Insularity and economies of density: analyzing the efficiency of a logistic network using an econometric simulation-based approach

ABSTRACT

This paper explores the challenges posed by insularity to economic development and overall welfare from a novel viewpoint. Using a multidisciplinary approach, we investigated the additional burden that this permanent geographical condition poses to retailers whose profit-maximizing strategy relies upon the exploitation of the economies of density. The analysis results show that a retailer finds it convenient to develop its network on the mainland, exploiting the proximity of his stores and distribution centres. Further, it shows that insularity, an unlikely similar condition such as peripherality and remoteness, prevents retailers from expanding their network on an island, thus lowering competition and affecting consumers’ welfare.

KEYWORDS:

• insularity
• economies of density
• retail sector
• simulation

JEL:

• C63
• L10
• L81
• R12

ACKNOWLEDGEMENTS

The authors thank Fabiano Schivardi and Concetta Rau for useful suggestions made during the early stages of this research project.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. For example, this refers to key development indicators, such as gross domestic product (GDP) per capita, migratory balance, unemployment rate, educational attainment, research and development (Spilanis, Kizos, Vaitis, & Koukourouvli, 2013; ESPON, 2010).

2. Licio and Pinna (2013), in this regard, have noticed that distance and discontinuity are interconnected dimensions that contribute to define three distinct groups of islands characterized by different degrees of insularity: fully insular (island states), partially insular and non-insular countries.

3. For more details, see Krugman (1991), Venables (1996), Ottaviano and Thisse (2005), and Ottaviano, Tabuchi, and Thisse (2002).

4. Smallness is a reinforcing condition to remoteness. Small islands might face higher transport costs compared with larger islands due to the reliance on small and fragmented cargoes or the exclusion from major sea and air transport and the consequent delays (Briguglio, 1995).

5. Empirical papers investigating the effect of insularity on economic development provide mixed evidence (Armstrong & Read, 1998, 2004; Armstrong, Ballas, & Staines, 2006; Bertram & Karagedikli, 2004), while the theoretical models (i.e., New Economic Geography) investigate the role of smallness and peripherality without explicitly handling the distinctive features of the island economies and their consequences in terms of firm localization and performance (Krugman, 1991; Ottaviano et al., 2002; Ottaviano & Thisse, 2005).

6. The fact that in reality Esselunga does not exploit economies of density (see the analysis of the Esselunga real network in Appendix C in the supplemental data online) might be due to management or administrative/political constraints (i.e., political barriers that prevented Esselunga from opening new stores below Emilia-Romagna). However, this is not significant to the present analysis.

7. Among the most relevant contributions, see Krugman (1991), Ottaviano et al. (2002), and Ottaviano and Thisse (2005) as well as the recent contribution of Allen and Arkolakis (2014) that emphasizes the role of remoteness and thus trade (trade over space is costly) in determining disparities of economic development over time.

8. Pinna and Licio (2013) measure different states of insularity (considered as a ‘state of nature’), finding that island states have a worse performance than countries with islands.

9. Klaesson and Öner (2014) provide a detailed literature review on this topic.

10. According to the McKinsey Global Institute (2001), Walmart alone is responsible for a large aggregate productivity gain realized over the past quarter century.

11. Following Holmes (2011), and in the absence of any prior knowledge and not having access to more detailed data for the estimation of transport costs, we assumed a linear relationship for simplicity.

12. Due to data limitations, the analysis assumed amortization cost to be a linear function of revenues. We did not model the new stores opening as a function of the previous stores’ full capacity.

13. We explored the sensitivity of the results using alternative radii, specifically 25 and 35 km. We eventually found that the opening sequence remains the same. The greater the radius, the lower is the sensibility of the model because this radius selects the potential consumers of a store and then, through the probability $p(t{)}_{j,l}$, the number of potential consumers decreases as we move towards greater densities and distances, thus capturing the negative effect of population density and distance on demand.

14. The variable $\text{dis}{\text{t}}_{jl}$ is distance (km) between store location $j$ and location $l$.

15. This is because a consumer at location $l$ usually buys at a store located near their house rather than at a more distant similar store. The term $t_{O,j}<t_{O,j^{\mathrm{\prime }}\in l}$ indicates that the opening date $t_{O,j}$ of store $j$ is earlier with respect to that of store $j^{\mathrm{\prime }}$, $t_{O,j^{\mathrm{\prime }}}$.

16. This probability captures indirectly the effects of distance decay and depends on the distance between location $\mathit{l}$ and store $\mathit{j}$ and on the population density of municipality $\mathit{l}$.

17. As in Holmes (2011), the formulation of the demand model takes prices as given and does not consider strategic and dynamic interactions between Esselunga and its competitors, since this is not the purpose of this paper. While such a specification might not encompass all the factors affecting the choice of the store location, this is unlikely to affect the results.

18. Estimating the degree of market overlap precisely is beyond the purpose of this paper, and the cannibalization phenomenon due to the Esselunga competitors is taken into account in the demand model.

19. Esselunga did not follow the economies of density in its real diffusion, as illustrated in detail in Appendix C in the supplemental data online, so the analysis reported in this section concerns only the hypothetical network.

20. The estimation of the degree of market overlap is beyond the scope of this paper. Thus, the results represents a rough estimate.

21. The year 2013 is when the hypothetical diffusion of Esselunga starts, while 2020, 2027 and 2035 are when new distribution centres at Sala Consilina, Latina and Messina respectively open. These are the years when the distance between store $\mathit{j}$ and the distribution centre can vary.

22. $\overline{\mathit{C}}$ is computed as the average of the distance $\overline{\mathit{D}}$ multiplied by parameter $\mathit{\tau }$, which is €2805 per kilometre per year.

23. From an economic perspective, peripherality and insularity impact the demand model differently due to the wealth of the inhabitants. In the demand model the different wealth of the inhabitants enters through the variable gdp, which represents the regional-level GDP per capita (see equations 3–5 in Appendix B in the supplemental data online). In addition, in our specific case study, the Italian context, there is a socioeconomic gap between South and North which is stronger than the peripherality–insularity gap. In other terms, Italian islands are similar, in terms of socioeconomic indicator, to the southern regions. In the analysis, we control for cross-regional differences in household spending capacity controlling for regional-level GDP per capita (see the value of gdp in Table D4 in Appendix D in the supplemental data online).

Additional information

Funding

The authors acknowledge financial support from the Regione Autonoma della Sardegna (L.R. 7/2007 – project Monetary and Non Monetary Costs of Insularity: Measuring Issues and Policy Implications). The funding source has no involvement in any phase of the research.