January 14, 2013
UN Photo/Kibae Park
If a person suddenly becomes poor — for example, due to an unexpected death or illness in the family — they will have a rather different experience and understanding of poverty than someone who has been impoverished almost their entire life. Importantly, these individuals most likely face quite different prospects for the future.
Standard measures of poverty, however, are only “snapshots”. They are based on people’s current income status and do not account for their history of poverty. Should they?
This is the question examined in the June 2012 special issue of the Journal of Economic Inequality, the culmination of a four-year research project initiated by the guest editors, Luc Christiaensen and Tony Shorrocks, while at UNU-WIDER. It was launched at a joint UNU-WIDER-World Bank event in Washington DC on 14 June 2012 chaired by Tony Addison, Deputy Director and Chief Economist at UNU-WIDER, and Branko Milanovic, member of the Journal of Economic Inequality’s editorial board.
Judging by the over 200 participants in total, both in person and online, from countries across the world — for example, China, Ethiopia, Mexico, Pakistan, Rwanda — and the lively debate which ensued, the topic clearly struck a note.
The progress made towards reducing worldwide poverty is undoubtedly encouraging. World Bank estimates, for example, suggest that the percentage of the population in the developing world living below US$1.25 a day declined from 52 percent in 1981 to 22 percent in 2008. The number of people living below this line in 2008 was estimated at 1.29 billion, down from 1.94 billion in 1981.
But this snapshot approach to measuring poverty, common to most standard measures of poverty, has shortcomings.
Without taking into account a person’s history of “poverty spells” — the timing, length and severity of previous poverty episodes — it remains hard to know how poor that person really is. People with long poverty spells in the past are, for example, less likely to escape from poverty. And if they do, they are far more vulnerable to becoming poor again. The timing of the spell also matters, with childhood poverty being especially detrimental. It can hurt children’s cognitive development, their prospects of marriage and lifetime earnings.
To overcome these shortcomings in poverty measurement, people’s lifetime experience of poverty should be reflected in them. The measurements should aim to present a “movie” rather than a snapshot of poverty.
But how best to account for the timing of previous poverty spells, their severity and their duration, as well as any potential trade-offs between them? For example, is a person who was briefly poor early in life “poorer” than a person who has been trapped in poverty for a long time as an adult? Furthermore, how to compare an episode of short-but-intense poverty with an episode of poverty that has been mild but protracted?
It is also difficult to find the necessary comparable panel data which tracks the same people and their poverty status over an extended period of time.
The profession has only just begun to grapple with the challenges of incorporating and formalizing more sophisticated, intertemporal perspectives in measures of poverty, let alone their practical application. Some may even argue that data and measurement limitations urge us to focus on fine-tuning practical poverty imputation methods instead, as even tracking poverty using snapshot measures (which require less data) continues to be a challenge.
For the purposes of advancing the debate, several ways to account for poverty spells and their implications for poverty rankings are explored in the special issue mentioned above.
Bossert, Chakravarty and d’Ambrosio, for example, propose to count the number of consecutive periods that a person has been poor, and to use this number to weigh the different spells in poverty. A person who was poor during two consecutive periods will thus be poorer than a person whose two-period poverty experience was interspersed with a period out of poverty. By this measure, non-whites in the USA would be considered much poorer than whites from 1967 to 1992, because they stayed in poverty much longer.
In addition to the duration of the poverty episode, Hoy and Zheng also account for the timing of the poverty spell, with childhood poverty considered much more detrimental than poverty episodes later in life.
Documenting Spain’s experience during the 1990s, Arranz and Canto empirically confirm that episodes of poverty in the past reduce the likelihood of escaping poverty, while also increasing the likelihood of falling back. In addition, the longer the time spent in poverty, the larger the effects. This provides support for the notion that poverty measures should capture a person’s lifetime poverty experience.
Such studies, including those drawing on subjective assessments of poverty, which examine the effects of the timing, duration, and severity of poverty spells on the prospects of escaping or falling into poverty, will prove very useful in motivating whether the profession should shift towards more sophisticated lifetime measures of poverty. They will also help in determining which features of previous spells should be especially accounted for, and they can give empirical content to the adjudicating parameters the different measures propose.
A deeper understanding and documentation of poverty spells would not only allow a more accurate measurement of poverty, it is also relevant operationally. Countries can use studies on poverty spells, for example, in their analysis of the middle class (which some define as people who aren′t at risk of falling into poverty for a specific period of time). And if the length of poverty spells affects a person’s ability to escape from poverty, a person’s poverty history needs to be considered when deciding about the graduation from safety net benefits.
Nonetheless, as highlighted during the event, as better ”movies” are made, support to regular (cross-sectional) household surveys, which is like taking successive snapshot pictures of welfare in a country, must continue — especially in low-income countries and Africa, where more than half of the countries are still struggling to collect post-2008 poverty data. This also underscores the importance of developing and validating poverty imputation methods that make maximum use of existing data.
One paper in the special issue experimented with using components of consumption or even household assets (such as cars, TVs and refrigerators) and human capital (such as the level of education) to track poverty. Those data are more comparable and regularly updated than a full set of consumption data.
When compared with the actual numbers in Viet Nam and China, the methodology worked well, holding hope for the wider use of such poverty prediction methods. Application to Kenya and Russia proved also useful in adjudicating which price deflator to use when measuring poverty over time, a recurring challenge, with more often than not widely divergent results depending on the deflator chosen. Imputation methods can further help overcome missing and coarse data within surveys and some more recent studies even point to their potential to construct synthetic panels, needed for the lifetime poverty measures.
There has undoubtedly been substantial progress in reducing poverty. Nonetheless, many theoretical and practical issues remain unresolved in measuring it, not least when considering how poverty evolves over time.
This presents an important and fascinating agenda for future research and worldwide debate, including on imputation methods. Such debate can now also be facilitated by inexpensive video conferencing. Judging by the feedback received from virtual participants from all over the world during and after the event, the experience of being able to participate in a high-level discussion on a serious global matter without having to physically travel clearly has great appeal.
♦ ♦ ♦
This article was previously published in UNU-WIDER′s WIDER Angle newsletter.