David Ricardo’s concept of “comparative advantage” is one of the most famous and venerable ideas in economics. Dating to 1817, Ricardo’s proposal is that countries will specialize in making the goods they can produce most efficiently — their areas of comparative advantage — and trade for goods they make less well, rather than making all kinds of products for themselves.
As a thought example, Ricardo proposed, consider cloth and wine production in England and Portugal. If English manufacturers are relatively better at making cloth than wine, and Portugal can produce wine more cheaply than England can, the two countries will specialize: England will concentrate on making cloth, Portugal will focus on making wine, and they will trade for the products they do not produce domestically.
Neat as this explanation may seem, it is by definition hard to prove. If England does not make wine, and Portugal does not make cloth, it is very hard to say how efficiently they could produce those goods. The same applies to any country not manufacturing any given product. So does Ricardo’s idea resemble reality?
A recent paper by MIT economists Arnaud Costinot and Dave Donaldson uses a novel approach to suggest that Ricardo’s hypothesis is buttressed by real-world evidence.
“The basic insight of David Ricardo holds up pretty well,” says Costinot, the Pentti J.K. Kourri Assistant Professor in MIT’s Department of Economics. “As simple as the theory is, it still has substantial explanatory power in the data.”
Why nations specialize
To arrive at this conclusion, Costinot and Donaldson identified a data source that let them quantify nations’ potential productivity: The Food and Agriculture Organization (FAO), an arm of the United Nations, analyzes farming conditions globally, estimating potential agricultural productivity based on factors such as soil type, climate and water availability.
Costinot and Donaldson looked at the numbers from an FAO model of yields of 17 crops on 1.6 million plots of land in 55 countries to examine whether countries specialize in the way Ricardo believed. That is, if a country’s terrain allows it to grow wheat more productively than grapes, comparative advantage suggests that specialization will occur. So Costinot and Donaldson compared the predicted output of crops in each of the 55 countries (based on the FAO data and on prevailing prices) with the actual output of those crops.
The numbers show that Ricardo was right — to an extent, anyway. Costinot and Donaldson analyzed the results so that if the real world worked just as Ricardo supposed, the correlation between productivity and output would be 1.000. Instead, the logarithmic correlation they found was 0.212, with a margin for error of 0.057.
“We found a positive and statistically significant correlation,” Costinot says.
‘Somebody should have done it 50 years ago’
The paper, “Ricardo’s Theory of Comparative Advantage: Old Idea, New Evidence,” was published in the May issue of the American Economic Review. Pol Antras, a professor of economics at Harvard University, has already assigned a version of the Costinot-Donaldson paper as reading in an undergraduate class where students learn about Ricardo’s ideas on international trade.
“It’s an ingenious paper,” Antras says. “It’s not easy to shed new light on old ideas. They came up with a very smart way to try to test a theorem. It’s one of these papers where you think somebody should have done it 50 years ago, which is one of the highest pieces of praise you can give to an article.”
To be sure, other scholars, Donaldson notes, have scrutinized Ricardo’s idea empirically, if a bit more indirectly, especially in the 1950s and 1960s. Economists Daniel Bernhofen and John Brown tackled the issue in a 2004 paper examining how Japan’s production changed when it opened up to trade in the 19th century. But the current paper appears to be the first to surmount what Donaldson calls “the missing data problem.”
“In a sense, all of economics is about choice,” Donaldson says. “But in many settings economists know very little about the non-chosen elements of the choice set. For instance, we would like to study farmers making decisions, but with conventional datasets we can only see how good farmers are at doing what they chose to do, not what they could have done but chose not to do.” The FAO data set allows the researchers to circumvent that problem.
The nature of agriculture, Antras agrees, makes it ideal for scrutinizing Ricardo’s hypothesis. “Ordinarily, countries that would be bad at producing things are not going to produce them,” he says. “But in agriculture, we know how terribly things would go if you tried to grow bananas in Iceland.”
Caveats and future directions
That said, Antras suggests a couple of caveats to the paper. One is that agricultural productivity is not purely a function of environmental factors; technical know-how and the availability of equipment also influence which crops are grown where. Secondly, Antras notes, the less-than-total correlation indicates that additional factors affect international trade as well. “The results suggest the theory is validated, but it is also quite clear that there are many other things that drive trade patterns,” Antras says.
For their part, Costinot and Donaldson acknowledge these qualifications to their findings. In the paper, they try to account for the technological and economic factors that influence crop selection, but recognize such estimates are imperfect. And the MIT economists add a third caveat: The data consist of productivity estimates made by agronomists; if those estimates are a bit off, it would affect the bottom-line findings as well.
Still, Donaldson says, “I was surprised at how, even with all the complexity in the real world, there was still this positive correlation between the theory and reality.”
As a thought example, Ricardo proposed, consider cloth and wine production in England and Portugal. If English manufacturers are relatively better at making cloth than wine, and Portugal can produce wine more cheaply than England can, the two countries will specialize: England will concentrate on making cloth, Portugal will focus on making wine, and they will trade for the products they do not produce domestically.
Neat as this explanation may seem, it is by definition hard to prove. If England does not make wine, and Portugal does not make cloth, it is very hard to say how efficiently they could produce those goods. The same applies to any country not manufacturing any given product. So does Ricardo’s idea resemble reality?
A recent paper by MIT economists Arnaud Costinot and Dave Donaldson uses a novel approach to suggest that Ricardo’s hypothesis is buttressed by real-world evidence.
“The basic insight of David Ricardo holds up pretty well,” says Costinot, the Pentti J.K. Kourri Assistant Professor in MIT’s Department of Economics. “As simple as the theory is, it still has substantial explanatory power in the data.”
Why nations specialize
To arrive at this conclusion, Costinot and Donaldson identified a data source that let them quantify nations’ potential productivity: The Food and Agriculture Organization (FAO), an arm of the United Nations, analyzes farming conditions globally, estimating potential agricultural productivity based on factors such as soil type, climate and water availability.
Costinot and Donaldson looked at the numbers from an FAO model of yields of 17 crops on 1.6 million plots of land in 55 countries to examine whether countries specialize in the way Ricardo believed. That is, if a country’s terrain allows it to grow wheat more productively than grapes, comparative advantage suggests that specialization will occur. So Costinot and Donaldson compared the predicted output of crops in each of the 55 countries (based on the FAO data and on prevailing prices) with the actual output of those crops.
The numbers show that Ricardo was right — to an extent, anyway. Costinot and Donaldson analyzed the results so that if the real world worked just as Ricardo supposed, the correlation between productivity and output would be 1.000. Instead, the logarithmic correlation they found was 0.212, with a margin for error of 0.057.
“We found a positive and statistically significant correlation,” Costinot says.
‘Somebody should have done it 50 years ago’
The paper, “Ricardo’s Theory of Comparative Advantage: Old Idea, New Evidence,” was published in the May issue of the American Economic Review. Pol Antras, a professor of economics at Harvard University, has already assigned a version of the Costinot-Donaldson paper as reading in an undergraduate class where students learn about Ricardo’s ideas on international trade.
“It’s an ingenious paper,” Antras says. “It’s not easy to shed new light on old ideas. They came up with a very smart way to try to test a theorem. It’s one of these papers where you think somebody should have done it 50 years ago, which is one of the highest pieces of praise you can give to an article.”
To be sure, other scholars, Donaldson notes, have scrutinized Ricardo’s idea empirically, if a bit more indirectly, especially in the 1950s and 1960s. Economists Daniel Bernhofen and John Brown tackled the issue in a 2004 paper examining how Japan’s production changed when it opened up to trade in the 19th century. But the current paper appears to be the first to surmount what Donaldson calls “the missing data problem.”
“In a sense, all of economics is about choice,” Donaldson says. “But in many settings economists know very little about the non-chosen elements of the choice set. For instance, we would like to study farmers making decisions, but with conventional datasets we can only see how good farmers are at doing what they chose to do, not what they could have done but chose not to do.” The FAO data set allows the researchers to circumvent that problem.
The nature of agriculture, Antras agrees, makes it ideal for scrutinizing Ricardo’s hypothesis. “Ordinarily, countries that would be bad at producing things are not going to produce them,” he says. “But in agriculture, we know how terribly things would go if you tried to grow bananas in Iceland.”
Caveats and future directions
That said, Antras suggests a couple of caveats to the paper. One is that agricultural productivity is not purely a function of environmental factors; technical know-how and the availability of equipment also influence which crops are grown where. Secondly, Antras notes, the less-than-total correlation indicates that additional factors affect international trade as well. “The results suggest the theory is validated, but it is also quite clear that there are many other things that drive trade patterns,” Antras says.
For their part, Costinot and Donaldson acknowledge these qualifications to their findings. In the paper, they try to account for the technological and economic factors that influence crop selection, but recognize such estimates are imperfect. And the MIT economists add a third caveat: The data consist of productivity estimates made by agronomists; if those estimates are a bit off, it would affect the bottom-line findings as well.
Still, Donaldson says, “I was surprised at how, even with all the complexity in the real world, there was still this positive correlation between the theory and reality.”
