Too little knowledge is a dangerous thing
This page in:
Stefan Dercon’s wordle based on our data of the countries that economists work on led Chris Blattman and Tyler Cowen to wonder why there are more papers on Latin America relative to Africa in the Journal of Development Economics, a leading journal in the field of development economics. We looked at this issue in our paper onthe Geography of Academic Research; here are four figures to add to the discussion (two of them are in our paper).
Fact 1: “Just” Income: There is a strong correlation between GDP and publications—a doubling of GDP leads to a 37 percent increase in the number of publications on the country. The US is bang on the regression line relating GDP to publications—a lot more is produced on the US because it is big and rich. Surprisingly, most countries in Sub-Saharan Africa are also on the regression line! In fact, there is no “SSA penalty” in the production of empirical research—there is very little work on most SSA countries mostly because they are poor and small. That 36,649 papers were written on the US between 1985 and 2004 relative to 4 on Burundi, 5 on Benin or 20 on Niger is largely explained by income and population size. As Bill Easterly puts it “the poor get the worst of everything, including the worst economics”.
Fact 2: Data are perhaps not the most important issue: Data availability and quality are definitely associated with publications, as are measures of governance in the expected directions (Table 2 in our paper). But when we related publications to the release dates of the country’s first large household-based surveys like the Demographic and Health Surveys and the Living Standards Measurement Surveys, we found no association between release dates and publications (trend or levels). Here is a picture in event-time (0 is the year that the data were released, -10 is 10 years before and +10 is 10 years after). One interpretation is that data are collected once people start working on a country, rather than the other way around.
Fact 3: A lot of research is produced by researchers outside the country: The extent of research on a country depend both on how much research is produced by institutions within the country, and how much attention it receives from researchers based outside (in the US and Europe mostly—there is very little South-South research). Using two countries as examples, we find that in Ethiopia, around 30 percent of the research is produced in-house and there is no trend although there are large fluctuations, while in India, the amount in–house has climbed from 1990 onwards, although we are not sure why there was a dramatic decline in the share between 1985 and 1990. The research produced on both countries, like research around the world, has increased over time.
How do we want to think about the tradeoff between fostering local research, and incentivizing outside researchers?
Wild Speculation: How much research is driven by the audience? If researchers really care about who is going to read the paper they write, geographical patterns will reflect the tastes of the readership. Given that media content is carefully calibrated to the audience, we thought it would be interesting to see whether media reports for a prominent US newspaper are distributed in the same way as economic research across countries (given how wild this speculation is, the results are not in the paper!). If so, this could suggest that the readership for both comes from the same population. As the figure below shows, the countries that economists work on (Empirical Economics Times) and the countries that the New York Times writes about are the same. If this is more than just a random association, it could suggest that what’s missing in low-income countries is the demand for research rather than a market-failure in its production.
In summary
The paucity of research on and in low-income countries is a serious problem. To the extent that this is driven by low demand within those countries for research, there is a classic case for subsidizing the production of knowledge. In fact, this is partly the rationale for institutions like the research group within the World Bank—and one that the group is taking very seriously. We find, for instance, that the wealth elasticity of research (the percentage increase in research on a country for a percent increase in the country’s wealth) in academia has remained constant between 1985 and 2005, but that within the World Bank the elasticity is (a) much lower and (b) has been declining over time.
1. What policies can “incentivize” knowledge production in low-income countries?
2. Should we worry about this, or can policies to improve education in, say, Burkina Faso be based on research done in South Africa or the US?
3. Appendix 2 in the paper lists for all the countries the number of papers written on them. Do you have a specific story about economics research in your country? Write in, and let us know!
My question is: have you considered the existence of publication bias? There are selections in those academic journals. Research are done for sure but they might not be rigourous enough to be selected. What is sad also is the value developing countries give to research. Research is done and then they just pile up on the shelves.
One Scandinavian program encourages development professionals to research and create knowledge in their own countries and fields of expertise through a 2 year master's program: www.lumid.lu.se
The participants come from all over the world and research development related topics. Please spread the word and encourage talented young development professionals from the global south to apply! Scholarships offered by SIDA.
I'm interested in hearing the author's response to the first comment
I think the point of publication bias is very important. In fact, in our paper we document that the top-5 economic journals publish papers from the U.S. a lot more than those from other countries. However, for the entire body of economic research, we looked at the top 200 economics journals. Among these, the last 80 have a 0 citation count, suggesting that publication bias may not be as big a concern. At the same time, as we note in the paper, these numbers exclude reports, books and journals like the Economic and Political Weekly in India, which are all important outlets for research. The key question I think is whether publication in these other outlets is a substitute or a complement to publication in the top-200 economics journals, and my sense is the latter. That is, I think that the ratio of research in these journals to the total body of research is pretty similar and perhaps even declining with country income. This however, is based entirely on observation and not data--so would love to see what it actually looks like.
A lovely statistic! Thanks.
As to whether it is important: you give one question (do we need to study Burkina or can we base policy in burkina on resarch in S Africa?)
A second question is equally important. Is tehre any evidence in teh countries where there is no research that evidence has any influence on a) policy and b) more importantly, practice (ie 'de facto policy as opposed to a theroetical policy not designed for implementation)?