asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a...

47
Location choice determinants of new and relocated firms. Does accessibility still matter? Ioannis Baraklianos : [email protected] Louafi Bouzouina : [email protected] Ouassim Manout : [email protected] Patrick Bonnel : [email protected] Location choice determinants of new and relocated firms. Does accessibility still matter? Ioannis Baraklianos*, LAET, ENTPE, Vaulx-en-Velin, France Louafi Bouzouina, LAET, ENTPE, Vaulx-en-Velin, France Ouassim Manout, LAET, ENTPE, Vaulx-en-Velin, France Patrick Bonnel, LAET, ENTPE, Vaulx-en-Velin, France * corresponding author Contact : [email protected] Abstract The theory of the location choice of economic establishments consider accessibility one of the most essential determinants together with location externalities like agglomeration and urbanization effects. Nevertheless, modern advances (behavioral and structural changes of firms and transportation system, Information and Communication Technologies) are making the task of identifying the influence of accessibility challenging. In this paper we are willing to measure this effect that accessibility has for a firm location choice. Using descriptive statistics and modelling technics (Multinomial Logit) we are trying to quantify the effect of accessibility for firms created and relocated during the period 1

Transcript of asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a...

Page 1: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Location choice determinants of new and relocated firms. Does accessibility still matter?Ioannis Baraklianos : [email protected] Bouzouina : [email protected] Manout : [email protected] Bonnel : [email protected]

Location choice determinants of new and relocated firms. Does

accessibility still matter?Ioannis Baraklianos*, LAET, ENTPE, Vaulx-en-Velin, France

Louafi Bouzouina, LAET, ENTPE, Vaulx-en-Velin, France

Ouassim Manout, LAET, ENTPE, Vaulx-en-Velin, France

Patrick Bonnel, LAET, ENTPE, Vaulx-en-Velin, France

* corresponding author

Contact : [email protected]

Abstract

The theory of the location choice of economic establishments consider accessibility one of the most

essential determinants together with location externalities like agglomeration and urbanization

effects. Nevertheless, modern advances (behavioral and structural changes of firms and

transportation system, Information and Communication Technologies) are making the task of

identifying the influence of accessibility challenging. In this paper we are willing to measure this

effect that accessibility has for a firm location choice. Using descriptive statistics and modelling

technics (Multinomial Logit) we are trying to quantify the effect of accessibility for firms created and

relocated during the period 2005-2011 in the Lyon urban area in France. The results are showing that

there is a different appreciation of different dimensions of accessibility depending on the firm event

(creation or relocation) or on the economic activity of the firm.

Keywords

accessibility, location choice model, firms, Multinomial logit, Lyon urban area

Introduction

The importance of accessibility has been highlighted at a theoretical level in the very first works on

location choice determinants of economic activities. The bid-rent theory, developed by the works of

1

Page 2: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Von Thünen (1842) and extended by Alonso (1964), Mills (1967) and Muth (1969), reveals the role of

accessibility on the spatial distribution of economic establishments, using the most simple form, an

Euclidean distance to CDB. Also, the theory of the location choice of economic activities highlights

transportation infrastructure and accessibility as a traditional explanatory location attribute along

with other externalities like agglomeration (Marshall, 1890) and urbanisation effects(Jacobs, 1969).

Highly accessible areas with well-developed transportation infrastructure can potentially minimise

the transportation costs for suppliers (input), distribution (output), labour (production factor) and

clients (profit) (Maroto and Zofío, 2016). In that way, it can create cost efficiencies and can be

considered as a positive attribute of a location (de Bok and Oort, 2011).

Nevertheless, its influence on the location choices of firms is not easy to grasp, notably for two

reasons. First, accessibility is a multidimensional concept. It depends at the same time from the

transportation network, the spatial distribution of people and firms, the specific preferences of

businesses and individuals and the time constraints (Geurs and van Wee, 2004). But in the most of

empirical approaches, it is considered as a more simplified concept, like an Euclidean distance to CBD

without considering the influence of transport time (Melo et al., 2016) or like a proximity to transport

infrastructures. In that way, some of its’ dimensions are being omitted, possibly leading to partial

findings. Second, the modern advances are redefining the role of transport infrastructure and

accessibility for the location choices of economic establishments. The increased development and

performance of the transportation sector, have shifted the balance between a location choice and

the importance of accessibility. In developed countries, transportation networks are more

widespread and they are no longer concentrated to some principal axes. At the same time, they have

become more productive (Bodenmann, 2011). Those two factors have given more flexibility to the

location choices of firms, while creating polarisations around transportation hubs and motorway

junctions (Mérenne-Schoumaker, 2011). Moreover, Information and Communication Technologies

(ICT) have decreased the importance of the friction of distance (cost, time or effort to travel)

(Muhammad et al., 2008), allowed the firms to separate their activity in different establishments in

different locations based on a functional division (Ota and Fujita, 1993) and permitted the economic

establishments to be disconnected from the direct proximity with the clients or even with their

employees (telework) (Aguilera et al., 2016). If in the past, the firm had to be somewhere visible to

become well-known, nowadays is not the only way. Despite the fact that these factors seem to

decrease the importance of accessibility, evidence highlights that it stays a factor, but its importance

depends on the economic sector of the establishment and other firm specific characteristics (Arauzo-

Carod et al., 2010; de Bok and Oort, 2011; Muhammad et al., 2008).

2

Page 3: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

In such a context, searching whether accessibility is still important for a firm location choice is

essential. From an academic point of view, the role of accessibility in such processes and under those

constant changes rests still uncertain, from a methodological and an empirical point of view (Arauzo-

Carod, 2013; Geurs and van Wee, 2004). From a transport policy perspective, the clarification of the

role of the accessibility can facilitate the policy decision making and the design of programmes that

aim to attract new establishments. Furthermore, knowing about how firms appreciate accessibility

can have a contribution to the evaluation of new transportation infrastructure. In this paper, we are

trying to clarify and quantify the role of accessibility for a location choice of firms by taking into

account the most possible dimensions that are usually omitted in studies. More specifically are

questioning: What are the preferences for accessibility between different economic sectors? The

potential accessibility or the proximity to transport infrastructure is more important for a location

choice of a firm? New and relocated firms have different behaviour?

In order to respond to these questions, we have developed an empirical application for the Lyon

urban area in France. We are using descriptive statistics to trace general trends of creations and

relocations of firms during the period 2005-2011 by economic sector, to reveal any differences in

creation or relocation rates or differences in location choice patterns. Then, we are applying discrete

choice modelling to analyse the location choice behaviour of new and relocated firms in order to

quantify the sensitivity to accessibility in relation to other location attributes. In particular, we are

using a multinomial logit model where we have introduced location attributes like accessibility,

proximity to transportation infrastructures, economic and social environment as independent

variables.

This paper is aiming to contribute to the existing empirical literature in many levels. First, we are

comparing the location choice behaviour, focused on the impact of accessibility, between economic

sectors and between new and relocated firms, which is not thoroughly studied in the literature. We

are searching to quantify any differences deriving from these specific firm characteristics in the

location choices of firms with a particular attention to accessibility. Second, we are using accessibility

as a major explanatory variable and considering the most possible dimensions of accessibility

(transport infrastructure, land use distribution, individual dimension). Most studies in the literature

consider only some aspects of accessibility. Third, the analysis is made at neighbourhood level in the

limits of the urban area, which is not the norm in such studies due to data availability at a such

desegregated level (Bodenmann, 2011). Most studies are examining location choices at country level

(Graham, 2007; Holl, 2004; Maroto and Zofío, 2016) or even international level (Graham, 2007;

Siedschlag et al., 2013). Examining at a such level one can evaluate better the location attributes

driving firm location choices in fine since aggregating approaches can disguise possible

3

Page 4: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

heterogeneities even at the same metropolitan area (Beaudry and Schiffauerova, 2009; Dubé et al.,

2016). Forth, our study area is not a large metropolitan area like the region of Paris (Buczkowska and

de Lapparent, 2014; de Palma et al., 2008; Le Nechet et al., 2012; Padeiro, 2013), but a smaller

medium size. Most of the literature has concentrated on large metropolitan areas but the decision

making behaviour of firms can be different in smaller areas like ours. Evidence from this article can

contribute to the empirical literature of location choice determinants of economic establishments.

The paper is structured as following. Section 2 focuses on the role of accessibility from the point of

view of a firm location choice. Section 3 presents the study area, the urban area of Lyon and section

4 the data used in our analysis. Section 5 presents in detail the applied method and presents the

different variables and their measures. Sections 6 and 7 presents the descriptive statistics and the

modelling results respectively and section 8 summarises the findings along with the conclusions and

the perspectives of the paper.

1. The role and the dimensions of accessibility for a firm location choice

One location attribute traditionally considered to have a significant impact to the location choice of

firms, other than the agglomeration and urbanisation externalities, is the accessibility. Location

accessibility for firms can be defined by 4 dimensions, (i) the transport system, (ii) the spatial

distribution of land-use, (iii) the individual dimension and (iv) the temporal dimension (Geurs and van

Wee, 2004). These dimensions from the point of view of a firm, influence the location choice decision

at the same time:

2. Transport network have always had an important role for the location choices of firms. Even

at the beginning of the industrial revolution, industries were looking to be located near

railway stations or rivers (Mérenne-Schoumaker, 2011). Today, proximity to transportation

infrastructure like motorways or public transportation seems that it is something that

entrepreneurs take into account when they take decisions for a location choice (Mejia-

Dorantes et al., 2012a). This is because proximity to such infrastructures can increase the

potential clients and can facilitate the access for workers and other associate firms.

3. However, this potential interaction is conditioned by the relative spatial distribution of these

different agents. For agents who are distant from one another, the potential interaction is

low. Therefore, the spatial distribution of land-use is important. Especially for firms we can

distinguish 4 different levels relevant to their economic activity; the industry, the suppliers,

the labour and the client level. The industry level concerns the spatial distribution of own

industry firms or industries on other economic sectors. In this case, we can understand that

accessibility is closely related to agglomeration and urbanisation effects (de Bok and Oort,

4

Page 5: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

2011; Melo et al., 2016) and studies have integrated the impact of transportation to such

interactions (Melo et al., 2016). The supplier level, even though it can be somehow related to

the industry level, concerns the actual suppliers of the firm, and real interaction (not

potential) is needed. If suppliers are far away, this can cause increases in costs and time that

can impact efficiency and profit directly. Next, the labour level concerns the spatial

distribution of the active working population potentially available for the firm. Easy access to

pool of workers can increase the possibility of recruiting and can decrease commuting costs

and possible problems like absenteeism (Mérenne-Schoumaker, 2011). Finally, the client

level concerns the spatial distribution of potential clients which can be individuals or other

firms. These clients should be able to visit the firm, if the firm offers a service in its premises,

or the firm should be able to offer its services at their the client’s place if the firm offers a

service by distance. So, relative proximity between clients and firms is essential but its’

importance can vary depending on the activity of the establishment.

4. The individual dimension concerns individual preferences and abilities. For firms this

dimension can have 2 perspectives; the internal perspective from the point of view of the

firm and the external perspective which concerns all other agents external to the firm. The

internal perspective influences the ability of the firm to attract labour, clients or suppliers.

This ability depends on the characteristics of the firm like the size, the age and the economic

sector of the firm. For example, a firm in traditional manufacturing sector would need fewer

skilled workers than a firm in Finance and Insurance services. However, these internal

characteristics should be matched with the characteristics of the agents external to the firm,

the external perspective of the individual dimension of accessibility (Martín-Barroso et al.,

2017). For example, if high skilled workers are far away, an Insurance firm would have a

problem recruiting them. These external characteristics apply not only to workers, but also to

clients and suppliers and can influence the potential relation with the firm (Martín-Barroso et

al., 2017). Last, an aspect which should be considered in the individual dimension is the

competition between firms and these different agents (Geurs and van Wee, 2004). Firms

whose activities are in the same economic sector would potentially compete for a work force

with the same abilities. However, to account properly for competition effects, the study area

should be a closed area, meaning that there are almost no interactions with outside locations

(Bunel and Tovar, 2014).

5. The same 2 perspectives, internal and external, can be found on the last dimension of

accessibility, the temporal dimension. Internally, firm has its own working hours and

constrains. Externally, workers, clients and suppliers have their own temporal constrains.

5

Page 6: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Additionally, as an external factor the transportation system can impose as well temporal

constrains (congestion, time table of public transportation).

Figure 1 : The dimensions of accessibility for a firm, adapted from Geurs et Wee (2004)

From the description of the dimensions we can understand that accessibility can be considered as a

positive location attribute. It creates cost efficiencies for firms (de Bok and Oort, 2011) and increases

the potential for face-to-face interaction between different agents (Bentlage et al., 2013; Holl, 2012).

It facilitates the connexion and decreases costs, time, risk and uncertainty for suppliers, customers

and stuff of the firm (Leitham et al., 2000). It can also increase the potential market access helping

firms to be more specialised and to exploit better the economies of scale (Holl, 2012; Maroto and

Zofío, 2016). In that sense, high accessible areas are ideal especially for new firms, to develop a

network, meaning the relations with customers, suppliers and labour. When a new firm is entering

into a new area, it is not known to the clients and it doesn’t have a well-established suppliers’ and

distribution network. Thus, in order to minimise the cost and the risk of the creation of this “eco-

system”, new firms should in theory be more sensitive to accessibility. On the other hand, firms who

migrate can potentially have a smaller sensibility to the accessibility of the location. A relocated firm,

who stays in the same geographical area, has already developed a network. This fact can give a

flexibility to the firm to migrate to a less expensive location keeping the same relations with the “eco-

system”. However, this dependency to the pre-existing network poses a restriction, which is the

relation with the previous location. Firms want to keep the already developed network of clients and

suppliers and they want that their employees can keep their commuting habits. But in any case, even

the relocated firms should account for good accessible areas.

During the last years, this importance of accessibility seems to be shifting. Urban areas have faced

important mutations because of the dispersion of economic activities (Mejia-Dorantes et al., 2012b;

6

Page 7: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Mérenne-Schoumaker, 2011). Firms are avoiding high priced central areas and are searching for

locations at the periphery where rents are lower so they can have more margin for profit. Marshall

(1890) have described this process of the firms leaving high priced areas limiting their margin of

profit to explain the emergence of sub centres (Boiteux-Orain and Huriot, 2002). Moreover, central

areas are plagued by congestion problems (Sweet, 2014). The congestion of high accessible areas can

in fact neutralise the positive effects of accessibility and create location diseconomies. When these

diseconomies lead to a critical decrease of profits of the firm, the firm is willing to take to decision to

locate somewhere else where the activity can be profitable (Van Dijk and Pellenbarg, 2000).

Other than these push factors who are encouraging firm to leave central high accessible areas, there

can be identified some other causes. The phenomenon of the dispersion of economic activities is not

independent to changes of the transportation sector. Transportation was the accelerator of the rapid

suburbanisation of households and activities (Boiteux-Orain and Huriot, 2002; Czamanski and

Broitman, 2017) after the second Word War and it seems that is always a factor (Mejia-Dorantes et

al., 2012a). Transports have become more efficient. Innovations and technologies like telematics

have helped to optimise the capacity of the existing infrastructure (Graham and Marvin, 1996) and to

increase the travel speed (Padeiro, 2013). These changes have decreased the cost of transportation

for goods and people, a traditional location choice factor (Boiteux-Orain and Huriot, 2002) and have

given more flexibility to firms when they are choosing a location. At the same time, this flexibility

leads to polarisation of activities around transport hubs and junctions of the networks (Mérenne-

Schoumaker, 2011). Additionally, due to advances of ICT, the transmission of information is almost

costless (Ioannides et al., 2008). Taking advantage of this cost minimisation, firms are decentralising

completely their activities or only specific functions of their production process which have a more

routine character in order to decrease their expenses (for land and salaries), the so called “back

officing” of routine functions (Ota and Fujita, 1993). Moreover, businesses can sell and buy services

or products without being in close proximity with their clients (Czamanski and Broitman, 2017). Last,

firms can recruit employees who are living thousand miles away, the so called teleworking or

telecommuting.

2. Study area: Lyon urban area

The study area is the Lyon urban area which is the second largest urban area in France after Île-de-

France (the Paris region) in economic and population terms. In 2011, the urban area1 had a

population of 1.8 million people. The Gross Domestic Product of the metropolitan area in 2011 was

72,754 million euros (Eurostat), which places the urban area among the 25 top European

1 Limits of the urban area of 1999

7

Page 8: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

metropolitan regions in terms of total gross production. In total, the urban area had more than

900,000 jobs in 2011, of which more than 43% were concentrated in the area’s central municipalities

(Lyon-Villeurbanne) and almost 77% inside so called “Greater Lyon”, which is made up of the city of

Lyon and some suburbs. It is a dynamic area which have increased the number of jobs by almost 13%

and the number firms by almost 17% during the period 2005 – 2011. Concerning the economic

establishments, in 2011, the urban area had 142,500 establishments (self-employed excluded), of

which 72% are located in Grand Lyon area and 42% in Lyon-Villeurbanne (see figure 2). Another

attribute of the city is the polarisation between the poor east, where the low income working class is

mostly living, and the central and western areas concentrate the middle and upper classes.

Figure 2 : The density of firms in 2011 – presentation in quantiles

Despite the deindustrialisation process of the latest years, Lyon stays one of the most industrialised

cities of France (Carpenter and Verhage, 2014). Nevertheless, its economy has a more tertiary role

which is reinforced during the latest years. This diversity and strength of the local economy places

the city between the most dynamic European metropolitan areas of this size like Cologne, Turin,

Dublin, Helsinki etc. Evidence from this article can help understanding the behaviour of firm in such

urban contexts which can differ from the large European metropolitan areas like Paris (Buczkowska

and de Lapparent, 2014; de Palma et al., 2008; Le Nechet et al., 2012; Padeiro, 2013) or other

American cities (Melo et al., 2016; Sweet, 2014) on which the research is mostly focused.

3. Data

To apply our methodology, we combined different data bases from various sources.

8

Page 9: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

This work is principally based on the register of economic establishments (SIRENE database) which is

a disaggregated database that contains all the companies in France and it is provided by the INSEE

(Institut National de la Statistique et des Etudes Economiques - French National Institute of Statistics

and Economics Studies). We used this database for two time periods, the analysis year of 2011 and

the comparison year of 2005. The use of the same database in two time periods permits us to

identify the firms created or relocated during this period. For methodological reasons, we are

focusing only on firms with one establishment in the study area in 2011. We are using only the firms

with one establishment because it is a consistent method to distinguish between creations and

relocations. Each observation has a unique code for establishment and for firm. However, when an

establishment changes location, the unique code of establishment changes, while the unique code of

the firm stays the same (INSEE, 2016). So, if for example a firm has one establishment A in 2005 and

two A and B in 2011, but A has a different establishment code, we are not able to distinguish

between the new and the relocated one. A disadvantage of this method is the non-identification of

the inbound firms, which are considered as newly created. However, it is expected that the number

of such establishments is small and given that it is the first time for these firms to be located to this

area, their behaviour is expected to be close to the newly created firms. The advantage of this

method is that it can be applied to any time period for which there is available data.

The SIRENE database contains many information for each economic establishment like the economic

activity, the location of the firm, the size in number of employees etc. In order to group the firms into

economic sectors we have used the classification of the INSEE into 20 groups based on the economic

activity of the firm. However, this categorisation can create heterogeneities in some groups due to

the differences in location choice behaviour and thus misspecifications of the model. To bypass this

issue, we have recreated some of the groups using a bottom-up approach based on the activity of the

firm defining Front Office services the firms who need a face-to-face interaction and Back Office

services the firms who can provide their services from distance (Ota and Fujita, 1993), while we have

excluded some types from the analysis. In the table below we are presenting the categorisation of

the INSEE and the retained grouping of this article.

Table 1 : Classification of economic establishments by INSEE and modifications

Classification of INSEE ModificationsAgriculture -

Extraction industries -Manufacturing -

Production and distribution of electricity, gas, etc. Grouped to back office servicesProduction and distribution of water Grouped to back office services

Construction -Wholesale and retail activities Divided to retail and wholesale

Transports and storage Divided to back office (the majority) and front office

9

Page 10: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

servicesAccommodation and restaurant services -

Information et communication Divided to back office and front office services (the majority)

Finance and insurance -Real Estate -

Specialised activities, technical and scientific Divided to back office and front office services (the majority)

Services and activities of support and administration Divided to back office and front office servicesPublic Administration -

Education -Health -

Arts and recreative activities -Other activities and services -

Extra-territorial activities -

Other than the SIRENE database, for the estimation of the models we have mobilised other data

sources as well. For the calculation of the accessibility indicators, we used generalised times by car

and public transportation and the number of jobs or the population. The generalised times by private

vehicle and public transport along with the parameters for the estimation of the accessibility

indicators were calculated by a transportation model developed in LAET (Laboratoire, Aménagement,

Economie, Transport - Transport, Urban Planning, Economics Laboratory) for the Lyon urban area.

For the calibration of the model, the data of the household travel survey of 2015 was used. Even

though there might be some changes between 2011 and 2015, especially for public transportation,

they are considered to be marginal in terms of travel times (some new tram stations) and thus this

data is applicable to our case. The number of jobs per zone was calculated using the SIRENE

database. The other accessibility indicators, like the proximity to transport infrastructure, were

calculated using Geographic Information Systems.

Last, the calculation of the agglomeration and urbanisation effects was based on the SIRENE

database as well. For these indicators we have included all the stock of the establishments, even the

firms with multiple sites in the area of study. For the characterisation of the social environment we

have used the FILOSOFI database of INSEE for the year 2012, which gives the distribution of the

available revenues of each zone in deciles.

4. Modelling the location choices of firms: Model specification, determinants and measures

One of the most fundamental concept of the location choice theory of firms, is that a firm is choosing

the location which maximises its profits or minimises its expenses. In this framework, a firm is

evaluating all the available location possibilities (perfect information) and then makes the optimal

choice (Arauzo-Carod et al., 2010; Holl, 2004). Even though they seem unrealistic assumptions, and

10

Page 11: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

the firms act more as satisfiers than maximisers (Elgar et al., 2015), the framework of utility

maximisation is appealing from a theoretical and computational perspective. In order to model such

choices, the use of discrete choice modelling seems justified given the assumption that the firm is

choosing the location that maximises its utility. For each firm i, the utility U of location j is

conditioned by an observable part V ij and an unobservable stochastic term ε ij:

U ij=V ij+εij

The probability that a firm i chooses a location j is equal to the probability that the utility of this

location has the largest utility of all the other alternative locations Cn. The observable utility is given

by V ij=∑j=1

Cn

β X ij where Xij is a vector of characteristics of the firm and the attributes of the locations

(Siedschlag et al., 2013). Making the assumption that the error term ε ij is independently and

identically distributed (IID) with type 1 extreme value distribution (McFadden, 1977), the probability

of choosing a location takes the logit form:

P( j∨1…Cn)= eβ Xij

∑j=1

n

β X ij

Besides all the critics of the Multinomial logit model, especially for modelling spatial choices, it stays

an attractive method due to the ease of computation and to no a priori assumptions on the

correlation between alternatives, like the Nested Logit. In cases where we have many alternatives,

like in spatial modelling, it is possible to estimate the probability using a random sampling of

alternatives Dn for each firm (McFadden, 1977). This is the case if our study. The area is divided to

432 zones. In order to estimate the model, after testing for different random samples, we have

picked randomly 12 alternatives for each observation including the observed chosen zone.

In the developed model, the focus is the accessibility variables. However, in order to have a

consistent model, there is a need to integrate other location attributes which the location theory of

firms highlights. The selected locational attributes can be classified in three groups: economic

environment, accessibility and proximity to transport infrastructure, and social environment.

4.1 Accessibility and proximity to transport infrastructure

In order to measure the accessibility, we have selected two types of measures, the proximity to

transportation infrastructure which captures the effect of transportation and the potential indicator

which combines the ease to travel with the distribution of economic activities. It can be also adapted

to include individual preferences or even competition effects.

11

Page 12: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

The potential accessibility measures the potential employment or population potentially available for

a firm. A general form of the measure for origin and destination locations i and j respectively to

measure the accessibility for population P and travel time by a mode of transport t, with parameter β

to be estimated using local data:

A j=∑iPi e

−β tij (1)

The parameter β is estimated by local data of the trip behaviour of individuals and reveals the effect

of the time on the probability to make a trip. When we have multiple transport mode serving an

area, one should consider aggregating between modes in order to calculate a combined accessibility.

Usually, the aggregation is performed at the impedance, whereas in our case is the time. So, for the

calculation of the accessibility we are using a composite generalised time (Bhat et al., 1999). The

composite generalised time Tcij for every pair of origin i destination j is given by the equation 2,

where y tc is a dummy variable which takes value 1 if the zone is served by public transportation and

0 if not, Tvpij is the generalised time for the private vehicle, and Ttcij is the generalised time for the

public transport. This form of the composite generalised time has the advantage that if two areas A

and B have the same levels of private car accessibility but A has good public transportation service,

the accessibility for A would be higher because potentially, area A offers more mobility solutions.

Tcij=(1− y tc )Tvpij+ y tc(Tvpij

1+Tvp ijTtcij

) (2)

For the estimation of the model, we have estimated the accessibility to general population, as a

proxy for the potential market. This means that firms who need face to face contact should be more

sensible to accessibility. However, in general, accessibility should be considered as a positive location

attribute in all cases.

The transportation infrastructures considered in this study are the stations of public transportation

(metro, tram, railway) and the motorway. The proximity to these infrastructures is measured as a

binomial variable which takes the value 1 when this type of infrastructure is present into the

alternative zone. We have not used a continuous measure, like the distance to the motorway,

because we want to capture only the local effect of the infrastructure. Additionally, the potential

accessibility captures any sensitivity beyond the proximity.

Last, many studies have highlighted the importance of centrality of the location (Dubé et al., 2016;

Elgar et al., 2009). In our case, in order to capture this preference for central areas we have

introduced dummy variables. We have divided the area in 5 areas where we have: (i) the central area

12

Page 13: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

composed by the municipalities of Lyon and Villeurbanne, (ii) the eastern surrounding areas which

are considered as areas with low skilled workers, (iii) the western surrounding areas which are

considered as areas with high skilled workers, (iv) the 2nd suburban belt and (v) the 3rd suburban

belt. In that way, we are capturing not only the preference for the central areas, but the preferences,

if any, between those different zones.

4.2 Economic environment

Location externalities or external economies seem to be the most undeniable determinant for a

location choice of a firm, highlighted by the neoclassical approach (Hayter, 1997). They arise when

firms use other establishments as resources to their own productivity and from which a firm benefits

without any direct financial exchange (Mérenne-Schoumaker, 2011). These location externalities can

be divided in two different types, the agglomeration and the urbanisation effects (Glaeser et al.,

1992). As Marshal (1890) has pointed out, the agglomeration or specialisation externalities emerge

from the concentration of an economic sector to a specific geographical area. It is considered as a

positive location externality because proximity between firms can favour the labour market pooling,

input/output sharing and knowledge spill over (de Bok and Oort, 2011). In that sense, it increases the

performance of firms and reduces the risk for the implementation of new ones. In empirical

applications, agglomeration effects are measured either by using the location quotient by economic

sector or by the density of employment or firms (Beaudry and Schiffauerova, 2009). In our case, after

testing for all possible formulations, instead of using the density of employment, we have used the

density of firms by location and by sector.

The urbanisation or diversity externalities (Jacobs, 1969) are the result of the concentration of

diverse economic sectors into a geographic area. Literature has not concluded if it has a positive or a

negative influence on the location choice of a firm. It seems that its influence depends on the

characteristics of each specific industry (Rosenthal and Strange, 2003). There are economic sectors

which value more the diversity and the density of a location while others are searching for more

specialised locations. The urbanisation effects can be measured by the employment density, the Gini

coefficient or the Hirschman-Herfindahl index (HHI) (Beaudry and Schiffauerova, 2009). Since it is

difficult to capture these effects, we have opted for 2 measures, the employment density and the

HHI, modifying the latter as 1-HHI in order to have more intuitive results.

4.3 Social environment

Other than the accessibility and the economic environment of the location, we have included also

the social environment of the location. Studies are not including social environment variables very

13

Page 14: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

often into the analysis. However, we are arguing that they can influence the location choice of a firm.

Firms who offer high quality services and they need face-to-face contact are expected to be located

to areas where the revenues of the households are high (Elgar et al., 2009). Additionally, it is

expected that firms should in general avoid areas why low-income households due to any possible

social problems that can hurt the productivity of the firm. Also, high income households are

attractive from a market potential point of view. We are taking into account the effect of the social

environment by introducing into the model the percentages of the population belonging to the 1st

quantile (the poorest) and the 5th quantile (the richest) of the revenue of the area. Some sectors

should be positive to the 5th quantile while all sectors should be negatively influenced by the 1st

quantile.

4.4 Other considerations

More recent research have shown that these aforementioned factors depend on the individual

characteristics of the firms (Arauzo-Carod et al., 2010). These characteristics are internal to the firms,

like the economic sector, the size or the age of the firm (Van Dijk and Pellenbarg, 2000). In our case,

in order to capture the effect of the economic sector of the firm, we have divided the firms in 10

economic sectors; manufacturing, construction, wholesale, retail, accommodation and restaurant

services, finance and insurance, real estate, front office services2, back office services3 and health. As

an individual firm characteristic we have tested the size of the establishment. Empirical analysis has

shown that small firms are more prone to the owner’s choices and preferences like home location

(Elgar et al., 2009), while larger firms are driven by more “objective” economic factors (Arauzo-Carod

et al., 2010). Nevertheless, small and young firms can be more sensitive to location externalities

while larger firms can use their own scale advantage to their profit (de Bok and Oort, 2011). In our

study, we didn’t find any significant impact on the appreciation of accessibility between different firm

sizes, so we didn’t include any relative variables.

Another attribute that influences location choices of is the stage of life of the firm. New firms who

chose for the first time a location have different behaviour than the relocated ones (Elgar et al.,

2009). Newly created establishments can be more sensible to agglomeration effects in order to

minimise the related risks while relocated firms rely on their already developed network (clients,

labour, suppliers). In that sense, relocated firms are dependent to the previous location and it must

be included into the analysis. This is why, for the relocated firms, we are including the distance to the

previous location. However, firms can benefit from their already developed network and relocate to

a less accessible less expensive area.

2 Business services that need face to face interaction3 Business services that can carry out their activity by distance

14

Page 15: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

5. Descriptive statistics analysis

Before proceeding to location choice modelling it is useful to make a descriptive analysis of the data

in order to characterise our dataset. For the analysis, we are using the data from the SIRENE

database in aggregated form, which is enhanced by some estimations like the relocation distance or

the accessibility of locations.

Table 2 : Selected firms for the analysis by sector (In green the sectors included into the analysis)

Sector Total establishments Share Number of firms with only one establishment

Share of firms with only one establishment

Agriculture 5332 4% - -Extraction industries 72 0% - -

Manufacturing 7684 5% 6402 83%Construction 13292 9% 12359 93%

Wholesale 8902 6% 7689 86%Retail 14193 10% 10586 75%

Accommodation and restaurant services 7391 5% 5770 78%Finance and insurance 7451 5% 5822 78%

Real estate 5330 4% 4614 87%Front Office Services 20819 15% 18802 90%Back Office Services 9238 6% 7522 81%Public administration 2748 2% - -

Education 3910 3% - -Health 14110 10% 10515 75%

Art and recreation activities 5659 4% - -Other activities and services 15114 11% - -

Extra-territorial activities 21 0% - -

First, we present the data that we have included to our analysis. Sectors which we have not included

to our analysis are the agriculture, extraction industry, public administration, education, art and

recreation activities and other activities non-classified. The reason we have excluded these categories

is because they don’t follow a traditional economic reasoning to their location choices (public

administration and education), they have a totally different economic reasoning (agriculture, art and

recreation activities and other non-classified activities), or they have very few establishments

(extraction industries). These excluded activities account for 24% of the total number of economic

establishments of the study area. Additionally, as we mentioned, our analysis is focused only on firms

with one establishment into the study area. So, we have not included establishments of firms which

have more than one into the study area. The excluded establishments depend on the activity sector

(table 1). The share of firms with one establishment depends on the economic sector and it varies

between 75% (for retail and health) and 93% for construction.

Table 3 : Newly created firms by activity sector

Sector Created after 2005

Rate of new establishments by

sectorManufacturing 2270 36%Construction 6611 54%

Wholesale 3713 48%Retail 5202 49%

Accommodation and restauration services 3045 53%

Finance and insurance 3079 53%

15

Page 16: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Real estate 2695 59%Front Office Services 9075 48%Back Office Services 3878 52%

Health 3586 34%

Manufacturing

Construction

Wholesale

Retail

Accommodation and restauration services

Finance and insurance

Real estate

Front Office Services

Back Office Services

Health

0% 20% 40% 60% 80% 100% 120%

Share of firms to 25% most accessible areas Share of firms to 25% - 50% most accessible areasShare of firms to 50% - 75% most accessible areas Share of firms to 75% - 100% most accessible areas

Figure 3: Preference of new firms for accessibility to population

Between the analysed activity sectors, we can observe that the majority of the firms are services,

with front office services have the highest share with 21%. This shows the tertiary character of the

local economic activity. However, the creation of new establishments differs between the activity

sectors. Real estate and construction activities seem to be the more dynamic sectors of the urban

area. These two sectors can be related to the one another. On the contrary, Manufacturing and

Health seem to be the less dynamic sectors with less creations than the other analysed sectors. Also,

we can observe a strong variation between the mean values of accessibilities between the sectors.

The sectors who are choosing the most accessible areas are the Front Office services and the

Accommodation and Restauration services. By definition, those two sectors depend on face to face

contact and the new establishments in these sectors follow an expected behaviour. On the other

end, Construction and Back Office services seem to be the less sensitive to the accessibility of the

location, which are economic sectors who are not dependent to face to face contact.

Table 4 : Relocated firms by activity sector and migration distance

CategoryRelocations

during 2005-2011

Relocation rate

Mean relocation distance in km

Manufacturing 693 11% 8.25Construction 1375 11% 8.33

Wholesale 984 13% 8.29Retail 764 7% 8.57

Accommodation and restauration services 242 4% 7.61

Finance and insurance 647 11% 6.92

Real estate 511 11% 6.93

16

Page 17: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Front Office Services 2722 14% 5.43Back Office Services 904 12% 8.28

Health 1559 15% 6.59

Manufacturing

Construction

Wholesale

Retail

Accommodation and restauration services

Finance and insurance

Real estate

Front Office Services

Back Office Services

Health

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Share of firms to 25% most accessible areas Share of firms to 25% - 50% most accessible areas

Share of firms to 50% - 75% most accessible areas Share of firms to 75% - 100% most accessible areas

Figure 4 : Preference of relocated firms for accessibility to population

Concerning the relocations of the firms, we can observe that in general the relocation rate during the

analysed period is around 11%. As a general observation, we can say that even though firms do not

migrate in long distances, the mean migration distance varies based on the economic sector, with

some sectors being more sensitive that other. More specifically, taking for example two sectors, the

Front office services and Wholesale, they have both relatively high rates of relocation with 14%-15%

but they have different relocation behaviour. Front office services are looking for areas with good

accessibility not very far from the previous location (in 5,43 km in average). It is the most sensible

sector to the distance of the previous location and in terms of accessibility they are searching the

most accessible in average in comparison to the other analysed sectors, with 59% of the relocated

establishments choosing an area which belongs to the 1st quartile of the most accessible areas.

Wholesale on the contrary, is moving to locations which are in average 8,33 km from the previous

location and in terms of accessibility it is not a very demanding sector.

Comparing the location choices of relocated firms with the newly created firms, notably the share of

firms who are choosing the most accessible areas (tables 3 and 4), we can observe that the relocated

firms choose in average areas which have lower levels of accessibility. This is an indication that in fact

firms who relocate have other criteria when they select a location, confirming possible the

assumption that they are staying in the same area more or less for network reasons, but they are

moving to peripheral areas to decrease their costs. However, this observation is not true for some

17

Page 18: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

sectors, like the Front Office services, the Finance and Insurance and the Real Estate. These sectors

share a common characteristic; they offer services to clients and personal contact is very important.

The same observation can be made from the figure 5 as well. Here, we are searching any differences

in the location preferences of firms for different areas between the new and the relocated firms with

a more qualitative approach than before. Rather than using the accessibility levels, we have divided

the area in 5 large zones in order to capture any polarising effects between east and west. New

establishments prefer mostly the centre of the urban area. On the contrary, the relocated

establishments prefer less than the average the central areas. Depending on the sector, they prefer

other surrounding areas. For example, Wholesale relocated establishments prefer the east

surrounding areas or the 2nd suburban belt while Real estate prefers the west surrounding areas.

This observation can be related to the differences between the socio-economic characteristics

between those areas and the different sensitivity of economic sectors to these characteristics. Last

moved firms of Front Office services prefer central areas, confirming the findings in the previous

tables.

Figure 5: Zone preference for new and relocated firms

6. Modelling results

From the descriptive statistics analysis, where we have focused on accessibility preferences, we have

found that there are not only differences between economic sectors but also between the new and

relocated firms. In this chapter, using a logit model, we are quantifying this effect for all the analysed

economic sectors. We have estimated ten models (for each economic sector) for the newly created

firms, using as explanatory variables the measures presented earlier. For the relocated firms, we

have estimated a series of models using the same explanatory variables as for the newly created, to

18

Page 19: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

see if there are any differences between those two types of firms, and a series where we have

included the distance to the last location. The reason is to quantify this effect of the distance to the

last location which literature highlights (de Bok and Oort, 2011; Elgar et al., 2009; Van Dijk and

Pellenbarg, 2000) and to analyse any emerging trade-offs. The tables 6, 7 and 8, are summing the

results of the models. We are presenting only the significance of the estimators and the marginal

effects. For the estimation of the marginal effects, we have estimated the utility difference for each

variable between the 95th and the 5th percentile, keeping all other variables to their mean values

(50th percentile) (Lee et al., 2010). Next, we are giving the most noticeable results by variable

groups. We have to note that concerning the relocated establishments, for some sectors, especially

for Accommodation and Restauration, we do not have many observations. So the results are not very

reliable.

As a general remark, we should mention that the addition of the distance to the previous location for

the relocated firms, increases the quality of the model dramatically, reaching to values of the ρ2 from

around 0.1-0.25 to 0.4-0.5. This results reveals the importance of the last location for the explication

of the behaviour of relocated firms.

6.1 Accessibility and proximity to transport infrastructure

Given that accessibility and transport infrastructure is in the centre of our analysis, we are starting by

analysing the results of this group of variables. Beginning with accessibility to population, we can see

that in general it has a positive influence, especially for newly created firms. However, there are

important differences between economic sectors and between firm even (creation or relocation).

Focusing on the differences between economic activities, we can observe that for some sectors (Back

Office and Manufacturing), even for new firms, accessibility is not significant. This result can be

related to the fact that we have used accessibility to population, and not to employment.

Nevertheless, this observation confirms the work of Ota and Fujita (1993) on the decentralisation of

firms with back office character, even though in their theoretic model they are considering only back

office activities of one firm with multiple sites. On the contrary, for the other sectors of economic

activity the accessibility has a positive effect but the real influence tends to vary considerably

between them. For the newly created establishments, for sectors like Retail and Accommodation and

Restauration, accessibility is essential, and has the highest influence between all other variables.

These sectors value potentially crowded locations because they are dependent to the local market

and the flow of people, which can be potential customers. The other sectors seem to appreciate as

well accessibility but not to this extend. Nevertheless, these results are not valid for relocated firms.

When distance to the last location is not included, accessibility is for the most sectors not significant,

19

Page 20: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

meaning that it is not a decisive location attribute, except for the Retail and the Health firms. Even

for those two sectors, the marginal effect of accessibility is smaller for the relocated firms comparing

to the newly created ones. Interestingly enough, when we include the variable of the distance to the

previous location, the estimator of accessibility becomes negative and significant. The negative and

not intuitive effect of accessibility can be caused by omitted variables or by a real negative influence.

However, other studies have found similar results. Elgar et al. (2009) found for Toronto and for

relocated office firms that when they include only distance to CDB (a proxy for accessibility) the

influence is negative but when they add the distance to the last location the influence of the distance

to the CBD becomes positive. Another study from De Bok and Oort (2011) found in the South Holland

(a Dutch region) that accessibility to employment for relocated firms has not a significant effect for

the new location choices except for the business and general services and the transport and

distribution, even after accounting for the distance to the previous location. It seems that for

relocating firms, distance to the previous location dominates the decision of the new (de Bok and

Oort, 2011; Elgar et al., 2009; Sweet, 2014), due to possibly risk aversion of the firm (Van Dijk and

Pellenbarg, 2000). A location away for the last location where the firm have already developed its

“eco-system” involves some sort of risks especially for the mobility habits of clients, labour and

suppliers.

Passing now to the influence of the transportation infrastructure, and more precisely to the

motorway, we can see that the presence of a motorway increases the probability that a firm chooses

a location. The results are almost identical between the new and the relocated firms even after

accounting for the distance to the last location. Only Accommodation and Restauration and Real

Estate don’t seem sensitive to proximity to motorways. In the Paris region, Buczkowska and de

Lapparent (2014) found also a positive effect of the proximity almost for all economic sectors.

Passing to the proximity to public transportation, the results seem to be less clear. Metro stations,

for Accommodation and Restauration has a positive influence, while for the rest of the sectors has a

negative, while there are some sectors for which proximity is not significant. This is strictly related to

the character of the economic activity of the firm. Proximity to tram stations and even more

proximity to train stations seem to have in general a positive influence. This effect confirms the

positive influence of such public transportation infrastructures, highlighted in the literature (de Bok

and Oort, 2011; Elgar et al., 2009). Globally, Accommodation and Restauration seems to be the most

sensitive to accessibility, because getting there especially without a car is essential for the economic

activity. Our results are in agreement with related research, which have shown that in fact transport

infrastructure can increase the potential attractiveness of a location for all economic sectors, but

20

Page 21: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

especially for business services (Bodenmann and Axhausen, 2012; Buczkowska and de Lapparent,

2014; de Bok and Oort, 2011) like in our case the Finance and Insurance and Front Office services.

Last, concerning the preference for centrality or other locations of the study area, it seems that in

fact, firms have different behaviours. We have to note that the omitted area is the centre, so results

concerning all the other areas, should be interpreted in relation to the centre. In general, the results

are confirming the observations of the other accessibility variables. Sectors, which the economic

activity is based on face to face contact, prefer to locate in the central area and alternative locations

have negative influence. On the contrary, activities which don’t rely on personal contact, like Back

Office services, avoid central and prefer mostly peripheral areas. In any case, the suburban areas far

from the economic centre have negative effect for all economic sectors, with a very strong influence.

6.2 Economic environment

In order to characterise the economic environment, we have used to measure two effects, the

agglomeration and the urbanisation effects. In accordance with the theory, results are showing that

agglomeration effects have always a positive significant effect. Even for the firms who migrate, with

or without accounting for the distance to the previous location, agglomeration effects seem to play

an important role for the location choice. An exception is the Retail for which agglomeration effects

are not significant. Le Nechet et al. (2012) for Paris also found the smaller for agglomeration effects

for the Wholesale-Retail sector. In general, this result confirms the theory that there are positive

externalities when firms of the same sector are grouped together and in general firms are seeking

locations with high density of firms of the same industry. The real effect of agglomeration varies

between the industries and between new and relocated firms. An emerging pattern in that Front

Office services are very sensible, with the relocated firms being even more. On the other hand,

Accommodation and Restauration are less sensitive, and from relocated Retail firms, agglomeration

has no significant effect. Last, one would expect that migrating firms should value more

agglomeration effects and less diversity. As Duranton and Puga (2001) point out, migrating firms are

searching for specialised areas to take advantage of the location externalities and to avoid

competition with other activity sectors (Holl, 2004). However, there is no pattern between the new

and the relocated firms confirming that. This result can be caused by the absence of variables on

local policies which can influence the concentration of industries.

Concerning the diversity effects, they also seem to have a positive influence for all the analysed

economic sectors. For the new firms, the HHI index is always positive and significant. For the

relocated ones, the results of the HHI index are more or less the same, with some non-significant

estimators. On the other hand the employment density gives more unclear results. However,

21

Page 22: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

generally we can observe that firms who are seeking economically diverse areas are the ones who

are not very sensitive to agglomeration effects. Back Office services for example, are not searching

for diverse while being sensitive to agglomeration. On the contrary, Wholesale and Retail seem to

have the opposite behaviour. Once again, we were not able to find any patterns between the new

and relocated firms in terms of sensitivity to diversity.

6.3 Social environment

The impact of the local social environment is reflected by the variables of the percentages of the

population belonging to the 1st and the 5th quantile of revenue. We can observe that in general

firms avoid locations where there is a high percentage of low revenue population. The results are

more or less the same between the new and the relocated firms, with the new being more reluctant

to locations with low revenue. This result confirms our initial assumption that firms would in general

avoid areas with high percentage of low revenue population, with an exception of the migrating firms

from health industry. Concerning the areas with high levels of high income population, we can see

that firms have different behaviour based on the economic sector. Firms which offer services

appreciate neighbourhoods with high income population (Insurance and Finance, Real Estate, Front

Office services). This can be related to the image that the firm wants to promote. Also high revenue

population has higher purchase power and firms can have higher margins of profit. On the other

hand all other firms are negatively influenced by the high revenue locations, possibly due to high land

prices. Our results are different from Elgar et al. (2009) who found that Health is positively influence

by high revenue and other services negatively.

22

Page 23: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Table 5 : Significance and marginal effects of the newly created firms

Manufacturing Construction Wholesale Retail Accommodation & Restaurant

Finance & Insurance Real Estate Front Office

servicesBack Office

services Health

Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E.

v_dens_same_sector *** 0.73 *** 0.82 *** 0.38 *** 0.49 *** 0.40 *** 0.77 *** 0.45 *** 0.86 *** 0.64 *** 0.80

v_dens_emp - -0.03 - 0.03 - 0.05 *** 0.13 *** 0.17 - 0.06 *** 0.19 * 0.05 - -0.04 - -0.03

v_diversity HHI index *** 0.36 *** 0.33 *** 0.50 *** 0.47 *** 0.36 *** 0.42 *** 0.45 *** 0.45 *** 0.23 ** 0.15

v_acc_pop _2015 * 0.47 - 0.21 *** 0.71 *** 1.61 *** 2.19 ** 0.62 ** 0.69 *** 0.90 - 0.30 *** 1.31

v_motorway *** 0.39 *** 0.32 *** 0.32 - 0.07 - 0.02 *** 0.42 *** 0.38 *** 0.24 *** 0.41 - 0.01

v_metro - -0.03 *** -0.21 - -0.01 - -0.01 *** 0.26 *** -0.26 ** -0.20 *** -0.13 ** -0.14 * -0.11

v_tram - 0.15 *** 0.22 *** 0.21 - 0.00 - -0.01 *** 0.24 - 0.12 *** 0.28 ** 0.14 *** 0.23

v_rail_station * 0.13 *** 0.18 *** 0.25 *** 0.21 *** 0.39 *** 0.34 *** 0.21 *** 0.23 *** 0.17 - -0.04

v_zones5_cour_east ** 0.29 *** 0.25 *** 0.28 *** 0.34 - 0.08 - 0.05 *** -0.44 *** -0.46 ** 0.21 - 0.06

v_zones5_cour_west * 0.21 *** 0.25 ** 0.22 *** 0.31 - -0.04 - 0.13 ** 0.29 - 0.05 - 0.14 *** 0.54

v_zones5_2cour *** 0.50 ** 0.26 - 0.14 *** 0.39 * 0.27 ** -0.36 ** -0.37 *** -0.52 - 0.12 ** 0.36

v_zones5_3cour * -0.36 *** -0.51 *** -0.94 - -0.23 - 0.10 *** -1.30 *** -1.24 *** -1.35 *** -0.78 *** -0.56

%_q1 *** -0.27 - -0.03 *** -0.38 * -0.12 ** -0.24 *** -0.37 *** -0.49 *** -0.17 - -0.07 ** -0.20

%_q5 *** -0.82 *** -0.71 ** -0.21 *** -0.56 *** -0.64 ** 0.27 ** 0.29 *** 0.39 *** -0.76 *** -0.54

Observations 2270 6611 3713 5202 3045 3079 2695 9075 3878 3586Likelihood zero -5601 -16302 -9155 -12848 -7508 -7592 -6648 -22396 -9564 -8846Log of Likelihood -5180 -14979 -8131 -11063 -6037 -6168 -5449 -17282 -8737 -7550Adjusted ρ2 0.078 0.082 0.113 0.140 0.198 0.189 0.182 0.229 0.088 0.148*** Significant at 99%, **Significant at 95%, *Significant at 90%, - not significant

23

Page 24: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Table 6 : Significance and marginal effects of the relocated firms

Manufacturing Construction Wholesale Retail Accommodation & Restaurant

Finance & Insurance Real Estate Front Office

servicesBack Office

services Health

Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E.

v_dens_same_sector *** 0.60 *** 0.46 *** 0.66 - 0.17 * 0.32 *** 0.89 *** 0.48 *** 1.15 *** 1.04 *** 0.96

v_dens_emp - -0.06 - 0.03 - 0.09 - 0.04 *** 0.47 - 0.12 ** 0.29 ** 0.12 * -0.21 - -0.05

v_diversity HHI index - 0.22 *** 0.63 *** 0.64 ** 0.36 * 0.56 - 0.25 *** 0.57 *** 0.54 - 0.05 * 0.19

v_acc_pop _2015 - -0.20 - 0.03 - 0.16 ** 1.24 - -0.05 - 0.75 - 0.85 - 0.44 - -0.42 ** 0.73

v_motorway *** 0.68 *** 0.38 *** 0.85 *** 0.36 - -0.02 *** 0.58 - 0.22 *** 0.55 *** 0.49 - -0.08

v_metro - -0.06 ** -0.32 - -0.24 - 0.15 - 0.23 - -0.22 - -0.22 - -0.12 * -0.29 ** -0.23

v_tram - -0.11 - -0.08 - 0.23 ** -0.43 - -0.15 - 0.01 - -0.02 *** 0.32 - 0.07 *** 0.38

v_rail_station ** 0.28 ** 0.24 - 0.10 * 0.21 - 0.15 *** 0.52 - 0.19 *** 0.49 - 0.07 - 0.05

v_zones5_cour_east *** 0.78 *** 0.64 *** 0.61 - 0.26 - -0.44 - 0.22 - -0.38 *** -0.39 *** 0.51 - 0.04

v_zones5_cour_west - 0.25 - 0.23 *** 0.52 - 0.28 - -0.13 * 0.38 - 0.22 - -0.03 - 0.12 *** 0.68

v_zones5_2cour - 0.42 ** 0.53 * 0.40 - 0.02 - -0.74 - -0.21 - -0.35 *** -0.98 - 0.18 - 0.25

v_zones5_3cour ** -0.95 *** -0.75 *** -1.26 * -0.65 ** -1.68 *** -1.44 ** -1.27 *** -2.31 *** -1.30 *** -0.90

%_q1 *** -0.51 *** -0.48 *** -0.82 - -0.13 - -0.45 *** -0.73 - -0.08 ** -0.24 *** -0.46 ** 0.26

%_q5 *** -0.80 *** -0.97 *** -0.68 ** -0.48 * -0.65 - 0.01 *** 0.82 ** 0.26 *** -0.60 - -0.11

Observations 693 1375 984 764 242 647 511 2722 904 1559Likelihood zero -1712 -3391 -2427 -1885 -596 -1592 -1264 -6713 -2230 -3846Log of Likelihood -1550 -3137 -2065 -1692 -495 -1227 -971 -4583 -2020 -3221Adjusted ρ2 0.103 0.079 0.155 0.110 0.193 0.238 0.243 0.319 0.100 0.165*** Significant at 99%, **Significant at 95%, *Significant at 90%, - not significant

24

Page 25: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Table 7: Significance and marginal effects of the relocated firms – distance to the last location

Manufacturing Construction Wholesale Retail Accommodation & Restaurant

Finance & Insurance Real Estate Front Office

servicesBack Office

services Health

Signif. M.E. Signif. M.E. Signif. M.E. Signif

. M.E. Signif. M.E. Signif

. M.E. Signif. M.E. Signif. M.E. Signif. M.E. Signif. M.E.

v_dens_same_sector *** 0.54 ** 0.40 *** 0.73 - 0.16 - 0.27 *** 0.93 *** 0.58 *** 1.14 *** 0.96 *** 0.83

v_dens_emp - -0.10 - 0.10 - 0.10 - 0.08 ** 0.49 - 0.09 * 0.25 * 0.11 - -0.15 - -0.04

v_diversity HHI index - 0.19 *** 0.63 *** 0.69 ** 0.32 * 0.52 - 0.26 *** 0.61 *** 0.50 - 0.06 - -0.01

v_acc_pop _2015 *** -2.68 *** -2.03 *** -1.88 - -0.82 ** -1.91 ** -1.52 * -1.23 *** -1.61 *** -2.37 ** -0.96

v_motorway *** 0.55 *** 0.38 *** 0.97 ** 0.28 - 0.02 *** 0.64 - 0.24 *** 0.55 *** 0.48 - -0.18

v_metro - 0.28 - -0.16 - -0.10 * 0.30 - 0.45 - -0.07 - -0.03 - 0.01 - -0.11 - 0.00

v_tram - -0.08 - -0.09 - 0.24 ** -0.35 - -0.11 - 0.13 - 0.19 *** 0.40 - 0.11 ** 0.31

v_rail_station *** 0.45 *** 0.42 - 0.18 * 0.24 - 0.31 *** 0.64 - 0.19 *** 0.54 - 0.19 - 0.04

v_zones5_cour_east ** 0.59 ** 0.43 *** 0.55 - 0.24 - -0.29 - 0.19 - -0.42 - -0.08 * 0.35 - 0.06

v_zones5_cour_west - -0.06 - -0.05 * 0.32 - 0.14 - -0.11 - 0.17 - 0.11 - -0.15 - -0.18 *** 0.52

v_zones5_2cour - 0.15 * 0.45 ** 0.61 - 0.18 - -0.52 - -0.13 - -0.37 ** -0.47 - 0.24 ** 0.56

v_zones5_3cour - -0.05 - -0.05 - -0.19 - 0.28 - -0.51 * -0.77 - -0.80 *** -0.72 - -0.22 - 0.44

%_q1 ** -0.52 *** -0.45 *** -0.78 - -0.22 - -0.30 ** -0.60 - 0.10 *** -0.33 *** -0.47 * 0.24

%_q5 ** -0.60 *** -1.08 *** -0.71 ** -0.50 - -0.41 - 0.24 *** 0.88 - 0.17 ** -0.52 - 0.03Distance to the previous location (km) *** -8.99 *** -9.23 *** -8.17 *** -8.45 *** -7.23 *** -8.44 *** -8.56 *** -9.39 *** -8.73 *** -12.30

Observations 693 1375 984 764 242 647 511 2722 904 1559Likelihood zero -1712 -3391 -2427 -1885 -596 -1592 -1264 -6713 -2230 -3843Log of Likelihood -993 -1885 -1490 -1165 -382 -912 -681 -3506 -1323 -1799Adjusted ρ2 0.429 0.449 0.392 0.390 0.384 0.437 0.473 0.480 0.413 0.536*** Significant at 99%, **Significant at 95%, *Significant at 90%, - not significant

25

Page 26: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Conclusions and perspectives

The conclusions of this paper are deriving from the context of the study and essentially from the

empirical application. In this constantly changing behavioural context, quantifying the influence of

accessibility is an important issue. In order to quantify its influence, one have to approach

accessibility as a multidimensional concept that is. Additionally, it is essential to analyse the influence

of accessibility in comparison to other location attributes influencing the location choice of firms

highlighted in the literature.

As a global comment we can say that the results of the analysis are reflecting the economic

behaviour of the firms and are conforming to the behaviour of the economic sector. The descriptive

statistics analysis have shown that the tendency of creation or relocation of establishments is not

homogenous between the sectors, but there are sectors who are more dynamic in terms of new

establishments while others provoke more relocations. Concerning the influence of accessibility, in

fact accessibility to population is appreciated differently based on the economic sector and the firm

event (birth or relocation). New and face-to-face activity oriented establishments are more prone to

choose a location with good accessibility.

These results are confirmed by the models as well. Summarising the results, we can distinguish 3

types of economic sectors based on the influence of accessibility. The highly sensitive (Retail,

Accommodation and Restauration, Health), for which highly accessible areas are essential for their

activity, the somewhat sensitive (Wholesale, Real Estate, Finance and Insurance, Front Office

services), which appreciate good accessibility but it doesn’t seem that it drives their location choice

and last the no sensitive (Manufacturing, Construction, Back Office), which are not searching for

areas with good accessibility because their activity is not related to face-to-face contact. Additionally,

the models have revealed different sensitivities not only to population accessibility but also to the

proximity to transportation infrastructure. While the proximity to motorway is generally appreciated

by the firms, new or relocated, proximity to public transportation is conditioned by the economic

sector of the firm. Last, for the relocated establishments, distance to the previous location is crucial

and dominates the location decisions. Migrating firms, prefer locations with smaller accessibility

given that they are not far from the last location and thus, not far from the already established “eco-

system”. Possibly, these locations are less expensive as well. Last concerning the other variables for

the economic and social environment, the modelling results have confirmed our initial assumptions.

Nevertheless, the present work is still ongoing and there are many possible extensions that we are

willing to confront in the future. First of all, we want to integrate some important omitted variables

which can increase the explicative power of the model and help with the interpretations of the

26

Page 27: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

results. The dimensions that we are willing to integrated are essentially two: (i) the possible influence

of the macro-agent through local policies in terms of planning like industrial and economic zones

which can have a great effect especially in a French context and (ii) the price of premises which can

reveal a traditional trade-off between accessibility or other location attributes. Additionally,

specifically, for accessibility, the applied potential measure needs to integrate better the labour

dimension and the individual dimension matching the firm’s internal and external level. Moreover,

the methodology of the identification of the creation and relocation of firms must be extended to

include the firms with multiple sites in the study area. After the inclusion of these elements, more

attention should be paid to the analysis of the results. The trade-offs between accessibility and the

other location attributes should be analysed more in depth in order to understand how firm are

choosing their locations. If the results are encouraging, we are willing to extend the empirical

application in more than one time period, in order to track changes though the time.

Bibliography

Aguilera, A., Lethiais, V., Rallet, A., Proulhac, L., 2016. Home-based telework in France: Characteristics, barriers and perspectives. Transp. Res. Part A Policy Pract. 92, 1–11. doi:10.1016/j.tra.2016.06.021

Alonso, W., 1964. Location and land use : toward a general theory of land rent. Harvard University Press, Cambridge, Mass.

Arauzo-Carod, J.M., Liviano-Solis, D., Manjon-Antolin, M., 2010. Empirical Studies in Industrial Location: an Assessment of Their Methods and Results*. J. Reg. Sci. 50, 685–711. doi:10.1111/j.1467-9787.2009.00625.x

Beaudry, C., Schiffauerova, A., 2009. Who’s right, Marshall or Jacobs? The localization versus urbanization debate. Res. Policy 38, 318–337. doi:10.1016/j.respol.2008.11.010

Bentlage, M., Lüthi, S., Thierstein, A., 2013. Knowledge creation in German agglomerations and accessibility - An approach involving non-physical connectivity. Cities 30, 47–58. doi:10.1016/j.cities.2012.07.003

Bhat, C., Carini, J., Misra, R., 1999. Modeling the Generation and Organization of Household Activity Stops. Transp. Res. Rec. J. Transp. Res. Board 1676, 153–161. doi:10.3141/1676-19

Bodenmann, B.R., 2011. Location choice of firms with special emphasis on spatial accessibility.

Bodenmann, B.R., Axhausen, K.W., 2012. Destination choice for relocating firms: A discrete choice model for the St. Gallen region, Switzerland. Pap. Reg. Sci. 91, 319–341. doi:10.1111/j.1435-5957.2011.00389.x

Boiteux-Orain, C., Huriot, J.-M., 2002. Modéliser la suburbanisation. Rev. d’Économie Régionale Urbaine, 73. doi:10.3917/reru.021.0073

Buczkowska, S., de Lapparent, M., 2014. Location choices of newly created establishments: Spatial patterns at the aggregate level. Reg. Sci. Urban Econ. 48, 68–81. doi:10.1016/j.regsciurbeco.2014.05.001

Bunel, M., Tovar, E., 2014. Key Issues in Local Job Accessibility Measurement: Different Models

27

Page 28: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

Mean Different Results. Urban Stud. 51, 1322–1338. doi:10.1177/0042098013495573

Carpenter, J., Verhage, R., 2014. Lyon city profile. Cities 38, 57–68. doi:10.1016/j.cities.2013.12.003

Czamanski, D., Broitman, D., 2017. Information and communication technology and the spatial evolution of mature cities. Socioecon. Plann. Sci. 58, 30–38. doi:10.1016/j.seps.2016.10.003

de Bok, M., Oort, F. Van, 2011. Agglomeration economies, accessibility and the spatial choice behavior of relocating firms. J. Transp. Land Use 4, 5–24. doi:10.5198/jtlu.v4i1.144

de Palma, A., Motamedi, K., Picard, N., 2008. Étude des décisions de localisation des emplois en Île-de- France.

Dubé, J., Brunelle, C., Legros, D., 2016. Location Theories and Business Location Decision : A Micro-Spatial Investigation of a Non-Metropolitan Area in Canada. Rev. Reg. Stud. 46, 143–170.

Elgar, I., Farooq, B., Miller, E.J., 2015. Simulations of firm location decisions: Replicating office location choices in the Greater Toronto Area. J. Choice Model. 17, 39–51. doi:10.1016/j.jocm.2015.12.003

Elgar, I., Farooq, B., Miller, E.J., 2009. Modeling Location Decisions of Office Firms. Transp. Res. Rec. J. Transp. Res. Board 2133, 56–63. doi:10.3141/2133-06

Geurs, K.T., van Wee, B., 2004. Accessibility evaluation of land-use and transport strategies: review and research directions. J. Transp. Geogr. 12, 127–140. doi:10.1016/j.jtrangeo.2003.10.005

Glaeser, E.L., Kallal, H.D., Scheinkman, J.A., 1992. Growth in cities. J. Polit. Econ. 100, 1126–1152.

Graham, D.J., 2007. Variable returns to agglomeration and the effect of road traffic congestion. J. Urban Econ. 62, 103–120. doi:10.1016/j.jue.2006.10.001

Graham, S., Marvin, S., 1996. Telecommunications and the city. Taylor & Francis, Abingdon, UK. doi:10.4324/9780203430453

Hayter, R., 1997. The dynamics of industrial location. The factory, the firm and the production system. John Wiley and Sons, Chichester, UK. doi:10.1002/9781118786352.wbieg0444

Holl, A., 2012. Market potential and firm-level productivity in Spain. J. Econ. Geogr. 12, 1191–1215. doi:10.1093/jeg/lbr030

Holl, A., 2004. Start-ups and relocations: Manufacturing plant location in Portugal. Pap. Reg. Sci. 83, 649–668. doi:10.1007/s10110-004-0218-y

INSEE, 2016. SIRENE - Base de données: Comprendre le dessin “L.”

Ioannides, Y.M., Overman, H.G., Rossi-Hansberg, E., Schmidheiny, K., 2008. The effect of information and communication technologies on urban structure. Econ. Policy 23, 201–242. doi:10.1111/j.1468-0327.2008.00200.x

Jacobs, J., 1969. The economy of the cities, Vintage. ed. New York.

Le Nechet, F., Melo, P.C., Graham, D.J., 2012. Transportation-Induced Agglomeration Effects and Productivity of Firms in Megacity Region of Paris Basin. Transp. Res. Rec. 21–30. doi:10.3141/2307-03

Lee, B.H.Y., Waddell, P., Wang, L., Pendyala, R.M., 2010. Reexamining the influence of work and nonwork accessibility on residential location choices with a microanalytic framework. Environ. Plan. A 42, 913–930. doi:10.1068/a4291

Leitham, S., McQuaid, R.W., D. Nelson, J., 2000. The influence of transport on industrial location choice: A stated preference experiment. Transp. Res. Part A Policy Pract. 34, 515–535.

28

Page 29: asrdlf2017.comasrdlf2017.com/.../envoitextefinal/auteur/.../195.docx  · Web viewIn such a context, searching whether accessibility is still important for a firm location choice

doi:10.1016/S0965-8564(99)00030-0

Maroto, A., Zofío, J.L., 2016. Accessibility gains and road transport infrastructure in Spain: A productivity approach based on the Malmquist index. J. Transp. Geogr. 52, 143–152. doi:10.1016/j.jtrangeo.2016.03.008

Marshall, A., 1890. Principles of Economics. Online Libr. Lib. 1–627. doi:10.1057/9781137375261

Martín-Barroso, D., Núñez-Serrano, J.A., Velázquez, F.J., 2017. Firm heterogeneity and the accessibility of manufacturing firms to labour markets. J. Transp. Geogr. 60, 243–256. doi:10.1016/j.jtrangeo.2017.03.008

McFadden, D., 1977. Modelling the choice of residential location. Spat. Interact. Theory Plan. Model.

Mejia-Dorantes, L., Paez, A., Vassallo, J.M., 2012a. Transportation infrastructure impacts on firm location: The effect of a new metro line in the suburbs of Madrid. J. Transp. Geogr. 22, 236–250. doi:10.1016/j.jtrangeo.2011.09.006

Mejia-Dorantes, L., Paez, A., Vassallo, J.M., 2012b. Transportation infrastructure impacts on firm location: The effect of a new metro line in the suburbs of Madrid. J. Transp. Geogr. 22, 236–250. doi:10.1016/j.jtrangeo.2011.09.006

Melo, P.C., Graham, D.J., Levinson, D., Aarabi, S., 2016. Agglomeration, accessibility and productivity: Evidence for large metropolitan areas in the US. Urban Stud. 42098015624850. doi:10.1177/0042098015624850

Mérenne-Schoumaker, B., 2011. La localisation des industries. Enjeux et dynamiques, Presses Un. ed. Rennes.

Mills, E.S., 1967. Transportation and patterns of urban development: An Aggregative Model of Resource Allocation in a Metropolitan Area. Am. Econ. Rev. 57, 197–210. doi:10,1007/978-1-4684-4001-0

Muhammad, S., de Jong, T., Ottens, H.F.L., 2008. Job accessibility under the influence of information and communication technologies, in the Netherlands. J. Transp. Geogr. 16, 203–216. doi:10.1016/j.jtrangeo.2007.05.005

Muth, R.F., 1969. Cities and housing: the spatial pattern of urban land use. University of Chicago Press, Chicago, IL.

Ota, M., Fujita, M., 1993. Communication technologies and spatial organization of multi-unit firms in metropolitan areas. Reg. Sci. Urban Econ. 23, 695–729.

Padeiro, M., 2013. Transport infrastructures and employment growth in the Paris metropolitan margins. J. Transp. Geogr. 31, 44–53. doi:10.1016/j.jtrangeo.2013.05.007

Rosenthal, S.S., Strange, W.C., 2003. Geography, Industrial Organization, and Agglomeration. Rev. Econ. Stat. 85, 377–393. doi:10.1162/003465303765299882

Siedschlag, I., Smith, D., Turcu, C., Zhang, X., 2013. What determines the location choice of R&D activities by multinational firms? Res. Policy 42, 1420–1430. doi:10.1016/j.respol.2013.06.003

Sweet, M.N., 2014. Do firms flee traffic congestion? J. Transp. Geogr. 35, 40–49. doi:10.1016/j.jtrangeo.2014.01.005

Van Dijk, J., Pellenbarg, P.H., 2000. Firm relocation decisions in The Netherlands: An ordered logit approach. Pap. Reg. Sci. 79, 191–219. doi:10.1007/s101100050043

Von Thünen, J.H., 1842. Der isolirte Staat in Beziehung auf Landwirthschaft und Nationalökonomie, 2nd ed. Rostock.

29