It is a commonplace that there are large differences in living standards across countries. For example, in 2014 purchasing-power-adjusted GDP per worker in India and China was only 15% and 25%, respectively, of per-worker GDP in the EU28. Even within clubs of relatively similar countries, marked differences persist – adjusting for purchasing power, the per-worker GDP in the poorest EU country (Bulgaria) is just 30% of that in the richest (Luxembourg).
For some time, economists have used an approach known as ‘development accounting’ to ascertain what portion of these differences can be attributed to directly observable differences in countries’ endowments of production factors. For example, the size of the per-worker capital stock of Bulgaria is only 20% of the per-worker capital stock in Luxembourg. Development accounting maps these differences in endowments into differences in incomes. Using this method, studies have shown that differences in countries’ endowments of production factors can only account for a small fraction of international income differences (Caselli 2005, Hsieh and Klenow 2010, Jones 2015). Traditionally, the large remainder is attributed to differences in countries’ total factor productivity (TFP).
Since TFP is measured indirectly as the residual determinant of incomes once the contribution of all measurable aggregates have been accounted for, it also captures all drivers of income differences which elude quantification. It is thus possible that the large TFP differences which have been found in the abovementioned studies partly reflect our ignorance about what makes some countries rich, and others poor. In a recent paper, we highlight that earlier exercises in development accounting crucially omit the effect of international linkages on countries’ incomes (Cuñat and Zymek 2017). As a result, they may have overstated the role of TFP in explaining international income differences.
Development accounting has tended to proceed on the assumption that countries are closed. This is a useful simplification, but one that is increasingly difficult to defend – not only do we live in a world of internationally integrated markets, but numerous empirical studies have already documented a relationship between the extent and pattern of regions’ access to other markets and their income levels (e.g. Frankel and Romer 1999, Redding and Venables 2004).
We show that, according to standard trade theory, a country’s purchasing-power-adjusted income in a globalised world is not only determined by its factor endowments and TFP, but also by the effect of its international linkages on its terms of trade and ‘trade proneness’ with other countries (which in turn may reflect differences in bilateral trade costs, technologies, and preferences). Having established this point, we demonstrate that the differing magnitude of the linkages-related portion of countries’ incomes can be gleaned from data on the input-output structure of international trade.
Figure 1 International value-added linkages in 2006
New data sources such as the World Input–Output Database (WIOD) (Timmer et al. 2013) allow us to trace how the value added generated by the production factors of one country is distributed to final consumers in other economies through international value chains. Figure 1 provides a graphical overview of the patterns of international trade in value added in the year 2006.1 Each dot represents the share of value added from the vertical-axis country used in final expenditure of the horizontal-axis country, with the size of the dot indicating the magnitude of the share. The figure reveals significant heterogeneity in countries’ use of foreign value added. For example, the share of Luxembourg’s value added in Germany’s final consumption, investment, and government spending is twice the share of Bulgarian value added.
Using this information in conjunction with insights from standard trade models, we can describe how countries’ incomes depend not only on their own production factors, but also on their bilateral trade determinants and, indirectly through their international linkages, on the production factors and expenditure patterns of all other countries. This generalises development accounting to a world of open economies.
Figure 2 Implications of international value-added linkages for measured TFP
Figure 2 gives a flavour of the findings obtained when the open-economy development accounting framework outlined in our paper is used to analyse income differences among the 40 countries for which the WIOD contains the requisite data. The grey bars represent countries’ 2006 TFP levels (relative to the US) implied by conventional development accounting under the assumption that countries are closed – so only domestic endowments of production factors are used to explain countries’ incomes. The black bars represent TFP levels (relative to the US) once international value-added linkages are also accounted for. As can be seen from the figure, implied TFP differences between countries are considerably smaller in the latter case.
Why does the introduction of international linkages into development accounting reduce the need to rely on TFP to explain international income differences? Again, the example of Bulgaria and Luxembourg is illustrative. We find that Luxembourg’s relatively strong linkages with large neighbouring markets result in more favourable terms of trade than those experienced by Bulgaria. This raises the purchasing power of the overall output produced by Luxembourg’s production factors, and results in higher real per-worker GDP for a given level of factor endowments. Unable to capture this mechanism, closed-economy development accounting would attribute the resulting difference in the two countries’ incomes simply to low TFP in Bulgaria and high TFP in Luxembourg.
The new distribution of TFPs obtained by using information about international value-added linkages is conditional on a key parameter we need to specify in order to perform open-economy development accounting – the elasticity of trade flows with respect to trade frictions, or ‘trade elasticity’ for short. For Figure 2, we have used a trade elasticity of 2. However, in our paper we show that accounting for international-value added linkages significantly reduces ‘residual’ TFP differences between countries for any reasonable value of the trade elasticity. For a trade elasticity in the range of 0-8, the variance in TFP levels required to account for countries’ incomes is reduced from 30% to between 3% and 20% of the variance of incomes.
These results suggest that unspecified productivity differences play a smaller role in explaining the observed income gaps between countries than previously thought. At the same time, they raise new questions to be explored in future research. Our analysis remains agnostic about the fundamental determinants of the observed variation in value-added trade proneness across country pairs. The question of whether trade costs, technologies, or preferences can best explain such variation remains central in international economics. If we accept the relevance of trade linkages as a source of income variation across countries, this question should also be at the heart of future endeavours to understand international income differences.
Caselli, F (2005), “Accounting for cross-country income differences”, in P Aghion and S Durlauf (eds), Handbook of Economic Growth, 1(1), chapter 9: 679-741.
Cuñat, A and R Zymek (2017), “International value-added linkages in development accounting”, Edinburgh School of Economics, University of Edinburgh, Discussion Papers 281.
Frankel, J A and D H Romer (1999), "Does trade cause growth?" American Economic Review 89(3): 379-399.
Hsieh, C-T and P J Klenow (2010), “Development accounting”, American Economic Journal: Macroeconomics 2(1): 207-223.
Jones, C I (2015), “The facts of economic growth”, NBER, Working Papers 21142.
Redding, S J and A J Venables (2004), “Economic geography and international inequality”, Journal of International Economics 62(1): 53-82.
 All subsequent figures and statistics are based on data from the year 2006, but our findings remain the same if data from other years are used.