deletions | additions       diff --git a/If_you_wait_to_do__.tex b/If_you_wait_to_do__.tex  deleted file mode 100644  index 3899785..0000000  --- a/If_you_wait_to_do__.tex  +++ /dev/null 
...
“If you wait to do everything until you’re sure it’s right, you’ll probably never do much of anything.” – Win Borden  Paraphrasing the quote of Winston Churchill, courage of writing thesis is going from failure to failure without losing enthusiasm (Original quote: “Courage is going from failure to failure without losing enthusiasm.” – Winston Churchill)  Introduction  1. Tezy i cel pracy - 1 strona, dlaczego to jest wazne i warte zachodu.  2. Analiza literaturowa (wez z papierow lub raportow + ksiazki wazne)  3. Opisac pokrotce co bedzie w rozdzialach (2 strony)  4. Rozdzialy  1. Location choices of newly created establishments: Spatial patterns at the aggregate level  2. Euclidean versus network distance in business location: A probabilistic mixture of hurdle-Poisson models  3. Location choices under strategic interactions: Interdependence of establishment types  4. Locational strategies of multi-store firms.  GENERAL INTRODUCTION  1 XXX GENERAL XXX  XXX  This thesis is breathing new life into the location choice models of establishments. The need for methodological advances in order to model more realistically the complexity of establishments' decision-making processes, such as their optimal location choices is the key motivation of this thesis. A clear distinction between an establishment and a firm should be made in the first place. An establishment is defined as a distinct economic unit that produces goods or services at a single physical location. In contrast, a firm is a legal entity that consists of one or more establishments or plants under common ownership and control (van Wissen, 2000).  1.1 KEY QUESTIONS  A number of key questions will be addressed along this thesis. First, location choice models use geo-referenced data, for which choice sets have an explicit spatial component. It is thus critical to understand how to represent spatial aspect in location choice models. Second, what makes these discrete choices particularly interesting and challenging to analyze is that decisions of a particular establishment are interrelated with choices of the others. These thorny problems posed by the interdependence of decisions generally cannot be assumed away, without altering the realism of the model of establishment decision making. The conventional approaches to location selection fail by providing only a set of systematic steps for problem-solving without considering strategic interactions between the establishments in the market. One of the goals is to explore how to correctly adapt location choice models to study establishments' discrete choices when they are interrelated. Third, a firm can open a number of units and the market from multiple locations. Once again traditional theory and methods may not be suitable to situations wherein individual establishments instead of locating independently from each other, form a whole large organization, such as a chain facing in addition a fierce competition from other chains. There is a necessity to incorporate interactions between units within the same and competing firms. Illustrative questions that can be answered are: What is the nature and degree of competition for each of the analyzed chain? How fast do the firm profit and the market power decline when the number of firms and their outlets in the market increases or when the distance to their rivals decreases?   %How does a firm perceive a rival store located in a very close neighborhood and how if it is located far away?   An intensified research effort along the lines of location choices is still desirable to bring the answers to many questions of this type.  1.2 WHY SHOULD WE BOTHER?   1.2.1)   In this thesis, we concentrate our research on the Paris region, a vibrant and innovative region with over 5,6 million jobs, 37 percent of French executives, and 40 percent of national workforce in research and development. It is the first R&D hub in Europe and the third worldwide. 11.7 million people or over 19 percent of the country's population reside in the area which occupies only 2.2 percent of the surface of France. The Paris region is the third World touristic destination (in 2013) (Global Destination Cities Index 2015) with 16 millions of visitors from abroad. The GDP of the region amounts to 29 percent of total French GDP (IAU IdF, 2014) %31 percent of total French GDP (http://www.grand-paris.jll.fr/fr/paris/chiffres-cles/) that is 612 milliards euros (2012). It is 1st European city considering the number of firms classified in Fortune 500 (July 2014).   Yet, the Paris region’s economy is spatially unbalanced (Combes et al., 2011). The region is highly heterogeneous, especially regarding economic activity. While few municipalities host a large number of new establishments, others struggle to be chosen by any, and a large group of municipalities is left with no new creation. When the observed data display a higher fraction of zeros than would be typically explained by the standard count data models, two types of models can be suggested: the hurdle (Mullahy, 1986) or the zero-inflated model (Lambert, 1992). The hurdle model reflects a two-part decision making process. It relaxes the assumption that the zero observations and the positive observations come from the same data generating process.  Much work has been done in the domain of location choice models, however, several issues arise when analyzing involved phenomena, which scholars have yet to fully inquired: 1) addressing the excess of zeros problem in the location choice model in highly heterogeneous geographic areas and 2) determining an appropriate way to accommodate spatial effects in location decisions. We respond to the complaint voiced by Liviano-Solis and Arauzo-Carod (2013) and Bhat et al. (2014) that heretofore the hurdle model technique has not been well investigated when analyzing location patterns. These are the first challenges that we face in the first chapter of this thesis.   1.2.2)   %When selecting the appropriate location in which to set up in the market, an establishment may consider not only the characteristics of a particular area, but also the characteristics of neighboring zones.   The second chapter extends the research on the hurdle model and the study presented in the first part of this thesis. The final decision of an establishment seems to be related to the surrounding economic landscape. When accounting for the linkage between neighboring observations, the decision on the spatial weight matrix specification should be made. Yet, since there exist no solitary claim on the concept of space, the form of the weight matrix is largely debated. One of the problems hides in the definition of distance usually based on the straight-line segment connecting two locations. Euclidean distance is typically used in the empirical literature and has been utilized also in the first chapter of this thesis to account for spatial spillovers in location choice model. However, Euclidean distance is believed to be only one simplistic possibility out of an infinite number of shortest path relations. Other alternative distance metrics may be proposed when building the spatial distance weight matrices.  Geographic factors such as terrain, land cover, infrastructure, and traffic congestion may cause agents not to follow pure Euclidean relations. The Euclidean distance might thus not always be the most relevant one depending on a given problem. Interest in this question dates at least to the 1960s and research on network models in geography (Haggett 1967). There are insights to be gained by mindfully reconsidering and measuring distance. The second chapter investigates establishments location decisions in the Paris region where high congestion, speed limits, or physical uncrossable barriers, such as rivers or industrial corridors can diminish or totally eliminate the linkage between neighboring areas. Rather than imposing a restrictive structure of the weight matrix, this research proposes a flexible toolkit to point which distance metric is more appropriate to correctly account for the surrounding economic landscape. A probabilistic mixture of two ”mono-distance” hurdle-Poisson models is developed. Each model’s latent class uses a different distance representation to incorporate spillover effects in location choices of establishments from several activity sectors. Seven distance metrics are considered: Euclidean distance, two road distances (with and without congestion), public transit distance, and the corresponding travel times. This methodology allows to capture the diversity of agents’ behavior, i.e., to distinguish establishments which are more time- or more distance-oriented given location.   %In spite of recognizing the importance of incorporating spatial effects in establishments location decision processes, the literature is still scarce on previous attempts.   1.2.3)   In the third chapter we further enhance the literature on the location choices, this time incorporating strategic interactions among establishments. We shed light on strategic interactions, fundamental in establishments’ location choices, yet largely unheeded in the empirical literature. If establishments acted in isolation, it would be a relatively simple task to adapt existing discrete-choice models. Yet, being non-strategic means that an establishment ignores other players’ decisions. Less is known about how to correctly adapt location choice models to study establishments’ discrete choices when they are interrelated. In very sparse empirical applications, when locational choice models are developed for several activity sectors, each of the model is typically run independently.  What makes these discrete choices particularly interesting and challenging to analyze is that decisions of a particular establishment are interrelated with choices of the others because an establishment accounts for the actions of other agents when making its own decisions (Draganska et al., 2008). These thorny problems posed by the interdependence of decisions generally cannot be assumed away, without altering the realism of the model of establishment decision making (Berry and Reiss, 2007). The conventional approaches to location selection, i.e., traditional theory and methods, fail (Thill, 1997) by providing only a set of systematic steps for problem-solving without considering strategic interactions between the establishments in the market. A appropriately specified model of simultaneous entry or location decisions needs to recognize this interdependence of  profits (Berry and Reiss, 2007).  There is a need for more realistic studies of complex establishment’s decision-making processes. Even though the computational burden imposed by these models considering strategic interactions is relatively high, it seems that the costs imposed are more than offset by the benefits that accumulate/accrue (Draganska et al., 2008).  Strategic interactions have been largely unsung in the empirical analyses since the year 1929 when Hotelling (1929) brought the discussion in the industrial organization literature. Most of the papers are less than a decade old (Bajari et al., 2013). This literature is in its infancy, in part, due to the complexity of expressions for the probabilities used in the models which increases along with the number of locations and establishment types (Draganska et al., 2008).   We estimate a static discrete game of incomplete information to obtain a Bayesian Nash Equilibrium at the group level using data at the aggregate level. We permit asymmetries across establishment types in the impact of interaction effects and exogenous market characteristics. We develop one location choice model which embraces seven individual models for seven establishment types run simultaneously to account for interactions from all the types on each other.   1.2.4) The motivation of the last chapter comes from the fact that most previous discussion on locational decisions has one common feature of making unrealistic and restrictive assumptions and perceives the industry in terms of independent stores. The analysis of multi-store competition has started already with the trailbraking work of Teitz (1968) who introduced the idea that a firm can open multiple facilities in the context of Hotelling’s linear city model and serves the market from a number of locations. Yet, the subject of location of competing firms with multiple component units seems to have been largely unsung/unheeded in the spatial location literature (Peng and Tabuchi, 2007). This gap is inquisitive/inquiring for the systems which dominate in the market (Karamychev and van Reeven, 2009; Iida and Matsubayashi, 2011 ; Janssen et al., 2005 ; Pal and Sarkar, 2002 ; Peng and Tabu- chi, 2007). The conventional single-store location theory may not apply to situations wherein individual stores are part of larger organizations under common strategy, intuition, and control, where a centralization is applied to reach global goals and consider the interest of a firm as a whole (Thill, 1997). Conceptually, a firm selects a distribution of locations instead of choosing a point location (Chu and Lu, 1998).  Our main motivation for this chapter is to combine a number of novel solutions into one model and to rectify several mathematical and methodological misconceptions made in numerous existing store-location papers. We incorporate strategic interactions between stores within the same firm and stores that belong to different chains. We consider spatial competition, business stealing and learning effects. We pay a particular attention to correctly capture market segments and to select potential customer groups by observing their characteristics, their mobility patterns, their trip chaining behavior, and activities’ purposes during the day and during the week. A clear distinction between a daytime and nighttime population present in a particular area is needed, and therefore a more appropriate distance measure to a store traveled by a potential customer is carefully proposed and applied in a more realistic manner than it has been done in the existing literature. Further, we consider the markets as being interdependent. A combination of all these elements can create a more authentic/more realistic and original model.  Literature   Chapter I introduces the reader to the location choice models. The list of the key factors that potentially influence the locational decisions has been created based on the research of Maoh (2005), Strotmann (2007), Liviano-Solis and Arauzo-Carod (2011), Rocha (2008), Maoh and Kanaroglou (2005, 2007), Bondomi and Greenbaum (2007), Bodenmann (2011), Duvereux et al. (2007), Neumark and Kolko (2010), De Bok (2004), Bodenmann and Axhausen (2012, 2010), and the review of Arauzo-Carod et al. (2010). Chapter I provides also a discussion on the first attempts to incorporate spatial effects into location choice models starting with Bhat and Guo (2004) on modeling spatial dependence in residential locations using a mixed logit. Sener et al. (2011) propose the generalized spatially correlated logit and Miyamoto et al. (2004) the mixed logit with the error autocorrelation and an autocorrelated deterministic component of utility to model the residential behavior. Garrido and Mahmassani (2000) discusses a multinomial probit with spatially and temporally correlated error structure to analyze and forecast the distribution of freight flows.   Nguyen et al. (2012) discusses a tree-stage firm relocation model wherein spatial correlation between zones has been implemented in the error term and spatial interactions among firms in the deterministic part. Klier and McMillen (2008) provide a description of the generalized method of moments spatial logit to model the clustering of the auto supplier establishments.   Chapter II continues with a discussion on further attempts to incorporate spatial effects into establishments location choice models. It highlights the fact that, whenever a distance metric has been used in the weight matrix to implement a spatial aspect into the establishments location models (see the research of Dube et al., 2016; Bhat et al., 2014; Liviano-Solis and Arauzo-Carod, 2013; Lambert et al., 2010; and Klier and McMillen, 2008), the Euclidean distance was employed. The debate in this chapter demonstrates that there are insights to be gained by mindfully reconsidering and measuring distance depending on a given problem, such as a high congestion, speed limits, or physical uncrossable barriers which can diminish or totally eliminate the linkage between neighboring areas. Alternative distance metrices can be proposed and tested. Several articles, including the ones by Miller (2004, 2003) and the sudies by Combes and Lafourcade (2005), Graham (2007), Duran-Fernandez and Santos (2014), Weisbrod (2008), Faber (2014), Kwon (2002), Boscoe et al. (2012), Chalasani et al. (2005), or Rietveld et al. (1999) encourage the use of real distance metric based on the transport network over geographical distance metrics.  %Several studies, such as the ones by Bodsen and Peeters (1975), Aten (1997), Duran-Fernandez and Santos (2014), Conley and Ligon (2002), Slade (2005), Le Gallo and Dall'erba (2008), and Fingleton (2008) have considered metrics not purely based on topography, including network distances and transport costs.   Most of the analyses on establishments location choices presented in the first two chapters treat only one selected activity sector at a time, typically industrial or retail activities. Very few empirical studies, i.e., Dube et al. (2016) or Buczkowska et al. (2014) develop models for a number of various sectors, yet still, these models are analyzed independently without considering strategic interactions between the sectors. Literature revised in Chapter III and especially works of Zhu and Singh (2009), Seim (2006), Jia (2008), Berry (1992), Bresnahan and Reiss (1991a), Ellickson and Misra (2012), Singh (2009), Draganska et al. (2008) ((, Chatman et al. (2016))) helps understand why it is worth considering within- and inter-industry interactions. An excellent review of Draganska et al. (2008) gives a classification of possible modeling choices along plural dimensions, in the informational, temporal contexts, considering the timing of moves, and the discrete, continuous, or mixed decisions of establishments.   Our context is a static discrete game of incomplete information. An establishment's payoff from choosing a particular location depends on its expectation of the optimal location choices of its competitors and exogenous market characteristics (Zhu and Singh, 2009). Based on the expected distribution of other agents across market locations, each establishment selects the location that maximizes its payoff given its own type (Seim, 2006).   %Establishments base their decisions on what they expect other players will do, where they will locate (Draganska et al., 2008).   %The influence of push and pull factors might be different depending on the establishment type considered. One can distinguish revenue-demend oriented establishments, such as the retail, service, and profesional activities sectors, and cost focused establishments, for example warehouses or goods-producing units.   The literature of Chapter IV starts with the the trailbreaking work of Teitz (1968) who introduces the idea that a firm can serve the market from multiple locations in the context of Hotelling's linear city model. Since then other researchers, such as Thill (1997), Chu and Lu (1998), Pal and Sarkar (2002), Janssen et al. (2005), Karamychev and van Reeven (2009), Iida and Matsubayashi (2011), Takaki and Matsubayashi (2013), Neven (1987), Peng and Tabuchi (2007), Granot et al. (2010), or Nishida (2015) have been proposing multi-store firms location models.  The light is then put on the research by: Yang (2015) and Igami and Yang (2015) who study Canada's fast food hamburger industry of five substituable with one another chains; Toivanen and Waterson (2005) who analyze fast food restaurants in the UK; and Thomansen (2005) who focuses on two fast food chains in California. For comparison, three other studies are mentioned in chapter IV, these are: Nishida (2015) on convenience-store chains in Japan, Schiraldi et al. (2013) and Holmes (2011) on the UK supermarket industry, such as Walmart and other supercenters.   Chapter IV pronounces the fact that location-store models typically do not consider competing rival-chain stores and the strategic interactions between stores within the same chain. The profitability is a function of player i's entire network and the competitors' networks (Nishida, 2015). The literature on business stealing and learning effects, cannibalization effects and economies of scale described in Chapter IV (see, e.g., Toivanen and Waterson, 2005; Aquirregabiria and Suzuki, 2015) is not consistent in specifying which of these effects has a prevailing influence on location decisions and depends on the tore type. In addition, Nishida (2015) and Toivanen and Waterson (2005) incorporate some form of spatial competition, demonstrating that the store's profitability is influenced by stores in the same location and and by those in the adjacent locations, incorporating the interdependence of markets.  %Nwogugu (2006) distance   X discusses   X et al. provide a closely related discussion on xxx.  X writes about the role of...   xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx  Chapter I  1. Location choices of newly created establishments: Spatial patterns at the aggregate level  Much work has been done in the domain, however, several issues arise when analyzing involved phenomena, which scholars have yet to fully explore: 1) addressing the excess of zeros problem in the location choice model in highly heterogeneous geographic areas and 2) determining an appropriate way to accommodate spatial effects for location decisions.   When the observed data display a higher fraction of zeros than would be typically explained by the standard count data models, the zero inflated or hurdle models can be suggested. In this paper, we respond to the complaint voiced by Liviano-Solis and Arauzo-Carod (2013) and Bhat et al. (2014) who notice that heretofore scholars have not fully explored the hurdle model technique when analyzing location phenomena.  Spatial effects can be incorporated in location choice models when modeling the observable explanatory variables and the unobservable components. However, most often, these spatial effects are either not properly treated or are completely ignored in the analysis of the establishment’s location.   Furthermore, even if spatial effects are present they are not incorporated in traditional discrete choice models.   Traditional discrete choice modeling methods are often based on the assumption of independence among choice alternatives, which according to Sener et al. (2011) is not appropriate. Furthermore, Sener et al. (2011) claim that the estimations of the parameters of the standard logit models are biased and inconsistent.  We wish to combine the methods used to produce location choice models with spatial econometric techniques by examinig the role of space in in these models. This is a new and challenging field of research.  As seen in the review by Arauzo-Carod et al. (2010) and the literature review, the focus has been, up to this point, on the manufacturing/industrial and high- tech or R&D sectors. We here consider several other sectors, such as construction, hotels and restaurants, and real estate activities. Thus, the existing literature can be treated as a guide to motivate factors that may be studied in the econometric models.   Def. of establishment and firm:   We define an establishment as a distinct economic unit that produces goods or services at a single physical location. In contrast, a firm is a legal entity that consists of one or more establishments or plants under common ownership and control (van Wissen, 2000).  Paris region:   we concentrate our research on the Paris region, called as Ile-de-France - a vibrant and innovative region with over 5,6 million jobs, 37 percent of French executives, and 40 percent of national workforce in research and development. 2.2 percent of the surface of France, 11.7 million, that is over 19 percent of the country’s population reside in this area. The GDP of the region amounts to 29 percent of total French GDP (IAU IdF, 2014).   The Paris region’s economy is spatially unbalanced (Combes et al., 2011).   The Grand Paris Project, up to 2030 and will cost around 32.4 billion euros.  The Grand Paris is a development project for the whole of the Paris metropolitan area. It is designed to improve residents' quality of life, address regional inequalities and build a sustainable city (https://www.societedugrandparis.fr/english).   The Paris region is highly heterogeneous16, especially regarding economic activity. While few municipalities host a large number of new establishments, others struggle to be chosen by any, and a large group of municipalities is left with no new entries.   Liviano-Sol ́ıs and Arauzo-Carod (2014) based on the analysis of the Catalan data. The authors state that the distribution of entries is heavily skewed: a small group of municipalities meet the largest number of entries, while more than a half receive no entries at all. Municipalities range from small isolated villages in rural areas to huge and densely populated cities.  Depending on the analyzed sector, the percentage of municipalities left with no new creation ranges from 34 percent up to 69 percent. The number of municipalities left with zero newly created establishments in the industry sector equals to 734, in construction 439, commerce 440, transport 837, financial activities 799, real estate activities 738, hotels and restaurants 792, information and communication 771, special, scientific and technical activities 569, education 890, health and social activities 794 out of 1300 possible municipalities.  --> In addition, when the observed data display a higher fraction of zeros than would be typically explained by the standard count data models, two types of models can be suggested: the hurdle model (Mullahy, 1986) or the zero-inflated model (Lambert, 1992).   The hurdle model, also called the two-part model, reflects a two-part decision making process. It relaxes the assumption that the zero observations and the positive observations come from the same data generating process.   The two-stage decision-making process is reflected through the hurdle model interpretation. A zero-inflated model assumes that there are two sources of zero counts (not just one as in the hurdle model).   We concluded that an establishment does not act in isolation during its decision-making processes and that it is likely to be influenced by other establishments located nearby.   The paper finds that the models tested with the distance matrix indicate that the incorporation of spatial spillovers leads to an enhancement in the models’ performance.  we make use of the distance matrix to characterize spatial patterns. Currently, there are two basic categories that define neighbors: contiguity (shared borders) and distance. Contiguity-based weights matrices include rook and queen matrices. Distance-based weights matrices include distance bands and k nearest neighbors.  Agglomeration etc.:  Rocha (2008) and Liviano-Soĺıs and Arauzo-Carod (2011) concur, finding that local employment density attracts new entrants in related sectors and has a positive impact on the establishment’s productivity. However, when this density is too high, the effect becomes negative due to congestion costs, including for example, high land prices and costly commuting.   Maoh and Kanaroglou (2005, 2007) state that agglomeration effects tend to be more significant in particular activity sectors, such as retail and services.  Chapter II  2. Euclidean versus network distance in business location: A probabilistic mixture of hurdle-Poisson models  As noticed by McMillen and McDonald (2004) and emphasized by Rincke (2010), Bill ́e and Arbia (2013), and Vega and Elhorst (2013), the use of an arbitrary matrix is often the starting point to specify the linkage between neighboring observations followed by the sensitivity analysis based on models estimations using alternative, equally arbitrary, matrices. However, this arbitrary choice has a disadvantage of imposing a restrictive structure that can bias results when inappropriate.   several studies have considered alternative distance measures that are not purely based on topography (e.g., Conley and Ligon 2002; Slade 2005), including network distances and transport costs. However, there is little comparison with the geographical distance (Euclidean or great circle depending on the spatial scale), and when there is, it is based on the relative performances of two models, one based on the alternative distance and the other on the geographical distance.  This research proposes a new, flexible approach, where several distance measures may coexist and be combined instead of being systematically opposed. The methodology is based on a mixture of ”mono- distance models” which allows us to capture the diversity of agents’ behavior, and provides a more direct and integrated way of comparing various distance measures with each other. We address the criticism of Rincke (2010), Vega and Elhorst (2013), and other authors that the choice of the spatial weight matrix is usually quite arbitrary, while it refers to the choice of the distance measure.  Yet, whenever the distance measure was used in the weight matrix to implement the spatial dependencies or spatial spillovers in location choice models, no discussion was provided on the choice of the distance measure itself and the Euclidean distance was utilized.   However, whenever a distance measure was used in the weight matrix to implement the spatial effects in the location choice model, no discussion was provided on the choice of the distance measure itself.  they do not concentrate on the distance definition.   Hence, we find it necessary to open up a discussion on distance definition to be used in the location  choice models.  Definition of Euclidean distance:   Euclidean distance is the ”ordinary” distance between two points that one would measure with a ruler (Dattorro 2015).   The use of the Euclidean distance is widespread in economics (e.g., Duranton and Overman 2005; Partridge et al. 2008). This metric is known to all and experienced by all in everyday life, hence a prime candidate in economics. It is easily available to boot. Combes and Lafourcade (2003, 2005) claim that any Euclidean distance can only be regarded as a proxy for the actual physical distance. The curvature of the Earth is the first source of systematic error. The second source of systematic error comes from the fact that in practice, people (or goods) move along a transport network. For instance, car users may only drive on the existing road network. They rarely go from point A to point B along the straight line as assumed in the Fetter’s ”Law of Markets” (1924).   Interest in this question dates at least to the 1960s and research on network models in geography (Haggett 1967). According to Guy (1983), the use of air-line distance to represent a travel function is unsatisfactory, although it simplifies computation. In most cities transportation is along a network of roads and Euclidean metric is not appropriate for the study of intra-urban location (Eaton and Lipsey 1980; Perreur and Thisse 1974).  Actual driving distances over a road network and their corresponding travel times are perceived also by Boscoe et al. (2012) to be superior and substantially more precise than the straight-line distance.   In addition, heavy street use, road and parking congestion, speed limits, one-way roads, interstate highways with limited crossings, river with insufficient bridges, parks, and cemeteries may cause drivers to make detours in order to reduce their travel time, meaning that the shortest path may not be the fastest one.   Based on these considerations, several authors advocate the use of ”real” distance measures based on a transport network over geographical distance measures, Euclidean and great circle alike (Combes and Lafourcade 2005; Graham 2007; Duran-Fernandez and Santos 2014; Weisbrod 2008; Faber 2014; Kwon 2002). This point is especially cogent when it comes to the location choice of economic establishments, for which the role of a transport infrastructure is now well-known  Locations separated by rivers, lakes, mountains, steep hills, parks, cemeteries, golf courses, landmarks, highways, rail roads, train routes, industrial corridors often mark neighboring boundaries and have higher-than-expected travel times (see Fig. 3 and 4). Physical barriers are also formed by major single-purpose zones and major transport infrastructures that can only be crossed at the cost of substantial effort and tend to reduce the mobility of population living nearby (H́eran 2011).  The barriers tend to be difficult to cross and the whole zone is marked by the severance effect. The mobility number of journeys of residents declines in intensity. One can observe a reduction in neighborhood relations. Barriers require to make detours, expend additional energy. Access to employment and population becomes highly restricted (Motte-Baumvol et al. 2015). Jacobs (1961) finds that barriers usually make destructive neighbors by limiting interactions. Barriers mitigate neighbor externalities (Noo- nan 2005). Barriers that mitigate spatial externalities are expected to have important differential effects on neighborhood and land use patterns (Noonan 2005). Chakravorty (1996) mentions that physical features may imply a total non-contiguity.  When choosing an appropriate place in which to set up on the market, an establishment can take into account not only the characteristics of a particular area but also those of its surroundings. The degree of spatial correlations is expected to be greater among choice alternatives that are close to one another. Despite the existence of these spatial effects, they are most often completely ignored in the analysis of the unit location. There is little mention in the literature of previous attempts to incorporate spatial effects in establishment or firm location decision-making processes (Bhat et al. 2014; Buczkowska and Lapparent 2014; Liviano-Sol ́ıs and Arauzo-Carod 2013; Liesenfeld et al. 2015; Lambert et al. 2010; Klier and McMillen 2008).  In this paper, we respond to the complaint voiced by Liviano-Sol ́ıs and Arauzo-Carod (2013) and Bhat et al. (2014) who notice that heretofore scholars have not fully explored the hurdle model technique when analyzing location phenomena. Consequently, the empirical evidence (for comparisons purposes) is still scarce. We will try to fill this gap in the business location modeling literature limited to two recent papers 1) of Liviano-Soĺıs and Arauzo-Carod (2013) and 2) of Buczkowska and Lapparent (2014).  ((Liviano-Sol ́ıs and Arauzo-Carod (2013) find that the hurdle approach fits their industrial sector loca- tion data better than the zero-inflated approach. The authors compare several models: Poisson, negative binomial, zero-inflated versions of these models, hurdle-Poisson (HP) and hurdle negative-binomial (HNB). They show that the hurdle models (HP and HNB) are the models whose expected number of zero counts match the observed zero counts, and that the distribution of the HNB model is the one that best fits the data under study. They account for the excess of zeros problem and the overdisperssion (the excess of conditional variance over the conditional mean). They conclude that the use of a HNB clearly improves the explanatory power of the econometric estimations, and they suggest that the analysis of firm location behavior should consider the following factors: 1) the existence of a threshold that allows a site to be chosen by at least one firm and 2) the number of times that this site is chosen by the total population of plants during the analyzed period.))  ((Buczkowska and Lapparent (2014) test various count data models: Poisson, zero-inflated Poisson, zero-inflated (tau) Poisson, negative binomial, zero-inflated negative binomial, and hurdle-Poisson models. Hav- ing estimated 84 nested and non-nested count data models for various activity sectors, the authors demon- strate that the hurdle models are preferable for taking into account the presence of excess zeros. Hurdle models offer greater flexibility in modeling zero outcomes than the zero-inflated models and relax the as- sumption that the zero observations and the positive observations come from the same data generating process.))  We modify the modeling framework of Buczkowska and Lapparent (2014) in order to consider alternative ”transport distances” in addition to the Euclidean distance, namely: two road distances (with or without congestion), the public transit distance, and the corresponding travel times.   Chapter III  3. Location choices under strategic interactions: Interdependence of establishment types  The need for methodological advances in order to model more realistically the complexity of establishments’ decision-making processes, such as their optimal location choices is the key motivation of our present paper. We shed light on strategic interactions, fundamental in establishments’ location choices, yet largely unheeded in the empirical literature. If establishments acted in isolation, it would be a relatively simple matter to adapt existing discrete-choice models. Yet, being non-strategic means that a firm ignores other players’ de- cisions. Less is known about how to correctly adapt location choice models to study establishments’ discrete choices when they are interrelated. In very sparse empirical applications, when locational choice models are developed for several activity sectors, each of the model is typically run independently.  Many key strategic decisions establishments make, such as where to set up in the market, involve discrete choices (Draganska et al., 2008). These decisions are fairly complex, yet particularly important since, unlike other marketing mix elements, they are less adjustable in the short-run without incurring significant costs (Zhu and Singh, 2009). In the literature on location choices of establishments, usually only one activity sector, typically an industrial or a re- tail sector, is considered at a time. In very sparse empirical applications, when locational choice models are developed for several activity sectors, each of the 10 model is typically run independently (see, e.g., Chatman et al., 2016; Bucz- kowska and Lapparent, 2014).  What makes these discrete choices particularly interesting and challenging to analyze is that decisions of a particular establishment are interrelated with choices of the others because an establishment accounts for the actions of other agents when making its own decisions (Draganska et al., 2008).  These thorny problems posed by the interdependence of decisions generally cannot be assumed away, without altering the realism of the model of establishment decision making (Berry and Reiss, 2007). The conventional approaches to location selection, i.e., traditional theory and methods, fail (Thill, 1997) by providing only a set of systematic steps for problem-solving without considering strategic interactions between the establishments in the market. 30 Being non-strategic would mean that an establishment ignores other players’ decisions (Toivanen and Waterson, 2005). A properly specified model of simultaneous entry or location decisions needs to recognize this interdependence of  profits (Berry and Reiss, 2007).  However, less is known about how to correctly adapt location choice mo-  35 dels to study establishments’ discrete choices when they are interrelated (Draganska et al., 2008). Strategic motives are widespread phenomena and they are fundamental in establishments’ location choices, in particular. Yet, the small amount of attention the topic has received is surprising.   Strategic interactions have been largely unsung in the empirical analyses since the year 1929 when Hotelling (1929) brought the discussion in the industrial organization literature. Most of the papers are less than a decade old (Bajari et al., 2013). This literature is in its infancy, in part, due to the complexity of expressions for the probabilities used in the models which increases 45 along with the number of locations and establishment types (Draganska et al., 2008).   There are several relevant reasons why incorporating strategic interactions may turn out to be necessary. If strategic effects matter and are ignored, other 115 factors included in the payoffs will be estimated with a bias (Draganska et al., 2008). The magnitude of such bias will depend on the degree to which strategic effects matter.   There is a need for more realistic studies of complex establishment’s decision-making processes. Even though the computational burden imposed by these models considering strategic interactions is relatively high, it seems that the costs imposed are more than offset by the benefits that accumulate/accrue (Draganska et al., 2008).  The presence of competitive establishments seems to have a negative effect on payoffs and therefore a negative influence on new establishments’ formation (Seim, 2006 ; Zhu and Singh, 2009 ; Chatman et al., 2016). The fact that the number of establishments in the same industry category tends to negatively predict the number of establishment births in that industry suggests that the own-industry competition effect overrides any localized agglomeration economies (Chatman et al., 2016).  Zhu and Singh (2009) observe significant asymmetries across players in their response to market conditions and interactions.  We estimate a static discrete game of incomplete information to obtain a Bayesian Nash Equilibrium at the group level using data at the aggregate level. We permit asymmetries across establishment types in the impact of interaction effects and exogenous market characteristics.  We report results for location choice models run simultaneously for seven establishment types.  influence of variables might be quite different depending on the establishment type considered. As highlighted by Min (1987), the retail, service sector, and professional location analyses are likely to be highly sensitive to the demand and revenue-generating factors. Whereas cost factors should play a critical role rather in the locational strategies of, e.g., warehouses or goods-producing establishments.  While favorable market conditions, such as large base of consumers or users  and low cost characteristics, e.g., low cost labor, encourage establishments’ choices, such markets are also likely to attract competitors. In the long-run, the interaction of described different push and pull factors determines the observed location patterns by establishments (Zhu and Sing, 2009). Opposing forces drive profits. The trade-off between the proximity to competitors and the desirability of certain location characteristics should be pondered (Zhu and Singh, 2009 ; Seim, 2006).   According to Chang and Hsieh (2014), a proportion of rent expenses against sales should be carefully calculated. As the authors claim, "value for money" locations are usually hard to find due to the fact that most of this kind of sites are already being rented or that rent expenses of good locations, such as in the downtown or newly developed areas, are extremely high eliminating the profit. Aquirregabiria and Suzuki (2016) highlight the fact that attractive locations are typically expensive and can be associated with stronger competition.  Establishments should thus consider the trade-off between being accessible to many potential consumers or users, higher land prices, and ferocious competition when deciding where to open a new store.  Chapter IV  4. Locational strategies of multi-store firms  Most previous discussion on locational decisions has one common feature of making unrealistic and restrictive assumptions and perceives the industry in terms of independent stores.   The subject of location of competing firms with multiple component units seems to have been largely unsung/unheeded in the spatial location literature (Peng and Tabuchi, 2007). This gap is inquisitive/inquiring for the systems which dominate in the market (Karamychev and van Reeven, 2009; Iida and Matsubayashi, 2011 ; Janssen et al., 2005 ; Pal and Sarkar, 2002 ; Peng and Tabu- chi, 2007).   Chu and Lu (1998) note that most previous discussion on locational decisions has one common feature of making unrealistic and restrictive assumptions and perceives the industry in terms of independent stores. The conventional approaches to location selection, i.e., traditional theory and methods, fail (Thill, 1997)  The conventional single-store location theory may not apply to situations wherein individual stores are part of larger organizations under common strategy, intuition, and control, where a centralization is applied to reach global goals and consider the interest of a firm as a whole (Thill, 1997).   Conceptually, a firm selects a distribution of locations instead of choosing a point location (Chu and Lu, 1998).  Major modern international brands sell their goods through chain stores (Takaki and Matsubayashi, 2013 ; Peng and Tabuchi, 2007).  The analysis of multi-store competition has started with the trailbraking work of Teitz (1968) who introduced the idea that a firm can open several facilities in the context of Hotelling’s linear city model and serves the market from multiple locations.   ((The store format and its strategy is supposed to fit well the targeted customer groups identified by their desires, interests, buying requirements, and their expectations (Thill, 1997).))  Our main motivation is to combine these novel solutions into one model and to rectify several mathematical and methodological misconceptions made in numerous existing store-location papers. We incorporate strategic interactions between stores within the same firm and stores that belong to different chains. We consider spatial competition, business stealing and learning effects. We pay a particular attention to correctly capture market segments and to select potential customer groups by observing their characteristics, their mobility patterns, their trip chaining behavior, and activities’ purposes during the day and during the week. A clear distinction between a daytime and nighttime population present in a particular area is needed, and therefore a more appropriate distance measure to a store traveled by a potential customer is carefully proposed and applied in a more realistic manner than it has been done in the existing literature. Further, we consider the markets as being interdependent. A combination of all these elements can create a more authentic/more realistic and original model.  Toivanen and Waterson (2005) emphasize that while traditional industrial organization theories cannot explain the positive effect of rival presence on own entry, learning models can.   Toivanen and Waterson (2005) show that the rival presence increases the probability of entry due to the firm learning, yet, profits are decreasing in the number of rival stores and are increasing in the number of own outlets. The Toivanen and Waterson’ (2005) results further suggest that learning effects are strong enough to dominate any negative effects that competition between firms may have on entry decisions.  Nwogugu (2006) finds the issue with distance measure particularly disturbing. The author claims that existing store-location models do not incorporate the distance element appropriately, erroneously assuming that all residents of each community travel the full length of the distance between their community and the store location each time they visit a store. However, in reality most potential buyers do not work in their immediate communities and tend to stop over a store on their way to other destinations and so will prefer locations that are on the route to destinations that they regularly visit. Most people drop in on the store before going to work, during lunch, after work, or on weekends.   Fransen et al.’ (2015) method adds a behavioral realism to the existing metrics that are typically based on static, nighttime representations of population.  Nishida (2015) and Toivanen and Waterson (2005) incorporate some form of spatial competition, showing that a store’s revenue is influenced not only by other stores in the same location but also by those in adjacent locations.   Degree of local competition is keen between neighboring firms, but weak between remote ones (Peng and Tabuchi, 2007).  Schiraldi et al. (2013) allow business stealing and cannibalization effects to operate not only within the boundary of a particular location, but also across locations which may be the source of the spatial competition. Igami and Yang (2015) notice the future need to include the shops’ distances within geographical market stating that closer shops seem to compete more fiercely.   Firms are expected to carefully choose locations to ease the access to the highest number of spatially dispersed potential customers.  Igami and Yang (2015) observe that fast food stores seem to compete within relatively small markets concentrating their efforts on a micro- level location. Their results indicate a highly localized nature of competition among fast food stores. The authors state that a distance criterion equal or greater than a mile (1,61 kilometer) would appear to be useless for an empiri- cal analysis of competition among fast food stores. According to calculations of Thomadsen (2005) for the fast food market (McDonald’s and Burger King) in Santa Clara County, only outlets within 0.5 mile (800 meters) will compete as close substitutes, even in car-obsessed California.   Fransen et al. (2015) suggest that not the entire working population should be treated as potential customers for a simple reason, that not every person commuting to work make use of a particular facility (in our application, service or store). This can lead to an overestimation of the facility’s (service’s or store’s) demand.  one of the most common mistakes is targeting people instead of targeting money. A market segment may represent a large percentage of population, but a small part of the market. There is always a need to look at the money potential of market segments, not just the number of people in the segments”. One of the results of Chu and Lu (1998) seems not to tell the full story when indicating that more populated areas will encourage the player to establish more stores.  In addition, Gira Conseil evaluates that a potential customer is willing to walk 10-12 minutes or drive a car for 5-7 minutes to reach a particular restaurant.  Igami and Yang (2015) show that typically used population and income might not adequately capture the demand. Population statistics usually reflect the number of residents but not necessarily the real daytime population, which might be much more critical in the fast food industry.   Thomadsen (2007) shows that firms will choose to locate near large sources of demand.   Following the remark of Nishida (2015), a clear distinction between a daytime versus nighttime population present in a particular area is to be made.  As a consequence, a more appropriate distance between a potential customer and a store definition will be carefully proposed and applied in a more realistic manner than in most of the existing papers (see the arguments of Nwogugu, 2006).  Citing other papers is easy. Voilà: \cite{2012} or \cite{Holstein_2009}. Click on the \verb|cite| button in the toolbar to search articles and cite them. Authorea also comes with a powerful commenting system. Don't agree that $E = mc^{3}$?!? Highlight the text you want to discuss or click the comment button. Find out more about using Authorea on \href{https://www.authorea.com/help}{our help page}.