Data Corrigendum to “Unpacking Sources of Comparative Advantage: A Quantitative Approach” Davin Chor∗ February 2014
This short note corrects two errors in the description of the industry variables used in Chor (Journal of International Economics 2010), which I take sole responsibility for.1 Fortunately, neither of these errors affect the empirical results or the substantive conclusions reported in the paper. The corrections are nevertheless documented here, in order to clear up any confusion that the errors in the published paper might cause, particularly since similar versions of these results have been replicated and reported in more recent papers, for example in Nunn and Trefler (2013). 1. The first error concerns the industry variable used in Chor (2010) to test the hypothesis in Levchenko (2007). To recapitulate, Levchenko (2007) examines whether countries that have better rule of law have a comparative advantage in industries that are “institutionallyintensive”. The latter concept is captured by a measure equal to the negative of the Herfindahl Index of an industry’s input use (−HI), where the underlying idea is that industries that rely on a large number of suppliers to provide small quantities of each input – i.e., that have a less concentrated mix of inputs – are more exposed to holdup problems in their production process. In particular, Levchenko (2007) constructs this measure using U.S. Input-Output Tables from 1992. On this, there is a twofold error in Chor (2010). First, there is a data documentation error. The variable that I used in the empirical analysis in Chor (2010) was actually one minus the Herfindahl Index of an industry’s input use (1 − HI), and not equal to the Herfindahl Index itself (HI) as documented in Appendix B.3. (I constructed this using U.S. Input-Output Tables from an earlier year, 1987.) Thus, what is actually reported in the regressions in the paper are coefficients for the interaction term (1 − HI) × LEGAL, and not for HI × LEGAL, where LEGAL is the country rule of law measure from Gwartney and Lawson (2004). The variable name used in the left-most columns of Tables 1, 2 and 4 should therefore be (1−HI)×LEGAL, and not HI ×LEGAL. In this regard, the positive and significant coefficient that I consistently obtained for the effect of (1−HI) interacted with country rule of law is entirely consistent with ∗ Department of Economics, National University of Singapore, 1 Arts Link, AS2 #06-02, Singapore 117570. Email:
[email protected] 1 I thank Pol Antr` as for drawing these to my attention.
1
the interpretation in Levchenko (2007): Countries with better rule of law are better able to export in industries that feature a less concentrated mix of inputs (or greater input “variety”). For the record, a corrected version of Appendix B.3 should read: Input “variety” (1 − HI). Constructed following Levchenko (2007) using the 1987 US Input-Output (IO) Use Table. The IO-87 6-digit categories map cleanly into the SIC-87 4-digit categories; this correspondence table is available from the Bureau of Economic Analysis (BEA) website. The value of one minus the Herfindahl index of input use is calculated for each SIC 4-digit industry, and the value for each SIC 2-digit industry is then calculated as the median across its constituent SIC 4-digit categories. Second, my corresponding interpretation of this Levcehnko (2007) effect in the second paragraph of Section 3.2.3 in the main text was incorrect. The corrected text should instead read: The next few columns turn to the role of the contracting and legal environment in facilitating production. Levchenko (2007) argued that industries that rely on a large number of suppliers to provide small quantities of many inputs are more vulnerable to holdup problems, and are hence more dependent on the legal system to enforce contracts. Column 4 examines this mechanism by interacting one minus the Herfindahl index of input-use concentration in each industry (1 − HI) calculated from US Input-Output Tables against a measure of the strength of legal systems in each country (LEGAL) from Gwartney and Lawson (2004). The positive and significant coefficient obtained (θβlm2 ) indicates that countries with stronger legal systems are in a better position to specialize in goods with a less concentrated input mix. To reiterate, the error here was not in the construction of the variable (1 − HI), but in the data documentation of this variable and in the original interpretation of the findings in Chor (2010). I report on the next page the full list of the values of 1 − HI by industry that were used in the regressions in the paper:
2
SIC 20: 21: 22: 23: 24: 25: 26: 27: 28: 29: 30: 31: 32: 33: 34: 35: 36: 37: 38: 39:
Name Food and Kindred Products Tobacco Products Textile Mill Products Apparel and other Finished Products made from Fabrics and similar materials Lumber and Wood Products, except Furniture Furniture and Fixtures Paper and Allied Products Printing, Publishing, and Allied Industries Chemicals and Allied Products Petroleum Refining and Related Industries Rubber and Miscellaneous Plastics Products Leather and Leather Products Stone, Clay, Glass, and Concrete Products Primary Metal Industries Fabricated Metal Products, except Machinery and Transportation Equipment Industrial and Commercial Machinery, and Computer Equipment Electronic and other Electrical Equipment, except Computer Equipment Transportation Equipment Measuring, Analyzing, and Controlling Instruments (Photographic, Medical and Optical Goods; Watches and Clocks) Miscellaneous Manufacturing Industries
1 − HI 0.823 0.724 0.792 0.828 0.792 0.922 0.775 0.883 0.833 0.791 0.834 0.796 0.904 0.812 0.903 0.943 0.927 0.911 0.937 0.904
2. The second error concerns the description in Appendix B.4 of the “Input relationship-specificity (RS)” variable from Nunn (2007), specifically how this was aggregated up from the SIC 4-digit to the SIC 2-digit industry level. In the Appendix, it is reported that this aggregation was done “by taking a weighted average, using the share of total input consumption of each 4-digit industry as weights”. The results reported in Table 1 of Chor (2010) actually use the median value of RS across the SIC 4-digit industries that are constituents of the SIC 2-digit industry code. Our results are nevertheless very similar when we use a measure that aggregates up to the SIC 2-digit level using the weighted average procedure instead. This is largely because the correlation between the weighted-average RS measure and the median-based RS measure is very high (0.925). In the regression table on the next page, I report the corresponding results that would be obtained when instead using the weighted-average RS measure in the specifications in Column 5 of Table 1 and Column 1 of Table 2 from the main text of the paper.
3
Corrigendum Table OLS Regression Results using Weighted-Average RS Measure k Dependent variable = ln Xni
Table 1 Col. (5)
Table 2 Col. (1)
−1.165*** (0.038)
−1.162*** (0.038)
θβd2 : Common Language
0.504*** (0.069)
0.502*** (0.069)
θβd3 : Colony
0.769*** (0.108)
0.766*** (0.107)
θβd4 : Border
0.194 (0.149)
0.191 (0.149)
0.284*** (0.072)
0.287*** (0.072)
0.211 (0.237)
0.221 (0.242)
θβf 1 : log(H/L)k × log(H/L)i
3.530*** (0.156)
1.274*** (0.256)
θβf 2 : log(K/L)k × log(K/L)i
0.188*** (0.019)
0.159*** (0.020)
Distance and Geography: θβd1 : Log (Distance)
θβd5 : RTA θβd6 : GATT
Heckscher-Ohlin:
Institutional: θβlm1 : CAP DEP × F IN DEV
1.175*** (0.088)
θβlm2 : (1 − HI) × LEGAL
13.176*** (1.749)
θβlm3 : RS × LEGAL
11.193*** (0.562)
6.353*** (0.612)
θβlm4 : COM P L × LEGAL
2.979*** (0.446)
θβlm5 : COM P L × log(H/L)i
1.449*** (0.437)
θβlm6 : SV OL × F LEX
8.805*** (2.234)
Exporter fixed effects: Importer-industry fixed effects: Number of obs. R2
Yes Yes
Yes Yes
45034 0.606
45034 0.612
Notes: Robust standard errors, clustered by exporter-importer pair, are reported; ***, **, and * denote significance at the 1%, 5%, and 10% levels respectively.
4
References Chor, D., 2010. Unpacking sources of comparative advantage: A quantitative approach. Journal of International Economics 82, 152-167. Gwartney, J., Lawson R., 2004. Economic freedom of the world: 2004 annual report. Vancouver: The Fraser Institute. Levchenko, A., 2007. Institutional quality and international trade. Review of Economic Studies 74, 791-819. Nunn, N., 2007. Relationship-specificity, incomplete contracts and the pattern of trade. Quarterly Journal of Economics 122, 569-600. Nunn, N., Trefler, D., 2013. Domestic institutions as a source of comparative advantage. NBER Working Paper No. 18851. Forthcoming, Handbook of International Economics, 4th edition.
5