Poor Numbers: How We Are Misled by African Development Statistics and What to Do about It by Morten Jerven (review)

The title of Morten Jerven’s engaging and valuable new book, “Poor Numbers: How We Are Misled by African Development Statistics and What to Do about It,” is unfortunate. Even Jerven seems aware that such a title could raise hackles. After dedicating his book to African civil servants working in statistics offices, he writes, “I do realize that the title of this book may seem like an undisguised insult to these statisticians, and for that I apologize” (xv).

Nonetheless, perhaps because of the book’s inflammatory title and perhaps because of the serious implications of his findings, Jerven has faced a backlash from at least one African statistician. Due to pressure from Pali Lehohla, South Africa’s statistician-general, Jerven’s September 2013 keynote speech at the United Nations Economic Commission for Africa (UNECA) was abruptly canceled. Interviewed in the South African paper Daily Maverick (Sept. 26, 2013), Lehohla remarked, “We shall not be labeled a pitiful sight. … Jerven says he’s trying to help Africa. That kind of help we detest.”

It is rare for academic writing to generate such hostility, perhaps especially academic writing on something as seemingly dry as, in Jerven’s own words, an “ethnography of national income accounting in Africa” (xii). Yet the implications of Jerven’s carefully researched book are nothing short of controversial. By focusing on national accounts and gross domestic product (GDP) statistics, Jerven not only describes in detail how the data are collected and constructed, but also analyzes the political pressures that shape the statistics. As he points out, “while these data are presented as facts, they are better considered products” (5). Jerven argues that African GDP data are unreliable, cautioning that any cross-country regressions relying on these growth numbers are “likely to yield nonsense or misleading findings” (57). Without reliable GDP data, assessing whether, and by how much, an entire country is growing, or determining the impacts of a particular policy regime such as structural adjustment, is not nearly as straightforward as previously thought.

In chapter 1 Jerven finds few consistent patterns across different GDP databases, and lucidly explains the statistical reasons for such differences. He argues that challenges in gathering data and use of outdated base years tend to result in an underreporting of income. In chapter 2 he examines how statistical offices’ approaches to data collection have changed over time, noting logistical complexities associated with compiling national accounts data—such as capturing “subsistence” production and informal economies—and arguing that a weakening of the state, following structural adjustment programs, has led to decline in the quality of national accounts data. He concludes at the end of both chapters that these issues indicate strongly that income and growth have been underestimated. In chapter 3 he develops his argument that data quality matters through three case-studies: population in Nigeria, crop-production in Nigeria, and growth rates during the era of structural adjustment in Tanzania. In chapter 4 and in his conclusion he lays out several possible ways to improve both data quality and our understanding of how the data are produced.

So, what is to be done? Some researchers, aware of the challenges of using this sort of macro-level data, have championed household surveys and other micro-level data sources (see A. Deaton, Analysis of Household Surveys [Johns Hopkins, 2000]; B. Milanovic, Worlds Apart [Princeton, 2005]). Others see figures such as GDP as part of a broader geopolitical project rather than as indicators in and of themselves (see T. Mitchell, Rule of Experts [California, 2002]). Nonetheless, Jerven argues that “GDP is too important to be ignored and the numbers are too poor to be trusted blindly” (115). In addition to arguing for more funding for statistical offices, he claims that to improve data reliability statistical...