On the numerical structure of local and nationwide government spending multipliers: what can we learn from the Greek crisis?

ABSTRACT

It has been recognised that the fiscal multiplier is a function of structural features of the economy and policy reaction parameters. Moreover, the debate on the magnitude of the multiplier along the business cycle has also been the subject of disputed debates. On these grounds, we look at the Greek case by calibrating a NUTS-2 interregional general equilibrium model using data for distinct states of the Greek economy during the development of the recent crisis. Whether this matters for local and nationwide multipliers depends on qualitative differences in the numerical structure of the model. As governments have to decide how to conduct fiscal policy, our results on the Greek case provide insights to policymakers on the reach of their actions in a spatially integrated system.

KEYWORDS:

• interregional computable general equilibrium model
• government spending multiplier
• regional impacts of macroeconomic policies
• sensitivity analysis
• Greece

JEL:

• D58
• E17
• E62
• R13

ACKNOWLEDGEMENTS

The authors thank Jaqueline Visentin, Julio Fournier and Thiago Simonato for useful comments and suggestions. The authors are thankful for the comments and helpful suggestions collected through the University of Sao Paulo Regional and Urban Economics Lab (NEREUS) working paper.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1 ‘An empirical economic model  …  embodies three types of information: analytical, functional and numerical. The analytical structure is the background theoretical material which identifies the variables of interest and posits their causal relations. The functional structure is the mathematical representation of the analytical material and consists of the algebraic equations which make up the actual model. The numerical structure consists of the signs and magnitudes of the coefficients in the equations which form the functional structure’ (McKitrick, 1998, p. 545).

2 Hereafter, we narrow down the definition of local multipliers by excluding other fiscal policy instruments.

3 There is also a growing literature on the use of spatial econometric methods. Of particular interest to our discussion, the work by Marquez et al. (2015) provides an excellent example. The existence of spatiotemporal regional spillovers of growth output captures some of the spatial dynamics found in our paper with important implications for the choice of regional policy goals and regional policy instruments; however, the contributions of specific channels are more difficult to unravel.

4 For detailed information on the estimation process of the 13 NUTS-2 regions input–output systems for Greece, see Haddad et al. (2018). Moreover, the input–output database can be downloaded at https://doi.org/10.13140/rg.2.2.25151.41124

5 In the CGE model, changes in household income generated by government stimulus are directed to consumption. In spite of some evidence that households may use stimulus payments to retire debt rather than expand spending, in our simulations we consider that all income changes are distributed according to the structural coefficients of the model, without explicitly affecting household debts. Since the model is solved in percentage changes – household consumption moving with household disposable income – it is implicitly assumed that the average propensity to consume and the average propensity to save do not change over the simulated scenarios. One of the implications for our results, given there is considerable spatial heterogeneity in industrial structure, is that the household consumption channel ‘activates’ sector-specific shocks, as highlighted by Atalay (2017). We thank one of the referees for raising this point.

6 For additional information of notation and specification, see Appendix A in the supplemental data online.

7 Notice that the variation in the regional outcomes is small for the total effect but relatively large for each of the two components (i.e., intra- and interregional). The small variation in the nationwide, source-specific shock, multipliers is heavily influenced by the method used to generate the interregional input–output system, which assumes that regional demands for domestic and imported products follow the national pattern for all users (including government demand). In other words, economic agents share the same technology and preferences. However, it is important to note that different trade matrices are used for each sector, which allows having different regional sourcing for intermediate inputs and final products, which generates the larger variation in the two components in Table 3 (Haddad et al., 2018).

8 Appendix B in the supplemental data online provides model-based empirical evidence for the arguments in the forthcoming discussion.

9 The strength of the substitution effect depends on the structure of interregional trade and on the size of the regional trade elasticity [$\sigma 1_{\left(i\right)}^{\left(u\right)r}$ in equation (A1) in Appendix A in the supplemental data online, as the use of nested Constant elasticity of substitution (CES) structures may exacerbate regional differences in the absence of modelling agglomeration economies (Elliott et al., 2012; Haddad & Hewings, 2005). Table A8 in Appendix B in the supplemental data online shows the sensitivity analysis of results presented in Table 2 on setting the parameter value twice as high. In such a case, the absolute magnitude of the negative interregional multipliers is reduced at the expense of the local multipliers, favouring the interregional trade balance effect over the international trade balance effect.

10 The 2010–13 changes in the intra-regional component ranged from −1.9% (South Aegean) to 9.1% (Central Greece), and the changes for interregional component from −25.9% (South Aegean) to 24.2% (Central Greece).

Additional information

Funding

This work was partially supported by INCT-Climate Change Project Phase 2 [grant numbers FAPESP 2014/50848-9, CNPq 465501/2014-1, and CAPES/FAPS 16/2014]; Conselho Nacional de Desenvolvimento Científico e Tecnológico [grant number CNPq 406105/2022-7]. Eduardo A. Haddad acknowledges financial support from CNPq [grant number 301684/2022-7], and Vinicius A. Vale acknowledges financial support from Fipe.