The Scottish Government

« Previous | Contents |

Listen

Social Justice
a Scotland where everyone matters
Indicators of Progress 2003

ANNEX A
Measures of Low Income amongst scottish households

Introduction

Poverty is a complex issue, and as such there is no single way to measure it. The Social Justice strategy contains 29 milestones, which cover a range of issues related to poverty and social exclusion in Scotland. These include worklessness, low income, educational attainment, health and care, and housing. The Executive's view is that poverty encompasses not only poverty of income, but poverty of opportunity.

However, commentators regularly focus their attention on low income and, therefore, this article aims to explain the issues around the measurement of low income, and provide advice on their interpretation. The measurement of low income is not straightforward and interpretation of the results should be carried out with care.

Measurement of Low Income

Income data come from the Family Resources Survey, an annual survey run by the Department for Work and Pensions (DWP). The measure of income used in DWP's Households Below Average Income (HBAI) analysis, and this report, is weekly net (disposable) household income, and includes the income of all members of the household, including dependants. Two measures of household income are used, before housing costs (BHC) and after housing costs (AHC) ¹.

Household income is used to represent material living standards, or, more precisely, the level of consumption of goods and services that people could afford with the disposable income of the household in which they live. Household income is used rather than individual income because the living standards of an individual depend not only on their own income but also on the incomes of other members of the household. For example, a non-working person living with a high earning partner may have a very low individual income but a high standard of living.

To allow a fair comparison between households, incomes are adjusted to take account of household size and composition. The adjustments are made using a technique known as equivalisation which reflects the common-sense notion that, in order to enjoy a comparable standard of living, a family of five will need a higher income than a person living alone.

Low income is defined in terms of thresholds of mean ² or median ³ income. Any household with an income below a given threshold is classed as a low income household. Five thresholds are included in this report - 50% and 60% of mean income, and 50%, 60% and 70% of median income.

The most appropriate choice of threshold was debated by the Statistical Program Committee (SPC) of the European Union in 1998. They concluded that for international comparisons of low income the 60% median threshold should be used as a headline indicator, but recommended that analysis be carried out on a wide range of thresholds to provide a fuller picture. The Scottish Executive, along with the UK Government have adopted this recommendation; there are a total of 20 measures of low income for each of the three relevant milestones based on the criteria discussed above, with the 60% median (AHC) threshold highlighted as the headline measure.

Low income thresholds are defined in both relative and absolute terms. The relative measure compares household income against the low income threshold for that year, whilst the absolute measure compares incomes against the threshold in a fixed year (in this case 1996/97 is used as this represents the position when the present UK Government came into power) after adjusting for inflation.

Issues around Relative and Absolute Measures

The relative measure compares household income against the threshold for that year, and is therefore an indicator of how the living standards of poor families compare against society as a whole at that time. It is a measure of inequality and tracks whether the incomes of poorer families are keeping pace.

Poverty is generally felt to be a relative concept and based on the ability to participate in society. The current affluence of the society in which you live is therefore relevant, as expectations rise over time. For example, being unable to afford a television would not have classed you as poor 50 years ago, but would do so today.

There are problems in using relative measures in isolation as they can be misleading if they are not interpreted in the context of other economic indicators. In times of economic growth, relative measures can mask real improvements in living standards. This has been the case in recent years in Ireland, where rapid economic growth led to considerable increases in incomes but rising levels of relative income poverty despite real improvements in living standards of the poorest in society.

In the same way, when economies are performing poorly, indicators of relative low income can show a reduction, falsely indicating an improvement in the situation of poor families. This was the case in several post-Soviet economies where overall living standards had collapsed but relative measures showed no increase in 'poverty' despite large numbers living at levels well below what would generally be considered as poverty.

Absolute measures of low income go some way to solving this problem. By comparing incomes against a fixed threshold (adjusted to remove the effects of inflation), absolute measures provide a clear indicator of the changes in living standards of the poorest. The effects of economic growth are removed from the equation to allow a clear picture of how the incomes of the poorest families have changed over time.

However, the downside is that absolute measures have a fixed shelf-life. As society's expectations change over time, absolute thresholds become less relevant. They may indicate that poor families are benefiting from increased living standards, but do not measure whether they are keeping up with, or indeed closing the gap with more affluent families.

The following charts illustrate these issues using recent figures for Scotland. The first chart shows that despite small year-on-year fluctuations, there has been little change in the proportion of people experiencing relative low income since 1996/97. However, there is a notable decrease in the proportion of people experiencing absolute low income over the same period. The second chart clearly shows the reason for this pattern. The absolute low income threshold has remained fairly stable over time, with the slight increase representing inflation. However, the value of the relative threshold has increased substantially - by 19% (or 26 per week) after removing the effects of inflation.

This suggests that the poorest families have experienced real increases in their living standards since 1996/97 but strong income growth across society has pulled up the relative threshold, creating a moving target and leading to no clear improvement in relative low income.

These figures provide support for the EU SPC's conclusion that a range of indicators should be examined. Focusing solely on the relative or the absolute measure only shows part of the picture and can be misleading.

Other issues

Housing Costs

In addition to considering both relative and absolute measures, the issue of housing costs is also important in the analysis of low income. Housing costs are important as they are a necessary expenditure which households cannot choose to do without. Bearing in mind that income is used to represent standard of living, using income before housing costs (BHC) can overstate the living standards of those whose housing costs are high relative to the quality of their accommodation.

For example, housing costs in Edinburgh are generally higher than in some other parts of Scotland. Using a BHC measure of low income will artificially inflate the living standards of people in Edinburgh as they need to spend a larger proportion of their income to attain a similar standard of accommodation as those living in areas where housing is cheaper.

BHC measures can also be problematic when measuring low income amongst pensioners. Many owner-occupiers will have paid off their mortgages at this age and will therefore have very small housing costs; by contrast, the poorest pensioners often rent their homes and receive housing benefit, which is included in BHC income. The BHC measure may therefore overstate the living standards of the poorest pensioners.

Using a measure of income after housing costs (AHC) helps to address these problems to some extent. However, the AHC measure doesn't allow for the fact that some people choose to pay more for a higher standard of accommodation and as a result, benefit from a better quality of life. In these circumstances, the AHC measure will understate their standard of living.

Some situations support the use of a BHC measure whereas others suggest the use of an AHC measure. There is no clear rule as to which is more universally appropriate, and therefore the low income analysis presented in this report is carried out on two bases, before and after housing costs.

Equivalence Scales

Equivalisation is the process whereby household income is adjusted to take account of the size and composition of the household. Each household is assigned an equivalisation factor based on the number of people in the household and their relationships ⁴. Equivalised income is then calculated by dividing cash income by the equivalisation factor, which has the effect of increasing the incomes of single person households and reducing the incomes of larger households.

A number of equivalence scales exist, with differing factors applied to each individual in the household. Generally, a higher weight is applied to older children than younger children as they are considered to incur greater expenditure in the household, but the degree of this additional weight differs across scales. Some scales introduce an additional weight for disabled household members to reflect the increased financial needs associated with disability.

The construction of an equivalence scale is, to some extent, subjective and it is generally accepted that no single scale is perfect. The HBAI analysis is carried out using the McClements scale, but results are sensitivity tested by replicating key estimates using a range of alternative equivalisation scales. Any estimates which are particularly sensitive to the choice of equivalisation scale are highlighted in the publication.

The low income milestones in this report are also calculated using the McClements scale to maintain comparability with GB results. The McClements scale does not include an element for disability and therefore the results potentially overstate the living standards of households which contain disabled people. The low income milestones in this report have not been broken down by disability for this reason.

Choice of Threshold

Figures are shown on a range of different thresholds of mean and median income to provide a fuller picture of changes in the extent of low income over time. Due to the benefits and tax credits systems large groups of households will have very similar incomes. It clearly has a major effect on the low income figures if a 'cluster' of households is very close to one of the low income thresholds.

For example, if there is a cluster of households with incomes around the 60% median threshold, a slight change in the value of the threshold would have a major effect on the proportion of people below 60% median. If considered in isolation, one might draw the conclusion that this represented a great success or failure in reducing the level of income poverty. In reality, these households will not have noticed any significant change in their circumstances.

The range of income thresholds provides additional information on which to establish trends. If trends across the range of thresholds are moving in the same direction, this gives more weight to any overall conclusions of improvement or decline. If the trends differ across the indicators, it may suggest that policies are impacting on specific groups of people rather than having a more general impact.

Persistent Low Income

The current measures of low income are based on the situation at one point in time. They cannot distinguish between families who are experiencing short-term low income due to a temporary loss of employment but who may have high levels of savings, and those who have experienced low income persistently for a number of years.

Persistent low income is generally considered to be a far more serious problem than short-term low income, but panel study data (which tracks households from year to year) is required to measure this. Annual analysis of the British Household Panel Survey (BHPS) is carried out to assess the extent of persistent low income in the UK and is published in DWP's Opportunity for All Report. Households are defined as experiencing persistent low income if they were below the low income threshold in three out of the last four years.

Until recently, this analysis has not been possible for Scotland because the data source did not have a large enough Scottish sample. However, since 1999 the BHPS has been boosted for Scotland and Wales and data are now available for 3 years (1999, 2000 and 2001) based on the larger sample. Whilst four years of boosted data are required before the UK calculation can be mirrored, it does allow some analysis to be carried out on movements in and out of low income for Scotland, and this year's report provides some of the initial results.

This analysis of persistent low income complements the single year measures by providing further insight into the extent and severity of low income in Scotland. As further data become available, this measure can develop and will become an important tool in developing policies to combat poverty.

Sampling Variation

With any analysis carried out on survey data, there is always an issue of sampling variation ⁵. Any estimates produced from a sample survey come with advice about the degree of sampling variation surrounding them. The extent of sampling variation associated with an estimate depends on many factors, but a general rule is that the smaller the sample used to generate an estimate, the larger the sampling variation associated with that estimate.

This is particularly important for Scotland. As the estimates are based on a sample size of roughly ten per cent of the GB sample, the Scottish estimates have a greater degree of uncertainty than the GB estimates. Small year-on-year changes in indicators should not be judged to be significant as there will be some degree of sampling variation. Several years of data are required to establish the existence of a trend in any indicator.

The Scottish Executive is funding a doubling of the Scottish sample of the Family Resources Survey from 2002-03 to improve the reliability of the Scottish low income estimates in future. This will result in more robust estimates with smaller sampling errors, but the fact that the figures are derived from a sample survey means that a degree of sample variation will remain in any estimates.

Improvements in Data Quality

Due to an improvement in the DWP methodology for deriving working-age adult and pensioner results, some of the figures published in this year's report for these groups differ slightly from those published previously. These differences are, however, very small and do not alter any of the general trends. Details of the how the methodology has been improved are described in the text accompanying the data on Milestones 14 and 19.

As part of an ongoing programme to improve data quality, there will be a further revision to the data series when the full results of the 2001 Census become available. Any changes resulting from this are expected to be small and will be as a consequence of improved measurement rather than a change in the trend.

Summary

It is essential when developing an understanding of low income that a range of indicators is considered. It can be misleading if focus is restricted to one single indicator without considering it alongside alternative measures of low income and external factors such as the prevailing economic circumstances.

The entire range of thresholds of mean and median income should be examined to identify changes in the household income distribution, whether the changes are focused on certain parts of the distribution, or whether the changes are being experienced by households in general.

The issue of housing costs should be addressed and the most appropriate measure for the group of interest should be considered, although in some cases both measures should be examined.

Indicators based on both relative and absolute definitions of low income must be analysed together to provide a full understanding of trends. The relative indicator provides a measure of inequality by comparing income against the current year, whereas the absolute indicator allows the measurement of progress against a fixed baseline and removes the effects of factors such as the performance of the economy.

The Department for Work and Pensions carried out a consultation exercise recently on the measurement of child poverty to identify the best approach for monitoring progress. On the basis of the responses DWP are considering the most appropriate way forward for measuring child poverty. No final decision has been reached, but whatever the outcome of the consultation, it is likely that low income statistics will continue to be monitored and the need for informed interpretation will remain.

The Social Justice strategy measures 29 different milestones covering all aspects of poverty and social exclusion. Within each of the three milestones concentrating on low income, a range of twenty indicators is covered. All of this information must be considered to provide an accurate interpretation of progress against targets.

References:
Households Below Average Income 1994/5 - 2001/02, 2003, DWP, Corporate Document Services
Indicators of Progress, A discussion of approaches to monitor the Government's strategy to tackle poverty and social exclusion, Report of the workshop held on 19 July 2000 organised by DSS and CASE, LSE
Measuring Child Poverty, A Consultation Document, 2002, DWP

Footnotes

1. A full list of the components of income can be found in the box at the end of this article.
2. The mean is a measure of average calculated by summing together the incomes of all people and dividing by the total number of people measured. The value of the mean can be distorted by a small number of people on very high incomes.
3. The median is a measure of average calculated by ranking all people in order according to their income and taking the income of the middle person. It is the preferred measure of average when considering income as it is not subject to the same distortions as the mean.
4. Couple households are generally taken as the reference point, with an equivalence value of one. Single person households receive a factor of less than one, although generally greater than 0.5, based on the assumption that sharing allows individuals to attain a higher standard of living than they would if living independently with the same cash income. Households with three or more persons receive a factor greater than one.
5. Any estimate derived from a sample survey is intended to represent the population as a whole. For example, the mean income calculated from the sample is used to represent the mean income of the population. Sampling variation is the extent to which the sample estimate is likely to differ from the true population mean.

« Previous | Contents |