Summary

This paper takes a look at three questions relating to the provision of more data via surveys. First, we ask whether surveys in poor countries have produced results, second, what types of users demand what types of data and third, how much it would cost to close the remaining gaps in household survey provision.

The International Household Survey Network (IHSN) database provides the most comprehensive information on surveys and censuses in low- and middle-income countries. The pace of survey data collection has accelerated rapidly across all regions, as has the trend towards making data open. Poorer countries actually produce significantly more household surveys each year and are more likely to put their data into the public domain.

The demand for such data, and also for it to be openly available, is likely to come from citizens and international aid donors, and the evidence tends to support this. Countries receiving more foreign aid tend not to conduct more household surveys but are somewhat more likely to publish open data. More democratic countries are also more likely to publish their data.

Jerven estimates a figure of $1 billion a year to produce an adequate data package to cover the MDGs. This seems to us to be a reasonable approximation, but what it neglects is that many of the middle-income countries included are wealthy enough to fully fund their own statistical services. Countries such as Kuwait, South Korea and Chile, for example, are included in the figure. Recognising the need for socioeconomic data, we would expect international development assistance to fund a substantial fraction of statistical costs for countries below a cut-off level and only a small share of costs above that level.

The cut-off point should be somewhere in the range $2,000-5,000 GDP per capita. Total data package costs would be $275 million for the lower limit or $510 million for the upper one. 36 out of 52 countries with GDP per capita below $2,000 are in sub-Saharan Africa, and total annual survey costs for this region would be $276 million. We therefore suggest that the total amount of international aid needed to support this basic survey program is about $300 million. We should note that IDA already provides a substantial proportion of the funds, so the $300 million does not represent a marginal increase in aid required, but a total figure including some current funding.

We argue that greater efforts to ensure data openness are as important as supporting data production. There are already a number of laudable data access models, including Afrobarometer and the International Integrated Public Use Microdata project.

International statistics must also have the right goals. The main value of data is not for monitoring international targets but to generate knowledge for policy decision-making in each country. High-frequency, disaggregated data are needed for domestic users. Finally, although household survey data will be useful for monitoring the SDGs, actually achieving them will require greater focus on other types of data, including administrative systems.