Hotelling-Downs with Facility Synergy: The Mall Effect

Elliot Anshelevich, Jianan Lin, Noah Prisament · 2026

Read Hotelling-Downs with Facility Synergy: The Mall Effect by Elliot Anshelevich, Jianan Lin, Noah Prisament (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-26055.txt.

Introduction

Elliot Anshelevich, Jianan Lin, and Noah Prisament add synergy to the classic Hotelling-Downs location game. In the classic model, customers visit the nearest facility. Real shoppers often prefer a mall, where many facilities cluster, even at extra distance. Their variant lets clients weigh cluster size against travel distance together. Facility placement becomes a richer strategic puzzle.

Existence results come first. Nash equilibria always exist in the synergy model, unlike the classic one. One equilibrium even matches the cost of the optimal placement exactly. The headline bound caps the price of anarchy, the worst-case cost of selfish play versus the optimum. That ratio stays below roughly three and a half. Client behaviour is more complex and often more realistic here. Selfish choices still cannot drift far from the social optimum. The mall effect, it turns out, does not wreck efficiency.

Catalog record

CREATORElliot Anshelevich, Jianan Lin, Noah Prisament
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE90 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.26055

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Hotelling-Downs with Facility Synergy: The Mall Effect Elliot Anshelevich, Jianan Lin, Noah Prisament Abstract We consider a variation of the classic Hotelling-Downs model with the addition of facility synergies. Unlike in the classic model, where clients always use the facility closest to them, we study clients who prefer locations with many facilities to those with few facilities while simultaneously attempting to minimize their distance as well. We show that, in contrast with the classic model, Nash equilibria for our setting always exist, and, in fact, there always exists a Nash equilibrium such that the sum of client costs equals the cost of the optimal solution. Our main result is a bound of 225 64 ≈ 3.516 \frac{225}{64}\approx 3.516 on the Price of Anarchy for our model, showing that, although the client behavior is more complex in our model (and often more realistic depending on the application), the cost of Nash equilibrium solutions still cannot be much worse than the cost of the optimal facility placement. 1 Introduction Beginning with Hotelling’s seminal work almost a century ago [ 22 ] , the Hotelling-Downs model has grown to become one of the classics of game theory. Originally defined using the story of ice cream vendors choosing their placement along a beach, it has been since used to model a variety of applications, including the strategic placement of facilities and firms, as well as the formation of political parties and the ideological placement of strategic political candidates. In most existing work on the topic, there are clients located along some interval X , and a set of k facilities or firms. Each facility can individually choose a location x_{i}\in X where it will be placed; it is allowed to choose any location in X . In the classic model, each client simply utilizes the closest facility to them, i.e., a client located at x would shop at the position x_{i} with a facility such that |x-x_{i}| is as small as possible. Note that in political and social choice applications, this corresponds to a voter/client supporting the party/candidate which is closest to them. Knowing this client behavior, the goal of each facility is typically to position itself to maximize the total fraction of clients using their facility. If multiple facilities choose the same location, then they typically share the clients equally. More precisely, let u_{i} represent the fraction of clients which use location x_{i} , i.e., the fraction of clients for whom x_{i} is the closest location of any point with a facility at it; and let n_{i} be the number of facilities at x_{i} . Then, the utility of a facility located at x_{i} is defined to be u_{i}/n_{i} : it is the total fraction of clients which use that location, divided by the number of facilities located there. Many important variations of the above classic model have been studied for many diverse applications, as we discuss in the Related Work section. However, an important aspect missing from existing work is that of facility synergy . In many realistic applications, a client does not simply wish to go to the closest facility but prefers to go to a location with multiple facilities if possible. Consider, for example, someone who wishes to go shopping for a luxury good, such as designer shoes. They could go to a nearby store, but they may not find something that suits them. Or they could instead go to a collection of stores all near each other, even if these stores are farther away. Similarly, when people need to buy goods without knowing exactly what they are looking for, they are much more likely to go to a shopping mall with its many options than simply to the nearest store. Such behavior does not only apply to shopping, of course. When a group of friends wants to go to a restaurant or for entertainment but does not want to decide on a place in advance, they are likely to go to a place with a large collection of options, even if it is farther away. When joining a group (such as patients choosing a hospital or a medical practice, or graduate students choosing a research group), there is a benefit to joining a larger group since there will be many options and more support, even if there is a smaller group which is technically closer to your interests. This also applies to political parties: while a small party may exist which is closer to your views, people will often join larger parties due to the benefits which come from having a large number of powerful representatives who are members of your party. In this paper, we begin the study of facility synergy by considering the simplest model possible, which we believe captures the basic fundamental differences with the classic model without synergy. To model the fact that clients prefer to go to a location with more facilities but still care about the distance as well, we define the client cost function c(x,x_{i})=|x-x_{i}|/n_{i} to be the cost that a client positioned at x experiences for going to location x_{i} with n_{i} facilities located there. Each client then uses the location that minimizes their cost c(x,x_{i}) , not simply the distance. Thus, clients prefer to use closer facilities, but at the same time prefer locations which contain more facilities. The utility of the facility u_{i} is still defined as before: it is the total fraction of clients who use location x_{i} , and thus the utility that each facility receives is still just u_{i}/n_{i} . 1.1 Our Contributions As done previously for the classic Hotelling-Downs model, our goal in this paper is to understand the structure of Nash equilibrium solutions which result from the self-interested behavior of the facilities. We are especially interested in the relationship of Nash equilibrium solutions with the optimal facility placement, as quantified by Price of Anarchy and Price of Stability measures. 1 1 1 The Price of Anarchy [ 23 ] is the ratio between the cost of the worst Nash equilibrium and the cost of the optimal facility placement, while the Price of Stability [ 3 ] is the same ratio but for the best Nash equilibrium instead of the worst one. In other words, we are interested in quantifying how much the self-interest of the facilities hurts the clients as compared to facilities being placed by some central and altruistic decision maker. Here, the cost of a solution is defined as the total cost of all the clients, as in most previous work. 2 2 2 Note that the total sum of the utilities of the facilities for our model is always the same since the game is constant sum, due to the fact that every client uses some facility location. This is in contrast to previous work such as [ 17 ] . However, the cost of the clients can vary greatly depending on the solution. Although similar to the classic model, the difference in client behavior in our model results in very different properties of Nash equilibria and behavior of the facilities. This mainly stems from the fact that, unlike in the classic model where a set of clients using a particular

Read the complete text