the data collectors for their efforts. The PSE study was funded through the Joseph Rowntree Foundation, involving teams based principally at Universities in Bristol, Loughborough and York. The BHPS is ESRC-funded through Essex University, Institute for Social and Economic Research. None of these organisations bears any responsibility for the analysis.

Authors

Sharon Collard is a research fellow, and Stephen McKay a senior research fellow, at the Personal Finance Research Centre, University of Bristol.

Recommendations on question selection

Overview

There is strong existing literature on the use of deprivation indicators. These appear to be a widely accepted way of analysing deprivation, having strong face validity. They both command academic support, and are explicitly based on a public basis of support (Section 2).

Panel analysis from FACS shows that families that have low incomes, but do not appear to be deprived, are often captured in a temporary downturn in fortunes – or may have incomes that are mis-measured including through imputation. In such cases it seems that deprivation indicators – and some of other kinds of questions we propose – are at least as good a guide to living standards as income (Section 8).

It is not possible to include in FRS the range of deprivation questions that would be used in a special purpose one-off study, such as PSE which included over 80 questions covering both adults and children. However, it is possible to identify as deprived most of the same people using a relatively short series of questions. That much is clear from the analysis conducted (Section 3). Moreover, the group identified as deprived using a shorter series of questions is not clearly different; average incomes and subjective assessments are very similar whether you use the longer or shorter series (Section 3.2).

We therefore recommend including around ten questions relating to adult deprivation indicators. The particular questions may be drawn from the kinds of questions asked in PSE, FACS and BHPS. By taking questions from these surveys we will be able to track progress, at least to some extent, for a longer time period. Some of these questions have been included in the BHPS since the mid 1990s, and in FACS from 1999.

It is current practice to ask respondents for the reason why they do not have particular goods or services. Is it because they "do not want" them or "cannot afford" them? The measure of deprivation is then based only on those saying they cannot afford them. It has been argued that this split can sometimes be misleading, and simply lacking the good or service is more important[1]. However, the existing practice remains widely accepted. Moreover, it is clearly still possible to identify that goods are missing from a three-fold coding structure (but not to extract "cannot afford" from a binary set of responses).

The questions we select should be relatively robust to technical innovations and changes in tastes and relative prices. In part they relate to having some slack in budgets – being able to save, being able to maintain stocks of goods. However it is certainly worth reviewing the set of such questions over time. The method of chain linking' (as used within the Retail Price Index) provides one model of having a single concept with varying components (see section 7.4). This is a possibility worth investigation if analysts are prepared to convert a range of questions into a single index of living standards. It would allow for varying items forming part of an index of wellbeing.

In a measure of child poverty, we believe it is important to include a series of questions that are, indeed, specifically about children. Questions should try to capture details of their poverty and deprivation, not just those of the surrounding family unit. In fact fewer families cite problems affording child items than say they cannot afford adult items. It appears that child deprivation is predominantly a subset of adult deprivation, not a whole new group. Children may be protected' from some effects of low income through the priorities of their parents and the choices that they make. This point has routinely been made in qualitative work, and now finds support in quantitative analysis. Hence too exclusive a reliance on child-related questions may tend to give a more narrow definition of poverty among families with children. However, we recommend including at least 6 questions of this kind, specifically for families with children.

There is, we think, less experience with child-related deprivation questions. The distribution of responses is fairly uniform, unlike for adults, which may indicate different priorities within families. The reliability' of the child questions is also somewhat lower, and the different statistical methods do not provide consistent answers as to which questions are, overall, best'. Further work on the child-related questions is likely to be important, and the selection of particular questions involves a larger element of judgement than for the adult-related questions. At present the questions are arguably too much focussed on children of school age, and not sufficiently relevant to pre-school aged children.

By having a reduced set of questions, there may be a significant proportion of families unable to afford a large proportion of them. On a practical level it might be sensible to include a couple of questions that almost all people will be able to answer positively', particularly if they have a direct bearing on poverty. This must be consistent with not occupying too much space, of course, and providing meaningful information in its own right.

There are different ways in which these subsets of questions may be selected. We have discussed the use of two relatively naïve' approaches (Section 4). That is, to use those items that families were most commonly unable to afford, or to select those questions with the greatest association with the outcome (the overall index). We have also investigated the correlations between individual questions and income, and how widespread is the agreement that they really are necessities (Section 5). Using more powerful statistical methods we have used factor analysis to consider if there are particular dimensions of poverty, as was used to select questions for the consistent poverty measure in Ireland. A relatively exploratory use of latent class analysis (akin to cluster analysis) was able to indicate a useful set of adult deprivation questions as well (Section 4.1).

Ultimately the different methods mostly point to the same questions as having greatest relevance. A mix of statistical evidence, survey question design experience, and reasonable speculation about the future development of family incomes is needed to arrive at the sets of recommendations used here.

We believe the deprivation indicators should be supplemented in two further ways. First, using subjective assessments of people's living standards. Subjective questions about living standards have been regularly used in large-scale surveys and are well correlated both with income and sets of deprivation indicators. They may play a useful role in qualifying families' living standards when income appears to be very low. We therefore propose including at least two such questions (Section 9.3).

We argue that questions on debt provide another particularly useful indicator of living standards (Section 9.2). If people are unable to meet their current spending within their current income, having to borrow – or running down savings – are the logical result. Given that few lower income families have savings of any real kind, then having to borrow is a good indicator of stress in making ends meet. There are important conceptual questions about how to measure borrowing, arrears, problem debt, and so on. We have tried to develop a fairly minimal set of questions based on the FACS series.

It seems clear to us that it is getting into arrears that is the key question, not simply the fact of having outstanding commitments. Indeed the presence of some types of commitments (mortgages and to a lesser extent credit cards) may be indicative of a degree of affluence rather than poverty. We also focus on the current situation, rather than asking questions relating to previous periods. This is therefore fully in line with FRS practice.

However, it is clear that asking about debt consistently and coherently means a series of at least six questions. This is extensive, but we believe both useful and meaningful to respondents. It is possible to replace these with a single question, similar to that used in PSE[2]. If the current suggestions occupy too much space that would be a possible alternative. It might also be possible to reduce the FACS series to a shorter series but this might affect comparability with the FACS questions.

We next section we present the first draft of the questions we proposed for consideration. Tabulations with the proportion with each type of deprivation are shown in Annex B, for each country of Great Britain. The set of questions decided upon by DWP are listed in Annex C. This takes forward the main deprivation questions, and includes a single question on debt situation. We understand there is currently insufficient space to incorporate questions relating to subjective well-being.

Proposed question list

Do you and your family have... /

Are you and your family able to afford

SHOW CARD

1. "We have this",
2. "We would like to have this, but cannot afford it at the

moment"

3. "We do not want/need this at the moment"  Source

Adult  ** A holiday away from home for one week a year, not with  PSE/FACS deprivation  relatives

** represents a  ** Replace any worn out furniture  BHPS/PSE core of the six  ** A small amount of money to spend each week on yourself,  PSE

most key  not on your family

questions.  ** Regular savings (of £10 pounds a month) for rainy days or  PSE

retirement

** Insurance of contents of dwelling  PSE Have friends or family for a drink or meal at least once a  BHPS month

A hobby or leisure activity  PSE

** Replace or repair broken electrical goods such as  PSE refrigerator or washing machine

Keep your home adequately warm  BHPS/PSE Two pairs of all weather shoes for each adult  PSE/FACS Enough money to keep your home in a decent state of  PSE decoration

Child  A holiday away from home at least one week a year with his  PSE/FACS deprivation  or her family

Swimming at least once a month  PSE

A hobby or leisure activity  PSE Friends round for tea or a snack once a fortnight  PSE Enough bedrooms for every child over 10 of different sex to  PSE

have his or her own bedroom

Leisure equipment (e.g. sports equipment or a bicycle)  PSE/FACS Celebrations on special occasions such as birthdays  PSE/FACS Christmas or other religious festivals

Play group at least once a week for pre-school aged children,  PSE ELSE: Going on a school trip at least once a term for school

aged children

Subjective  How often, would you say, do you have money over at the  FACS assessment  end of the week, or if you budget by the month, at the end of

the month?

Would you say it was ... READ OUT ...

Always

Most weeks/months

More often than not

Sometimes

Hardly ever

Or never?

Don't know/too hard to say/varies too much to say

Taking everything together, which of the phrases on this card  FACS best describes how you and your family are managing

financially these days?

"manage very well",

"manage quite well",

"get by alright",

"don't manage very well",

"have some financial difficulties", "are in deep financial trouble"

[or BHPS question – FISIT[3]]

Debt  Sometimes families are not able to pay every bill when it falls  FACS

due. May I ask, are you up-to-date with the bills on this card,

or are you behind with any of them?

"behind with the electricity bill",

"behind with the gas bill",

"behind with other fuel bills like coal or oil",

"behind with Council Tax",

"behind with insurance policies",

"behind with telephone bill",

"behind with television/video rental or HP",

"behind with other HP payments",

"behind with water rates",

"not behind with any of these"

Do you use any of the different ways of buying things listed on this card?

"Credit cards {like Access, Visa etc}",

"Charge cards {like American Express, Diners Club}", "Shop or store cards {like Marks and Spencer, BHS etc}", "Catalogues / mail order schemes",

"none of these"

At the moment are you able to manage the repayments on (name of card mentioned). I mean, to meet the minimum amount you have to repay?

Yes

No

Over the past 12 months, have you used any of these ways to  FACS borrow money? CODE ALL THAT APPLY

"a bank overdraft",

"a fixed term loan from the Bank or Building Society (NOT MORTGAGE)",

"a loan from a finance company",

"a loan from a money lender or 'tally man'",

"a loan from a friend or relative",

"a loan, or advance on wages, from your employer",

"a Social Fund loan",

"none of these"

Are you able to keep up with the repayments for the (type of  FACS loans mentioned) or are you getting behind?

"keeping up",

"getting behind"

READ OUT Living comfortably / Doing alright / Just about getting by / Finding it quite difficult / or Finding it very difficult?

Is your rent/mortgage paid up to date at the moment, or do FACS, you have some arrears that will have to be paid? adapted

"Up to date",

"Some arrears"

Revisiting the list of proposed questions

In this section we present further analysis relating to the above proposed list of questions.

Having established this list, a number of more detailed questions were raised as part of the process of drafting a new section for FRS. We have listed these enquiries separately in this section, to enable the logic of the questions' development to be more clearly followed.

First, it was suggested that the two child questions relating to having a hobby and going swimming may be covering similar ground. Analysis showed, indeed, that there was a high degree of correlation between answers given on these two questions (=0.38). No other child-level question was so highly correlated with the swimming variable. There is therefore a case for trying to combine these into a single question.

Second, alternative questions relating to debt were proposed, from sources such as BHPS, some of which were also used in the ECHP. This approach would have provided greater time-series consistency. However, the questions referred to dealt only credit commitments (HP, loans) and not other kinds of commitment (utility bills, Council Tax), and therefore have more limited scope.

Third we were asked to consider the scope for having one rather than two questions about holidays. The adult deprivation questions refer to a holiday "away from home for one week a year, not with relatives", whilst the child deprivation questions ask about a holiday "away from home at least one week a year with his or her family". Is there scope to reduce this to a single question? Overall, 23 per cent of families with children said they did not have and could not afford a holiday (adult-version), whilst 18 per cent could not afford a holiday (child-version). The overlap between these two groups was substantial. Of those unable to afford an adults holiday', some 65 per cent also could not afford a child holiday (though 30 per cent were able to provide a child holiday, presumably staying with relatives). Conversely, only two per cent of those lacking a child holiday said they had an adult's holiday (two respondents out of 97). Those families with children lacking holidays of both types, compared to those lacking just the adult holiday' have slightly lower incomes and rather larger families. The child holiday question is therefore picking up the depth of poverty perhaps more than the incidence of poverty.

It was commented that the answers to some questions assume a degree of access – facilities such as swimming pools may be less easily reached by those in rural areas. This could be one of cause of deprivation. Users with access to detailed survey data, not generally available in public-use version, may be able to explore the consequences of location on deprivation indicators in more detail.

1 Introduction

1. Objectives

The aim of this study is to suggest questions on deprivation to be included in the Family Resources Survey (FRS) from April 2004 onwards. The FRS contains the best available income data for a large national sample of the household population. It is the foundation of current statistics on households below average income, for which it replaced the smaller and more consumption focused Family Expenditure Survey (FES). At present, the FRS does not contain questions about deprivation and living standards measures more generally, except for a section on consumer durables. The new questions proposed will mostly replace the section on durable goods.

This investigation is conducted within a number of other constraints. The analysis is based on existing questions that have been implemented in at least one major survey. We make no attempt to design new sets of questions. We do, however, comment on some important issues of question wording, where it varies between surveys or where we think it may be improved. The agencies charged with conducting the fieldwork (ONS and NatCen) may also find ways to enhance issues of question wording and ordering still further. It would be entirely appropriate that they advise on such issues. Piloting and perhaps other forms of testing (such as cognitive testing) may also prove helpful in framing the new section on deprivation.

Questions in the FRS have a distinctive character. Where possible questions relate to facts, not judgements or opinions. They are based, for the most part, on information for the current time rather than being retrospective. Any questions proposed must fit within this broad framework. They must also fit within the household, benefit unit and individual levels of questioning currently used. The existing durables questions are asked within the household block.

There are two main aims for these questions:

• to provide a measure of material deprivation;

• to validate the income data at the bottom of the income distribution.

2. Methods

In this investigation, we have not presupposed that any particular methodology for measuring poverty has been selected. Instead we aim to avoid ruling out any options, as far as possible. Some people are likely to continue to believe that income lines, such as 60 per cent of the median, or some threshold set by a budget standards method, represent the ideal means of measuring poverty. Others may incline towards using deprivation indicators, but suggest that some weighting of such indicators is needed. Our aim is to analyse existing data to try to select those questions capable of providing greatest insight into people's living standards and the extent of deprivation. In doing this, we do not wish to remove any existing information that may be exploited within mainstream poverty measurement.

The impetus for this study is, of course, the consultation relating to the measurement of child poverty. The Government has committed to eradicating child poverty within a generation, with staging posts along the way. We aim to provide a measure that is sensitive to child poverty issues.

We are well aware that income measurement and often poverty measurement for "children" is treated as synonymous with "families containing children". A direct child focus is rare. Clearly many of the issues relating to child poverty are those of parental income – but by no means all of them. We are therefore likely to want to include questions that pick up the lived experience and possible deprivation of children. However, wherever possible it would be beneficial to include questions that may be answered by all family types. Otherwise the relative position of families with children compared to others will not be amenable to quantification.

The research is based on secondary analysis of a number of large national datasets, principally:

• the Millennium Study of Poverty and Social Exclusion (PSE), 1999.

• the Families and Children Study (FACS) (1999-2002).

• the British Household Panel Study up to wave 10 (BHPS) (1991-2001).

• the Family Resources Survey (FRS) (2001-02).

Annex A outlines their role and coverage for readers unfamiliar with these surveys. A very short description is that the PSE is a small [N=1500 respondents] one-off survey of poverty covering the population in 1999; FACS is a refreshed longitudinal study of families with children started in 1999 [N=7500 families]; BHPS is a household panel covering all individuals since 1991 [N=5500 households]; FRS is a repeated cross-sectional survey covering all population groups, and focussing on incomes [N=25000 households].

Each of these studies interviews only those in private households. By definition those in institutions (such as prisons, care homes) and some others in temporary accommodation are excluded. These groups may be particularly prone to poverty, though they are hard to reach and so excluded from most surveys.

From 2002 the FRS was extended to Northern Ireland. However, most of the analysis conducted in this report is based on data collected for Great Britain. This analysis pre-dated the publication of new survey data in Northern Ireland covering deprivation (Hillyard et al 2003), which found higher rates of poverty than in Great Britain. The methodology employed followed that of the PSE, with strong similarities on items regarded as necessities.

2 Approaches to poverty measurement

1. Introduction

There are different ways of measuring poverty. Poverty may be measured in an indirect' way, using income, or in direct' ways using questions that measure living standards (Ringen 1988). Using particular income cut-offs to measure poverty, typically based on proportions of the mean or median income, can appear arbitrary in identifying the poor. Moreover, measured income may not always correlate well with people's standards of living (Perry 2002). Measures based on direct indicators of poverty, or living standards more generally, may be used either instead of, or in combination with, income.

Several publications in Britain have recently raised interest in direct' measures of poverty, based on different deprivation indicators. These new publications have included, most notably, the Poverty and Social Exclusion Survey of Britain (PSE), conducted in 1999 by three university-based teams (Gordon et al 2000) and funded by the Joseph Rowntree Foundation. At the Policy Studies Institute, a measure of hardship among families with children has been developed largely based on enforced lack of necessities' (Marsh et al 2001, Vegeris and McKay 2002). At the Centre for the Analysis of Social Exclusion (CASE), various studies have used indicators of social exclusion, using data from the British Household Panel Study (Burchardt et al 1999). Researchers at CRSP have explored a large number of variables, direct and indirect, to identify poor children in BHPS and PSE (Adelman, Middleton and Ashworth 2003).

There has also been clear policy interest in the direct measurement of poverty. The Government's target is to reduce and ultimately end child poverty, and the Department for Work and Pensions' public consultations on measuring child poverty have included some emphasis on deprivation indicators (DWP 2002).

There are, of course, many different ways of measuring poverty. They include the use of budget standards, income lines, self-assessments, and other methods. Roll (1992) and Alcock (1997) provide accessible introductions to the definition and measurement of poverty. One important way that poverty may be measured is through the use of (consensual) deprivation indicators'.

2. Using deprivation indicators to measure poverty

The idea that indicators of living standards may be used to capture poverty has been associated with the work of Townsend (1979). In that study, a list of 12 items (selected by Townsend) was used to identify a threshold point in the income distribution that could be regarded as indicating poverty.

In this method, people are asked to judge which items all families should be able to afford. People are then poor if they are unable to afford items that the majority in society say are necessary – an enforced lack of socially perceived necessities' (Mack and Lansley 1985: 39). The approach may be broken down into a number of different steps. The method generally employed by social scientists to construct deprivation indicators today is as follows:

1. Ask a large group of people which of a number of items they believe to be essential, which everyone should be able to afford.
2. Ask a large sample which of these items they have, which they lack because they don't want them, and which they lack because they can't afford them.
3. Add up the number of items that people can't afford, from the list of those that at least 50 per cent said were essential (alternative [higher] thresholds are possible, though used infrequently).
4. Establish a threshold point at which a shift in the experience of deprivation seems to occur. This may be set using statistical criteria, graphical approaches, or perhaps in other ways.

In arriving at the more modern formulation above, two principal innovations were introduction by Mack and Lansley (1985), over the earlier work of Townsend (1979).

First, rather than using some expert or researcher judgement on which items people should have, they asked people for their opinions about what items everyone should be able to afford. Second, instead of looking simply at non-possession of those items, they checked whether people were lacking items through inability to afford, or choice. These refinements may be seen, at least in part, as a response to various criticisms that had been made of Townsend's approach (such as by Piachaud 1981). In particular, that choosing to do without items could not be regarded as poverty – hence the later emphasis on being unable to afford items, rather than electing to do without them. In more recent studies, people are therefore asked to say either that they don't want items, or alternatively that they cannot afford them (Gordon et al 2000, Vegeris and McKay 2002).

3. The 1999 PSE

The study of Gordon et al (2000) found that 35 items were considered necessities' by more than half the sample, from a list of 54 presented to them. Among the goods and services described as necessary were having a damp-free home, owning a refrigerator, having home contents insurance, having a television, and having a dictionary[4] – all of which were classed as necessities' by 50 per cent of people of more. The list presented to respondents also included a number of items that fell short of the 50 per cent threshold, including new, not second-hand clothes', a car, a tumble dryer, annual foreign holidays and access to the internet.

Six of the 35 items classed as necessities' were dropped on statistical grounds from the list used to measure who was poor. A threshold of lacking two items was selected, since that seemed to maximise income differences across the two selected groups (and minimise income differences within the groups). It is possible to discuss the overall statistical strategy taken in dropping these six items, but the effect on the final measure is so small it is not worth a detailed debate. Of those lacking 2 or more items from the 35-item list, 99.3 per cent lacked 2 or more items from the shorter 29-item list. For this reason, and on the grounds of simplicity and transparency, in the analysis that follows we include the full 35-item list – but no results would be affected by using the alternate shorter list.

The great strength of the PSE lies in the wide range of poverty and deprivation-related information collected. It was able to use the GHS information to collect income data, and the subsequent PSE could then be used to concentrate on a range of questions related to living standards, social exclusion and poverty.

These datasets have since been made available at the ESRC Data Archive at Essex University, with serial numbers 4349 (PSE) and 4384 (ONS Omnibus).

2.3.1  MEASURING CHILD POVERTY USING THE PSE

The PSE contained questions about 30 child-related items, asking families whether they had or could afford a range of items. These ranged from swimming once a week, to celebrations on special occasions, to stocks of clothing.

In the published report, the data is recast at child-level. Then, a few questions are ignored if they do not relate to particular age ranges. Answers at the family level, which is how the questions were asked, are assumed to relate to each child in a given family.

Unfortunately, it has not been possible from the published data to replicate this approach. So the analysis in this section is based at the family level. There are differences between the family-level distribution and the figures in the PSE report, which is based on child-level analysis. This is to be expected. Larger families tend to be poorer on average than smaller families (Willits and Swales 2003), so a higher proportion of children are poor than families.

The list of 30 items was reduced to 27, by excluding three questions that families did not regard as essential. These were computer games, a computer for school work and 50p a week for sweets. The numbers of families unable to afford the remaining items are shown in Figure 2.1, in order.

One item, being unable to afford a holiday of at least a week away from family, was reported by 18 per cent of families, or nearly three times as often as any other. The other items that families most commonly could not afford were having bedrooms for each child older than ten of a different sex (seven per cent), swimming at least once a month (six per cent), and the child having a hobby or leisure activity (six per cent). Even so, it is noticeable that only one question reached double figures, and for only a further three activities were more than five per cent of families unable to afford it. These are rather lower than the figures for adults, where there were six items that could not be afforded by 10 per cent or more of respondents.

Figure 2.1  Child-related items that families were unable to afford

Holiday away from home for one week a year 18.1% Bedrooms for every child >10 of different sex 6.6%

Swimming at least once a month 5.7%

A hobby or leisure activity 5.4%

A garden to play in 4.4%

Leisure equipment such as a bicycle 3.9%

Friends round for tea or a snack once a fortnight 3.9%

Celebrations on special occasions (eg birthday) 3.4%

Meat, fish or vegetarian equivalent twice a day 3.2%

A bike new or second hand 3.2%

Educational games 3.1%

Play group weekly for pre-school aged children 2.5%

School trip once a term for school aged children 2.4%

At least four pairs of trousers  2.4%

Some new, not second-hand clothes 2.3%

Construction toys such as Duplo or Lego 2.2%

At least four jumpers cardigans or sweatshirts 2.1%

New properly fitted shoes 2.0%

A warm waterproof coat 2.0%

All the school uniform required by the school 1.9%

Fresh fruit or vegetables at least once a day 1.9%

Seven underpants or knickers in good condition 1.5%

A carpet in their bedroom 1.3%

Three meals a day 1.0%

Toys such as dolls and teddies 0.8%

A bed and bedding to her/himself 0.7%

Books of her or his own 0.4%

0% 5% 10% 15% 20%

Source: PSE, family-level analysis

A classification at 1+ items yields 31 per cent of families with children deprived; or at 2+ produces half that number (15.5 per cent). These compare with figures of 34 per cent and 18 per cent for the PSE report based on children. Most of those lacking just one item were going without a holiday.

Figure 2.2  The number of child items that families could not afford

80%

69.0% 70%

60%

50%

40%

30%

20% 15.5%

10% 5.4%

2.1% 2.7% 2.5% 2.7% 0%

None 1 2 3 4 5 6 or more

Number of child items lacking (family level)

Source: PSE, family-level analysis

In the PSE analysis the authors had to consider where to put the threshold to define child poverty. Some relatively formal statistical analysis suggested that lacking one or more items was most appropriate. However, with one particular question dominating the list of items that could not be afforded, there were also grounds for looking at the group lacking two or more items. The authors of the report therefore looked both at those lacking one or more items, and those lacking two or more.

3 Effects of having shorter sets of questions

1. Introduction

The traditional approaches to measuring poverty using deprivation indicators require using a relatively long set of questions – in the PSE study 29 separate questions for adults and 27 questions for children. It is not feasible to include such lengthy lists in the FRS where competition for space is fierce. How much of the usefulness of the approach is retained using fewer questions? Alternatively, how many questions are needed to capture most of the same people as poor? We seek answers to these questions in this section.

2. Adults

The question we now consider is how far a shorter list of questions may adequately reflect the longer list. Or, how many questions are needed to give a defensible coverage of the main list.

The simplest test of this is to take various subsets of the full scale, and investigate how many of the truly' deprived (lacking 2+ of 29 items) would be identified as deprived using a threshold of 2+ of a shorter list of items. The subsets we selected initially were based on the frequency of people being unable to afford them. This ordered' set of questions takes in questions relating to:

• Having a comfortable home environment, including being able to replace worn out furniture and replacing or repairing broken electrical goods.

• Taking part in social and hobby/leisure activities.

• Having adequate food and clothing.

• Being able to go on holiday.

• Having spare money to spend on oneself or to put into savings.

• Having home contents insurance.

As shown below, using two questions is sufficient to identify just over half of those who appeared to be poor with the full scale. Increasing this to six questions extends the reach to almost 80 per cent, with progressively smaller gains from increasing the list of questions asked.

Figure 3.1  Overlap between full deprivation scale and reduced subsets of 2-10 questions

100

90 90.5 91.7

86.5 87.9

82.7

80 79.9 Ordered 'Tweaked'

72.2

70

65.8

60

50 50.1

40

0 2 4 6 8 10 12

2+ items lacking from list of N

It is, however, possible to improve on the coverage by tweaking' the set of items included. For example, by including for the 10 question option the question dryhome' rather than fammeal'. Having a dry home correlates less well with the other items than does family meal. It appears to improve the fit between the shorter subset and the full index because it is better at picking up groups not already identified in the ordered' measure.

The above analysis suggests that just six questions can accurately reflect 80 per cent of the full' picture. We should also inquire whether those identified from fewer questions are a biased set of those identified from the full set. Some tests of this are shown in Table 3.1. This looks at overlapping groups, defined as deprived on the basis of differing numbers of questions – but with the same threshold of lacking two of those items on the grounds of being unable to afford them.

By definition, the shorter lists will identify fewer people as deprived if we retain the same threshold. So, the full list of 29 questions is associated with a deprivation rate of 27.5 per cent. With just six questions this falls to around 23 per cent, with figures in between for varying number of questions.

Another way of testing the robustness of the selected question is to take the ten selected questions, and plot their distribution against that of the remaining questions – 19 from the set used in the PSE report. As shown in Figure 3.2, the greater the number of items lacking from the main set of ten, the greater the number of items that people could not afford from the remainder of the list (19 questions from the set used in the PSE report). This confirms the high degree of overlap and hence consistency between the shortened set of questions and the overall set.

Figure 3.2  Main and remaining PSE deprivation questions

0 1 2 3 4 5 6 7 8 9 10

N lacking from main 10 questions

The average incomes of those unable to afford 2+ items are relatively similar whether we base this on 29 questions, 20 or 10. Using just six, the average income of the deprived group is actually lower than this. We also explored the effect on subjective assessments of poverty. Again there seems to be no or relatively little information loss by adopting a shorter set of questions in preference to a much longer set. If anything, poverty is more severe when a more minimal set of questions is used. It would seem that the most efficient' reduced set of questions would be between six and ten.

Table 3.1  Validity of reduced scale in PSE: lacking two items

Column percentages

Full scale  Reduced scales  All

 (29 items)  20 items  10 items  6 items  respondents

Do you think you could genuinely say you are poor now

All the time  21  22  23  25  7 Sometimes  44  44  44  43  20 Never?  35  33  33  32  73

Equivalised income per week  £250  £248  £251  £238  £381 Per cent deprived' on this  27.5  26.5  25.4  22.7  -

measure

Unweighted base  595  575  552  512  1534 Source: PSE

It is also possible to explore any differences in the composition of those deprived using different sets of questions. Given the high degree of overlap between shorter and longer sets, we would expect the composition of deprivation under each definition to be relatively similar. Some key characteristics are shown in Table 3.2. There were some differences, but these tended to be rather minor. If anything, the shorter series of questions identified a higher proportion of families with children as being deprived, and a few more with a long-standing illness or disability.

Table 3.2  Composition of deprivation (lacking two items) using shorter series of

questions

Column percentages

Full list (29  10 items  6 items  All

items)  respondents

Lone parent  10  11  12  6 Couple with children  36  38  39  31 Single person  21  20  20  17 Couple no children  33  32  29  47

Has no educational qualifications  36  35  35  24 Has a long-standing illness  46  48  49  38

Average number of adults  2.1  2.1  2.1  2.1 Average number of children  1.0  1.1  1.1  0.7

Unweighted base  595  552  512  1534 Source: PSE

3. Children

The coherence of the child deprivation scale is, perhaps, not so clear as for adults. Work analysing the reliability of the scale, in particular, gives less grounds for confidence than it did among adults.

Subsetting the questions gives reasonable coverage of the whole scale, at least once six or more questions are included. Those lacking two or more of the six items constitute 72 per cent of those lacking two or more items from the full 27-item list.

This chart is based on taking the most frequently missing items. The bowed-in' elements of the line indicate that there is scope to improve on the overall overlap by using the component variables in a different way. This feature arises through the various links between lacking different of the groups of items.

Figure 3.3  Overlap between full child deprivation scale and reduced subsets of 2-10 questions

100

90

85.9 80 78.9

76.1 76.1

70 72.2

64.8 60 61.1

50

40 38

30

20

13.9

10

0

0 2 4 6 8 10 12

2+ items lacking from list of N

3.3.1  FACS

The FACS data contains a range of deprivation indicators with seven questions specific to children, or mostly relevant to them. In 2002 the proportions of families unable to afford them ranged from nearly one in five (money for trips, outings, etc) to one in fifty (weatherproof coat for each child). The first two questions (Table 3.2) are potentially helpful to measuring deprivation, in that they are not concentrated in a small group of the population. However, the first question contains rather too many elements to carry complete confidence. It would be better expressed in more than one question – perhaps one relating to holidays, another to school trips, another for day trips, and so on.

None of the questions have exact parallels in PSE. However, for comparison in PSE 2.0 per cent of families could not afford a warm, waterproof coat' (the same as here), whilst the same proportion could not afford new, properly fitted shoes' (much lower than the alternative version used in FACS). Some 3.6 per cent, in PSE, could not afford "celebrations on special occasions".

Table 3.3  Proportions of families unable to afford child-related items

Cell percentages

Source: FACS 2002

4. Connecting parental and child deprivation

Families who were not in poverty (on the basis of the adult list of goods and services) occasionally lacked child items through not being able to afford them, although this was rare. Among families lacking none or only one of the adult necessities, just one per cent lacked two or more of the child items. Conversely, some 26 per cent of those not experiencing child deprivation were lacking two or more of the adult-based items. A mosaic' plot of this data is shown below (Figure 3.4). For those more accustomed to traditional scatterplots, the same overall patterns are shown in Figure 3.5 in the annex at the end of this chapter.

The conclusion that may be drawn is that families only tend to lack child items, on the grounds of inability to afford, once they are lacking several adult items. Even among those parents lacking between two and eight adult items, the upper end equating to severe hardship, still 77 per cent were not suffering from child deprivation (defined as being unable to afford two or more of the child items). However, when adult poverty reaches a very severe extent then child poverty very often accompanies it.

Figure 3.4  Distribution of parental and child poverty in families, using deprivation

indicators (mosaicplot' for banded groups)

Adult and child deprivation

Not poor Lack 2-8 Lack 9+

Lack 2+ Poor 1 lack

Not poor

Adult deprivation

A particular example are the different questions asked about holidays, relating to the whole family and then to children. Respondents were asked if they had A holiday away from home for one week a year, not with relatives'. Families with children were asked if children had A holiday away from home at least one week a year with his or her family'. A cross-tabulation appears in Table 3.4. Obviously the questions are not identical, and the question relating to children may perhaps be satisfied in a less costly way by staying with other family members. But for many families each question would presumably be satisfied by the same trip – the nuclear family participating in a holiday.

Among families who could not afford an annual holiday, away from the family, still one in three (31 per cent) did manage a holiday including their children. Similarly, for the small number of families not wanting a holiday, two thirds (62 per cent) went on a family holiday that included the children. The balance of evidence is of children receiving a holiday of some kind, when resources do not allow for a fully independent parents and children break.

Table 3.4  Holidays for adults and children, in families with children

Column percentages

Note: not applicable' codes dropped.

Annex: detailed comparison and adult and child poverty measures

Figure 3.5  Adult and child deprivation scales (scatterplot with small jittering')

Adult and child deprivation

20 Adult deprivation

threshold

15

10

Child deprivation 5 threshold(s)

2 1

0

0. 2 5 10 15 20 N adult items CN afford

A small amount of random noise' or jitter has been added to each point. This is to prevent points over-plotting each other, since the data are categorical. Without adding this randomness the cloud of points at [0,0] would appear as a single point.

4  Reducing sets of questions – naïve approaches and latent class analysis

1. Adult poverty

An obvious or naïve approach to selecting questions might be to show which were most strongly associated with being deprived. In other words, among those classified as deprived, which items could they most often not afford? So, for instance, 74 per cent of those lacking two or more items could not afford regular savings for rainy days, whilst 66 per cent could not afford to replace worn out furniture, and 51 per cent were unable to find the money for an annual holiday (Table 4.1). Conversely, six per cent of the non-deprived could not afford savings, and four per cent each for the other two items just mentioned. The ordering of items in this list is related to, but slightly different from their frequencies of not being affordable. For instance, overall 12 per cent could not afford to replace furniture compared with 18 per cent unable to afford a holiday. This perhaps indicates that being unable to afford a holiday appeared as the single item lacking more often than some of the other questions.

Table 4.1  Unconditional analysis of adult poverty questions

Cell percentages

An alternative approach to ostensibly the same question is to use latent class analysis (Bartholomew et al 2002). Under this approach it is assumed that we do not observe the true variable, deprived'. It is hidden, or latent. Instead we observe a series of manifest or

observed variables – the deprivation indicators – which only correlate with each other because of the latent variable. Otherwise they are independent (known as local independence).

The variables we observe (the deprivation indicators) are categorical. Moreover, it is reasonable to assume that poverty is categorical and indeed dichotomous – certainly that is how most studies perceive it. Bartholomew et al (2002) suggest the following applications for different kinds of latent variable models, depending on the data structure. Using this schema suggests we use a latent class analysis approach. Later in the report we resort to the more familiar factor analytical approach – in line with other studies if not wholly in keeping with this particular statistical advice based on the data types.

Observed variables are:

Latent variable(s) is(are):  Interval/ratio scale  Categorical/ordinal

Interval/ratio scale  Factor analysis  Latent trait analysis Categorical/ordinal  Latent profile analysis  Latent class analysis

The results from a LCA are shown in Table 4.2. The analysis suggests that the deprivation group was around 29 per cent of the sample. The LCA builds this from the individual questions, and it is reassuring that this proportion is very similar to the proportions based on the thresholds identified by Gordon et al (2000). Given the subject matter a two class solution is appropriate.

Table 4.2  Latent Class Analysis – 2 class solution[5]

2. Child poverty

We may also analyse the child deprivation questions to check their usefulness against the overall scale. In Table 4.3, below, we show rates for those lacking just one item out of 27, and those lacking two or more[6]. The patterns suggest the authors were probably right to go with their judgement that two or more items was the appropriate threshold, despite some of the statistical analysis that suggests a threshold at just one.

Among those lacking two or more items, 70 per cent could not afford a holiday. This was also among the most important indicators for adult poverty, of course. Children having a hobby, and participating in monthly swimming sessions, were the next most important indicators of child deprivation. The relative contributions of the other questions were less important. Compared to adult deprivation, there was greater diversity in the kinds of items that the families were unable to afford.

Table 4.3  Unconditional analysis of child poverty questions

Cell percentages

Source: PSE 1999

5 Checking validity of individual questions

1. Introduction

Several sections in this report provide statistical grounds for selecting questions. This section provides more detail about individual questions. How far, among the deprivation indicators, do they command assent among different groups? And how well are they correlated with income?

2. Perceptions of necessities

AS outlined previously, in the first stage of the 1999 Poverty and Social Exclusion Survey, respondents were asked to identify which of a list of items and activities that they considered necessary, which all adults should be able to afford and which they should not have to do without'. Of 54 adult items and activities, 35 were thought necessary by more than 50 per cent of the population. Similarly, respondents were asked to distinguish items and activities that they considered necessary for their children. Of 30 items and activities, 27 were thought to be necessities by over 50 per cent of parents.

If some of these perceived necessities are to be included in a measure of deprivation, they should ideally be relevant to different sections of the population. The following sections therefore explore how much consensus exists between (a) families and the population as a whole; (b) different income groups; (c) different regions; and (d) different ethnic groups, about the items and activities that they consider to be essential. Analysis of this kind, including by age group, may also be found in Bradshaw, Williams and Middleton (2000) for child-related items.

1. FAMILIES COMPARED WITH THE POPULATION AS A WHOLE

On the whole, families with children held fairly similar views to the general population as to the type of items and activities that are essential.

For half the items and activities asked about, there was less than three percentage points difference between the proportion of people with children who considered these to be essential and the population as a whole. As Table 5.1 shows, for the remaining items and activities the difference was three or more percentage points and, for the most part, people with children were somewhat less likely to consider these items and activities to be essential than the population as a whole. Where people with children were more likely to consider items and activities to be essential, these tended to be child-focused, e.g. construction toys or visits to school such as parents evening (Table 5.1).

Table 5.1  Items and activities considered essential by people with families

compared with the general population (differences of 3% or more)

% % Difference

essential  % essential  between

- all  - families  families and all

Items considered less essential by families

Money to spend on self weekly  58.9 42.0 -16.9 Telephone  71.1 62.7 -8.4 A dressing gown  33.7 25.6 -8.1 A holiday away from home (child)  70.5 63.4 -7.1 Coach/train fares to visit family/friends  38.2 31.4 -6.8 Two pairs of all weather shoes  63.9 57.1 -6.8 Having a daily newspaper  30.3 23.7 -6.6 An outfit for social occasions  50.6 44.1 -6.5 Friends round for tea or a snack (child)  58.7 52.8 -5.9 Attending place of worship  41.5 35.8 -5.7 A evening out once a fortnight  38.9 33.5 -5.4 Going to the pub once a fortnight  19.9 15.2 -4.7 A roast joint/vegetarian equivalent weekly  55.6 51.3 -4.3 At least 50p a week for sweets (child)  49.2 44.9 -4.3 Computer games (child)  17.7 13.4 -4.3 Visits to friends or family  84.1 79.8 -4.3 A holiday away from home  54.8 50.6 -4.2 A hobby or leisure activity  78.3 74.2 -4.1 Insurance of contents of dwelling  79.0 75.0 -4.0 Swimming at least once a month (child)  74.6 70.8 -3.8 Presents for friends/family yearly  56.0 52.5 -3.5 New, not second hand, clothes  47.7 44.4 -3.3 Replace worn out furniture  53.9 50.6 -3.3 Attending weddings, funerals  80.2 76.9 -3.3 Celebrations on special occasions  83.0 79.8 -3.2 A dictionary  53.4 50.2 -3.2 Computer suitable for school work (child)  41.4 38.2 -3.2 Regular savings for rainy days  65.9 62.9 -3.0 A television  55.5 52.5 -3.0 Some new, not second-hand clothes (child)  70.0 67.0 -3.0

Items considered more essential by families

A bed and bedding to her/himself (child)  92.8 95.8 3.0 A home computer  11.4 14.8 3.4 Construction toys such as Duplo or Lego (child) 62.1 65.8 3.7 Carpets in living rooms and bedrooms  67.2 71.5 4.3 At least four pairs of trousers (child)  68.8 73.5 4.7 A bike, new or second hand (child)  54.4 59.7 5.3 Collect children from school (child)  74.6 81.2 6.6 A carpet in their bedroom (child)  66.8 74.5 7.7 Visits to school, e.g. sports day (child)  81.0 90.6 9.6

Weighted base*  1,782 481 - Unweighted base*  1,785 461

Source: ONS Omnibus Survey, 1999

* The base varied slightly between different items because of missing data

This seems to indicate that, for people with children, having disposable income and taking part in family and social life is rather less important than it is for the population as a whole. In particular, having money to spend on themselves every week is far less of a priority for people with children than it is for the general population. So, while nearly 60 per cent of the population consider having spare money for themselves to be essential, this is true of only 42 per cent of families. This is particularly significant when we consider that Gordon et al (1999) defined a necessity as any item or activity that more than 50 per cent of the population thought was essential. According to this definition, while the general population consider having spare money to be essential for all adults, this is not true of people with children. This also applies to having an outfit for social occasions, although the difference is not as pronounced.

2. COMPARISON ACROSS INCOME GROUPS

To facilitate comparison, respondents were divided into three groups of roughly equal size according to their gross household income. As Table 5.2 shows, household income seems to have a greater effect than family status upon people's views of necessities. In particular, there are several items and activities that the majority of low-income households consider to be essential that most households with middle-to-high incomes do not (Table 5.2, top section).

Table 5.2  Items and activities by gross household income (excluding refused/DK)

% diff  % diff

between  between

% med and high and

essential Low  Medium  High  low  low

-all  income* income* income*  income  income

Items classed as necessities

A television  55.5 64.7 53.9 41.3  -10.8 -23.4 An outfit for social occasions  50.6 55.0 47.5 43.6  -7.5 -11.4 Replace worn out furniture  53.9 61.2 51.7 46.6  -9.5 -14.6 A bike, new or second hand  54.4 60.6 54.0 49.9  -6.6 -10.7 At least 50p a week for sweets  49.2 59.8 52.2 33.8  -7.6 -26.0 Computer suitable for school work  41.4 50.7 40.2 32.4  -10.5 -18.3 Other items

A dressing gown  33.7 46.9 27.3 27.5  -19.6 -19.4 Two pairs of all weather shoes  63.9 70.5 63.3 58.5  -7.2 -12.0 A roast joint/veg equivalent wkly  55.6 60.9 55.2 50.9  -5.7 -10.0 Carpets in living rooms and

bedrooms  67.2 72.5 67.2 61.5  -5.3 -11.0 A car  37.7 30.7 42.1 36.0  11.4 5.3 Having a daily newspaper  30.3 34.6 24.4 28.6  -10.2 -6.0 A evening out once a fortnight  38.9 43.5 35.3 31.3  -8.2 -12.2 Visits to school, e.g. sports day  81.0 73.9 86.4 84.8  12.5 10.9 Enough bedrooms for every child  77.5 83.3 78.4 72.0  -4.9 -11.3 Computer games  17.7 25.5 15.6 8.1  -9.9 -17.4 A garden to play in  68.1 77.5 68.7 55.1  -8.8 -22.4 Swimming at least once a month  74.6 85.3 75.1 68.1  -10.2 -17.2 A school trip at least once a term  73.7 80.7 73.3 65.9  -7.4 -14.8 Friends round for tea or a snack  58.7 67.6 53.6 51.5  -14.0 -16.1

Weighted base**  1,782 436 452 454  - - Unweighted base**  1,785 617 443 376

Source: ONS Omnibus Survey, 1999

* Low income: Less than £14,560 gross per year; Medium income: £14,560-£31,200 gross per year; High income: More than £31,200 gross per year

** The base varied slightly between different items because of missing data

The most pronounced differences relate to:

• Children having at least 50 pence a week for sweets, which six in ten low-income households considered to be essential, compared with around half the general population and only a third of high-income households;

• A television, regarded as a necessity by nearly two-thirds of low-income households in contrast to just over half of the population as a whole and four in ten of high-income households;

• A computer suitable for school work, considered to be a basic requirement by half of low-income households but only four in ten of the general population and just a third of high-income households.

The different attitudes towards children having money to spend on sweets probably reflects different cultural values between low and high-income households. Parents from higher income households, for example, may wish to discourage their children from eating sweets on health grounds, and instead encourage them to eat more healthy (and usually more expensive) snacks such as fresh fruit. Conversely, lower-income parents may regard sweets as the most affordable treat for their children, and one that is easily accessible.

As earlier research has indicated, for lower-income households who cannot afford to go out or pursue hobbies and leisure activities, television represents a relatively low-cost form of entertainment (Mack and Lansley, 1985, cited in Pantazis et al, 1999). In contrast, television may be regarded by some people on higher incomes as a dumbing down' of culture, which they frown upon. There does not seem to be an easy explanation as to why more low-income than high-income households consider a computer to be essential.

The bottom section of Table 5.2lists the items and activities where there was at least a 10 percentage point difference in opinion between the income groups as to what constitutes a necessity. In all but two instances, a higher proportion of low-income households regarded these items and activities as essential than either other households or the population as a whole. The exceptions were car ownership and school visits.

3. REGIONAL DIFFERENCES

There are some significant regional differences of opinion about essential items and activities. As earlier research has noted, people living in Wales are generally less likely than elsewhere to consider items as necessities (Pantazis et al, 1999). In the case of eleven items and activities, majority opinion about whether these are essential is divided across the regions (Table 5.3, top section), and the greatest differences are indeed between Wales and the rest of Britain. For example, only 48 per cent of people in Wales consider having money to spend on themselves to be essential, compared with 69 per cent in London and 59 per cent of the general population. In contrast, having a roast joint or its equivalent once a week is regarded as a necessity by over 70 per cent of Welsh people, but just 56 per cent of the population as a whole and only 41 per cent of people in Scotland.

Moreover, for over half of items and activities, the proportion of people who think they are essential varies regionally by 10 percentage points or more. Again, the differences are most marked between Wales and the other regions, and the bottom section of Table 5.3 gives some examples of this.

Table 5.3  Items and activities by region

Source: ONS Omnibus Survey, 1999

* The base varied slightly between different items because of missing data

4. DIFFERENCES BETWEEN ETHNIC GROUPS

Detailed analysis of how different ethnic groups perceive necessities is precluded by the fact that nearly all respondents to the ONS Omnibus Survey described themselves as white - 1,744 respondents out of 1,855, or 94 per cent of the unweighted numbers. The highest number of respondents from any other single ethnic group was only 20, and these were people who described themselves as Indian.

In order to be able to carry out some analysis by ethnic group, we created a single category for people from black ethnic minority groups, which comprised respondents who described themselves as Black Caribbean, Black African or from other Black groups. We also combined the two categories of Indian and Pakistani into one group. Even so, the number of people in these two combined categories remains small, with 23 people classified as Black and 36 as Indian or Pakistani (unweighted numbers). For this reason, we have simply looked at whether people who are Black, or who are Indian or Pakistani are more or less likely to consider things to be essential than the (white) population (Table 5.4). These findings should still be treated with caution.

As Table 5.4 indicates, there are a considerable number of items and activities that Indians/Pakistanis tend to be more likely to consider essential than either the population as a whole or respondents from Black ethnic groups. These include:

• new, not second hand, clothes;

• a car;

• regular savings for rainy days;

• Internet access;

• an evening out once a fortnight; and

• a range of consumer durables including video cassette recorder, home computer, microwave oven, tumble dryer, satellite TV and CD player.

In addition, there are a small number of items and activities that both Indians/Pakistanis and people from Black ethnic groups are more likely to regard as necessities than the general population, namely a dishwasher; holidays abroad once a year; attending place of worship; visits to school; computer suitable for school work and a school trip once a term. A higher than average proportion of people from black ethnic groups feel that having money for fares to visit their family and friends was essential; this is not the case for Indians/Pakistanis.

Of the six items and activities that Indians/Pakistanis are less likely to consider essential than the general population or people from Black ethnic groups, it is notable that four are child- focused (toys; enough bedrooms for every children over 10 of different sex; at least seven pairs of underpants; having friends round for tea or a snack). The remaining two comprise having meat, fish or a vegetarian equivalent twice a day and having a damp-free home.

In contrast, of the eight items and activities that people from Black ethnic groups are less likely to regard as necessities than other respondents, only two are specifically related to children (educational games and having at least four jumpers). Among other things, the remaining items and activities include having new rather than second hand clothes and having a car, both of which are felt to be of above-average importance by Indians/Pakistanis. Finally, there are a number of items and activities that both ethnic minority groups are less likely to deem essential than is the case generally– having a warm waterproof coat (adult), a washing machine, a holiday away from home, and going to the pub once a fortnight.

Table 5.4  Items and activities by ethnic group

Source: ONS Omnibus Survey, 1999, using recoded ethnicity variable

More likely than general population to consider this item or activity to be essential x Less likely than general population to consider this item or activity to be essential

3. Correlations with income

The kinds of questions we have been discussing all have significant correlations with income. This is a basic test of their validity as questions for measuring living standards. Some of the questions correlating most highly with income are shown in Table 5.5. Notable among them are: having a week's holiday not with relatives; money for trips, holidays, etc.; swimming at least once a month for children; and being two months in arrears with housing payments.

Table 5.5  Correlation between questions and various income concepts

Correlation Income  coefficients

Holidays and discretionary spending  variables  (signs ignored) BHPS 10  JHSCANB  Pay for annual holiday  JEQMINC  0.230 BHPS 10  JHSCNTB  Would like to pay for annual holiday  JEQMINC  0.200 FACS 4  DEXPHOL  One week holiday, not with relatives  DAHCINC  0.325 PSE  H6ABROAD  Holidays abroad once a year  MCCINC  0.261 PSE  H3HOL  A holiday away from home  MCCINC  0.222 PSE  J5HOL (CHILD)  A holiday away from home (child)  MCCINC  0.206 PSE  GONEWA9  Gone without: Holiday  MCCINCF  0.212 PSE  H7FARES  Coach/train fares to visit family/friends  MCCINCF  0.229

Money for trips, holidays, outings or going with

FACS 4  DEXPTRIP  gifts to parties  DAHCINC  0.294 PSE  G38MONEY  Money to spend on self weekly  MCCINCF  0.204 FACS 4  DEXPNIGH  Night out once a month  DAHCINC  0.225 PSE  H5MEAL  A meal in a restaurant/pub once a month  MCCINC  0.215 PSE  H10PUB  Going to the pub once a fortnight  MCCINCF  0.202 PSE  GONEWA7  Gone without: Visits to pub  MCCINCF  0.215 PSE  J3SWIM (CHILD)  Swimming at least once a month (child)  MCCINC50  0.327 Subjective views of financial situation

BHPS 10  JFISIT  Change in financial position last year  JEQMINC  0.270 BHPS 10  JLFSAT2  Satisfaction with income of household  JEQMINC  0.220 FACS 4  WORRY  Worried about money almost all time  DAHCINC  0.218 PSE  GENPOR  Consider to be poor  MCCINC  0.247 Food and clothes

Good quality brand name food for family meals

FACS 4  DEXPBFOO  most days  DAHCINC  0.217

New not second-hand clothes for family when

FACS 4  DEXPNEWC  needed  DAHCINC  0.203

Good quality new brand name clothes/shoes for

FACS 4  DEXPBCLO  children  DAHCINC  0.239 Car

FACS 4  DEXPCAR  Car or van  DAHCINC  0.206 PSE  G14CAR  Car  MCCINCF  0.204 FRS 2002  USEVCL  No. of vehicles owned/used  AHCINC ALL  0.206 Credit and debt

BHPS 10  JXPHSDB  Been 2+ months late with housing payment  JEQFINC  0.570 FACS 4  UTDEBT  Debt: utilities  DAHCINC  0.207 FACS 4  ANYHH  Any HH debts  DAHCINC  0.232 PSE  ANYDEBTS  Whether any debts  JEQINC50  0.220 PSE  ANYBORR  Whether any informal borrowings  JEQINC50  0.218 Savings

BHPS 10  JSAVE  Saves from current income  JEQMINC  0.220 BHPS 10  JSAVREG  Saves on a regular basis  JEQMINC  0.230 FACS 4  DSAVMM8  Whether a regular saver  DAHCINC  0.275 PSE  G20REGSA  Regular savings for rainy days  MCCINC  0.240 Use of financial products

FACS 4  NOACC  No current or savings account  DAHCINC  0.218 PSE  NOCURRAC  No current account  MCCINC50  0.213 PSE  G23INSUR  Insurance of contents of dwelling  MCCINC50  0.213

6 Exploring dimensions of deprivation using factor analysis

1. Introduction

There are different ways to reduce a large set of variables to a smaller set with nearly as much information. Exploratory factor analysis is one technique that can be used to investigate if a wider range of questions may be reduced to a smaller number, with the smaller set of questions explaining variation in the larger set. Put simply, it is a way of analysing the correlations between a large number of observed variables in order to reduce them to a smaller number of unobservable underlying dimensions or factors, and to determine the relationship of the original variables to each factor.

However, as Everitt and Dunn (2001) note, Factor analysis has probably attracted more critical comment than any other statistical technique'. The main reasons for this are, first, that the researcher has to employ a certain amount of judgement in order to carry out factor analysis. There are no hard and fast rules about what variables should be included in the analysis, the most appropriate method of factor analysis, the number of factors that should be used to explain the underlying structure of the data, or how those factors should be defined (Calandrino, 2003). Second, because factor analysis is entirely data driven, different solutions are likely to be obtained from different samples or from the same sample over time (ibid).

Factor analysis provided key results in the construction of the Index of Multiple Deprivation. Again, this aroused criticism – as indeed did many of the steps in its construction. Longford (2001) pointed out that the existence of a factor does not mean that what is measured is what is desired: why should that factor be exactly what we want it to be? I recommend Gould's (1987) answer, summarized as "There is no scientific reason." '  

Despite its limitations, factor analysis remains a useful tool for exploring the underlying structure of data. It was used by researchers in Ireland as part of a study that lead to the development of a measure of consistent poverty' (Callan et al 1993). Using data from the 1987 Survey of Poverty, Income Distribution and Usage of State Services, factor analysis was carried out which identified three underlying aspects of deprivation. These were called the basic', secondary' and housing' dimensions. There were eight questions in the basic deprivation domain which related to goods or activities that most people had. Along with relative income, these types of questions (including whether people had two pairs of shoes, new rather than second-hand clothes, a weekly roast' or similar each week) formed part of the consistent poverty measure that was monitored as part of the National Anti-Poverty Strategy.

Subsequent research examined trends in poverty in Ireland between 1987 and 1997 and found that, although relative income poverty had increased, there had been a decrease in deprivation (Layte et al, 2000). Confirmatory factor analysis was then carried out to discover whether this decrease was due to the fact that the deprivation measures had not been updated to take account of changes in living standards and expectations. The results indicated that this was not the case, and that the structure of deprivation in Ireland had changed little over the previous ten years (Table 6.1). The researchers emphasized, however, that the basic dimension of deprivation might have to be revised in the future if the structure of deprivation changed.

Table 6.1  Unconstrained confirmatory factor analysis, oblique three-factor

solutions (Layte et al, 2000)[7]

Using data from FACS, exploratory factor analysis has been carried out to examine whether the consistent poverty' approach employed in Ireland could be applied to Britain (Calandrino, 2003). Four factors were identified, consisting of food items; clothing items; consumer durables; and leisure/social activities. It was not possible, however, to distinguish any one dimension as basic deprivation'.

2. Methodology

As outlined in Section 1.1, the overall aim of this study is to identify a set of questions that measure material deprivation. In order to facilitate this, factor analysis was carried out using data from the PSE survey, FACS and BHPS to try to identify a small number of underlying dimensions of deprivation. If, as with the Ireland data, we could extract a single main dimension of deprivation we would be a large part of the way towards question selection. As well as questions relating to goods and activities, the analysis included questions about subjective financial well-being and indebtedness where these were asked.

Separate factor analyses were run using the following subsets of the PSE survey data[8]:

• Adult items and activities (54 questions).

• Adult items and activities considered as necessities by 50 per cent of the population (35 questions).

• Child items and activities (30 questions asked of all parents)[9].

• Questions relating to debt, borrowing to make ends meet, and items and activities that people had gone without because they could not afford them.

Like the PSE survey, FACS asks about a range of items and activities, although it includes a smaller number of child-focused questions. People are also asked whether they are able to keep up with their household and credit commitments, and how they feel they are managing financially. All these elements were included in the factor analysis carried out on the FACS data.

The BHPS asks people about their ability to afford a range of items (mostly consumer durables) and about one or two social activities. It also includes questions about money owed, typically on consumer credit commitments, and people's financial situations. It does not contain any strongly child-focused questions.

To assist anyone wishing to replicate this work more technical details about the type of factor analysis carried out are given in the annex to this chapter.

3. Findings

Although the results of the factor analysis are driven by the data used, there appears to be some common dimensions that can be identified from the different analyses. These relate in particular to family and social life, having disposable income, food and clothes, durable goods and financial difficulties.

Table 6.2 shows the factors extracted from analysis of the PSE survey, when all 54 adult items and activities used to measure material deprivation were included in the analysis. The first factor comprises two distinct elements: questions relating to family and social life (an evening out, going to the pub, having a meal out, visiting family or friends), and having disposable income to spend (going on holiday, having money to spend on oneself or to put aside in savings, being able to replace worn out furniture). When we include in the analysis only those items and activities that were considered necessities by 50 per cent or more of the population, disposable income remains a prominent factor (Table 6.3).

Analysis of FACS data produces similar results (Table 6.6), so that the first factor comprises aspects of family and social life (e.g. celebrations with presents at special occasions). Having spare money to spend also features in this dimension, and it includes several child-specific questions (e.g. a best outfit for children). Added to this, the PSE data on child items and activities indicates that parents consider social activities for their children to be an important part of non-deprivation (Table 6.5).

As requirements for everyday life, it is not surprising that food and clothing factors were extracted from both the PSE and the FACS data, in relation to adults and children alike. In some cases, these items appear separately (Table 6.5 and Table 6.6) while in other they appear together (Table 6.2) or combined with other factors (Table 6.3). It is notable that having an outfit to wear for social or family occasions loads particularly highly on the

clothing factors extracted from the PSE data (Tables 6.2 and 6.3) indicating again that being able to take part in social activities and family events is an important aspect of perceived non- deprivation.

Ownership of consumer durables emerges as a further factor that could be included in a measure of deprivation (Tables 6.2, 6.6 and 6.7). Not surprisingly, given the questions asked in the survey, durables feature particularly strongly in analysis of the BHPS (Table 6.7). Generally speaking, the factors relating to consumer durables include domestic consumer durables (washing machines etc.) as well as home entertainment equipment such as satellite TV and CD players.

As well as questions about particular items and activities, the PSE survey also asked a series of questions about how people had been managing financially in the past year, which included:

• whether they had been seriously behind in paying household and credit commitments;

• whether they or their children had had to go without items because they could not afford to buy them; and

• whether they had had to borrow from a pawnbroker, moneylender or from friends or family to pay for day-to-day needs.

In addition, respondents were asked to say whether they now considered themselves to be poor.

Factor analysis carried out on this data identified four underlying factors (Table 6.4). The first factor mostly concerns being behind with priority commitments such as utility bills, mortgage and Council Tax, while the fourth mainly relates to non-priority debts. The remaining two factors comprise questions about items and activities that adults and children have had to go without because of lack of money.

The FACS survey also asks people whether they are able to keep up with their household and credit commitments, and how they feel they are managing financially. As Table 6.6 indicates, three factors relating to these questions are distinguished by the analysis. Two of these (factors four and seven) relate to debt problems, while the third mostly comprises people's views about how they are coping financially.

The BHPS survey asks similar types of questions about whether people owe money on commitments including consumer credit, student loans and Social Fund loans, and about their financial situation - past, present and future. Two factors can be identified from the analysis - the first (factor 3) relates mostly to money owed on consumer credit commitments, while the second (factor 7) relating to people's financial situation (Table 6.7).

4. Implications

The findings from the factor analysis have a number of implications. Firstly, the analysis indicates that interpretable factors can indeed be derived from the datasets. None of these factors, however, equates to the type of basic deprivation measure that was found in the study carried out in Ireland (Callan et al, 1993). Second, the results suggest that there is a case for including questions about how people are managing financially and their subjective views about this. Finally, where questions are asked about debt and the ability to keep up with financial commitments, this tends to reveal an underlying dimension separate from other forms of deprivation.

Annex: Results from the factor analyses

Table 6.2  PSE survey, all items and activities  

Factor loadings

Factor 1: Family and social life, disposable income

An evening out once a fortnight  .939 Going to the pub once a fortnight  .898 A meal in a restaurant or pub once a month  .883 Holidays abroad once a year  .716 Holiday away from home for one week a year, not with relatives  .586 Coach or train fares to visit family or friends  .535 Small amount of money to spend each week on yourself  .433 Regular savings for rainy days or retirement  .412 Replace any worn out furniture  .338

Factor 2: Mixed

Having a daily newspaper  .873 Fresh fruit and vegetables every day  .796 Carpets in living rooms and bedrooms  .750 Visiting friends or family in hospital or other institution  .663 A warm, waterproof coat  .639 Friends or family round for a meal, snack or drink  .547 A hobby or leisure activity  .542 A dishwasher  .422 Insurance of contents of dwelling  .375 A car  .368 Replace or repair broken electrical goods  .325

Factor 3: Food and clothes

Meat, fish or vegetarian equivalent every other day  .596 An outfit to wear for social or family occasions  .561 Appropriate clothes for a job interview  .541 A roast joint or vegetarian equivalent once a week  .531 Two pairs of all weather shoes  .498 Two meals daily  .469 New, not second hand clothes  .440

Factor 4: Computer-related

Access to the internet  .836 A home computer  .827

Factor 5: Consumer durables

Mobile phone  .725 CD player  .635 Tumble dryer  .570 Satellite TV  .457 Kaiser-Meyer-Olkin Measure of Sampling Adequacy .934[10]

Table 6.3  PSE survey, items and activities considered as necessities

Kaiser-Meyer-Olkin Measure of Sampling Adequacy .878

Some cross-loadings are present (eg for meat, fish or vegetarian equivalent every other day)

Table 6.4  PSE survey, additional questions relating to deprivation

Kaiser-Meyer-Olkin Measure of Sampling Adequacy .843

Table 6.5  PSE survey, children's items and activities

Factor loadings

Factor 1: Social activities

Play group at least once a week for pre-school aged children  .850 Friends round for tea or a snack once a fortnight  .831 Going on a school trip at least once a term for school aged children  .630 Swimming at least once a month  .579

Factor 2: Toys and clothes

Construction toys such as Lego or Duplo  .815 Educational games  .735 Some new, not second hand or handed-on clothes  .668

Factor 3: Clothes

At least four pairs of trousers, leggings, jeans or jogging bottoms  .783 At least seven pairs of underpants or knickers in good condition  .716 All the school uniform required by the school  .675 At least four jumpers, cardigans or sweatshirts  .504 Leisure equipment such as sports equipment or a bicycle **  .320 Toys such as dolls and teddies  .314

Factor 4: Leisure

Leisure equipment such as sports equipment or a bicycle **  .819 A new or second-hand bike  .802 A warm, waterproof coat  .556 At least 50p a week to spend on sweets **  .302

Factor 5: Family and social life

Celebrations on special occasions such as birthdays  .908 A hobby or leisure activity  .877 A holiday away from home at least one week a year with family  .638

Factor 6: Food

Fresh fruit or vegetables at least once a day  .863 At least 50p a week to spend on sweets **  .642 Meat, fish or vegetarian equivalent at least once a day  .606

Factor 7: Computer-related

Computer suitable for school work  .867 Computer games  .844 Kaiser-Meyer-Olkin Measure of Sampling Adequacy .690[11]

** Cross-loading

Table 6.6  FACS survey

Factor loadings

Factor 1: Family and social life, disposable income

Celebration with presents at special occasions  .773 Toys and sports gear for children  .759 Good quality new brand name clothes or shoes for children  .707 Money for trips, holidays, outings or going with gifts to parties  .608 Have friends or relatives for a meal once a month  .590 Night out once a month  .586 Best outfit for children  .556 New not second-hand clothes  .555 Good quality brand name food most days  .468 One week holiday, not with relatives  .424

Factor 2: Consumer durables

Tumble drier  .687 Dishwasher  .684 Separate deep freeze  .621 Cable, satellite or digital TV  .557 Home computer  .483 Car or van  .480

Factor 3: Food

Fresh vegetables most days  .763 Meat or fish every other day  .707 Fresh fruit most days  .643 Cooked main meal every day  .640 Roast or similar at least once a week  .539 Cakes and biscuits most days  .536

Factor 4: Debt

Debt: TV/video rental or HP  .790 Debt: Loan  .550 Debt: Council Tax  .515 Debt: Other HP payments  .504 Debt: Phone bill  .469 Debt: water rates  .467 Have problems with debt almost all the time  .360

Factor 5: Money problems

Never have money left over  .858 Always run out of money  .819 Has overdraft and uses it  .574 Worried about money almost all the time  .394 Not managing financially  .361 Not a regular saver  .350 Financial situation has got worse in past 12 months  .316

Factor 6: Clothes

Weatherproof coat for each child  .740 Two pairs all-weather shoes for each child  .679 Two pairs all-weather shoes for each adult  .604

Kaiser-Meyer-Olkin Measure of Sampling Adequacy .934

Table 6.7  BHPS survey

Kaiser-Meyer-Olkin Measure of Sampling Adequacy .788 ** Cross-loading

Annex: Technical details of the factor analyses conducted

The method of extraction was principal components analysis. An oblique rather than orthogonal rotation was used (promax with kappa=4), as the dimensions of deprivation are very likely to be correlated. Different methods of extraction and rotation were tried, including orthogonal rotations (such as varimax). These did not yield substantially different results.

Selection of the numbers of factors presented in Tables 4.2 to 4.6 was based upon:

• consideration of the eigenvalues (necessary condition, >1, the so-called Kaiser criterion');

• inspection of the scree plots (taking factors before any elbow');

• overall interpretability (meaningfulness of the variables in any factor).

The results reported are taken from SPSS output. Some of the key analyses were also re-run with stata and R to confirm their overall veracity. The results were very similar – however, to quote from the R documentation, "There are so many variations on factor analysis that it is hard to compare output from different programs" (Ripley, factanal' in package mva'). .

7  Longitudinal analysis

1. Introduction

Most work on measuring living standards has been based on snapshot pictures from cross- sectional data. When looking over time, it is apparent that ownership of some goods grows rapidly (eg mobile phones), while most people already own other goods, such as VCRs, and so growth is much slower.

Rogers' (1983) model of diffusion suggests that products follow an s-shaped pattern during their life-span, as shown in Figure 7.1. Over time coverage starts low, has a period of rapid rise, before tailing off in maturity – perhaps with coverage even falling at this point. The adoption of, say, video recorders and their maturity phase with the increasing take-up of DVD players may be an example of such a pattern. Products are said to follow this life cycle pattern for a number of reasons, including technological changes leading to reduced prices, and personal interactions between people[12].

Figure 7.1  Model of product diffusion

Start Innovators Early adopters Laggards

Time

When choosing a set of deprivation questions durable goods and some other products will be at different stages of their life-cycle. If the measure is not to be overtaken by events, some consideration needs to be given to this. Of course, in a period of rising living standards we would anyway expect people to be buying more products. We may also consider if there are questions which are not linked to goods and hence not quite so sensitive to being at different stages of their product life cycle.

2. Aggregate trends over time

In this section we analyse trends in deprivation questions over time. Has deprivation and poverty been declining in recent years, as the income data tends to show?

The picture for lone parents is illustrated in Table 7.1, since for this group we have consistent coverage in FACS for all four years. Over this period the financial resources available to lone parent families increased substantially, in particular with the replacement of Family Credit by Working Families' Tax Credit and increases in Income Support scale rates. Child Benefit has also increased in real terms.

All measures have shown improvement over this time, whether based on enforced lack of goods/services, debt (arrears) situation, or subjective view about living standards. It is promising that each type of measure shows the same basic picture.

Table 7.1  Trends over time in deprivation questions - FACS

Cell percentages

Change 1999-

Lone parents  2002

Proportions that can't afford items over

time  1999  2000  2001  2002

expmain  Cooked main meal every day expmeat  Meat or fish every other day

Roast meat joint or similar at exproas  least once a week

expveg  Fresh vegetables most days expfrui  Fresh fruit most days

expcake  Cakes and biscuits on most days

Good quality brand name food expbfoo  for family meals most days

excoata  Weatherproof coat for each adult excoatc  Weatherproof coat for each child

2 pairs all-weather shoes for each exshoea  adult

2 pairs all-weather shoes for each exshoec  child

New not second-hand clothes for expnewc  family when needed

expbest  Best outfit for children

Good quality new brand name expbclo  clothes/shoes for children

Celebration with presents at expcele  special occasions

exptoys  Toys and sports gear for children

Money for trips, holidays,

outings or going with gifts to exptrip  parties

One week holiday, not with exphol  relatives

expnigh  Night out once a month

Have friends/rels for meal once a expfrie  month

exptv  Colour TV

expcabl  Cable, satellite or digital TV expfrid  Fridge (inc. fridge freezer) expfree  Separate deep freeze expwash  Washing machine

exptumb  Tumble drier

expphon  Phone, inc. mobile

expdish  Dishwasher

expvide  Video recorder

expcent  Central heating

expmicr  Microwave

expcar  Car or van

7.5  5.7

17.9  15.7

20.4  18.9

7. 13.1
8. 14.4
   8. 16.5

40.0  36.5

23.4  20.0

9.0  6.9

32.8  27.8

24.6  18.9

1. 35.1
   1. 19.0
      2. 39.0

0. 22.7

24.4  21.1

58.6  52.4

74.0  68.5

45.6  44.0

33.9  29.0

6. 1.0

8. 34.4

1.6  1.1

17.9  16.1

4.2  3.6

30.7  27.1

0. 6.7

4. 34.8
   1. 8.8

8.5  7.5

11.8  8.8

33.7  30.4

5. 3.0  -4.5

8. 10.0  -7.9

14.1  11.4  -9.0

10.8  9.0  -7.7

10.7  8.6  -8.3

1. 9.5  -10.3
   2. 25.5  -14.5
      2. 14.1  -9.3
         3. 5.0  -4.0

22.9  21.2  -11.6

3. 13.3  -11.3
   4. 24.9  -16.2

15.2  12.8  -7.3

32.1  30.7  -14.5

17.1  14.0  -13.0

15.1  11.9  -12.5

0. 40.6  -18.0
   1. 58.1  -15.9

34.8  29.0  -16.6

22.5  19.8  -14.1

0.3  0.5  -1.1

28.9  27.9  -10.8

0.6  0.6  -1.0

9. 9.2  -8.7
   0. 2.0  -2.2

3. 21.8  -8.9
   1. 4.0  -5.0

30.6  29.8  -6.5

6.9  6.1  -5.0

6.3  5.3  -3.2

6.5  4.4  -7.4

25.9  23.8  -9.9

Change 1999-

Lone parents  2002

Proportions that can't afford items over

time  1999  2000  2001  2002

expmusi  Music system (tape or CD)  11.5  9.0  6.5  4.7  -6.8 expcomp  Home computer  50.3  46.3  37.7  31.9  -18.4

Average change deprivation

indicators 1999-2002  -9.5

debtt1  Debt: electricity

debtt2  Debt: gas

debtt3  Debt: other fuel bills

debtt4  Debt: council tax

debtt5  Debt: insurance policies debtt6  Debt: phone bill

debtt7  Debt: TV or video rental or HP debtt8  Debt: other HP payments debtt9  Debt: water rates

debtt10  Debt: cards or MOC

debtt12  Debt: loan

average change, debts

debtprob  Probs with debts almost all time nomoney  Never have money left over runout  Always run out of money

Worried about money almost all worry  time

manage  Not managing financially Situation worse in past 12

finsit  months

Average change, subjective statements

7.9  7.3

14.9  12.1

0.4  0.5

6. 14.9
   0. 0.8
7. 14.5

3.2  3.0

5.5  5.2

17.4  14.9

4. 4.8

2. 8.6

6. 13.3

47.9  40.4

1. 24.1

44.7  38.1

34.7  30.1

29.0  27.1

6.0  6.3  -1.6

8.8  8.1  -6.8

0.3  0.3  -0.1

12.3  10.9  -4.7

0.7  0.7  -0.3

1. 11.4  -4.3
   2. 2.2  -1.0

5.0  5.0  -0.5

7. 12.6  -4.8
   6. 3.4  -3.0
      8. 5.0  -3.1 -2.7

10.2  12.2  -2.3

34.2  17.4  -30.5

0. 19.3  -7.8
   6. 29.8  -14.9
      1. 18.3  -16.4
      2. 22.6  -6.4 -13.1

For the population as a whole, we may consider trends from the BHPS (Table 7.2). The pattern of improvement is not as clear. Obviously whilst average incomes have been rising, it is likely that policy changes have been most effective at increasing lower incomes among families with children.

Table 7.2  Trends over time in deprivation questions – BHPS

Change 2001-

1996 1997 1998 1999 2000  2001 1996

Keep home adequately

wHSCANA warm  3.2 2.5 1.6 1.7 1.6  1.7 -1.5

Would like to keep

wHSCNTA home warm  1.9 1.9 1.1 1.1 0.9  1 -0.9 wHSCANB Pay for annual holiday  31.7 27.8 25.1 24.6 23.7  27.8 -3.9

Would like to pay for

wHSCNTB  annual holiday  19 16.4 14.5 13.6 13.4  16.7 -2.3 wHSCANC Replace furniture  24.4 20.3 17.5 17.9 17.5  14.8 -9.6

Would like to replace

wHSCNTC  furniture  12.6 11.3 9.9 9.9 9.7  8.4 -4.2 wHSCAND Buy new clothes  9.6 8.1 7.6 6.8 7  6 -3.6

Would like to buy new

wHSCNTD clothes  4.9 4.5 4.3 3.5 3.3  3.4 -1.5

Eat meat on alternate

wHSCANE days  11.8 9.9 8.8 8.9 7.7  7.9 -3.9

Would like meat on

wHSCNTE  alternate days  2.7 2.4 1.6 1.6 1.6  1.9 -0.8

Feed visitors once a

wHSCANF  month  32.5 27.4 25.9 23.9 24.5  25.9 -6.6

Can't afford visitors

wHSCNTF  once a month  6.3 5 4.5 4.5 4.8  5.3 -1

Financial situation (quite

wFISIT  or very bad)  8.7 8 6.9 6.9 6.9  7.4 -1.3

3. Individual-level analysis

As incomes have grown, the different measures of living standards in FACS have reflected this improvement at the aggregate level. In this section we consider how well some particular measures respond to improved fortune at the level of the individual family. Four such measures are show in Table 7.3.

At the individual level, we examine those moving into and out of receipt of Income Support during 2001-02. This creates four groups. In the first row we then show the proportion of those with a poor' score who continued to have a poor' score after than IS transition. The second row shows the proportion of those with a good' score who moved to have a poor' score, after the IS transition.

So, among those leaving IS in 2001 who were not managing, 21 per cent were also "not managing" in 2002. This is the lowest figure among the transition groups. It indicates the kind of effect of leaving Income Support that we would expect. Conversely, those moving to IS were the most likely to have persistent problems (of not managing) and least likely to leave such problems behind – again in line with expectations.

The next three panels of Table 7.3 examine other measures of living standards. They behave much as we would expect.

Table 7.3  IS dynamics and living standards measures

Cell percentages

4. A case for chain-linking'?

As part of the theory of index numbers, a series of short-run indicators is sometimes converted to a single longer-run series. This is known as chain linking'. When different indices are changed, it is vital there is a link year', where both bases are known. This provides a link factor' to use in producing new indices by multiplying rates of change. A simple example is shown in Table 7.4. Effectively each index is recalculated with new weights each year.

We would recommend statistical advice on whether these concepts would be applicable to a weighted index of living standards. Index numbers are based around quantities and prices. A welfare index would consistent of quantities (proportions lacking through inability to afford) and weights', where such weights could be based on public perception of necessities, proportion owning, etc. This is somewhat different. However, such a methodology if acceptable would permit re-basing the index on a regular basis, even as often as annually though less frequently would probably be appropriate.

Table 7.4  Chain-linking' different index series

8 Validating low incomes

1. Introduction

In most cross-sectional studies there is a group of low income families that do not appear to be poor'. That is, despite their low incomes they have living standards rather higher than most low income groups – avoiding debt, able to afford more goods and services, stating satisfaction with their standards of living. Often, indeed, the lowest income 10 per cent of the population appears to be better off, in some senses, than the second decile.

Various explanations have been put forward for this. Often it is self-employed people in this situation, and some evidence suggests they have spending patterns more typical of the average than the poorest. This is one reason that the self-employed are often dropped from the base of income statistics – their incomes are not reflective of their true' living standard.

There is a range of other explanations. These relate to mis-measurement of income, use of savings/credit, money management, and longer-term status.

It is possible that some incomes have not been properly measured. Perhaps there are sources of income not accounted for, or sources of income not even asked about (perhaps money from family). As partial evidence of this, there was a higher rate of imputation of missing income data for the non-deprived group (16 per cent, compared with three per cent for the deprived group). This could indicate that there is disproportionate measurement error for this group.

Another possible explanation is that some low-income families are able to live off accumulated savings, or maintain a better standard of living through using credit. The credit explanation is less convincing, because this group generally has lower credit use. However, people in the non-deprived group often do have higher savings than those in the deprived group, who tend to have little if any money saved.

Those living on a low income with apparently adequate living standards may have very good budgeting skills, superior to the deprived group. This may explain, for instance, the apparent success of low-income pensioners in maintaining their living standard. It is less likely to be a useful explanation among families with children. It is worth noting that the low income (measured AHC) but not deprived group have rather different family characteristics. A much higher proportion of the non-deprived group were owner-occupiers, compared to the deprived subset of those on a low income (63 per cent compared to 27 per cent). And twice as many were couples rather than lone parents (64 per cent compared with 31 per cent). Both points may be indicative of higher longer-term income among the non-deprived group, which we explore below in section 8.2.2.

A final explanation, perhaps linked to the availability of savings, is that the non-deprived group on low incomes are in transition. They are accustomed to higher incomes in the past enabling them to build up savings, and stocks of durable goods. In due course they will move out of low incomes into higher incomes.

2. FACS panel analysis

We can shed some light on the previous discussion using the FACS panel families. This section is based on those in the bottom fifth of incomes (AHC) in 2001. We use the penultimate wave to investigate how circumstances had changed by late 2002, the most recently available information.

We define as deprived' those lacking three or more items from the full list of deprivation indicators, through being unable to afford them. This represents about one-third of families with children in 2001, and closer to two thirds among the lowest income quintile[1].

1. MONEY MANAGEMENT AND SAVINGS

Those families in the bottom quintile, but not deprived' on our broad definition, were more likely than deprived families to have a bank account, and to be regular savers. Among those with savings they had appreciably more saved.

More than half (56 per cent) of the non-deprived group said they were managing quite or very well, contrasted with 15 per cent of the deprived group. They were also much more likely to say they had money left over each week or month, though 41 per cent hardly ever or never did so (compared, however, with 74 per cent of the deprived group).

These figures suggest that various measures of material deprivation provide consistent results. Some families with incomes in the bottom quintile do not report high levels of hardship, and this may be identified by a range of different questions and pieces of information.

Table 8.1  Saving, managing and deprivation status: bottom FACS quintile

Column percentages

Base: bottom 20 per cent of equivalised incomes, AHC.

2. LOW INCOME – TRANSITORY OR LONG TERM?

It is possible that the non-deprived group among those on low incomes are observed as poor only for a short time. Instead, they generally have higher incomes and they are observed in a short dip. Whether that dip is real or measurement error is harder to say. To investigate this we take those on low incomes in 2001, and then look at the incomes of the deprived' and non-deprived' in the previous year, and following year[2].

Those in the bottom quintile, but not deprived, appeared to be just visiting low income temporarily. Their incomes were somewhat higher in the previous year (where known) and substantially higher in the following year. By contrast, those both deprived and on a low income had quite flat income trajectories over the same period. For whatever reason, the deprivation indicators appear to be distinguishing shorter-term from longer-term poverty. This may be a helpful feature, although in any case the persistence of poverty is best measured by a panel study than through cross-sectional surveys

Figure 8.1  Incomes histories of those in the bottom quintile in 2001

£150

£140

Not deprived in 2001 £130

Deprived in 2001 £120

£110

£100

£90

£80

2000 2001 2002 Year