Morten Jerven
About
Economist; researcher on the unreliability of African GDP statistics
Cast within
No topic-region cast yet — this appears once Morten Jerven's compiled claims are aligned into a topic region's argument tree.
Claims by Morten Jerven (20 of 50)
Data availability is the core problem in African GDP
The main problem in compiling African GDP estimates is data availability: statistical offices have little information about food production and the informal/unrecorded economy, while having better data on export crops, large manufacturers, and government activities, and this gap varies considerably across countries and across time since the 1950s.
World Bank brand laundering of national data
Although African GDP estimates are originally national statistical data—collected by local offices that became delayed in the 1970s, prompting the World Bank and IMF to produce their own data—people are surprised to learn the figures are national in origin and trust them only because of the World Bank or IMF brand name, so quoting a GDP statistic is implicitly a statement about how reliable you think that local office is.
Gatekeeper states collect data at borders not domestically
Because land was historically abundant in sub-Saharan Africa and taxes were not levied on private property, states collected information and taxes mainly at borders and ports—what Frederick Cooper calls 'gatekeeper states'—which means African states on average have both less information about their domestic economies and less incentive to collect it.
Validity vs reliability distinction in GDP measurement
GDP data should be assessed on two separate dimensions: 'validity' (whether the level estimate is correct, which in rich countries is off only a few percent but in sub-Saharan Africa can be plus-or-minus 50-100%) and 'reliability' (whether the methodology stays consistent over time); the bathroom-scale analogy shows that a consistently miscalibrated scale is fine for tracking change, but a scale someone secretly recalibrates in the night ruins time-series comparison, while comparing yourself to a neighbor using a different scale ruins cross-country comparison.
Three approaches to GDP and why two fail in Africa
The system of national accounts requires deriving GDP three independent ways: the expenditure approach (the Keynesian C+I+G+net exports, but consumption is unknown without reliable household surveys so it is always derived as a residual), the income approach (wages+profits+rents, uncomputable because most operators have no formal wage and barter), and the production approach (summing industrial sectors from agriculture down to government and NGOs); in sub-Saharan African statistical offices only the production approach is actually used.
South Sudan oil swings dominate measured growth
South Sudan was projected to be both the fastest-growing economy (about 60% growth) and the slowest-growing (about 50% contraction) in the world depending on the year, driven entirely by whether the oil pipeline is turned on—illustrating that petroleum export swings can matter more than statistics and that export-driven growth tells us little about the domestic economy.
Malawi overstates food production to keep aid
Governments have incentives to overstate food shortfalls to qualify for aid and to overstate food production to show aid is working; Malawi has for years seriously overstated maize production to justify continued funding for fertilizer and maize subsidies, yet Malawians are neither gaining weight rapidly nor exporting large amounts of maize, so the statistics must be wrong.
GDP rankings determine concessional lending eligibility
GDP statistics carry real stakes because whether a country is classified poor or middle-income determines eligibility for concessional lending: Ghana's GDP revision nearly doubled its measured economy, reclassifying it as middle-income and making it ineligible for the World Bank's IDA concessional lending, which makes comparative rankings against unrevised countries like Tanzania or Nigeria a mockery.
Data error makes structural-adjustment growth studies untrustworthy
The famous econometric debates comparing strong versus modest reformers to tease out an average GDP growth effect of structural adjustment/liberalization are undermined because the underlying data availability is so poor that there is enough error to make these analyses not trustworthy—effectively meaningless.
Liberalization reduced states' administrative data
Before the 1980s African states were directly involved in production, transport, and buying/selling food through marketing boards, generating abundant administrative data; post-1980 structural adjustment liberalized these states, so they now have access to less administrative data and less incentive to collect it, compounding the data problem alongside the 1980s economic shock that constrained statistical office budgets.
Food production estimated by circular FAO proxy
Because statistical offices lack data on domestic food production, in the 1960s they estimated it by taking the FAO's per-capita estimate for a country in their income bracket and multiplying it by their estimate of rural population and population growth—effectively assuming the answer (income) to derive food consumption, then adding it to another poorly-estimated quantity (population).
Three major data sets disagree on country rankings
Comparing year-2000 GDP per capita rankings of sub-Saharan economies across the Maddison, World Development Indicators, and Penn World Tables data sets, Jerven found they agreed on the ranking of only one country—DR Congo, the poorest—and disagreed on all 44 others, with discrepancies as large as the Penn World Tables ranking Liberia second-poorest while Maddison ranked it among the ten richest, and a standard deviation of about 7 ranks.
Statistics is a state knowing itself
The word 'statistics' refers to the state, and having a valid GDP measure is fundamentally a state knowing something about itself, so improving data is not merely about informing external users but about building a state's capacity to understand its own economy.
Reliability band of 50-75% renders regressions meaningless
Given the disagreement across data sets—Jerven estimated a 30% reliability band in 2010, revised upward to plus-or-minus 50-75% after Ghana's revision—all the econometric estimations using African GDP per capita are for all practical purposes meaningless, and the problem is the metric itself being wrong rather than merely capturing the wrong things.
Recent high growth comes from reclassifying informal economy
Much of the record-high growth recently reported in countries like Ghana and forthcoming Nigeria should be taken with caution because it largely reflects including previously unrecorded economy in the figures rather than genuine new economic activity.
Tanzania's growth recovery was a statistical artifact
Tanzania's GDP series until the 1990s assumed proportionality—that when the recorded economy declined the unrecorded economy declined too—but when the World Bank rebased the series in the 1990s it adopted the opposite assumption that the informal economy grew when the formal economy declined; this methodological switch settled a deep academic question at the statistical office and produced a sudden apparent growth recovery, implying the 1980s decline was overestimated and the 1990s liberalization recovery was overestimated.
Economists were too pessimistic about African growth
The literature on African growth not only ignored GDP data problems but was overly pessimistic—The Economist's 2000 'Hopeless Continent' cover and Collier's 2007 The Bottom Billion contrast with GDP series showing African economies growing rapidly since the mid-1990s, which took economists a decade and a half to notice, suggesting the governance deficit explanation has been overemphasized and revealing a knowledge problem in explaining why economies are 'not growing' when in fact they are.
Data disseminators should label projections and meta-data
Data disseminators like the World Bank should label their products correctly with meta-data—marking with a star data that are actually projections rather than observations, and providing a product declaration noting, for example, that Ghana's revised level estimate is reliable while Nigeria's unrevised-since-1990 estimate may be off by 100%.
Cross-country regressions rose due to computers and data availability
Development economists have shifted from writing area-expert monographs to running cross-country regressions, driven by the availability of computers and the very data sets being criticized, which makes it easy to research development from one's office but causes mistakes when researchers do not understand changes in the underlying data series.
My Notes
Loading notes...