Skip to content

New ways of measuring poverty Professor Mike Brewer describes ISER’s innovative approaches to analysing poverty data

19262656455 f66ae15826 o

Over the past 25 years, substantive and methodological research at ISER, as well as advances in survey design, have made major contributions to the study of how best to measure household living standards, of why the distribution of income in the UK is changing and how this compares to other European countries, and of income and poverty dynamics.

Such work continues, forming an important part of the new research programme for the ESRC-funded Research Centre on Micro-Social Change. Here we highlight three contributions: on how to measure persistent poverty when there is attrition or missing data, on cross-country comparisons of labour market volatility, and on whether we can rely on very low values of income reported to household surveys.

Measuring persistent poverty with incomplete data

Poverty has more adverse effects when experienced for a long period, rather than as a transient phase.

For this reason, policy makers in developed and developing countries try to monitor levels of persistent poverty, with the EU adopting a measure of persistent poverty as part of its programme of monitoring progress on social inclusion, and the Child Poverty Act committing UK governments to track the fraction of children in persistently poor households. Typically, poverty persistence is measured with data from household panels in which a large group of randomly-selected households are interviewed repeatedly to assess whether they fall below a poverty line in each year. The simplest estimate of poverty persistence is then the proportion of sampled households found to be in poverty for more than a certain number of years during some period of time. But missing data can make it very hard to measure persistent poverty reliably: if we are unable to observe a household in every year, it is not possible to say whether it has been in persistent poverty.

The usual way to deal with missing data is to use only the subset of households observed in every period and correct any non-representativeness by giving greater weight to types of household that appear to be under-represented in the reduced sample. Professor Stephen Pudney and ISER PhD student Yadira Diaz Cuervo develop an alternative approach that constructs a lower and upper bound for persistent poverty by considering all possible outcomes that the missing poverty observations could have taken. These bounds are very wide, demonstrating that missing data introduces a very large element of uncertainty into the estimation of persistent poverty measures. But they then show that introducing some very mild assumptions allows much more refined estimates. The researchers apply these methods to a panel survey carried out by the Peruvian National Institute of Statistics and show that the standard re-weighting method is badly biased, tending to result in an over-estimate of the rate of persistent poverty in the population.

Volatility in earnings in the UK and US

The chief value of longitudinal data is in tracking how outcomes change over time for individuals, and MiSoC Research Associate Professor Stephen Jenkins (togetherwith Professor Lorenzo Cappellari) has recently used data from the BHPSto assess whether the well-documented rise in earnings inequality in the UK is mirrored by a rise in earnings volatility. The researchers measure volatility in the following way: for each working-age individual, they calculate how much earnings change between one year and the next, and then the measure of volatility given by how spread out the distribution of individual earnings changes is. The researchers also show how one can measure labour market volatility, defined as the volatility that exists when one takes account of earnings changes for all individuals, including those who move in and out of work. The key findings are that earnings volatility inBritain remained constant between 1992 and 2008 for both men and women, but that there was a marked decline in labour market volatility. This latter finding is different from the US experience, and is caused mainly by a decline in theproportions of workers moving into and out of work or not having a job at all, reflecting the steady growth in the British economy after the early-1990s recession and before the impact of the 2007/8 financial crash was felt.

Should we believe low values of income reported to household surveys?

Researchers commonly measure poverty by comparing household income to a poverty line, because a household’s total income is seen as a good proxy for its standard of living. An alternative, long-favoured by economists on conceptual grounds, is to use a household’s consumption as a proxy for its standard of living. A measure of consumption starts from what a household spends, but then adds the benefits that a household derives from durable goods like housing or car ownership. Additionally, some have argued that households with low resources often under-report their income to household surveys but seem to report their level of spendingwith greater accuracy, and this provides a practical reason to use consumption rather than income to assess who is truly the poorest in society.

Our ongoing work, with the Institute for Fiscal Studies, has been looking closely at these issues for the UK. They show that households who report to have an income in the bottom 1% of the UK population also report levels of spending that are close to that of the median household, strongly suggesting that many of these households are not truly poor. This mismatch is found in data that spans many years, and can be seen for many different types of household. The researchers go on to show that the size of the mismatch between income and spending is considerably greater than would occur just through temporarily poor households smoothing their consumption by running down their savings or taking on debt. Thus they conclude that some of these households must be under-reporting their income.

In related work, ISER research has looked at how our impression of who is poor changes if we use consumption, rather than income, to assess living standards. A key finding is that the age profile of poverty changes markedly when moving from income to consumption, with the elderly appearing a lot less poor under a measure of consumption, reflecting the far greater rates of home ownership amongst the old compared with the young.

This article was first published in ISER’s annual review, Taking the Long View.

Come to our event on Measuring Incomes: Poverty, Dynamics, Persistence - as part of the ESRC Festival of Social Science.

References

Yadira Diaz Cuervo and Stephen Pudney. “Measuring poverty persistence with missing data with an application to Peruvian panel data”, ISER WP 2013-22

Lorenzo Cappellari and Stephen P. Jenkins (2014), “Earnings and labour market volatility in Britain, with a transatlantic comparison”, Labour Economics 30, pp201-211,

Mike Brewer, Ben Etheridge and Cormac O’Dea, “Why are households that report the lowest incomes so well-off?”, forthcoming in Economic Journal (Features)

Mike Brewer and Cormac O’Dea, “Measuring living standards with income and consumption: evidence from the UK”, ISER WP 2012-05,