By Morten Jerven, Associate Professor at the School of International Studies, Simon Fraser University. He is the author of Economic Growth and Measurement Reconsidered in Botswana, Kenya, Tanzania, and Zambia, 1965-1995, available on Oxford Scholarship Online.
How do we interpret growth and income data? How do make sense of the evidence. I started out doing my PhD in economic history at the London School of Economics, and I was first of all interested in associating economic growth with changes of policy in poor countries. I started looking at Tanzania, and I could not make sense of the data. One source would tell me high growth, another slow growth, and even if I looked at specific years in UN or World Bank databases, they had all different growth rates. Since then I have been trying to make sense of it.
On April 7, 2014, the Nigerian Bureau of Statistics declared that their GDP estimates were being revised upwards from 42.4 trillion naira to 80.2 trillion naira (US$510 billion), an 89% increase. Additional GDP revisions have recently been conducted, many of which are forthcoming in the African regions. A few years earlier on November 5, 2010, Ghana Statistical Services announced new and revised GDP estimates. The size of the economy was adjusted upwards by over 60%, suggesting that in previous GDP estimates economic activities worth about US$13 billion had previously been missed. The news, and the reconsideration of the quality of the data needed to evaluate basic trends in growth and poverty in LICs that followed in the development community, led the then World Bank Chief Economist for Africa to declare “Africa’s Statistical Tragedy”.
Most scholars of economic growth treat the downloaded data from international databases as primary evidence, although in fact it is not. The data available from the international series have been obtained from governments which have statistical bureaus, and has then been modified to fit the purpose of the data retailer and its customers. These alterations create some problems, and, as my research shows, the conclusions of any study that compares economic performance across several countries depends on what source of growth evidence is used. The international databases provide no detailed sources for their data, and provide no data that would enable analysts to understand why the different sources disagree about growth. See for example the disagreement in economic growth series reported by the national statistical office, from Penn World Tables, The World Bank, and the Maddison dataset for Tanzania, 1961-2001 which I have published in an earlier article. The figure below shows time on the X axis and GDP growth on the Y axis. I have plotted the highest and lowest growth rate recorded in any data of the four datasets for each year. The gap between the maximum and minimum line represents the range of disagreement on growth.
The average annual disagreement between 1961 and 2001 is 6%. It is not evenly distributed; there is serious dissonance regarding growth in Tanzania in the 1980s and 1990s, and how the economic crisis and structural adjustment affected the economy depends on which source you consult.
The problem is that growth evidence in the databases cover years for which no official data was available and the series are compiled from national data that use different base years. The only way to deal satisfactorily with inconsistencies in the data, and the effects of the revisions, is to consult the primary source: the official national accounts. The advantage of using the national accounts as published by the statistical offices is that they come with guidelines and commentaries. When the underlying methods or basic data used to assemble the accounts are changed, these changes are reported. The downside of the national accounts evidence is that the data is not readily downloadable. The publications may have to be manually collected.
I have suggested since the 1990s that, in the study of economics, the distance between the observed and the observer has increased. When international datasets on macroeconomic variables became available, such as the Penn World Tables, and the workhorse of the study of economic growth became focused on cross-country growth regressions, the trend turned away from carefully-considered country case studies toward larger country studies interested in average effects. The danger of such studies is that they do not ask the right kind of questions of the evidence. As an economic historian, I approach the GDP evidence with the normal questions in source criticism – how good is this observation?; who made this observation? And; under what circumstance was this observation made? When such studies of growth are done carefully, they offer reconsiderations of what used to be accepted wisdom of economic growth narratives.
My latest monograph does precisely that. Economic Growth and Measurement Reconsidered in Botswana, Kenya, Tanzania, and Zambia, 1965-1995, presents a study based on my research visits to the statistical offices of these four countries. In each country, I collected reports and handbooks on methodology, and have supplemented this with consultations with representatives of the respective central statistical offices.
In Poor Numbers: How We Are Misled by African Development Statistics and What to Do About It, I argued that we know less than we would like to think about growth and development in Africa based on the official numbers. I showed how the problem starts with the basic input: information. The fact of the matter is that the great majority of economic transactions, whether in the rural agricultural sector and in the medium and small scale urban businesses, go by unrecorded. In a study of the production and use of African economic development statistics, I emphasized that this is just not a matter of technical accuracy, and that the arbitrariness of the quantification process produces observations with very large errors and levels of uncertainty. This ‘numbers game’ has taken on a dangerously-misleading air of accuracy, and international development actors use the resulting figures to make critical decisions that allocate scarce resources. Governments are not able to make informed decisions because existing data is too weak or simply does not exist.
The book offers a reconsideration of economic growth in Africa in three respects. First, it shows that the focus has been on average economic growth and that economic growth has not failed. In particular, the gains made in the 1960s and 1970s have been neglected. Second, it emphasizes that for many countries the decline in economic growth in the 1980s was overstated, as was the improvement in economic growth in the 1990s. The coverage of economic activities in GDP measures is therefore inaccurate. In the 1980s, many economic activities were increasingly missed in the official records thus the decline was overestimated (resulting from declining coverage), and the increase in the 1990s was overestimated (resulting from increasing coverage). The third important reconsideration is that there is no clear association between economic growth and orthodox economic policies. This is counter to the mainstream interpretation, and suggests that the importance of sound economic policies has been exaggerated, and that the importance of the external economic conditions have been devalued in the prevailing explanation of African economic performance.
My findings on the measurement of performance have led me to reconsider some of the arguments about African economic growth. The study underlines the importance of looking beyond the averaged aggregate growth rates because of, rather than despite of, the issues of data quality. I hope that my findings will stimulate and pave the way for new research that will suggest new evidence and methods to explain long-term economic and social change and (by implication) the current predicament of African economies.
Discover more: the 'Introduction' in Economic Growth and Measurement Reconsidered in Botswana, Kenya, Tanzania, and Zambia, 1965-1995 is now free and available to read until August 31st. Get access to all of this book, as well as almost 600 other Oxford Economics and Finance titles, by recommending OSO to your librarian today. For further reading, read Morten's 'Why measurement matters' article on the OUPblog.