Exploring the role of accessibility in shaping retail location using space syntax measures: A panel-data analysis in Lisbon,1995

Abstract

Although location theory is now almost two centuries old, the firms' location choice processes areyet to be fully understood. And while accessibility, in some form, has long been used as an explanatory factor, spatial configuration measures (space syntax) have hardly been used in location models, and longitudinal analyses have also been infrequent. Therefore, a panel multinomial logit model is implemented in Lisbon to explore the role of this specific type of accessibility measures in shaping firm location, throughout a 15-year period, using data from 1995, 2002 and 2010. The analysis is focused on retail activity (five retail categories), restaurants and cafes. The results show that firm location can persistently be related to Local integration and Choice, although the magnitude of the relationship changes depending on the commercial category. These results reinforce the general idea that good urban design can potentially counteract information costs and help commerce locate in new areas of the city while allowing it to continue to succeed in the city’s older, more central areas.

Keywords

Firm location models retail location longitudinal analysis space syntax

Introduction

While the existence of modern cities has been attributed to the revival of trade in medieval times (Pirenne, 1925), retail has been acknowledged as a promoter of social interaction since at least the 1960s (Jacobs, 1961). Although cities had existed for millennia, before the Middle Ages they did not possess two fundamental attributes of modern cities: a middle-class population and a communal organisation (Pirenne, 1925). That community of people came to connect at retail, thus creating the sidewalk life we associate with cities (Jacobs, 1961). Apart from fostering liveability, retail is also a vital economic sector. Since Britain once being called a ‘nation of shopkeepers’ (Smith, 1776), retail has become the largest private-sector employer in Europe (wholesale and retail trade, transport, accommodation and food services) (Eurostat, 2017) while being worth $5570.4 billion in the USA (U.S. Census Bureau, 2022).

The ‘sidewalk life-creating retail’ of the 1960s was not unlike the small businesses we turned to during the COVID-19 pandemic when travel restrictions and the fear of contracting the disease turned many of us into customers and even supporters of local stores (Gordon-Wilson, 2022; Musselwhite et al., 2020). However, during the same time, e-commerce increased from 13.8% of worldwide retail sales in 2019 to a predicted 21% in 2022 (Statista, 2022). To some extent, the enforced relationship with local stores may counteract the ‘retail apocalypse’ that large digital-based companies were expected to bring upon traditional physical retailers (Hänninen et al., 2021; Helm et al., 2020). Local stores (or at least some) may thrive by catering to residents who want to buy food and daily necessities without leaving the safety of their neighbourhoods (Fei et al., 2020).

Nevertheless, and although stores are also increasingly becoming a ‘showroom, museum, and warehouse and fulfilment centre’ (Sachdeva and Goel, 2015) to add value to customer experience in the face of online competition, location is still a critical factor in determining an outlet’s success (Lin et al., 2018; Zambrano-Rey et al., 2019). Location theory has long informed the choice of an optimal location for economic activity, starting with Von Thünen’s ‘proto’ bid-rent theory (Von Thunen, 1826). Land rent falls as the distance to a central market increases to offset transportation costs (Von Thunen, 1826). Hence, those less dependent on proximity to a centre to sell their goods can locate farther from it. Therefore, accessibility measures first entered location models as distance to a centre. Distance to a centre is still a frequent explanatory variable in firm location models (Elgar et al., 2009), although more sophisticated accessibility measures have since been devised.

Consequently, and because the relationship between the presence of economic activities and urban configuration is relevant for urban planning, accessibility has entered retail location modelling through network centrality and accessibility measures such as betweenness, closeness and degree centrality (Porta et al., 2006). Spatial configuration measures such as integration, connectivity and choice although equivalent to closeness, degree and betweenness (Hillier and Iida, 2005; Omer and Goldblatt, 2016), have hardly been used in location modelling. Nevertheless, they have proven significant at predicting economic activity, and specifically retail location (Hillier, 1996; Lin et al., 2018; Martí et al., 2021; Mora, 2003; Ortiz-Chao and Hillier, 2007). The poor assimilation of spatial configuration measures by location modelling can eventually be attributed to the fact that, as measures become more abstract, their association with traffic flow modelling becomes less obvious, and this has been one of the main goals of location modelling (Batty, 2009).

Hence, this paper sets to analyse the impacts of spatial configuration on retail location in the framework of an (adapted) firm location model. The model is a multinomial logit with random effects that estimates the utility of a location for hosting a specific category from between five categories of retail, plus restaurants and cafes. To the best of our knowledge, the use of spatial configuration – or space syntax measures (Hillier, 1996; Hillier and Hanson, 1984) – on a multinomial panel model is unique. The most similar firm location models have been those of Sevtsuk (2014), while not being multinomial, and Mejia-Dorantes and Martín-Ramos (2013), while not using space syntax or centrality measures. Neither uses a panel specification.

As our main research hypothesis, we will test whether Choice and Integration can predict retail location persistently. Choice (betweenness) has proven a significant estimator of retail location (Buzzacchi et al., 2020; Sevtsuk, 2014) since it accounts for movement between any origin-destination pair of locations (or axial segments, in our case). Stores located in segments with a high value of choice (betweenness) benefit from the visibility of through traffic (Lin et al., 2018). Local integration (r=3) is heavily connected with pedestrian movement (Hillier, 1996). Considering the role of local businesses during the COVID-19 pandemic, testing the relationship between local integration and store location may provide relevant insights for urban design and the regulation of economic activity in years to come.

The following section presents the specific research framework of the paper, with a short literature review focused on space syntax measures and their relationship with retail location models while also adding some context on other explanatory variables and firm location models in general. This is followed by presenting the case study and data and the methodology and results. The paper closes with the conclusions and further research.

Research framework

Space Syntax was first introduced by Hillier and Hanson as a method to describe the built environment in a rigorous mathematical way (Hillier and Hanson, 1984). Integration, Connectivity and Choice are the three essential measures of the space syntax methodology. These measures are expected to be directly related to movement and, thus, to economic activity seeking to maximise the opportunities provided by movement (Hillier, 1996; Hillier et al., 1993). Summing up the concepts, Integration indicates the distance of any given space to all others in a system; Connectivity shows the number of nodes directly accessible from one node; and Choice (global) indicates the number of times a specific axial line lies on the shortest topological paths between any pair of axial lines in the system (Hillier and Hanson, 1984). Local measures of integration and choice can be obtained by defining a specific number of axial steps. Local integration of 3 axial steps (int r=3) is strongly related to pedestrian movement (Hillier, 1996). Additional information about the Space Syntax method is presented in the Supplementary Material.

After the space syntax framework was settled, Kim and Sohn (2002) are some of the first using space syntax measures to predict economic activity. A high correlation between space syntax indicators is found in two areas of Seoul, ultimately leading to using integration alone to predict office location (Kim and Sohn, 2002). This high correlation between space syntax measures has also been noted by other researchers (Berghauser Pont et al., 2019; Scoppa and Peponis, 2015). In Barcelona (Mora, 2003) and Mexico City (Ortiz-Chao and Hillier, 2007), both integration and connectivity are found to be significant predictors of the presence of economic activity. Assuming that integration and choice are equivalent to closeness and betweenness (Hillier and Iida, 2005), some analyses based on those measures can be brought in to support the discussion.

Hence, Sevtsuk (2014) finds that the relationship among the betweenness of a location and the presence of retail is significant for most store categories. Buzzachi et al. (2020) also find that betweenness is a significant predictor of retail location for convenience and shopping goods, although dependent on the trip’s purpose. Other researchers focus mainly on firm location, although similar conclusions can be drawn about retail (as a type of firm). For example, Ravulaparthy and Goulias (2016) find that betweenness can be related to retail location with a positive impact in smaller influence areas and a negative impact in larger ones. The conclusion is that a larger influence area will imply more competition (Ravulaparthy and Goulias, 2014, 2016).

Agglomeration effects were also introduced very early in the discussion about firm location (Marshall, 1920), considering that firms and businesses benefit from localisation economies (e.g. access to a skilled labour force and information spillovers). A different form of agglomeration comes from urbanisation economies since businesses locating close to one another to benefit, for example, from better infrastructures (Lösch, 1940). Agglomeration effects have also been emphasised in retail location models since stores tend to cluster, either for competitive purposes (Hotelling, 1929) or to benefit from information spillovers and multipurpose shopping (Caplin and Leahy, 1998; McLafferty and Ghosh, 1986). Stores selling goods that need to be compared or even ‘experienced’ before the purchase (comparison or ‘shopping’ goods (Copeland, 1923; Nelson, 1970)) will benefit more from locating in proximity, and hence (e.g.) antiques or clothing items tend to present a more noticeable clustering pattern (Krider and Putler, 2013).

Population density has also been a crucial factor for the location of firms and businesses, representing both the labour force (Dijk and Pellenbarg, 2000; Elgar et al., 2009; Weber, 1909) and the demand (Christaller, 1933; Lösch, 1940; Ravulaparthy and Goulias, 2016; Waddell et al., 2003). Using firm internal attributes (e.g. age, size, growth rate) as independent variables has also proven to be significant for estimating entire life-cycle location events, such as birth, death, growth and relocation (Van Wissen, 2000; Wever, 1983). Ultimately, variables that are hard to collect (proximity to the business owner’s residence) or more abstract (prestige of the location as perceived by the business owner) may ultimately explain a good deal of firm location (Elgar and Miller, 2010).

Therefore, we propose to model our research considering explanatory variables related to market area characteristics (e.g. population density, age of dwellings) along with different measures of accessibility and agglomeration to analyse the interaction between retail location, spatial configuration and, to some extent, land use. The framework is similar to the one used by Mejia-Dorantes et al. (2012) while focusing on different categories of retail instead of firms and using panel data (1995, 2002 and 2010) on commercial establishments to which all other variables (socio-demographics, spatial configuration) are related.

Case study and data

Case study and dependent variable

The analysis is carried out in Lisbon, using (point) data about stores’ locations and categories from a commercial census implemented by the municipality between 1995 and 2010 (Câmara Municipal de Lisboa, 2020). Axial data comes from two existing axial maps from 1994 and 2003 (Serdoura, 2006) and an axial map purposefully created for this analysis using spatial data from 2012. To make the panel, we used commercial activity data from 1995, 2002 and 2010 (Câmara Municipal De Lisboa, 2020) and population and housing Census data for 1991, 2001 and 2011 (INE, 2020). The final dataset contains 11.964 operating commercial establishments for 1995, 12.554 for 2002 and 13.216 for 2010. For simplicity, all data will be referred to as 1995, 2002 and 2010 data from this point on (even though, for example, commercial establishments data from 2010 is related with 2011 housing and population census data and 2012 space syntax data).

Considering the scope of research, we address only the stores located in the street (‘street commerce’ (Sevtsuk, 2020)) since shopping mall stores’ location dynamics is different. In shopping malls, footfall and clustering are the results of a store location that is designed to maximise profits for both tenants and the mall developer by the later (Carter and Vandell, 2006). Nevertheless, the presence of shopping malls in the vicinity of stores enters the models as an explanatory variable since the presence of a mall may impact stores located around it (Maoh, 2005).

As for the categories being considered, and while the nature of the product influences the choice of location, this nature can be defined as generalist as convenience and shopping (comparison) goods (Copeland, 1923; Nelson, 1970) or as thorough as using the three-digit NAICS (North American Industry Classification System) (Sevtsuk, 2014). Using fewer categories leads to a more marked distinction between the goods. For example, population density is found to be positively related to convenience goods and negatively to comparison goods by Buzzacchi et al. (2020). In comparison, Sevtsuk (2014) finds negative or non-significant relationships between proximity to population and all types of goods (except for miscellaneous).

The seven categories used in our models consider the goods’ nature and the frequency of purchases. The classification is adapted from the one used by the municipality in its commercial census (Câmara Municipal De Lisboa, 2020). The municipality’s classification is itself an adaptation of the official Portuguese Classification of Economic Activities (‘Classificação das Atividades Económicas’ (CAE)) (Institututo Nacional De Estatística, 2007), and thus, not unlike the NAICS classification. Although we assume there may be a less marked distinction between the goods than the one we would find using fewer categories, we expect that the insights about different goods’ location patterns will make up for this.

Hence, we considered stores to be grouped as Foodstuffs (supermarkets, bakeries, groceries and similar establishments), Personal Use Items (mainly clothing, clothing accessories and shoes), Household Articles (mostly furniture, home appliances and home decoration items), Leisure items (sporting goods, bookshops, music stores, etc.), Miscellaneous (all other retail items not included in the remaining categories), Restaurants (and similar establishments) and Cafes (and similar establishments). Store category poses as the dependent variable. The analysis is implemented within the municipality of Lisbon, that between 1995 and 2010, occupied an area of 100.5 km² (Direção Geral do Território, 2020).

Finally, since the commercial census (Câmara Municipal De Lisboa, 2020) did not include information about vacant stores, we were left with no obvious base category for our model. Hence, we defined non-operating stores as all stores that were ever part of the panel and that were empty at a specific year. A practical alternative would have been to use any category as the base outcome since it does not affect the validity of the results. However, it affects its interpretation since comparing a store (of any category) against a non-operating store says something about the location. A non-operating store is perceived as having less utility, at least by store owners, at a given time. Alternatively, comparing a store category against any another says ‘only’ that a store is perceived as having more (or less) utility for a specific purpose (category), which we found less interesting.

Therefore, non-operating stores were defined as mentioned, with exceptions being made for stores located in buildings that did not yet exist or no longer existed at a specific year, which entered the panel only after construction or were removed if already demolished. The implication is that a building where a store will open has some utility for commercial purposes, although less than other stores at that time. In the other way, keeping stores that are closed in the panel may correspond to stores that are indeed vacant. We found that 981 stores that were closed in 2002 had reopened in 2010. Moreover, we find that this approach may be closer to a real-life scenario than considering all buildings as having the potential to host a store, which is a common assumption in binomial models (Buzzacchi et al., 2020; Sevtsuk, 2014). Not all buildings can host a store due to regulations, or the buildings' characteristics, for example.

Spatial configuration, sociodemographic and other explanatory variables

To estimate our outcome, we considered seven explanatory variables dealing with spatial configuration (integration and choice) and commercial agglomeration, the presence of shopping malls and subway stations, population density, and density of buildings with less than a decade. The number of variables is minimal since using point data can amplify collinearity. We tested for collinearity using bivariate analysis to ensure that the correlation between our variables was minimal. The process of calculating the variables is described in the Supplementary Material, which also includes the Pearson’s correlation matrices for the independent variables at the beginning and end of our analysis (Table S1). Local Integration is presented in Figures S2, S3 and S4, and Choice is presented in Figures S5, S6 and S7, also in the Supplementary Material. The summary statistics of the explanatory variables is presented in Table 1.

Table 1.

Summary statistics of the explanatory variables.

Dependent variable
Commercial category			0	1	2	3	4	5	6	7
Label: CATG	Year	Total	Total	Total	Total	Total	Total	Total	Total	Total
	1995	16,284	4320	2154	1878	1180	1223	2385	1963	1181
	2002	16,292	3738	2048	1876	1248	1329	2473	2214	1366
	2010	16,929	3713	1739	1786	1085	1367	2674	2729	1836
Independent variables
Commercial category			0	1	2	3	4	5	6	7
		Year	Mean*	Mean*	Mean*	Mean*	Mean*	Mean*	Mean*	Mean*
Agglomeration effects (label: AGG)		1995	6.461 (6.95)	3.077 (2.82)	9.676 (11.47)	2.964 (3.18)	2.572 (3.18)	4.577 (5.43)	3.945 (4.58)	1.350 (1.38)
(#stores of the same type per block)		2002	5.898 (6.66)	2.902 (2.63)	10.803 (13.39)	3.045 (3.18)	2.868 (3.89)	4.835 (5.86)	4.175 (4.91)	1.449 (1.49)
(#stores of the same type per block)		2010	5.623 (5.35)	2.340 (2.23)	9.352 (10.85)	2.367 (2.59)	2.306 (2.64)	5.060 (6.12)	4.608 (5.66)	1.667 (1.65)
Local integration (r=3) (Label: INT3)		1995	2.403 (0.65)	2.251 (0.60)	2.701 (0.58)	2.603 (0.63)	2.551 (0.61)	2.468 (0.61)	2.431 (0.54)	2.408 (0.64)
		2002	2.472 (0.65)	2.259 (0.58)	2.700 (0.55)	2.599 (0.60)	2.570 (0.60)	2.475 (0.59)	2.424 (0.54)	2.394 (0.63)
		2010	2.438 (0.61)	2.263 (0.60)	2.734 (0.58)	2.583 (0.60)	2.555 (0.61)	2.495 (0.60)	2.407 (0.57)	2.350 (0.66)
Above average choice (label: CHCaa)		1995	0.288 (0.41)	0.20 (0.32)	0.44 (0.49)	0.31 (0.43)	0.36 (0.46)	0.29 (0.41)	0.285 (0.41)	0.24 (0.37)
(1 if the block is intersected by at least one axial segment with above average choice)		2002	0.297 (0.42)	0.193 (0.31)	0.451 (0.50)	0.310 (0.43)	0.370 (0.47)	0.296 (0.42)	0.285 (0.41)	0.239 (0.36)
		2010	0.252 (0.38)	0.204 (0.32)	0.460 (0.50)	0.301 (0.42)	0.331 (0.44)	0.309 (0.43)	0.289 (0.41)	0.238 (0.36)
Population density (label: POP)		1995	407.333 (268.31)	516.346 (263.18)	438.043 (253.37)	470.418 (244.52)	421.073 (260.06)	463.337 (266.34)	432.777 (263.38)	466.355 (267.64)
(#residents per block)		2002	350.873 (217.79)	416.278 (204.58)	348.410 (204.18)	379.827 (194.98)	329.779 (202.87)	367.047 (206.48)	338.744 (209.00)	379.470 (212.51)
(#residents per block)		2010	333.326 (181.98)	375.378 (192.71)	312.499 (181.71)	327.235 (166.57)	306.393 (180.77)	334.261 (187.75)	291.163 (184.65)	340.035 (188.80)
Percentage of new buildings (label: NEWBD)		1995	0.031 (0.05)	0.031 (0.05)	0.018 (0.03)	0.022 (0.03)	0.024 (0.04)	0.023 (0.03)	0.020 (0.03)	0.028 (0.04)
(#buildings <10yrs/100 per block)		2002	0.027 (0.04)	0.020 (0.03)	0.018 (0.02)	0.021 (0.03)	0.021 (0.02)	0.021 (0.03)	0.023 (0.03)	0.022 (0.03)
(#buildings <10yrs/100 per block)		2010	0.009 (0.01)	0.013 (0.03)	0.012 (0.03)	0.016 (0.04)	0.013 (0.03)	0.012 (0.03)	0.014 (0.03)	0.017 (0.04)
Proximity to subway (label: PSUBW)		1995	0.363 (0.46)	0.263 (0.39)	0.518 (0.50)	0.419 (0.49)	0.401 (0.48)	0.39 (0.48)	0.373 (0.47)	0.341 (0.45)
(1 if a subway station is located in the same neighbourhood)		2002	0.495 (0.50)	0.328 (0.44)	0.612 (0.47)	0.483 (0.50)	0.528 (0.50)	0.475 (0.50)	0.471 (0.50)	0.414 (0.49)
		2010	0.474 (0.50)	0.394 (0.48)	0.672 (0.44)	0.497 (0.50)	0.551 (0.49)	0.540 (0.50)	0.510 (0.50)	0.453 (0.50)
Proximity to shopping malls (label: PMALL)		1995	0.555 (0.49)	0.486 (0.50)	0.647 (0.46)	0.61 (0.47)	0.577 (0.49)	0.551 (0.49)	0.563 (0.49)	0.536 (0.50)
(1 if a shopping mall is located in the same neighbourhood)		2002	0.656 (0.45)	0.535 (0.50)	0.741 (0.38)	0.647 (0.46)	0.670 (0.44)	0.645 (0.46)	0.621 (0.47)	0.592 (0.48)
(1 if a shopping mall is located in the same neighbourhood)		2010	0.596 (0.48)	0.556 (0.49)	0.783 (0.34)	0.651 (0.45)	0.681 (0.43)	0.652 (0.45)	0.602 (0.48)	0.581 (0.49)

Note: 0 – non-operating stores, 1 – Foodstuffs, 2 – Personal Use Items, 3 – Household Articles, 4 – Leisure items, 5 – Miscellaneous, 6 – Restaurants, 7 – Cafes. *The number in parenthesis is the standard deviation.

Methodology and results

This research aims to estimate the relationship between spatial configuration and store location through 15 years and considering 16.929 commercial locations (amounting to an uneven panel of 49.505 individual observations). Moreover, we propose estimating store location considering five retail categories plus restaurants and cafes. Since our outcome is categorical and with no natural ordering, we find that a multinomial logit (MNL) will be appropriate for modelling the results. The MNL is used within the random utility framework (McFadden, 1973) that takes on the following formulation for a panel model (Grilli and Rampichini, 2007; Hartzel et al., 2001)

U_{i j t} = x_{i t} β_{j} + μ_{i j} + Є_{i j t}

where U_ijt is the utility given to the outcome j by the firm i at time t, the observed component of utility is x_itβ_j where x_it is a row vector of covariates and β_j is a column vector of coefficients for outcome j. The unobserved component of utility includes two error components: μ_ij to account for the panel level heterogeneity and Є_ijt, which is the observation level error term (Grilli and Rampichini, 2007; Hartzel et al., 2001). In this formulation, μ_ij can stand both for random and fixed effects. Our model considered random effects since more than half the stores do not change their category over time. Assuming a standard Gumbel distribution for Є_ijt and considering a base category j=1 (to allow for model identification), the logit probability for outcome j can be presented as follows

P_{j} = \frac{e β_{j}^{'} x}{1 + \sum_{j = 2}^{J} β_{j}^{'} x}

Our base category is non-operating stores (coded as 0 in our dataset). The model estimates are presented in Table 2. The odds-ratio Exp(B) is presented along the model coefficients B to help interpret the results. The model is significant at predicting the outcome (Wald test). The model’s likelihood-ratio test (LR) against an MNL with no effects shows that considering a specification with effects is also appropriate. The pseudo-R² of 0.058 is modest, although in line with the fit of other retail location models, eventually due to the very fine disaggregation of the categories (Sevtsuk, 2014).

Table 2.

Multinomial logit panel model. Parameter estimates by business category sector (retail, restaurants and cafes).

Variables
Dependent
CATG	1		2		3		4		5		6		7
Independent
	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)
constant	0.346*** (3.92)	1.413	−2.844*** (−27.65)	0.058	−2.134*** (−19.70)	0.118	−1.270*** (−12.51)	0.281	−0.520** (−6.03)	0.594	0.195** (2.26)	1.215	0.339 (3.53)	0.713
AGG	−0.184*** (−34.60)	0.832	0.011*** (4.08)	1.011	−0.213*** (−32.24)	0.808	−0.235*** (−35.46)	0.790	−0.068*** (−20.96)	0.935	−0.083*** (−23.74)	0.920	−0.460*** (−46.42)	0.632
INT3	−0.308*** (−9.17)	0.735	0.657*** (18.02)	1.928	0.516*** (13.16)	1.676	0.324*** (8.65)	1.382	0.108*** (3.40)	1.115	−0.049 (−1.52)	0.952	0.071** (1.96)	1.074
CHCaa	0.173** (3.26)	1.188	0.425*** (9.28)	1.529	0.288*** (5.40)	1.333	0.502*** (9.90)	1.652	0.208*** (4.79)	1.231	0.180*** (4.05)	1.197	0.298*** (5.68)	1.347
POP	0.002*** (21.64)	1.002	0.001*** (6.63)	1.001	0.001*** (12.79)	1.001	0.000*** (4.93)	1.000	0.001*** (9.07)	1.001	0.000 (−1.15)	1.000	0.001*** (15.66)	1.001
NEWBD	−4.249*** (−7.72)	0.014	−3.562*** (−5.41)	0.028	−3.223*** (−5.09)	0.040	−4.242*** (−6.65)	0.014	−3.958 (−7.69)	0.019	−4.462*** (−8.46)	0.012	−4.252*** (−7.57)	0.014
PSUB	0.084* (1.21)	1.087	0.099** (2.14)	1.104	0.284*** (5.60)	1.328	0.383*** (7.80)	1.467	0.245*** (5.89)	1.278	0.269*** (6.33)	1.308	0.444*** (9.25)	1.559
PMALL	−0.022 (−0.36)	0.978	0.232*** (5.13)	1.262	0.208*** (4.29)	1.231	0.196*** (4.15)	1.216	0.088** (2.24)	1.092	0.026 (0.65)	1.026	0.057 (1.25)	1.058
Pseudo-R²	0.058
Log-likel.	−93,197.42
Wald (χ²)	7616.46***
LR (χ²)	866.39***

Note: 0 – non-operating stores, 1 – Foodstuffs, 2 – Personal Use Items, 3 – Household Articles, 4 – Leisure items, 5 – Miscellaneous, 6 – Restaurants, 7 – Cafes. 0 - non-operating is the baseline. Number of observations = 49.505. Number of groups = 16.929. *p < 0.1, **p < 0.05, ***p < 0.01.

Agglomeration (AGG) is significant in predicting every category. Contrarily to models using binary outcomes (Buzzacchi et al., 2020; Sevtsuk, 2014). AGG is not always positive. This is because our baseline is non-operating stores, revealing a strong agglomeration (clustering) effect. Non-operating stores cluster more than every other type, except for Personal Use Items. Considering that non-operating stores are not yet open or are vacant, the model shows that the ‘supply’ of buildings with characteristics perceived as desirable (while being able to host retail) is very concentrated, which we find plausible while also being supported by the subsequent analysis.

As for Personal use items (mainly clothing, clothing accessories and shoes) clustering more than any other store (the only category to have a significant and positive relationship with AGG), this may be explained by these being goods that rely heavily on comparison and hence, on clustering. Sevtsuk (2014) also finds Clothing and Clothing Accessories Stores presenting a very high clustering effect, with Food Services and Drinking Places presenting the highest clustering coefficient in his models. We find that this does not disagree with our results since we are dealing with panel data from a period of changing shopping habits. Eating out became a form of entertainment and status display during this period (Warde and Martens, 2000). Businesses such as restaurants could eventually risk opening outside commercial districts while hoping to find business. This combination leads to a less clustered pattern of Restaurants in our model, although still very clustered. Restaurants are just 8% less likely to cluster than non-operating stores.

Considering spatial configuration variables, local integration (INT3) is significant at predicting every category except Restaurants and positive with all except for Foodstuffs and Restaurants. Foodstuffs include supermarkets, bakeries, groceries and similar establishments, which have been found not to reveal competitive clustering, eventually being ‘the most undifferentiated convenience goods’ (Sevtsuk, 2014). Foodstuffs locate in proximity to their consumers, in neighbourhoods with different spatial configurations. As for Restaurants, the explanation may be the same as presented before. These businesses became widespread in this period and started relying (less) on pedestrian movement. Eventually, some became destinations on their own, which is better predicted by choice (CHCaa).

Above Average Choice (CHCaa), as was defined, represents location on a block that is easily reachable coming from any point in the city. The variable is significant and positive with every category, although presenting a higher coefficient in categories that can be seen as destinations, such as Personal Use items and Leisure items. The positive effect on categories such as Foodstuffs or Cafes is less obvious, with the explanation being eventually found by looking at the remaining variables. Hence, proximity to a subway station (PSUB) is also positively and significantly related to store location for all categories. Proximity to shopping malls (PMALL) predicts positively and significantly all categories except Foodstuffs, Restaurants and Cafes.

Looking at the Pearson correlation matrix of independent variables in the Supplementary Material, we find that CHCaa, AGG, INT3, PSUB and PMALL were all significantly related in 1995 and 2010 (although with a correlation low enough not to compromise the models' results). This is because the subway network, although expanding, is still relatively small and, in 1995, covered essentially the city’s most central area. Afterwards, it expanded to areas that have, in general, an above average choice, a high integration, or both. Figures S2–S7 as for PMALL, the variable considers both the larger shopping malls that started to compete with the centre, and the smaller ones that were, to some extent, integrated into the general commercial structure (Colaço and De Abreu e Silva, 2021). Many of the latter were also located near streets with above average choice.

The generally positive and significant relationship between stores' location and CHCaa, , INT3, PSUB and PMALL ultimately suggest a commercial ‘structure’ of the city. This structure contains a ‘central area’, which in 1995 corresponded to the historic CBD and its expansion area (Cachinho, 1994), where the main avenues, the subway, the shopping malls, and the majority of street commerce located and hence, related. Since the relationship is found in a panel model, the results suggests that, even after 1995, street commerce has kept, to some extent, a similar relationship with these variables. smaller clusters can also be related to oneof these characteristics (location in areas CHCaa, INT3, PSUB and PMALL). The coefficients related to the percentage of new buildings (NEWBD) support the idea that stores are less willing to open in new areas, eventually because of information costs (Caplin and Leahy, 1998). Nevertheless, the model results show that these effects can be, to some extent, counteracted by the presence of one or all the remaining features (CHCaa, INT3, PSUB and PMALL).

Finally, the relationship with population density (POP) is significant but with a very low coefficient, suggesting that stores rely on residents only to some extent, in Lisbon. The threshold population (Berry and Garrison, 1958) of a store is eventually found outside a block, which is plausible.

Conclusions and further research

Retail activity is an essential feature of urban life. It creates lively streets and a sense of belonging (Jacobs, 1961; Salgueiro and Erkip, 2014) while promoting sustainable mobility and local employment (Sevtsuk, 2014). It is also an important sector of the economy, leading retailers to permanently devise new ways of approaching the consumers. From the ‘classic store formula’ to the omnichannel distribution (Belu and Marinoiu, 2014), the business model has suffered an intense change in the last decades, eventually since the introduction of the first shopping centres and big-box retail competing with town centre retail already in the 1960s (Fernandes and Chamusca, 2014). Although some predicted a ‘retail apocalypse’ (Helm et al., 2020), the local store may survive due to the opening of physical stores by digital-native brands (Dougherty, 2019), with the consumer-local store relationship having also been reinforced during COVID-19 (Gordon-Wilson, 2022; Musselwhite et al., 2020).

Nevertheless, although the city centre has declined in most European cities since the 1990s (Dolega and Lord, 2020), the centre still concentrates a good part of the activity due to the cost of creating information spillovers (Caplin and Leahy, 1998). These information spillovers have been mostly assumed by big box and large chain stores outside better-known commercial districts (Caplin and Leahy, 1998). And since retail location models have mostly been a ‘static snapshot of retailers’ distribution’ (Buzzacchi et al., 2020), the information about location choices of stores located outside malls has been limited to the insights deriving from those snapshots. Therefore, we set at analysing retail location (including also restaurants and cafes) using panel data and considering 15 years to try and capture the persistent effect spatial configuration (and accessibility) measures on retail location, which have not been given sufficient attention in location modelling.

Our model shows that Choice and Local Integration are significant predictors of commercial location through time. This is true both in the centre and in more peripheral areas. Nevertheless, a high percentage of new buildings (<10 years) impacts significantly and negatively the chances of a store being open in a specific location. This is in line with what was stated in the previous paragraph (being the first to locate comes at a cost). However, high values of choice and local integration, along with the proximity to a shopping mall or a subway station may counteract this effect. This has consequences for policymaking since policies related to urban design, the regulation of economic activities and the planning of transportation infrastructures can be explicitly designed to support the location of street commerce.

As a practical example, our choice variable (CHCaa), which is directly related to the ease of reaching a block, shows that some street segments may be essential for vehicular traffic since they are ultimately helping to connect the entire city. When the destination is reached, the car can be dropped, and the final path made on foot, which will be especially easy if the local integration is also high. We find that this idea can be further explored in commercial districts to address the everlasting car vs pedestrian debate. Moreover, this relates directly to the concept of neighbourhood units and local commerce that has recently been reintroduced in the 15-minute cities (Moreno et al., 2021).

Finally, we cannot hope that older, less educated and less affluent residents will all locate in dense commercial areas or be able to shop online (Beckers et al., 2018; Clarke et al., 2015; Colaço and De Abreu e Silva, 2022; Farag et al., 2006). Some may be excluded from the e-shopping revolution . Hence, we hope this paper contributes to informing urban planning and policymaking in a way that favours local commerce. Local commerce pays service to the population, which was more evident than ever during the COVID-19 pandemic. Good urban design and the strategic planning of transportation infrastructures may ultimately favour local businesses e in a very effective manner by creating externalities that will benefit street commerce in a permanent way. Drafting affordable housing policies to contravene the possible uptake in rents due to the provision of amenities and accessibility may also be necessary and is crucial to the debate about sustainability and the cities we want to live in.

Further research will address digital transformation impacts (e-shopping) on location choices. Qualitative assessments of small business owners’ decisions to locate may also add to the fit of location models (Elgar and Miller, 2010) within a mixed-methods framework. Access to more recent data will also lead to improved model versions and a better understanding of retail location phenomena in the face of technological change.

Supplemental Material

Exploring the role of accessibility in shaping retail location using space syntax measures: A panel-data analysis in Lisbon, 1995–2010

Supplemental Material for Exploring the role of accessibility in shaping retail location using space syntax measures: A panel-data analysis in Lisbon, 1995–2010 by Rui Colaço, João de Abreu e Silva in Environment and Planning B: Urban Analytics and City Science

Footnotes

Acknowledgements

The authors gratefully acknowledge Prof. Francisco Serdoura for providing axial maps for 1994 and 2003. The authors would like to thank Fundação para a Ciência e a Tecnologia (FCT) for Rui Colaço’s PhD grant SFRH/BD/136003/2018, as well as for providing financial support to the project PTDC/EGE-ECO/28805/2017 - Transport Infrastructure and Urban Spatial Structure: Economic, Social and Environmental Effects (TiTuSS).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This work also received financial support from FCT as part of the project PTDC/EGE-ECO/28805/2017-Transport Infrastructure and Urban Spatial Structure: Economic, Social and Environmental Effects (TiTuSS).

Data availability

The commercial establishments' location data (raw) is available from . Axial map data is available from the authors with the permission of Prof. Francisco Serdoura. The remaining data is available from the authors upon reasonable request.

ORCID iD

Rui Colaco

João de Abreu e Silva

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Rui Colaço is a PhD Candidate in Territorial Engineering at Instituto Superior Técnico (IST), Universidade de Lisboa, Portugal. His research (at CERIS) is presently focused on retail location, e-shopping, and urban planning. Before, he worked for 15 years in the private and public sectors as a civil engineer and urban planner. His background is also in Territorial Engineering at the IST. Since Fall 2022, he is also a Fulbrighter (UCSB).

João de Abreu e Silva is an Associate Professor at Instituto Superior Técnico (IST), Universidade de Lisboa, Portugal, and a researcher at CEris. His main research interests are focused on travel behavior and its interactions with land use patterns and the role of transport in urbanization patterns. He is a member of two TRB standing committees, and of the International Steering Committee for the Travel Survey Conference (ISCTSC). and chair of the World Society for Transport and Land Use Research (WSTLUR).

References

Batty

(2009) Accessibility: In search of a unified theory. Environment and Planning B: Planning and Design 36(2): 191–194. DOI: 10.1068/b3602ed

Beckers

Cárdenas

Verhetsel

(2018) Identifying the geography of online shopping adoption in Belgium. Journal of Retailing and Consumer Services, 45: 33–41. DOI: 10.1016/j.jretconser.2018.08.006.

Belu

Marinoiu

(2014) A New Distribution Strategy : The Omnichannel Strategy. Romanian Economic Journal 17(52): 117–133.

Berghauser Pont

Stavroulaki

Bobkova

, et al. (2019) The spatial distribution and frequency of street, plot and building types across five European cities. Environment and Planning B: Urban Analytics and City Science 46(7): 1226–1242. DOI: 10.1177/2399808319857450

Berry

BJL

Garrison

(1958) The Functional Bases of the Central Place Hierarchy. Economic Geography 34(2): 145. DOI: 10.2307/142299

Buzzacchi

Leveque

Taramino

(2020) Using betweenness metrics to investigate the geographical distribution of retailers. Environemnt and Planning B: Urban Analytics and City Science 0(0): 1–18. DOI: 10.1177/2399808320971303

Cachinho

(1994) O comércio a retalho na cidade de Lisboa. Lisboa, 29, pp. 119–144.Reestruturacão económica e dinâmicas territoriaisFinisterra57.

Câmara Municipal de Lisboa (2020) Geodados - Recenseamento Comercial. Available at: https://geodados.cm-lisboa.pt/datasets/6c40669f14c941779b48e2861767ead4_13 (accessed 11 September 2020).

Caplin

Leahy

(1998) Miracle on Sixth Avenue: Information externalities and search. Economic Journal 108(446): 60–74. DOI: 10.1111/1468-0297.00273

10.

Carter

Vandell

(2006) Store Location in Shopping Centers : Theory and Estimates Store Location in Shopping Centers. Journal of Real Estate Research 27(3): 237–266. DOI: 10.1080/10835547.2005.12091160

11.

Christaller

(1933) Central Places in Southern Germany (Translated by Carlisle W. Baskin in 1966. Englewood Cliffs (N.J.): Prentice-Hall.

12.

Clarke

Thompson

Birkin

(2015) The emerging geography of e-commerce in British retailing. Regional Studies, Regional Science 2(1): 371–391. DOI: 10.1080/21681376.2015.1054420

13.

Colaço

De Abreu e Silva

(2021) Commercial Classification and Location Modelling: Integrating Different Perspectives on Commercial Location and Structure. Land 10(6): 11.

14.

Colaço

De Abreu e Silva

(2022) Exploring the e-shopping geography of Lisbon: Assessing online shopping adoption for retail purchases and food deliveries using a 7-day shopping survey. Journal of Retailing and Consumer Services, 65: 11. DOI: 10.1016/j.jretconser.2021.102859

15.

Copeland

(1923) Relation of consumers’ buying habits to marketing methods. Harvard Business Review 1(3): 282–289.

16.

Dijk

J Van

Pellenbarg

(2000) Firm relocation decisions. The Netherlands 219: 191–219.

17.

Direção Geral do Território (2020) Carta Administrativa Oficial de Portugal. Available at: https://www.dgterritorio.gov.pt/cartografia/cartografia-tematica/caop (accessed 14 February 2021).

18.

Dolega

Lord

(2020) Exploring the geography of retail decline: A case study of the Liverpool City Region. Elsevier 96: 102456. DOI: 10.1016/j.cities.2019.102456.

19.

Dougherty

(2019) From Clicks to Bricks: The Impact of Digital-Native Consumer Brands on Retail Real Estate. Massachusetts: Massachusetts Institute of Technology. Available at: https://dspace.mit.edu/handle/1721.1/123613

20.

Elgar

Miller

(2010) How office firms conduct their location search process?: An analysis of a survey from the greater Toronto area. International Regional Science Review 33(1): 60–85. DOI: 10.1177/0160017609331398

21.

Elgar

Farooq

Miller

(2009) Modeling location decisions of office firms: Introducing anchor points and constructing choice sets in the model system. Transportation Research Record(2133): 56–63. DOI: 10.3141/2133-06

22.

Eurostat (2017) Which sector is the main employer in the EU Member States? Available at: https://ec.europa.eu/eurostat/web/products-eurostat-news/-/DDN-20171024-1 (accessed 25 May 2022).

23.

Farag

Weltevreden

van Rietbergen

, et al. (2006) E-shopping in the Netherlands: Does geography matter? Environment and Planning B: Planning and Design 33(1): 59–74. DOI: 10.1068/b31083

24.

Fei

Santini

(2020) Local food systems and COVID-19: an insight from China Resources, Conservation and Recycling. 162: 170–177. DOI: 10.1016/j.resconrec.2020.105022

25.

Fernandes

Chamusca

(2014) Urban policies, planning and retail resilience. Cities 36: 170–177. DOI: 10.1016/j.cities.2012.11.006

26.

Gordon-Wilson

(2022) Consumption practices during the COVID-19 crisis. International Journal of Consumer Studies 46(2): 575–588. DOI: 10.1111/ijcs.12701

27.

Grilli

Rampichini

(2007) A multilevel multinomial logit model for the analysis of graduates’ skills. Statistical Methods and Applications 16(3): 381–393. DOI: 10.1007/s10260-006-0039-z

28.

Hänninen

Kwan

Mitronen

(2021) From the store to omnichannel retail: looking back over three decades of research. International Review of Retail, Distribution and Consumer Research 31(1): 1–35. DOI: 10.1080/09593969.2020.1833961

29.

Hartzel

Agresti

Caffo

(2001) Multinomial logit random effects models. Statistical Modelling. 1(1): 81–102.

30.

Helm

Kim

Van Riper

(2020) Navigating the ‘retail apocalypse’: A framework of consumer evaluations of the new retail landscape. Journal of Retailing and Consumer Services, 54: 101683. DOI: 10.1016/j.jretconser.2018.09.015.

31.

Hillier

(1996) Space Is the Machine : A Configurational Theory of Architecture. Cambridge: Cambridge University Press.

32.

Hillier

Hanson

(1984) The Social Logic of Space. Cambridge: CambridgeshireCambridge University Press.

33.

Hillier

Iida

(2005) Network and Psychological Effects in Urban Movement. In: Cohn

A.G.

Mark

(eds), Spatial Information Theory. COSIT 2005. Lecture Notes in Computer Science. Berlin, Heidelberg: Springer, pp. 475–490.

34.

Hillier

Penn

Hanson

, et al. (1993) Natural movement: or, configuration and attraction in urban pedestrian movement. Environment & Planning B: Planning & Design 20(1): 29–66. DOI: 10.1068/b200029

35.

Hotelling

(1929) Stability in Competition. The Econometric Journal 39(153): 41–57.

36.

INE (2020) Censos - Página de download de informação geográfica. Available at: http://mapas.ine.pt/download/index2011.phtml (accessed 1 August 2020).

37.

Institututo Nacional de Estatística (2007) Classificação Portuguesa de Atividades Económicas. Available at: http://www.sicae.pt/Legislacao.aspx (accessed 17 August 2020).

38.

Jacobs

(1961) The Death and Life of Great American Cities. New York, NY, USA: Random House. DOI: 10.1002/9781119084679.ch4

39.

Kim

Sohn

(2002) An analysis of the relationship between land use density of office buildings and urban street configuration: Case studies of two areas in Seoul by space syntax analysis. Cities 19(6): 409–418. DOI: 10.1016/S0264-2751(02)00071-9

40.

Krider

Putler

(2013) Which birds of a feather flock together? Clustering and avoidance patterns of similar retail outlets. Geographical Analysis 45(2): 123–149. DOI: 10.1111/gean.12005

41.

Lin

Chen

Liang

(2018) The location of retail stores and street centrality in Guangzhou, China. Applied Geography. 100. 12–20. DOI: 10.1016/j.apgeog.2018.08.007

42.

Lösch

(1940) The Economics of Location. Woglom with the Assistance of Wolfgang F. Stopler in 1954. New Haven: Yale University Press. (Translated by William H .

43.

Maoh

(2005) Firm Demography in Urban Areas: An Agent-Based Approach. Hamilton, ON: McMaster University.

44.

Marshall

(1920) Principles of Economics : An Introductory Volume. London: Macmillan.

45.

Martí

Serrano-Estrada

Nolasco-Cirugeda

, et al. (2021) Revisiting the Spatial Definition of Neighborhood Boundaries: Functional Clusters versus Administrative Neighborhoods. Journal of Urban Technology (July). DOI: 10.1080/10630732.2021.1930837

46.

McFadden

(1973) Conditional Logit Analysis of Qualitative Choice Behavior. In: Zarembka

(ed), Frontiers in Econometrics. New York: Academic Press, pp. 105–142.

47.

McLafferty

Ghosh

(1986) Multipurpose Shopping and the Location of Retail Firms. Geographical Analysis 18(3): 215–226. DOI: 10.1111/j.1538-4632.1986.tb00094.x

48.

Mejia-Dorantes

Martín-Ramos

(2013) Mapping the firmographic mobility: A case study in a region of Madrid. Journal of Maps 9(1): 55–63. DOI: 10.1080/17445647.2013.778799

49.

Mejia-Dorantes

Paez

Vassallo

(2012) Transportation infrastructure impacts on firm location: The effect of a new metro line in the suburbs of Madrid. Elsevier Ltd, 22, p. 236. –250. DOI: 10.1016/j.jtrangeo.2011.09.006.Journal of Transport Geography.

50.

Mora

(2003) Land uses in the context of a perfect grid : the case of Barcelona. 4th International Space Syntax Symposium: 1–12. Available at: http://www.spacesyntax.net/symposia-archive/SSS4/shortpapers-posters/mora_barcelona.pdf

51.

Moreno

Allam

Chabaud

, et al. (2021) Introducing the “15-Minute City”: Sustainability, Resilience and Place Identity in Future Post-Pandemic Cities. Smart Cities 4(1): 93–111. DOI: 10.3390/smartcities4010006

52.

Musselwhite

Avineri

Susilo

(2020) The Coronavirus Disease COVID-19 and implications for transport and health. Journal of Transport and Health, 16. DOI: 10.1016/j.jth.2020.100853.

53.

Nelson

(1970) Information and Consumer Behavior. Journal of Political Economy 78(2): 311–329. DOI: 10.1086/259630

54.

Omer

Goldblatt

(2016) Spatial patterns of retail activity and street network structure in new and traditional Israeli cities. Urban Geography 37(4): 629–649. DOI: 10.1080/02723638.2015.1101258

55.

Ortiz-Chao

Hillier

(2007) In Search Of Patterns Of Land-Use In Mexico City Using Logistic Regression At The Plot Level. Proceedings of the 6th International Space Syntax Symposium. İstanbul.

56.

Pirenne

(1925) Medieval Cities. Princeton, NJ: Princeton University Press. Available at: http://www.jstor.org/stable/j.ctt6wpzw7

57.

Porta

Crucitti

Latora

(2006) The network analysis of urban streets: A primal approach. Pion Limited, 33, 705–725. DOI: 10.1068/b32045.Environment and Planning B: Planning and Design.

58.

Ravulaparthy

Goulias

(2014) Characterizing the Composition of Economic Activities in Central Locations. Transportation Research Record: Journal of the Transportation Research Board 2430(1): 95–104. DOI: 10.3141/2430-10

59.

Ravulaparthy

Goulias

(2016) Exploring the Role of Location in Dynamics of the Birth and Death of Business Establishments: Case Study Analysis of Santa Barbara County, California. Transportation Research Record: Journal of the Transportation Research Board 2564(1): 106–117. DOI: 10.3141/2564-12

60.

Sachdeva

Goel

(2015) Retail store environment and customer experience: a paradigm. Journal of Fashion Marketing and Management 19(3): 290–298. DOI: 10.1108/JFMM-03-2015-0021

61.

Salgueiro

Erkip

(2014) Retail planning and urban resilience - An introduction to the special issue. Cities 36: 107–111. DOI: 10.1016/j.cities.2013.01.007

62.

Scoppa

Peponis

(2015) Distributed Attraction: The Effects of Street Network Connectivity upon the Distribution of Retail Frontage in the City of Buenos Aires. Environment and Planning B: Planning and Design. Online ahead of print. DOI: 10.1068/b130051p.

63.

Serdoura

(2006) Espaço Público, Vida Pública: o caso do Parque das Nações. Instituto Superior Técnico.

64.

Sevtsuk

(2014) Location and Agglomeration: The Distribution of Retail and Food Businesses in Dense Urban Environments. Journal of Planning Education and Research, 34(4). 374–393. DOI: 10.1177/0739456X14550401.

65.

Sevtsuk

(2020) Street Commerce: Creating Vibrant Urban Sidewalks. Philadelphia: University of Pennsylvania Press.

66.

Smith

(1776) The Wealth of Nations. 2015th ed.. New York: Cambridge University Press.

67.

Statista (2022) E-commerce share of total retail sales. Available at: https://www.statista.com/statistics/534123/e-commerce-share-of-retail-sales-worldwide/(accessed 28 March 2022).

68.

U.S. Census Bureau (2022) U.S. Retail Sales Top $5,570 Billion. Available at: https://www.census.gov/newsroom/press-releases/2022/annual-retail-trade-survey.html (accessed 25 May 2022).

69.

Van Wissen

(2000) A micro-simulation model of firms: Applications of concepts of the demography of the firm. Papers in Regional Science 79(2): 111–134. DOI: 10.1007/s101100050039

70.

Von Thunen

(1826) The Isolated State. Oxford, London: Pergamon Press. Translated by Carla M. Wartenberg and Peter Hallon in 1966 .

71.

Waddell

Borning

Noth

, et al. (2003) Microsimulation of Urban Development and Location Choices: Design and Implementation of UrbanSim. Networks and Spatial Economics 3: 43–67. DOI: 10.1023/A:1022049000877

72.

Warde

Martens

(2000) Eating Out Social Differentiation, Consumption and Pleasure. Cambridge University Press.

73.

Weber

(1909) Theory of the Location of Industries. Chicago: University of Chicago Press.

74.

Wever

(1983) Cohort analysis in Economic Geography. Journal of Economic and Social Geography 74(3): 217–223.

75.

Zambrano-Rey

López-Ospina

Pérez

(2019) Retail store location and pricing within a competitive environment using constrained multinomial logit. Applied Mathematical Modelling 75: 521–534. DOI: 10.1016/j.apm.2019.05.040

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.50 MB

Exploring the role of accessibility in shaping retail location using space syntax measures: A panel-data analysis in Lisbon,1995–2010

Abstract

Keywords

Introduction

Research framework

Case study and data

Case study and dependent variable

Spatial configuration, sociodemographic and other explanatory variables

Methodology and results

Conclusions and further research

Supplemental Material

Exploring the role of accessibility in shaping retail location using space syntax measures: A panel-data analysis in Lisbon, 1995–2010

Footnotes

Acknowledgements

Declaration of conflicting interests

Funding

Data availability

ORCID iD

Supplemental Material

Author Biographies

References

Supplementary Material