What this covers

Morten Jerven of Simon Fraser University speaks with Russ Roberts about his critique of how economists have studied African economic history. The core argument is that economists have spent decades explaining a growth failure that never occurred. Growth datasets show African economies expanded substantially in the 1950s, 1960s, and 1970s, then contracted during the 1980s and early 1990s, then recovered strongly from the mid-1990s onward. The mistake was to average growth over entire decades or even 1960–1995 as a whole, which obscured this actual pattern and created the misleading stylized fact of permanent African stagnation. By using an African dummy variable in their regressions, economists treated the continent as a single unit with inherent flaws rather than examining how economies actually functioned.

Jerven challenges the econometric orthodoxy on multiple grounds. He argues that institutional explanations—particularly those of Daron Acemoglu and James Robinson—suffer from reverse causality, mismeasurement, and failure to account for local conditions. He shows how Policy Non-Transferability undermines the project of identifying universal institutional recipes: what works in one context, from tractors in American wheat fields to private land titling in Tanzania, may be economically irrational elsewhere. Land abundance in much of Africa, for instance, explains why techniques efficient in densely populated regions produce no incentive for private property formalization. Jerven argues economists should abandon the search for cross-country correlations and instead pursue Study Economies Not Economics—deep, archival engagement with how particular economies emerged and function. He contends the field's publication incentives reward broad generalizations over country-specific knowledge, and that approaches like randomized control trials, while methodologically clean, answer questions too narrow to inform macro policy choices.

Sharpest takeaway

Jerven argues that economists have misread African economic history by trying to explain a permanent growth failure that never happened; in reality African growth is recurring, and the econometric search for cross-country institutional explanations should be replaced with deep, context-specific study of how individual economies actually work.

  • African economies grew in the 1950s-1970s and again since the mid-1990s; only the 1980s-early-1990s 'lost decade' fits the failure narrative.
  • Averaging growth over 1960-1995 and adding an 'African dummy variable' framed Africa as inherently flawed and obscured the actual growth pattern.
  • Institutional explanations are intuitively appealing but mismeasured, often reverse causality, and ignore local conditions like land abundance.

The argument · threads8 threads · 29 claims
0.85

Institutions cannot be reduced to numerical scales; they are embedded in local political processes and do not transfer across contexts.

5 pointscentrality 5/5
  • Introducing private land titling across Tanzania would not automatically make it richer because the land plots are small and the transaction cost of titling exceeds the value of a year's crop, and even with low transaction costs there would need to be institutions to use the title for credit or a state able to tax it; absent those conditions, the rational outcome is not to import institutions that worked elsewhere.

    “The land plots are small; the transaction cost of titling the land is bigger than the value of the--over one year crop. And even if that transaction cost was low, it would still be some kind, have to be some kind of institution that could somehow use the land title so that we get credit.”

  • Institutions were defined by Nobel laureate Douglass North as the 'rules of the game' that may or may not be conducive to economic growth, and Acemoglu and Robinson built on the intuition that not only capital and labor matter but also how society is organized.

    “Institutions were defined most popularly by the Nobel Prize winner Douglass North as the 'rules of the game.'”

  • Acemoglu and Robinson attribute low agricultural productivity and plow use in Congo to weak property rights, but the real explanation is that land is abundant (so its price approaches zero and there is no incentive to secure private property rights), introducing a cow risks death from sleeping-sickness parasites, plowing shallow tropical soil ruins it permanently after rains, and slash-and-burn around trees is the efficient technique in a land-abundant context.

    “when land is abundant, the supply of it is abundant, then that determines the price, what will then be close to zero. Therefore there is no incentive for the state, nor the farmer himself, to work out some agreement how to agree upon the private property rights here.”

  • The intuition that institutions matter is almost universally agreed upon, but reducing institutions to a 0-to-5 scale (Germany 4.8, Tanzania 1.2) destroys the original North-derived insight that institutions are part of a particular political process suited to local conditions and not exportable; a policymaker reading those scores concludes 'make our institutions more German,' which is the exact opposite of the intuition.

    “you end up running a regression where institutional quality, institution has somehow lost its real definition but is now measured on a scale from 0 to 5... And then Germany somehow has 4.8 type of institution, whereas Tanzania has 1.2 institution... Which is completely the opposite of the intuition to begin with.”

  • A technique rational in one place does not travel well: tractors and threshers that are rational for wheat fields in America and Europe are not rational for a rice field in Bali, where a tractor would just make a big mess; thinking of institutions as concrete ways of doing things reveals there is no single method that works in all contexts.

    “That's not the same type of technological change or capital to labor ratios that is rational for a rice field in Bali. A tractor would just make a big mess of the rice field.”

0.79

Economists should study how economies actually work using archival and field knowledge rather than downloading global datasets.

4 pointscentrality 5/5
  • Economists should study economies instead of only studying economics: the distance between the observer (an economist with downloaded global datasets) and the observed (the actual economy) has steadily increased, and the discipline rewards fancy models and law-like statements over nitty-gritty archival and field knowledge of how a particular country's taxes, property rights, and institutions actually function over time.

    “I make call for, let's try to study economies instead of only studying economics... the distance between the observer--the economist attached to his computer with downloaded data sets--and the observed, the actual economy in question, has ever been increasing.”

  • A great benefit of the trend toward randomized control trials is that it recreated the field economist who does field research, though there remains work for the field economist who understands how central banks, statistical offices, and ministries of finance operate.

    “one of the great benefits of this tendency towards economists to do randomized control trials as well is that it has recreated the field economist--the one that does field research.”

  • It is paradoxical to claim history matters and then present Wikipedia-level history with regressions; if history and institutions truly matter, the domain should at least be shared with historians, so a paper on the effect of slavery on growth today should consult historians of slavery, not just test econometric robustness.

    “Branko Milanovic put it nicely, said that it's a paradox to be hearing that history matters and then be confronted with Wikipedia with regressions afterwards.”

  • The economics discipline's incentives reward overgeneralization: a paper studying Tanzania alone has roughly zero chance of publication in the Journal of Development Economics, while a paper generalizing about Africa as a whole can succeed, so publication and citation incentives push researchers toward broad, generalizing claims rather than country-specific knowledge.

    “if you do write something... about Tanzania, the likelihood that an article will be published on Tanzania is equal to zero in the Journal of Development Economics. But if you study something which is saying something about Africa as a whole, generalizing, there would be a chance of succeeding.”

0.73

Growth datasets show African economies grew in the 1950s–1970s and again since mid-1990s, not chronic failure.

3 pointscentrality 5/5
  • The most fundamental mistake mainstream economists made is spending two decades explaining a chronic failure of African growth that never actually happened: growth datasets show African economies grew in the 1950s, 1960s, and 1970s, and have been growing again since the mid-1990s, so the real puzzle is that African economies are growing, not failing.

    “Economists have, since the 1990s, been focusing on explaining why economic growth has failed in Africa... However, the problem and the puzzle is that... not only have African economies been growing for two decades, since the mid-1990s, but African economies have also been growing in the 1950s and the 1960s and the 1970s.”

  • The stylized fact of chronic growth failure should be replaced with one of recurring growth: most African economies grew on average in the 1950s-1970s, declined in the lost decade of the 1980s-90s, and have grown again from the 1990s to today, meaning growth comes and goes rather than being permanently absent.

    “we should approach African growth as recurring. And that means that most economies, on average, were growing in the 1950s, 1960s, and 1970s. And for most economies, growth has been coming back again in the 1990s until today.”

  • By taking chronic failure as a settled stylized fact, the literature moved into a 'second generation of growth literature' that no longer explains growth at all but instead explains differences in GDP per capita measured today, most famously in Acemoglu and Robinson's Why Nations Fail.

    “you get a new type of literature which is not focused on explaining growth, but explaining differences in GDP per capita, measured today. So then you've got this other type of literature which is most famously summarized in the recent book by Acemoglu and Robinson, in Why Nations Fail”

0.73

Cross-country regressions using averaged multi-decade growth data destroy meaningful variation and embed false causality.

3 pointscentrality 5/5
  • Many growth correlations fail the basic test that cause must precede effect: high aid dependence is a phenomenon of the 1990s, so it cannot plausibly have caused slow growth in the 1960s-1980s, and the causality may run the other way, with slow-growing troubled countries needing and therefore receiving more aid.

    “one basic kind of rule... about cause and effect, is that cause should precede effect. Not the other way around... high aid dependence in the 1990s has caused slow growth in the 1960s and the 1970s and the 1980s. It just doesn't work that way.”

  • Averaging economic growth over a three or four decade period from 1960 to the 1990s asks what is correlated with slow average growth, which is a fundamentally different and less meaningful research question than explaining rapid growth in the 1960s-70s followed by decline in the 1980s-90s; by averaging away the changes between years and decades, the analysis ends up measuring nothing.

    “the research question was: Find something that is correlated with slow average growth from 1960 to the 1990s. Which is a very different research question than to find out what can explain rapid growth in the in 1960s and 1970s and then a decline in the 1980s and 1990s. You would be looking for completely different types of variables.”

  • A 2005 review of the growth literature found 145 different explanatory variables that were statistically significant for growth, supporting no less than 43 conceptually different theories of economic growth, an embarrassment of riches that suggests the field may not actually understand growth at all.

    “a review of the economic growth literature thus far... found that there was so far found that there were 145 different explanatory variables that were found to be significant. And which then reckoned as supporting no less than 43 conceptually different economic theories of economic growth.”

0.59

The literature treats Africa as a single unit with shared character flaws instead of recognizing vast internal economic variation.

3 pointscentrality 4/5
  • Robert Barro's 1991 workhorse growth-regression model regressed average growth from 1960-1995 on labor, capital, human capital, currency overvaluation (black market premiums), and assassinations per capita; when these failed to fully explain the variation, he added an African dummy variable, which came out large, negative, and significant, interpreted as an unquantified characteristic of African countries causing slow growth.

    “Barro found that the African dummy variable was large; it was negative, and it was significant. And he interpreted that as that so far, there was something, particular characteristics of African countries that caused slow growth, that has not yet been properly quantified or captured.”

  • The subtraction approach puts rich countries on one side and poor countries on the other and concludes poorer countries are poor because they have different institutions, embedding a massive reversal of causality and a Marxian assumption that a rich country's present is the poor country's future; if history and institutions truly matter, this implies many different paths, so the cross-country correlation literature should be abandoned in favor of deeper contextual studies.

    “you put rich countries on one side and poor countries on the other; and then take [?] and found that the explanations for why poorer countries are not rich is that they are, have different institutional characteristics than the rich countries. And so that is, I think is, the subtraction approach.”

  • Treating Africa as a single unit with shared character flaws is a mistake because there is as much economic variation within Africa as there is between Africa and Europe; Africa is not a country, and much writing treating it as a unit derives from the econometric use of an African continent dummy variable.

    “I think it's fair to say that there is as much variation within Africa as there is, within Africa as there is vis a vis Europe and so forth.”

0.53

Land abundance and integration into world markets, not weak institutions, explain African economic patterns and agricultural choices.

2 pointscentrality 4/5
  • Far from being isolated, Africa is deeply integrated into the world economy: trade as a share of GDP rose from 15-20% in Western Africa 150 years ago to about 50% by 1950, stayed roughly 50% through 1975 even as GDP more than doubled, and reached almost 75% just before the 2009 financial crisis while total GDP nearly tripled, with exports being mostly raw materials.

    “in the past two decades, this share of trade in GDP has increased to almost 75% in the highest year just before the financial crisis, 2009. And that's particularly stunning when we know that total GDP has almost tripled in the same period.”

  • Land abundance and low population density are what make most African economies special, and they explain why the green revolution that succeeded in densely populated India has not been implemented in sub-Saharan Africa, because a green revolution only makes sense when land is relatively expensive relative to wages; exceptions like densely populated Ethiopia, Rwanda, and Burundi have different state structures and already use the cow and plow.

    “'Why hasn't the green revolution been implemented in sub-Saharan Africa yet, while it has been in India?' Most of that is explained by population density and land abundance. A green revolution only makes sense if, well, land is relatively expensive and particularly related to wages”

0.44

African economies are deeply integrated into world markets and people respond rationally to incentives within those constraints.

5 pointscentrality 2/5
  • A 2.5% GDP per capita growth figure means something different in different economies: in Germany it may reflect labor-market efficiency and technology adoption (total factor productivity), but in Equatorial Guinea, Botswana, or Tanzania where 75% of GDP is traded and growth is determined by quantities sold and prices on world markets that they cannot control, building sophisticated total-factor-productivity variables to make generalizable statements simply does not make sense.

    “a 2.5% GDP growth in Equatorial Guinea or Botswana or Tanzania where we have [?] where 75% of GDP is traded on the world economy, when most of GDP growth is determined by quantities sold on the world market”

  • One of the big discoveries in African economic history was that people in poorer countries are not irrational; they are sensible capitalists who make investments according to market opportunities, so the basic intuition that people respond to incentives holds.

    “One of the big discoveries in African economic history was to find out that people in poorer countries were not irrational. They were sensible capitalists who make investments according to market opportunities.”

  • Economists have long explained why a prairie is not a desert (why Africa is not rich) rather than how African economies actually work; we might learn more useful policy lessons by studying why certain African economies succeeded and others failed, and by recognizing that growth is recurring and Africa is not stagnant.

    “economists have been explaining for a long while why a prairie is not a desert. So, they've been trying to explain why a prairie is not something else. I think that's trying to explain how African economies work rather than to explain why they don't would be the key message from me.”

  • Economic growth is an increased ability to produce and/or consume, resulting from combining capital and labor, with the rate of technological change determining how fast that ability grows.

    “It's improved, increased ability to produce, and/or consume. We know that it's a result from combining capital and labor and the rate of technological change determines how fast our ability to do so.”

  • The finding that low governance quality correlates with slow growth is undermined because the governance ratings were subjective surveys of businessmen taken in selected countries in the 1980s, long after those economies had already experienced growth distress, so low institutional quality is more plausibly an effect of economic decline than its cause.

    “these observation were subjective surveys of businessmen taken in selected countries in the 1980s, long after these economies have experience similar economic growth distress. And therefore you would think it was an effect of economic decline and not a cause.”

0.34

Randomized control trials answer precise questions about interventions but cannot address macro policy questions that matter for development.

1 pointcentrality 3/5
  • Randomized control trials give very precise answers to the wrong questions: they satisfy peer-review demands for rigor, control groups, and clean causality and are in vogue in Washington and London, but they cannot answer the macro questions that matter (such as whether to build a fertilizer factory, how much to import, or whether to subsidize), and it is unclear how many village experiments must be run before a result generalizes universally.

    “Lant Pritchett has somewhere written or said that that kind of randomized control trials, the experiments... are giving very, very precise answers to the wrong questions.”