Aware that export success is the key to the success of industries and entire countries, national trade policies have been at the heart of recent policy debates. Donald Trump got elected in part by blaming the ails of many US manufacturing industries on trade deals like NAFTA, along with calls for government protection of manufacturing. In economic development debates, many call for national industrial policies to help pick winners, especially in order to help diversify away from reliance on commodity exports. Recent empirical research suggests both of these arguments overstate the importance of national trade policies in actual export outcomes by industry, and could actually make things worse, rather than better, by pulling back from trade deals and pursuing aggressive industrial policies.
A widespread view of trade stresses comparative advantage based on endowments of labour and capital, as well as natural endowments of non-renewable commodities or conditions suitable for renewable commodities. Trump's criticism of trade deals seems to reflect in part the longstanding fear that industries in high-wage countries will be undercut by a flood of imports from low-wage countries if trade deals are made with those countries. Recommendations for industrial policies in developing countries reflect the traditional view that commodity endowments will condemn the country to excessive reliance on one or two commodities with volatile or declining world prices – unless industrial policies succeed in diversification.
An influential paper that challenged the simple endowment-based explanation of export specialisation is Hausmann and Rodrik (2003). The authors showed that countries often specialise in very narrow categories (for example, exports of hats from Bangladesh, but not soccer balls, and the opposite in Pakistan), too narrow to be explained by cheap labour or commodity endowments, casting doubt on simple comparative advantage explanations based on these factors. Their work could still have led to the conclusion that industrial policy was the key to explaining these highly specialised successes or achieving them in the future, especially if the process of discovering what goods a country can export involves a positive externality, as they argue in Hausmann and Rodrik (2006).
However, our findings in a recent paper challenge even this view of a possible role for industrial policy (Daruich et al. 2016). We first revisit the well-known fact of hyper-specialisation in exports, namely that a few export goods account for the bulk of export value with each country. What is novel is that we find that hyper-specialisations are very unstable, making it unlikely they are explained by industrial policy or that industrial policy could even possibly work in the medium run.
For example, the average correlation of the ranks of top export products in 1998 with ranks of products in 2010 is only about 0.3. This finding is not consistent with stable patterns of endowment-based comparative advantage. The rank correlations drop to around 0.1 when we examine export flows, defined both by type of product and also by destination. As shown in Table 1, these results are not very sensitive to whether we examine top 20, top 50 or top 100 export products or flows, and whether or not we include extractables (mostly mining products) and commodities (mostly agricultural products like coffee and cotton). These rank correlations are only slightly higher for rich versus developing countries, and remain low for virtually all countries.1
Table 1. Rank correlations of top exports, 1998-2010 (average across all countries)
Notes: The table reports the average across all countries of the correlation between ranks of top export products and export flows (product-by-destination) that were exported in 1998 and their ranks in 2010.
Source: COMTRADE, 4-digit HS codes.
We illustrate instability graphically for the top 20 export products for two countries: Tanzania and the US. Figures 1 and 2 illustrate how pervasive instability is. It is striking that a similar degree of instability is observed in both a leading industrial economy and a developing one – despite very different composition of products.
Figure 1. Top exports churning in Tanzania
Notes: The figure reports the ranks and values of top ten exports in 1998 and in 2010, and their ranking and value in the opposite end of the sample, all in 2012 prices (thousands of US dollars). Source: COMTRADE, 4-digit HS codes.
Figure 2. Top exports churning in the US
Notes: The figure reports the ranks and values of top ten exports in 1998 and in 2010, and their ranking and value in the opposite end of the sample, all in 2012 prices (thousands of U.S. dollars).
Source: COMTRADE, 4-digit HS codes.
We then investigate what accounts for this instability. To do this, we develop a methodology that decomposes this instability into factors that lie at the product level within a country (comparative advantage), source-by-destination bilateral factors, destination demand factors, and every possible interaction of source, product and destination. Importantly, our methodology takes into account new and disappearing products. We do this for the sample of top 20 export flows for each country, and then for the entire range of exported products for each country.
Table 2. Results of export growth variance decomposition
Notes: Variance decompositions of export growth of export flows in 1998-2010 for the average country. There are two subsamples: top 20 exports, and all export flows. Columns do not sum exactly to 100 because other covariance terms are not reported here; these covariance terms account for small shares of overall variance.
Source: COMTRADE, 4-digit HS codes.
We find that only 19% of variation of success within a country is explained by country-specific product factors, undercutting explanations of national factor or commodity endowments, like the cheap labour feared by Trump or the commodity reliance feared by development policymakers. We find that only another 18% is explained by source-by-destination factors, which would include the trade deals blamed by Trump but also many other well-known bilateral factors, such as changes in transportation and other trade costs. Altogether, source-country related factors account for 37% of the variation in export success, even while bilateral forces are not determined solely by the source country. This drops to 29% when we examine all flows.
About 30% of the variance is idiosyncratic – not explained by any factors specific to products, sources, or destinations – which rises to 50% when examining all export flows. Trade seems to be like a casino in which it is hard to know what will work.
However, this surprising frequency of surprises should not imply shutting down the casino altogether. Participating in trade is important, as several studies have shown (e.g. Frankel and Romer 1999), but one must let the market and government support be nimble enough to support whatever is working at the moment. Government policies should not try to lock in support for one particular product. The government should not attempt to reverse failures of specialisations that were previously succeeding, but could try to cushion failure by generic policies such as trade adjustment assistance or bankruptcy protection. The best government export promotion policy would be just to create the business-friendly conditions for private entrepreneurs to exploit new opportunities as they arise.
Daruich, D, W Easterly and A Reshef (2016), “The Surprising Instability of Export Specializations”, NBER Working Paper No. 22869.
Frankel, J A and D Romer (1999), “Does Trade Cause Growth?”, The American Economic Review, 89(3): 379-399.
Hausmann, R and D Rodrik (2003), “Economic development as self-discovery.” Journal of Development Economics 72. 603-633.
Hausmann, R and D Rodrik (2006), “Doomed to Choose: Industrial Policy as Predicament”, working paper, Harvard University, John F. Kennedy School of Government.
 We show in our paper that these low correlations are not likely to be primarily driven by measurement error in the data. These results are also robust to using a year before the recent crisis instead of 2010.