Morten Jerven is Associate Professor at NMBU, Norway. He has published widely on African economic development, and particularly on patterns of economic growth and on economic development statistics. We are pleased to be welcoming Morten as the keynote speaker at our SIID Annual Lecture & Postgraduate Research Conference on 7th April 2016.
Numbers rule the world. All aspects of life – from health to the GDP are based on many different data sets. However, the measurement of population health or analysis of economic development is a challenging task and we all know that some of the data are “problematic”. GDP numbers and other statistics, which we often treat as facts, are not facts, they are products, and they are produced under difficult conditions. This is a knowledge problem which is doubly biased. We know less about income and growth in poor countries, and we know less of the economic condition of those who are the poorest in those poor countries.
The errors of measurement are of significant magnitude. A recent example comes from Nigeria where a change in the statistical material underlying the calculation of GDP meant that it was revised upwards to the extent that Nigeria’s recorded GDP per capita almost doubled.
The knowledge problem has implications for governance. Big decisions are made because of these numbers. Increasingly in economic governance, the trend is to let hard facts and evidence prevail where judgment and negotiation used to rule. So for instance, the decision to give a poor country access to concessional lending through the World Bank is based on these statistics. When Ghana rebased, they jumped out of the poor country bracket overnight. The re-classification made a mockery of the facts in this decision process. How could Ghana be classified as a poor country one day, and then move a lower-middle-income economy the next day? What then about the GDP levels of other poor countries?
The problem I describe in Poor Numbers focuses on the data needed for economic governance in Sub-Saharan Africa. When I did my survey of basic methods of estimating GDP in 2011, covering 34 countries, I found a very uneven level of knowledge. The IMF recommends that a base year for GDP estimation should be updated every 5 years. I found that only a handful of countries are able to fulfill that standard, and that for most countries the base years are about a decade or older. When Ghana had such a dramatic jump in income, it was a result of changing the base year from 1993 to 2006. For Nigeria, the current base year is 1990.
Thus we are approaching a quarter of a century since we have had a reasonably accurate picture of Nigeria’s economy, probably the biggest economy in Sub-Saharan Africa. Even before the revision, it is about 50 times the size of Malawi’s economy. That so much economic activity is not accounted for in our current data makes any factual statement of the aggregate trends of growth and poverty in Africa look very uncertain.
But baselines and methods is only the surface. What really matters is basic data availability. As is described in my book “Poor Numbers: how we are misled by African development statistics and what to do about it” the process of aggregating, adjusting and agreeing upon the final numbers of GDP is a process filled with arbitrary and discretionary decisions, which is based on negotiation and subjective interpretation. In the empirical social sciences it has been fashionable to use the binary categories of soft versus hard or subjective versus objective, to distinguish between interpretative qualitative scholarship and research based on inferential statistics. This distinction is imaginary. This is of course true for statistics from Europe, North America and elsewhere, but for poorer countries it is a problem of even bigger magnitude.
The problem is not isolated to GDP numbers –the knowledge problem in GDP is a symptom of the availability of reliable data from statistical systems in poor countries. What we are accustomed to treating as facts are indeed products. The supply of data has constraints – and basic factors such as manpower, budgets, infrastructure and the fundamental problem of converting complex social systems into simple statistical measures in societies where information is not regularly recorded, has serious implications for the quality of the data supply. This should be obvious, but it has not been. So far the treatment of this question has been alarmingly naïve in the development community. We need to wake up to this fact, because the difficulty of provision of statistics in poor countries must have implications for both how we use and demand numbers in the decision making process.
Perhaps the most visible public commitment to results based policy and evidence driven policy was the adaptation of the Millennium Development Goals in 2000 when UN members pledged to commit policy and funds towards reaching the 8 goals, measured by 18 targets and 48 indicators. In retrospect it was strikingly naïve to assert this extent of measurability without a systematic understanding of how data can and should be generated by these weak statistical systems. The current agenda of the Sustainable Development Goals, with 17 goals, 19 targets and 203 indicators, suffers from the same issue.
It is a classic mistake of failing to ask what can be measured, and a result of only considering what should be measured. The list of targets and indicators are based on wishes and demands rather than an appreciation of what could be supplied. One of the basic failings derived from inadequate understanding, for instance, is the distinction between administrative data and survey data. Administrative data are the statistics that are available due to the day to day operation of governments. In many richer countries, the data required to fill out the global data excel sheet to fulfill the data need of the MDGs are readily collectable from their government agencies. In most poor countries that is not the case. In order to fill the gaps in the global data agenda these countries need to collect these data as survey data. For a standard Household Budget Survey this filling out of a long questionnaire takes a day to complete for a couple of thousand households. The exercise costs a few million dollars and takes a couple of years to complete.
The problem of the hailed ‘data revolution’ and the influential calls for moving ‘paying for results’ in development is that it has not been accompanied by a sustainable statistical policy. A sustainable statistical policy means not only thinking about data demand, but also being concerned about data supply. The statistical offices in poor countries have long been neglected. We often talk about accountability in development, but we forget to invest in the ability of citizens to make their states accountable to economic and social policy. Currently, the demands for data are not regulated with a thought of how these demands for measurement coheres with the need for data for citizens in poor countries. The current system causes more confusion than enlightenment