Analysis of the offshoring network: Empirical evidence of the implied comparative advantage in offshoring

Abstract

Using a measure of the volume of offshoring conducted by 62 countries, the intrinsic properties of the global offshoring network are analyzed: size, density, and others. From these properties, we extract information about which countries and sectors of the economy are the main drivers of the network. We find a regularity through the network of all sectors, which we call a stylized fact, that yields an insight into the uniform trade integration of countries through time. Additionally, we construct the measure of implied comparative advantage (ICA) – proposed by Hausmann et al. (2019. Implied comparative advantage) – and empirically verify, for the offshoring conducted by these 62 countries, the hypothesis of a systematic correlation between pairs of industries across different countries. Finally, since the ICA measure is a predictor of the revealed comparative advantage for the offshoring, we verify that the ICA measure preserves the basic properties of the original offshoring network.

Keywords:

• Offshoring
• implied comparative advantage
• network
• product space

JEL Classifications:

• D85
• F13
• L81

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The list of the 62 countries and the OECD classification can be found in Appendix A.3.

2 Formal definitions are provided in Appendix A.1.

3 See Baldwin et al. (2017); Canals and Şener (2014).

4 The analysis of Canals and Şener (2014) is done for $l=$ United States.

5 The intensity of offshoring measure that we use differs from the foreign value-added measure in several ways. Following Hummels, Ishii, and Yi (2001), foreign value-added embodied in exports is defined as the value of intermediate imports ($a_{ijt}^l$) divided by domestic sales plus export and multiplied by exports. However, in our offshoring intensity measure (1(1) $O_{cit}^l=\sum _j{a}_{ijt}^l\cdot \frac{M_{cjt}^l}{C_{jt}^l}.$(1) ) we multiply by imports ($M_{cjt}^l$),not exports, and we divide by the industry consumption of intermediate goods ($C_{jt}^l$). To obtain foreign value-added in exports, we divide by production. The offshoring measure (1(1) $O_{cit}^l=\sum _j{a}_{ijt}^l\cdot \frac{M_{cjt}^l}{C_{jt}^l}.$(1) ) tries to capture a country's total offshoring volume, while Hummels, Ishii, and Yi (2001) refers to the subgroup of intermediate imports that are later translated into products.

6 A more recent release of the IO tables, which covers the period 2005–2015, is available on the website of the OECD. These IO tables are encoded in the classification ISIC rev 4, while the UN Comtrade Database is encoded in the classification SITC rev 4. We did not find a good match between these two correspondences. That is why we limit ourselves to the study of the period 1995–2011.

7 The database used in this work was built by Tamayo Plata, Chica-Castaño, and Canavire-Bacarreza (2018).

8 2011 is the last year for the data, and it will be seen later that this economic sector is the smallest in the offshoring network.

9 For a mathematical definition, see Appendix A.1.

10 For presentation reasons, we only present the graph for five of the sectors with the largest offshored value throughout all years. However, the rest of the graphs have a similar behavior.

11 As in Hausmann et al. (2019), we take $\underset{\_}{O}$ and $\overline{O}$ to be roughly 0.05% and 0.25%, respectively, of the total offshoring. We obtain similar results when varying these thresholds.

12 China, Canada, Germany, and Japan