Comment

Abstract

Previous articleNext article FreeCommentL. Rachel NgaiL. Rachel NgaiLSE, CEP, and CEPR Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMoreSpolare and Wacziarg (2009) construct a measure of genetic distance to proxy differences in customs, norms, and other ethnic traits. They establish a statistically significant empirical relationship between genetic distance and cross-country income differences. The current paper aims to provide a channel for this relationship through differences in adoption of technologies across countries. It estimates an empirical relationship between genetic distance and adoption of technologies across countries, where both measures are relative to a frontier country. More specifically, this is done by using data on genetic distance they used in their previous paper, and two data sets on technology adoption for the year 1500 and 2000. The frontier is the United Kingdom for the year 1500 and the United States for the year 2000.For the year 2000, the paper relates relative genetic distance to three layers of technology adoption: (1) 33 specific technologies in the Cross-Country Historical Adoption of Technology (CHAT) and 9 specific technologies in Comin, Easterly, and Gong (2010) (hereafter Comin et al.); (2) 4 sectors: agriculture, industry, communication, and transportation; and (3) Aggregate total factor productivity (TFP).I have two sets of comments. The first concerns the data and the methodology on estimating the relationship of adoption of specific technologies and genetic distance. The second is about aggregating the adoption of specific technologies into sectoral and aggregate economy level.I. Technology Adoption at Disaggregate LevelThe empirical framework that links relative genetic distance to the adoption of specific technologies across countries is simple. It computes the genetic distance of country i from the frontier country and the genetic distance of j from the frontier country. When a country consists of different ethnic groups, a country weighted index is constructed by using the share of population for each ethnic group. The key independent variable “relative genetic distance” between country i and j is then constructed as the absolute difference between the two genetic distances. The dependent variable of technology adoption index for each specific technology is taken directly from the two data sets CHAT and Comin et al.. The model in Section II provides a simple mechanism on how relative genetic distance is related to differences in technology adoption across countries. It studies both the cost and benefits of adopting a specific technology k innovated by the frontier country f. The idea is that if the genetic distance between country i and f is smaller than the distance between j and f (country i is more similar to the frontier country f in terms of customs, social norms, and ethnic traits), then it is more likely for country i to adopt technology k both at the extensive and the intensive margin.A. Data and MethodologyThe way the 33 specific technologies are chosen would have important implications for the results. Are the chosen technologies due to the availability of the data or because they are shown to be crucial to TFP? Suppose both technology A and B can be used to produce the same goods k (or goods that are highly substitutable), but technology A is among the 33 technologies but technology B is not, then neither the extensive nor the intensive margin of the adoption for technology A can give us a complete picture about technology adoption for the production of good k, which is what we really care about. There might be very little the authors can do if technology B is simply missing from the data set, but more could be done if technology B is in the data set.The model emphasizes how genetic distance between country i and the frontier country f acts as a barrier of technology adoption because it measures differences in customs, social norms, and ethnic traits. This partly addresses the issue of appropriate technology adoption since genetic distance can potentially capture how appropriate the technology innovated by the frontier country is for country i. But I have some comments on the implementation of this idea, or the linkage between the model and the empirical section.Take the empirical work for year 2000—the United States is used as the frontier country so the relative genetic distance between country i and j is computed as the difference between their genetic distance from the United States. However, to be in line with the model, the paper should use the country where each of these 33 technologies was innovated as the frontier country in the regression for each technology.On the other hand, when one thinks about how country i decides on whether to adopt technology k, maybe what matters to country i is not where technology k was innovated, but who is selling technology k when it comes to “trust or custom” that is captured in the measure of genetic distance. So as an alternative, the authors may want to use a weighted average of all countries that are exporting technology k as a “weighted frontier country” and compute the relative genetic distance accordingly.B. Other Related IssuesThe empirical exercise suggests that the measure of relative genetic distance did not do a good job in a few cases, such as Japan and medical technologies in table 7. I believe there are a few issues that could help to provide some explanations.First, the results in year 2000 may be affected by the history of technology adoption. For example, whether a certain technology was already adopted in the past will affect its intensive margin in year 2000. The concern is if such history dependency is due to factors other than genetic distance. For example, the policy in Japan during the Meiji Restoration in 1868 was to encourage adoption of western technologies. Ngai (2004) shows how this episode has potentially removed some of the barriers to technology adoption in Japan and helps to account for the take-off of Japan in the late nineteenth century. Combining this with the post–WWII institutional reform following the intervention of the United States, which dissolved the zaibatsu system and the deconcentration of many zaibatsu subsidiaries, they can account for a substantial part of the Japanese growth miracle in a model where technology adoption drives a country to transit from stagnation to sustained growth in income.Second, the involvement of international organizations, such as providing free vaccination to poor countries, removes the monetary cost of technology adoption. Controlling for this should strengthen the power of genetic distance and help to understand the current insignificant results for medical technologies in table 7.Finally, the issue of migration should be addressed. Professional repatriates from the frontier country f to country i may have positive effects on technology adoption in country i especially at the extensive margin—they simply bring back the technology from country f to practice in country i. This affects technology adoption positively, but does not show up in the measure of genetic distance between country i and f. One specific example is the establishment of Hong Kong Neurosurgery in 1968, which was mainly driven by the return of the founder from the United States with his own equipment. Similar examples can be found in other industries (e.g., the return of Indian trained professional workers from Silicon Valley). Controlling for the migration flows of the professions should strengthen the power of genetic distance.II. Technology Adoption at Aggregate Level Ultimately, the paper aims to provide a channel of how relative genetic distance is linked to cross-country income differences through adoption of technologies. The issue of aggregating the adoption of specific technologies is important. Explanations on how the sectoral and overall indexes of technology adoption are constructed are crucial.Within each sector goods are also good substitutes. So the issue about substitution that I discussed in Section A related to how the 33 specific technologies are chosen arises also when we look at goods that belong to the same sector. An explanation on how the substitution issue is handled when aggregating technology adoption indexes for specific technologies into sectoral index is needed.The issue of degree of substitutability is even trickier for aggregation into the aggregate economy. Here the main concern is complementarity. There are two potential channels for how complementarity could undermine the modeling of the costs or benefits in technology adoption in the model of Section II, thus casting doubts on the empirical results on genetic distance and aggregate TFP.First, suppose there is complementarity across consumption goods;for example, there is very low substitutability across food, clothes, and medical services. Previous literature has shown that sector-specific TFP growth can trigger the structural transformation as it generates movement in relative prices and the direction of resource reallocation depends on the degree of substitutability across goods produced in different sectors (see Ngai and Pissarides 2007). Faster TFP growth in sector i implies an increase in the relative price of good j relative to good i. If good i and j are complements, resources will reallocate from sector i to sector j. In terms of the model in Section II, the benefits of technology adoption in a particular sector depend on the speed of technology adoption in other sectors, in contrast to the assumption where benefits of technology adoption are independent across technologies.Second, complementarity can arise through the input-output linkages. Previous literature has shown sectors have asymmetric contribution to aggregate economic growth. One example is the investment-specific technical change (ISTC) in the growth-accounting of the United States (see Greenwood, Hercowitz, and Krusell 1997). Ngai and Samaniego (2009) show that once input-output linkages are taken into account, this asymmetry is amplified and the power of ISTC increases by 50% in a similar growth-accounting. This has two implication for the current paper. A simple average of technology adoption across sectors is not the best way to link to aggregate TFP, so discussion on how the aggregation is done is crucial. Moreover, in terms of the model in Section II, technology adoption in the intermediate goods sector lowers the cost of technology adoption in other sectors, in contrast to the assumption where costs of technology adoption are independent across technologies.Finally, once the aggregation issue is addressed, it will be interesting to see some quantitative work based on the estimates of the effect of relative genetic distances on the technology adoption index. What will be the cross-country income differences predicted by the estimates using the existing relative genetic distances for each country? Can the genetic distance deliver the magnitude and skewness in the world income distribution, where the income ratio between the top and bottom 5% of the world population is above 30 and half of the world population has less than 10% of US income?To conclude, this paper is potentially an important contribution to the literature on barriers to economic development, especially on its quantitative impact on cross-country income differences (see quantitative work by Parente and Prescott 1994, and empirical work by Hall and Jones 1999). NotesFor acknowledgments, sources of research support, and disclosure of the author’s material financial relationships, if any, please see http: // www.nber.org / chapters / c12487.ack.ReferencesComin, Diego, William Easterly, and Erick Gong. 2010. “Was the Wealth of Nations Determined in 1000 B.C.?” American Economic Journal: Macroeconomics 2 (3): 65–97.First citation in articleGoogle ScholarGreenwood, Jeremy, Zvi Hercowitz, and Per Krusell. 1997. “Long-Run Implications of Investment-Specific Technological Change.” American Economic Review 87 (3): 342–62.First citation in articleGoogle ScholarHall, Robert, and Charles Jones. 1999. “Why Do Some Countries Produce So Much More Output Per Worker Than Others?” Quarterly Journal of Economics 114:83–116.First citation in articleGoogle ScholarNgai, L. Rachel. 2004. “Barriers and the Transition to Modern Growth.” Journal of Monetary Economics 51 (7): 1353–83.First citation in articleGoogle ScholarNgai, L. Rachel, and Chris Pissarides. 2007. “Structural Change in a Multisector Model of Growth.” American Economic Review 97 (1): 429–43.First citation in articleGoogle ScholarNgai, L. Rachel, and Roberto Samaniego. 2009. “Mapping Prices into Productivity in Multisector Growth Models.” Journal of Economic Growth 14:183–204.First citation in articleGoogle ScholarParente, Stephen L., and Edward C. Prescott. 1994. “Barriers to Technology Adoption and Development.” Journal of Political Economy 102 (2): 298–321.First citation in articleGoogle ScholarSpolare, Enrico, and Romain Wacziarg. 2009. “The Diffusion of Development.” Quarterly Journal of Economics 124 (2): 469–529.First citation in articleGoogle Scholar Previous articleNext article DetailsFiguresReferencesCited by Volume 8, Number 12012 Article DOIhttps://doi.org/10.1086/663654 Views: 101Total views on this site © 2012 by the National Bureau of Economic ResearchPDF download Crossref reports no articles citing this article.