Effects of international trade on domestic employment: an application of a global multiregional input–output supermultiplier model (1995–2011)

ABSTRACT

The effects on domestic employment of international trade and the globalisation of supply chains are as politically controversial as they are empirically inconclusive. To estimate them we extend the global multiregional input–output framework by endogenising demand for both domestic and imported intermediates, private business investment and household non-durable consumption – or equivalently, we generalise the supermultiplier formula. The model accounts, in particular, for the employment consequences of economic integration and those channelled through integration. We estimate these foreign sector effects alongside those of domestic origin using a recursive hierarchical structural decomposition analysis and statistics from the World Input–Output Database and National Accounts that cover years 1995–2011. Focusing on Spain, Italy, France, Germany, the UK, the US, Japan and China we answer the following questions: To what extent did international linkages deriving from international trade affect domestic employment? Did domestic employment benefit from economic integration?

KEYWORDS:

• Employment
• international trade
• economic integration
• global multiregional input–output model
• supermultiplier

Acknowledgements

I thank the invaluable support of Miguel-Ángel García-Cestona and Anna Garriga Ripoll. I am also indebted to the GEAR Research Group (in particular to Óscar Dejuán Asenjo, Luís A. López-Santiago and Jorge E. Zafrilla Rodríguez), Ramon Boixadera Bosch, Isabella M. Weber, and two anonymous referees for their most helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹See the report for an overview of common methods used to estimate the employment effects of trade.

²These foundations are, of course, complementary and historically intertwined. Leontief's Input–Output Analysis has close ties with the Classical Surplus Approach (Kurz et al., 1998, pp. xix–xxxviii; Kurz and Salvadori, 2000; see also the special issue in this journal on ‘Input–Output Analysis and Classical Economic Theory’ edited by Kurz and Lager (2000)) and thus with Sraffa's contributions (Kurz and Salvadori, 2006). Furthermore, input–output frameworks have been extensively used to ground multisectorial Keynesian models (e.g. Goodwin, 1949; Pasinetti, 1981).

³These changes may respond to developments in, for example, prices, exchange rates, monetary policy, technology, productivity levels, the labour market, trade policy. Thus, we do not explicitly model the interaction of prices and quantities, nor do we explicitly account for monetary policy and the supply side of the economy.

⁴This is why the model presented in the next section does not posit any general and automatic mechanisms that correct external imbalances. Nonetheless, we recognise that prolonged current account deficits and the corresponding accumulation of external debt can lead to a severe crisis. In other words, if deficits must be corrected, income carries most of the burden of adjustment, in line with Harrodian trade theory (Harrod, 1933).

⁵Recent developments in Marginalist Trade Theory also find openness on employment to have an ambiguous effect (e.g. Davidson and Matusz, 2009; Helpman et al., 2010; Grossman, 2013; Baldwin and Robert-Nicoud, 2014).

⁶They also aim to account for a ‘reallocation effect’ of employment from growing trade with China. However, in accordance with our theoretical approach, they find no robust evidence of its existence.

⁷See Stirati (forthcoming) for an overview of different theoretical approaches to employment in the history of economic thought and Cesaratto et al. (2003) for an exposition of the approach that we follow more closely.

⁸Matrices are indicated with bold-faced capital letters, vectors are columns by definition and are indicated with bold-faced lower-case letters, and scalars are denoted by italicised lower-case letters. A prime (‘) indicates transposition, and a hat (^) signifies diagonalisation. The first subscript denotes the country of origin, and the second denotes the destination; a minus sign in front of a subscript stands for ‘all countries except’.

⁹We thank a referee for noting that in the disaggregated model each cell of the column of non-durable final consumption should be allocated in the cells of the same row according to the value added of each industry. Similarly, each cell of the column of private business investment should be allocated in the cells of the same row according to the capital intensity of each industry.

¹⁰The multiplier–accelerator mechanism has been fruitfully embedded in input–output tables and social accounting matrices, perhaps most notably by Óscar Dejuán (e.g. 2006).

¹¹Kaldor noted that ‘the doctrine of the foreign trade multiplier’ gains in generality if complemented with the Hicksian supermultiplier (Kaldor, 1970; 1983; see also McCombie, 1985; Palumbo, 2011).

¹²As noted by a referee, these complex flows are at the core of trade in value added literature, which ‘slices up’ global value chains by recourse to the Leontief inverse of a global input–output table (e.g. Johnson and Noguera, 2012; Timmer et al., 2014). (p. 7, footnote 12)

¹³ This idea unites the Balance-of-Payments-Constrained Growth literature with Centre–Periphery or North–South models (McCombie and Thirlwall, 1994, pp. 256–261; Thirlwall, 2013, pp. 108–111).

¹⁴ See Palumbo (2011) and Vernengo and Rochon (2001, pp. 90–94).

¹⁵Computing the average of all possible decomposition forms avoids accumulating the (small) errors that arise from taking the average of polar forms (Dietzenbacher and Los, 1998).

¹⁶For example, to estimate the effect on hours worked of changes in the trade structure of domestic intermediates (${\mathbf{\text{B}}}^{\left(\mathrm{r}\right)}$) following the recursive hierarchical SDA, we must introduce the two possible decompositions of B into the two possible decompositions of A belonging to the 120 possible decompositions of MESM, which, in turn, are inside the two possible decompositions of l. Therefore, we take the average of the 960 possible decomposition forms to estimate this effect for one country and for one year. Programming is easier once we take advantage of the fact that: (1) ‘for every factor [n], there are only 2^(n–1) different possibilities to calculate the coefficient preceding the Δ item of this factor’ and (2) ‘the weight of the coefficient is equal to (n–1–k)! * k!, being k the number of factors that are evaluated for the actual year’ (Seibel, 2003, p. 15).

¹⁷Of course, if the trade deficits lead to current account deficits, the resulting accumulation of foreign debt may eventually hinder employment by blocking the expansion of domestic demand, as highlighted by the Structuralist tradition (e.g. Prebish, 1959) and the Balance of Payments Constrained Growth Theory (e.g. McCombie and Thirlwall, 2004). Indeed, it has been at the root of the recent Eurozone crisis (e.g. Dejuán et al., 2013, Part II).

¹⁸These results are in line with those of Timmer et al. (2013).

¹⁹Provided that this last effect is overwhelmingly governed by the impact of outsourcing rather than the impact of changes in the geographical sourcing structure of imported inputs.

Additional information

Funding

This work was supported by the Spanish Ministry of Education under a grant of the programme Formación de Profesorado Universitario 2011.