The global goal to end extreme poverty lacks credibility without better data. The author argues that collective action is required to improve the reliability of poverty measurement.
Challenge
The last 25 years has seen a dramatic reduction in the number of people living in extreme poverty around the world. That success has motivated new commitments to eradicate this most egregious form of deprivation globally by 2030. Were this goal to be achieved it would represent a key milestone in human progress.
However, another kind of poverty—the poverty of data—acts as a stumbling block. The unreliability of global poverty estimates means that we risk being unable to assess with any confidence if and when extreme poverty has ended. That undermines both the credibility of the goal and the accountability of those who have committed to it, including the UN and World Bank. More importantly, it handicaps efforts to reduce poverty, which depend on being able to identify and target help to poor people and poor areas.
The Sustainable Development Goals agenda seems to recognize the urgency of this challenge with its mantra to “leave no-one behind” and its advocacy for a “data revolution”. But it remains unclear how these ideas can be operationalized.
Proposal
The workhorse of Poverty measurement is the household survey. These measure the level of income or consumption across a representative sample of a country’s population. The coverage and frequency of surveys have grown enormously over the past quarter-century but remain incomplete. More than a billion people live in countries with insufficient survey data[1] – exceeding the 700 million people who are believed to live in extreme poverty around the world. Coverage is especially poor in fragile states, where extreme poverty is increasingly concentrated. The gap in survey coverage needs to be closed, including in countries where surveys are hardest to conduct.
Even where surveys are administered, their variable quality presents fundamental problems for poverty measurement. Differences in survey design and implementation can generate drastically different poverty estimates for the same population.[2] Differences in survey methods occur frequently in practice, undermining the comparability of results between countries and within the same countries over time. Many countries use outdated questionnaires, designed predominantly to capture the welfare of poor people who rely on subsistence-based livelihoods. These are less effective for analyzing the welfare of some of today’s poor people who regularly eat outside the home and have access to a wider range of goods and services in the cash economy.[3] There is a need to review existing survey methods and to improve overall survey quality and comparability.
Poverty measurement doesn’t just rely on income and consumption surveys but a number of other data sources. These include population estimates from national censuses and measures of prices over time (inflation) and space (purchasing power estimates). Each data source introduces possible errors and reduces confidence in poverty estimates. The assembly of poverty numbers has consequently been compared to a “house of cards” with revisions to the underlying data creating “statistical earthquakes”.[4] Additional data sources are required to monitor non-income aspects of poverty such as access to services, asset ownership, hunger, health and education. The coverage and quality of these complementary data sources for poverty measurement also need to be improved.
There is some confusion and debate as to what stands in the way of advancing this agenda.
To date, a lot of attention has been focused on financing with calls for major new investments to unlock this constraint. While it is certainly the case that small and poor countries often struggle to muster sufficient funds for poverty measurement, it is a mistake to assume that national financing gaps translate into a global financing constraint. At a global level, the estimated administrative cost of achieving universal coverage of living standards surveys is a paltry $17 million a year.[5] The equivalent cost for expanding the coverage of surveys of non-income aspects of poverty[6] is an even smaller $6 million a year. If the global community is committed to resolving this issue, money shouldn’t stand in the way.
Another constraint that is regularly cited is politics, implying that governments conspire to undermine accurate data collection by meddling with results so they appear more favorable. While instances of government interference certainly do occur—and G-20 members are not immune from such scandals—it is hard to believe that these are the norm. It is more likely that governments harm poverty measurement through neglect than through interference. This is reflected in the lowly status of statistical agencies and the appointment of weak leaders to these institutions.
Poverty measurement is an inherently vexing exercise that entails various technical challenges. This is especially true for global poverty measurement, in which results are aggregated across countries where poverty is experientially very different. A Commission established by the World Bank has recently published a report advising the institution on how these technical difficulties can best be managed. Its recommendations have been largely taken on board. At a country level, the limited technical capabilities and project management skills of statistical agency staff can result in poor survey implementation.
While each of these factors bears consideration, the binding constraint to better poverty measurement is arguably coordination. Countries that are unable to measure poverty reliably on their own need more assistance from others. Credible global poverty measurement hinges on greater cooperation across countries, without subordinating country preferences in national poverty measurement.
As a powerful symbol of the merits of collective action and an advocate of coordination between the Bretton Woods institutions, the G-20 is uniquely placed to address this challenge. The G-20 should call for the following 3 actions:
- First, a systematic effort to close the data gaps that affect poverty measurement. In 2015, the World Bank made precisely this commitment with regard to income and consumption surveys, with a pledge to reach complete country coverage by 2020. A parallel target is now required to ensure full coverage of population censuses and surveys of non-income aspects of poverty. The organizations responsible for coordinating this effort are UNDESA (for censuses), UNICEF and USAID (for non-income surveys). Such commitments entail two changes in practical terms. The first is a more systematic approach to tracking survey schedules to identify and plug gaps. The second is to determine how data can be collected in countries whose governments are not able or willing to lead survey implementation.
- Second, the establishment of best practice principles in poverty measurement. This has three components. First, the General Data Dissemination System standards on data collection ought to be reviewed to ensure they are still relevant for modern poverty measurement requirements. Second, technical experts should be charged with defining criteria for good survey design and implementation, which can then be organized into an assessment tool and applied across all countries to raise the quality of poverty measurement. Third, poverty measurement needs to be incorporated into the open data movement to raise the scrutiny of data quality and promote poverty research.
- Third, the promotion of technology and knowledge sharing that supports poverty measurement. Innovations in data collection methods are making possible faster survey implementation in unstable environments, more accurate data collection through automation, and the generation of alternative datasets to verify official sources. Meanwhile, countries and regions are adopting different approaches to reform statistical agencies to protect their independence, secure their budgets and raise their capabilities. New initiatives to share successes and failures can help to accelerate learning.
References
- Chandy and Zhang 2015. These are countries that fail to meet the General Data Dissemination System standard of survey frequency.
To the document - See for instance
Beegle et al. 2012. - Smith et al 2014
To the document - Chandy 2013; Deaton 2010.
To the document - Chandy and Zhang 2015
To the document - Demographic Health Surveys and Multiple Indicator Cluster Surveys are the most common.
More Information
