ABSTRACT
Due to international fragmentation, production increasingly occurs in global supply chains (GSC). The common belief is that this leads to more specialization, which implies more concentration of imports and exports over time. In this paper, we empirically test this hypothesis by analysing the geographical and sectoral concentration of GSC over the period 1995–2011. We adapt the traditional Herfindahl’s concentration indexes to a multi-regional input–output framework. Taking the information on intersectoral and interregional linkages into full account gives the concentration indexes of GSC. The indexes are at different aggregation levels, which enables us to examine both geographical and sectoral concentration patterns. After that, we analyse the effect a country’s geographical and sectoral concentration on its gross domestic product (GDP) per capita. Our findings are: an increase of geographical and sectoral concentration of GSC from 1995 to 2011; a growing role in global production chains played by China and other Asian countries; less concentration for European Union countries; a significant positive effect of geographical concentration on GDP per capita; and a significant negative effect of sectoral concentration.
ACKNOWLEDGEMENTS
The usual disclaimers apply.
DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.
Notes
1 Specialization of countries in certain sectors, using multi-regional input–output tables (albeit not global) and based on extensions of the Balassa index, has received some attention in the literature (e.g., Oosterhaven, Citation1995; Van der Linden, Citation1999; Hoen, Citation2002). In the same vein, Van der Linden (Citation1999) looked at the sectors in which outputs or exports were concentrated, which is a different question than the one we will address in this paper.
2 The global MRIO tables in the WIOD provide information for 40 countries plus the rest of the world (as if it were a single country) and 35 sectors. The data are publicly available and can be downloaded for free from http://www.wiod.org.
3 Bold-faced lower-case letters are used to indicate vectors, bold-faced capital letters indicate matrices, italic lower-case letters indicate scalars (including elements of a vector or matrix). Subscripts indicate sectors and superscripts indicate countries. Vectors are columns by definition, row vectors are obtained by transposition, denoted by a prime (e.g., ). Diagonal matrices are denoted by a circumflex (e.g.,
).
4 The specific values of the index depend on the characteristics of the database and, in particular, on the number of sectors. Our analysis, however, focuses on the trends over time of the indexes, using the same database and sector classification across the whole period.
5 Alternatively, one might be interested in the concentration of the import bundle that goes from country r to sector j in country s. In this case, the shares would have been appropriate. It is also possible to look at the export shares and the concentration of countries or sectors of destination. The appropriate shares would be
and
, respectively.
6 The link between the import shares and
is that we can write
. The weights
depend on i and reflect the relevance of imports of intermediate product i (from any country r) by sector j in country s.
7 Detailed results of the analysis without China are available from the authors upon request.
8 In our analysis, we eliminated China from the matrices Z and Q and calculated the indices without this country. The analysis may be slightly biased in the case of Q because this matrix still partially captures the role of China. An alternative would have been to apply the hypothetical extraction method. This method, however, as pointed out by Dietzenbacher et al. (Citation2019), has other disadvantages and cannot be applied straightforwardly to world input–output tables, requiring making additional choices to redistribute the imports from China.
9 The null hypothesis is that the time series has a unit root; the alternative hypothesis is that the series is stationary.
10 The analysis without the RoW has done as in the case without China.
11 In another set of calculations, we took also the domestic deliveries into account, and found that the geographical trends were negative. However, the domestic parts represent around 80% of the indexes. A negative trend therefore reflects more and increasing openness of trade during the period under consideration.
12 Again, if the domestic inputs are included in the Herfindahl indexes, the indexes are higher, particularly for countries with a large share of domestic inputs. The average trend of decreasing concentration is, however, also found in this case.
13 Contrary to our observations for geographical concentration, the results for sectoral concentration do not differ between the global and the country level. This confirms, in general terms, the competitive character of intermediate imports. The input requirements are determined by the production function and for sectoral concentration the source country is not important.
14 We eliminate the RoW, Cyprus and Taiwan from the dataset due to a lack of appropriate data for GDPpc.