16 July 2015.
One of the stylised facts in debates on data quality and data availability is that over the past few years we have seen more and better data on many, if not most, aspects of development. But when it comes to economic statistics, and particularly statistics on economic transformation, there is a lack of good data.
Does the lack of high quality statistics matter? One of the things I used to say when I addressed this question in discussions following the publication of Poor Numbers: How We Are Misled by African Development Statistics and What to Do about It was that in the average low-income country in sub-Saharan Africa, even if a government wanted an industrial policy or plan for improvements in employment or agricultural performance, there would be no statistics to formulate or evaluate such a plan.
In Poor Numbers I focused on the highly visible gross domestic product (GDP) measure, but the problem of soft and outdated GDP benchmarks is just a symptom of paucity of data on economic production and consumption. Of the 77 countries that are classified as low-income countries, less than a handful are able to produce economic statistics of the quality and regularity required in order to be members of the International Monetary Fund’s (IMF’s) Special Data Dissemination System. A minority of these countries have benchmarks for calculating economic growth and inflation that are less than ten years old, and other data are simply missing. I surveyed the availability of labour statistics for African countries and found that only 5 or 6 countries (depending on the data source, IMF or the International Labour Organization (ILO)) have annual labour data. Perhaps more tellingly of our knowledge problem in labour statistics, ILO had no metadata for 19 of the 54 countries. So not only is there a lack of data, but there is even a lack of data on the size of that data gap. In agricultural statistics, there is a similar dearth of regular high quality data, and while we have more regular and reliable data on household budget expenditures, we know much less about production, particularly in small- and medium-sized enterprises.
What do we know?
It is beyond doubt that there is more growth, investment and trade in low-income countries now compared to a decade or two ago. The bottom line is that there is no longer any bottom billion. However, where does this growth come from? Is it associated with economic transformation, or are we seeing an intensification in external trade and activities in primary sectors? Our enthusiasm over ‘Africa Rising’ or catch-up growth in low-income countries should crucially depend on the extent to which this growth is sustainable and therefore accompanied by qualitatively observable differences in how goods and services are produced and how this production is organised.
What we do know is that some of these countries are richer than we thought. On 5 November 2010, Ghana Statistical Service announced new and revised GDP estimates. As a result, the size of the economy was adjusted upward by over 60%, suggesting that in previous GDP estimates, economic activities worth about $13 billion had been missed. While this change in GDP was exceptionally large, it turned out not to be an isolated case. On 7 April 2014, Nigeria’s National Bureau of Statistics declared that its GDP estimates also were being revised upward to $510 billion, an 89% increase from the old estimate for 2013.
So where does that leave us? It should have come as no surprise that the previous GDP numbers were a poor guide to levels of expenditures and income. The benchmark years in Ghana and Nigeria were updated from 1993 and 1990 to 2006 and 2010 respectively. It then appeared that a lot of new economic activity was missed. Research that made use of data from Demographic Health Surveys to correct for missing data in the national accounts showed that low rates of growth were at odds with higher rates of since the 1990s.
The new benchmark data do, among other things, also provide us with an updated view of the economic structures of these economies. Some analysts have taken the new numbers to mean that that there is structural change, but to compare an old benchmark with a new benchmark is like comparing apples and oranges. It is hard to know whether the new economic structures reflect different prices, weights, definitions and data availability, and to what extent they reflect real growth in some sectors. What we can say, is that the new estimates give us a picture of economies that are more diversified into manufacturing and service activities in the picture we had with the outdated benchmark years.
Overall, it is notable that tertiary sectors looks much larger in the new estimates – whereas the changes in agriculture and industry are too small to say that we know they are significant. These new GDP estimates were in many cases prepared without new data on production, and without specific surveys of business, industry and agriculture; and data were mostly drawn from surveys on household expenditure. It is symptomatic of the data availability problem that recent studies of industrialisation in sub-Saharan Africa are using data from health surveys in the absence of classic data sources on employment and production.
The way forward
It is easy to lament the status of knowledge and call for better and more data. Surely, more frequent data on economic activity would satisfy the needs of investors, central banks, commercial operators and some scholars and analysts, but these data demands have to be weighed against other priorities. Data collection is time-consuming and expensive, and capacity at statistical offices is stretched.
In the meantime, we need more hands at work. Part of the problem has been neglect of the study of economic change in many low-income countries. Since the 1990s, overwhelmingly the focus has been on poverty and poverty eradication. Since the 2000s, the focus broadened with the Millennium Development Goals. This focus is reflected in the statistical record. The Sustainable Development Goals do in part put industrialisation, employment and agricultural productivity back on the development agenda.
A lot can be done by triangulating data sources and unearthing new sources, through the careful work of the document historian. It is at the same time symptomatic and promising that the main from datasets meant to monitor health and demographic trends.
It remains true that levels of employment, or the share of employment of the labour force, will depend on the survey type and the questions asked. World Bank researchers found that for Tanzania, labour force participation rates vary by as much as 10 percentage points across four different surveys. In Nigeria, the reported unemployment rate just fell 75% as the definition of ‘employed’ was reduced from 40 to 20 hours per week. Similarly diverging results from one data source to another may be seen in the levels and trends in agricultural production. This underlines the importance of doing research close to the source of data. Comparisons of levels and rates of taxation suffer in similar ways – but recent work has been done to combine and compare all possible data sources and datasets, not only to fill the gaps in existing datasets but also to use different sources to decide upon the most plausible observation.
While waiting for the new data sources, researchers, analysts and policy-makers will have to make do with the data. The SET portal on Data and Statistics is promising in this regard.