World Bank Voices
Syndicate content

Making data work for everyone

Ana Revenga's picture
Also available in: العربية | Español | Français
When you think about ending poverty and promoting shared prosperity, what comes to mind?  For many people, it is building schools and roads, developing effective safety net programs, improving health facilities, making more and better jobs available, and so forth… for good reason.

But how do we know where to build these roads and schools? How do we find out who needs health facilities, and what kinds of skills exist in a particular country in order to design better employment programs? How do we know where and which kinds of deprivation exist, in order to design safety net programs that actually work?

The answer? Data. Good data. And lots of it. The World Bank Group’s 2014 Policy Research Report: A Measured Approach to Ending Poverty and Boosting Shared Prosperity, takes a carefully considered view of the progress and challenges of measuring and monitoring the twin goals, and pays special attention to data.  

Good data has three essential features. It must help us learn—which usually means that over time it must collect similar information in a similar fashion that is comparable from year to year. It must be timely, to aid fast moving decision making. And the quality of the information collected must not be doubted.

It may come as a surprise to many that for many countries in the world we still do not have good data to enable us to build a solid knowledge of how many people are poor, why they are poor, and what interventions could work for them. I recently traveled to Indonesia and the Philippines, and saw firsthand how good data can help us understand not only the number of people living below the poverty line at a given time, but also to identify those who have just moved out of poverty but are still quite vulnerable to falling back down, and the sources of their vulnerabilities.

In both countries, a mini-census of the bottom 40 percent of the population was used to build up a household registry of poor and near-poor families. These “universal databases” are now being used to target a variety of social assistance and human development programs to reach both poor and vulnerable households. We can and should focus on both groups if we want to succeed in improving the lives of the least well-off in every context—not just in the short-term, but with a view toward sustained upward progress and multigenerational improvements.

We across the development community talk a lot about data gaps, and how important it is to fill them. Consider this statistic: shared prosperity (income growth of the bottom 40 percent) during the last decade can only be measured properly for less than half of the World Bank Group’s client countries.
Most Recent Household Consumption Surveys Available, Africa
Source: World Bank Africa Region, Statistics Practice Team

The Policy Research Report gives us solid arguments as to why now is the time to move past discussion and take real, concrete action to improve the quality and frequency of data collection to measure progress toward the twin goals and to identify the drivers of progress in every country. More and better data are critical to build on our Systematic Country Diagnostics that are creating a solid evidence base to enhance the impact of our support to countries, while identifying data gaps that affect the quality of that evidence.

Improving the quality and frequency of data requires financing and innovation, and the collective will of the development community. It means working differently to make sure that we don’t just know how many people live in extreme poverty, or what the growth rate is among the bottom 40% in every country, but that we use these data to find what needs to be done to make markets, institutions, governments and development agencies work better for poor people. Making data an integral part of the development agenda also requires building a diverse community of stakeholders who “champion” and demand data. This is possible if the development community and governments step up their efforts in making data openly available, inviting scrutiny and free flow of ideas on the use and interpretations of data.

As the Policy Research Report points out, making data work for everyone will take efforts to improve countries’ capacity to collect and assess data, and innovations in statistical methods and data collection technologies. The combination of statistical methods and information technology offers opportunities that would have seemed implausible even a few years back. For example, World Bank teams are fielding surveys through mobile phones to collect real-time data on well-being in remote or fragile areas, and experimenting with systems to collect data, upload them to the “cloud” for validation and use in statistical models to obtain welfare estimates almost instantaneously.

That said, we must recognize that there is no short-cut – technology and statistical methods do not offer a silver bullet where capacity and institutions are lacking, which is why investments in country capacity and statistical systems must go together with expanding the frontiers of innovation in surveys.   The costs of such investments are modest relative to total aid. A recent estimate places resources needed to produce such good quality data for a large number of low-income countries to be around 300 million per year.

Now is the time to mobilize real and sustained resources to fill data gaps, improve data quality, and boost country capacity to keep those gaps filled moving forward. This will provide the evidence base to ensure that policies are timely, well-targeted, and effective for the poor and most vulnerable. What are we waiting for?


Good blog.

The above figure does a good job of conveying the unsatisfactory state of household consumption surveys in Africa measured against your second essential feature of good data: timeliness. However, it's worth pointing out that the collection of surveys contained within PovcalNet, and used to calculate the latest regional and global poverty estimates published by the Bank last week, performs even worse on this measure! For a number of countries (Burundi, DRC, the Gambia, Ghana, Guinea-Bissau, Liberia, Mauritania, Mozambique) the most recent surveys appear to be missing from PovcalNet. This means that the new regional and global poverty estimates rely on extrapolations from older surveys.

Assuming the above figure is correct, this highlights the importance of data not only being generated, but shared, in a timely fashion.

It would be great if you could confirm the status of the latest surveys for those 8 countries. Thank you.

Thank you for your comment, Laurence- it is true that the surveys that you pointed out are not yet in PovcalNet, and there are several reasons why recent data sets do not always appear in the latest release. One is that countries generally do not release data until they have vetted any lingering errors and produced a report on the data. At the time of the last PovcalNet release, Democratic Republic of Congo and Ghana were still working on their final and report preparation. Secondly,  sometimes field work has not been completed, or has just concluded and data entry has not yet started. At the time of the most recent PovcalNet release, Burundi, Mauritania, and Mozambique were in that position. Liberia had to be stopped after 6 months due to the Ebola outbreak, and the plan is to restart when that crisis is over. Finally, sometimes when a country releases and shares their data, we have to do some additional checks of the methodology that was used to create the variables used in the computation of poverty (such as consumption aggregate) in order to ensure that they are comparable to other countries in the global database. At the time of the release we had not completed our internal checks on the data from Gambia and Guinea-Bissau.
We hope to have all these surveys in the PovcalNet in the next update.

Submitted by David Rieff on

Your penultimate paragraph finally grapples with the real challenge, which is not one of data, but one of politics and of justice. And just who is the 'We' in this piece? You speak repeatedly of doing things 'for' poor people, in other words, you speak in the way the Bank always has spoken. What you don't do is speak of listening, let alone of heeding the wishes of poor people, who, to state the obvious, don't need surveys to know what they need. And when you say that "the quality of the information collected must not be doubted," this is an entirely post-political, and, with respect, naively statistical idea of what information is --- as if information too didn't have political and moral content, and exist in a specific moral and political context.

Thank you for your comment, David.  I (and I think I can speak for my colleagues) don’t consider the poor as the passive recipients of anything! On the contrary, our work is about helping countries create the conditions for poor people to have equal opportunity to live the kind of life they aspire to live, and contribute to economic growth and their families, communities, and countries.  With that said, if you don't know where and who the poor are, it is hard to correct the biases and obstacles that get in the way of their being in charge of their own lives.
The poor are indeed the main actor in getting out of poverty and our goal is to help the poor in helping themselves. The constraints that they face are enormous and complex, while the solutions are not obvious- the data can give insight what works and what does not. Data can help us understand, listen, and see what is needed, and can also prove crucial to the poor and vulnerable to help amplify their voices with policymakers in their respective countries.
Additionally, data is by no means just quantitative, but also qualitative. Data is collected through surveys where the poor and the disadvantaged can express their priorities, needs and views without reservation or fear of any adverse consequence. In that sense, good data is empowering for the respondents. Furthermore, unlike anecdotes and opinions, data collected the right way can be "questioned" only on the basis of scientific reasoning, which is what we mean when we say that data needs to be of high quality so that it is credible.

Add new comment

Plain text

  • Allowed HTML tags: <br> <p>
  • Lines and paragraphs break automatically.
By submitting this form, you accept the Mollom privacy policy.