Spatio-sectoral heterogeneity and population–employment dynamics: some implications for territorial development

ABSTRACT

Grounded in the general equilibrium framework of regional adjustment models, this paper studies how the spatial distribution of sectoral employment can affect the intra-regional spatial location of a population, and so affect territorial development. Although spatial interactions, spatial heterogeneity and sectoral heterogeneity have been introduced in these models, no empirical studies reveal how spatio-sectoral heterogeneity affects the intra-regional distribution of population and jobs. This paper explores this effect through a complex set of interdependencies among the regional population, employment in the regional economic sectors and their respective regional neighbours within the different regional typologies (urban, semi-urban and rural areas), as suggested by the concept of proximity developed by Torre in 2019. We use a system of simultaneous equations to focus on the phenomenon of rural depopulation for the 542 municipalities of the Valencian region of Spain. The results provide evidence for the relevance of spatio-sectoral dynamics, suggesting that reversing depopulation in rural areas depends strongly on the services sector.

KEYWORDS:

• depopulation
• neighbours
• population dynamics
• sectoral employment
• spatial effects
• territorial development

JEL:

• C3
• O18
• O21
• R1
• R23
• R3

View correction statement:

Correction

ACKNOWLEDGEMENTS

We are particularly thankful to two anonymous referees and the guest editor, André Torre, whose comments and suggestions have contributed to the overall improvement of the paper. An earlier version of this article circulated as a working paper of the Universitat Jaume I Department of Economics (#2020/24). The usual disclaimer applies.

DISCLOSURE STATEMENT

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

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/00343404.2022.2088725)

Notes

1. Such institutions include the Committee of the Regions (CoR), the Demographic Change Regions Network (DCRN), the Northern Sparsely Populated Areas network (NSPA) and other associations of European regions.

2. The issue of depopulation and how it affects territorial development is not confined to a specific geographical context. It also occurs in countries as relevant as China (Wang et al., 2019), Russia (Batunova & Gunko, 2018), the United States (Franklin, 2021; Heim LaFrombois et al., 2022; Peters et al., 2018), the Netherlands (Ubels et al., 2019) and the Czech Republic (Malỳ, 2018), to name just a few. However, it is more significant in some particular areas because of current demographic trends.

3. Namely, Cuenca, Evrytania, Lika-Senj, Soria and Teruel; http://sspa-network.eu/wp-content/uploads/Position-Paper-SSPA-EUROPE-2019.pdf/.

4. See https://avant.gva.es/es/avant.

5. Empirical evidence of the relevance of our approach will be provided by performing the appropriate statistical tests.

6. The matrix of weights will be defined in subsection 3.4.

7. As we shall see below, the Spanish territory is divided into regions (NUTS-2), provinces (NUTS-3) and municipalities (LAU-2, corresponding to local administrative units, former NUTS-5).

8. Its latest update corresponds to the 2011 population grid and the 2016 LAU boundaries; the next major update will be based on 2020 Census results; https://ec.europa.eu/eurostat/web/degree-of-urbanisation/background/.

9. Densely populated areas: at least 50% living in high-density clusters; intermediate density areas: less than 50% of the population living in rural grid cells and less than 50% living in a high-density cluster; and thinly populated areas: more than 50% of the population living in rural grid cells. For more details, see Dijkstra and Poelman (2014).

10. For descriptions of this cluster and related investigations, see Molina-Morales (2005) and, more recently, Molina-Morales and Martínez-Cháfer (2016).

11. See https://www.ine.es/en/index.htm; and https://pegv.gva.es/es/, respectively.

12. $L{Q}_{i,j}={\displaystyle \frac{firm{s}_{i,j}/firm{s}_i}{firm{s}_j/firms}}$,

where $firm{s}_{i,j}$ denotes the total of firms located in municipality $i$ from sector$j$; $firm{s}_i$ represents the total number of firms in municipality$i$; $firm{s}_j$ denotes the total firms from sector $j$ in the Valencian region; and $firms$ represents the total number of firms in the Valencian region.

13. Beyond that distance, Moran’s I is first negative (negative spatial autocorrelation for the distance between 36 and about 42 km) and then rises (being positive) and falls (being negative) irregularly.

14. The model was estimated by means of a GS3SLS since, considering that our system model is correctly specified, some sort of cross-correlations among the residuals of the five equations in the model is assumed due to the existence of common unobservable factors that are affecting the endogenous variables. Consequently, a GS3SLS estimator would result in better efficiency than the GS2SLS option. Furthermore, we tested the residuals of the two models for spatial correlation, and the results show that the GS3SLS residuals are more efficient (in general, the null hypothesis of spatial correlation is not rejected). Moreover, as the residuals are not spatially correlated, it is not necessary to incorporate additional spatial effects (substantive and nuisance spatial dependencies) in our model. In summary, the full set of five equations is jointly estimated using the GS3SLS method.

15. As one referee pointed out, the usual identification strategy is that population changes are affected by employment changes in the wider surrounding local labour market (municipality and neighbouring area), in such a way that the changes in the municipality itself are considered to be similar to those of its neighbours.

16. Following the terminology of Myrdal (1957) and Hirschman (1958), positive spatial spillovers from core (urban) areas to the (rural) peripheries are called spread effects, while negative spillovers are referred to as ‘backwash’ effects (see also Barkley et al., 1996). As Lavesson (2017) points out, because land prices and costs in rural areas are generally lower than in urban areas, these are factors that potentially cause spread and backwash effects.

17. We are grateful to an anonymous referee for this comment.

18. This result could be related to the characteristics of some construction workers such as, for instance, immigrant status (e.g., Champion et al., 2009), although the issue deserves specific investigation.

19. We are grateful to an anonymous referee for pointing this out.

Additional information

Funding

Luisa Alamá and Emili Tortosa-Ausina are grateful for grant number PID2020-115450GB-100, funded by the MCIN/AEI/10.13039/501100011033, as well as Universitat Jaume I [grant numbers UJI-B2020-27 and UJI-B2020-57] and Generalitat Valenciana [grant numbers HIECPU/2020.1 and PROMETEO/2018/102]. Miguel Á. Márquez is grateful for grant number PID2019-109687GB-I00, funded by the MCIN/AEI/10.13039/501100011033. Besides, Miguel Á. Márquez acknowledges the financial support of the Junta of Extremadura (Spain), European Regional Development Fund [grant/award number GR21089].