The role of distance in the peer-to-peer market for tourist accommodation

Abstract

This article provides evidence on the concentration of peer-to-peer tourist accommodations in the center of cities and the role of distance. On that basis, an explanatory model is proposed to understand the locating decisions of the different agents involved. The model is empirically implemented through a two-stage least squares regression, which allows estimating the elasticity of demand with respect to price and distance. Results for the Spanish cities of Barcelona and Madrid confirm similar price elasticity of demand in both (2.2 and 2.4, respectively) but greater sensibility of demand with respect to distance to the center in the former.

Keywords

accommodation Airbnb distance elasticity 2SLS

Introduction

The taste that tourists have for locating close to the places that they visit and the consequent central location of hotels in cities have long been documented (Arbel and Pizam, 1977). Nonetheless, the recent outburst of peer-to-peer platforms for tourist accommodations, such as Airbnb, added a new dimension to the phenomena that, in many aspects, presents some gaps of knowledge. In this context, the evidence on spatial distribution models of supply in the peer-to-peer market is still limited. Gutierrez et al. (2017) demonstrated a high correlation between hotels and new forms of accommodation for the case of Barcelona, even though the high flexibility of short-term touristic rentals allows for greater proximity to touristic attractions. In EY (2015)’s report, with data on 12 Spanish cities, the short-term rental supply is shown to be more concentrated in the old town relative to hotel supply (and, in the case of coastal cities, in maritime fronts). Hotels, as they cater to a higher proportion of business travelers, tend to be more located often in zones of high financial or economic activity, in which, in turn, the relative presence of Airbnb is smaller.

The objective of this article is to identify the variables that condition the spatial distribution of peer-to-peer accommodation within a city. The cities of Madrid and Barcelona constitute our reference framework. Both of them present different spatial dynamics (Dowling, 2016). Madrid, as the state capital, dominates the political, administrative and financial arenas partly favored by the radial transportation infrastructure of the country. Barcelona, situated at the Northeast coast of Spain, rivals with Madrid as a cultural and economic capital and underwent a phenomenal transformation in the last decades to become one of the tourists’ favorite European cities. In these Spanish cities, the development of new forms of tourist accommodation has sparked growing social response, as their marked geographical concentrations are aggravating the ground-level perception of the negative externalities of tourism.

This article is organized as follows. First, statistical sources are detailed. Then, we describe the evidence on the importance of distance to the city center in the spatial distribution of peer-to-peer accommodation supply. On that basis, we propose an explanatory model. Afterward, we offer an empirical approximation, to quantify the elasticity of demand to price and distance, based on the method of two-stage least squares (2SLS). Finally, we expose some conclusions and limitations of this work.

Data

We have taken the ads listed on the digital platform Airbnb.com as a measure of the supply of accommodation in the peer-to-peer market. Since its founding in 2008, Airbnb has experienced exponential growth which, according to the company itself, has led to its product being offered in more than 65,000 cities in over 190 countries around the globe and reach a number of guests exceeding 160 million (Airbnb, 2017), becoming de facto the leading virtual marketplace for tourist accommodation. Exchanges realized through this platform do not feature in official tourism and hospitality statistics, which renders its analysis complicated. Nevertheless, data collection is possible through web scraping of publicly available information, as performed by the website InsideAirbnb (http://insideairbnb.com/) since its founding in early 2015, with the goal of facilitating study and research.¹ Such data turns singularly useful for the execution of our intended empirical approximation. Among the cities of the world for which information is provided, we find Barcelona and Madrid, of which we will use their latest update at time of this writing (from April 8th of 2017). Besides, the City Halls of Barcelona and Madrid own Statistical Departments that compile data on multiple characteristics (sociodemographic, labor, economic, etc.) that appear not only disaggregated by districts and neighborhoods, but presented with greater detail, and have been used for variables relating to the sociodemographic, urban and real estate environment of the neighborhoods in which accommodations are to be found.

The aforementioned variables are presented in Table 1. Among them, we find three sociodemographic variables (aging, foreigners, and education), three real estate characterizers (house density, recency, and house price), and one economic indicator (unemployment). The rest of variables (P, p, Q, q, and distance) relates to Airbnb and is better described in the subsequent section.

Table 1.

Description of variables.

Variable name	Description	N (Barcelona)	N (Madrid)
P	Average accommodation price^a per night in euros	72	126
p	Logarithm of P	72	126
Q	Review-weighted sum of potential accommodates per neighborhood per inhabitant	72	126
q	Logarithm of Q	72	126
Aging	Ratio of 65-year-old or older population to population between 0 and 15	73	128
Foreigners	Share of population of foreign origin over total	73	128
Education	Share of population with a post-secondary education diploma over total	73	128
House density	Number of homes per hectare	73	128
Recency	Average construction year of buildings	72	128
House price	Average house price per square meter built in euros	66	115
Unemployment	Share of unemployed over total working-age population	73	128
Distance	Euclidean distance, in kilometers, from the average accommodation to the center of the city^b	73	128

Source: Compiled from www.insideairbnb.com (08/05/2017); Barcelona City Hall, http://www.bcn.cat/estadistica/castella/documents/barris/index.htm; Madrid City Hall, http://www.madrid.es/portales/munimadrid/es/Inicio/El-Ayuntamiento/Estadistica/Distritos-en-cifras/Distritos-en-cifras-Informacion-de-Barrios-?vgnextfmt=default&vgnextoid=0e9bcc2419cdd410VgnVCM2000000c205a0aRCRD&vgnextchannel=27002d05cb71b310VgnVCM1000000b205a0aRCRD.

^aPer guest plus additional accommodates included without surcharge.

^bPlaza de Cataluña for Barcelona and Puerta del Sol for Madrid.

Toward an explanatory model

Before building a theory, we first examine the descriptive empirics produced by our data. Upon quick observation, two stylized facts are noteworthy. First, the exponential growth that Airbnb has experienced since its recent emergence is clearly appreciable in Figure 1, which shows that supply of accommodations was practically nonexistent before 2010, in both Barcelona and Madrid, but steadily grew afterward, exceeding the striking figures of 20,000 ads in Barcelona and 15,000 in Madrid. These are equivalent, respectively, to 80% and 40% of the city’s traditional hotel supply (hotels and pensions).

Figure 1.

Evolution in the number of Airbnb ads in Barcelona and Madrid. Source: https://www.airdna.co.

The second remarkable fact is about the spatial concentration of supply. Choosing a center of importance in each city (Plaza de Cataluña in Barcelona and Puerta del Sol in Madrid), it can be quickly noted in Figure 2 accommodations happens to cluster around these. In fact, over 80% of supply of each city is to be found within a 3-km radius around them. Such unequal distribution of supply suggests analogous gaps in the neighborhoods conditions. It is worth noting that in two neighborhoods, Airbnb ads were not found in both cities. Barcelona’s mean ratio of accommodation per house nearly doubles that of Madrid (0.073 and 0.038, respectively). However, given the unequal administrative fragmentation of land, Madrid’s most accommodation-dense neighborhood (Sol, with 0.99 places per house) almost doubles Barcelona’s maximum (Barri Gòtic, with 0.58).

Figure 2.

Spatial distribution of the listings in Barcelona and Madrid coordinates crossing in Plaza de Cataluña (Barcelona). Latitude (abscissae), 41,386,935. Longitude (ordinates), 2,169,961. Coordinates crossing in Puerta del Sol (Madrid). Latitude (abcissae), 40,417. Longitude (ordinates), −3,703,552. Source: Compiled from www.insideairbnb.com.

For the subsequent analyses, we will no longer use supply, but rather, the total number of review-weighted accommodations’ guest capacity per capita in a neighborhood. We denote this value by Q and its logarithm by the lowercase q = log Q(similarly, we denote the level of price by P and its logarithm by p = log P). We weight accommodations proportionally to each one’s amount of reviews, in order to have a better measure of actually realized market transactions.² We include neighborhood population in the denominator because zones with more population might just supply more if each individual person has similar propensity to participate in peer-to-peer hosting than the rest of the population, so correlation of a variable with the absolute number of review-weighted accommodates may just be spurious; however, if we use review-weighted accommodates per capita, we discard what is just due to higher concentration of individuals.

As we can see in Table 2, most environmental variables, p and q, are highly interrelated. Therefore, a more throughout analysis is needed. An ordinary least squares (OLS) regression can help us control for the linear correlations between the variables to isolate the relationships of each variable to accommodation demand, given the others.

Table 2.

Pearson correlation matrix of neighborhood-level variables.

(a) Barcelona
	q	p	Aging	Foreigners	Education	House density	Recency	House price	Unemployment	Distance
q	1
p	0.327***	1
Aging	0.344***	0.0283	1
Foreigners	0.327***	−0.155*	0.0342	1
Education	0.390***	0.692***	0.0303	−0.417***	1
House density	0.485***	0.0208	0.528***	0.281***	0.0787	1
Recency	−0.719***	−0.181**	−0.681***	−0.472***	−0.0989	−0.601***	1
House price	0.572***	0.675***	0.215**	−0.252***	0.920***	0.200**	−0.393***	1
Unemployment	−0.291***	−0.635***	−0.0371	0.453***	−0.938***	0.0117	0.0151	−0.868***	1
Distance	−0.616***	−0.0984	−0.390***	−0.296***	−0.190**	−0.606***	0.588***	−0.371***	0.0385	1
(b) Madrid
	q	p	Aging	Foreigners	Education	House density	Recency	House price	Unemployment	Distance
q	1
p	0.573***	1
Aging	0.0469	−0.187	1
Foreigners	0.453***	0.0287	−0.238**	1
Education	0.608***	0.545***	−0.00845	0.0476	1
House density	0.135	−0.320***	0.381***	0.221*	−0.164	1
Recency	−0.491***	0.0291	−0.192	−0.654***	−0.277**	−0.194	1
House price	0.574***	0.850***	−0.148	0.134	0.881***	−0.161	−0.208*	1
Unemployment	−0.391***	−0.296**	−0.0684	0.136	−0.803***	0.0609	0.229*	−0.705***	1
Distance	−0.854***	−0.469***	−0.270**	−0.371***	−0.641***	−0.311***	0.576***	−0.617***	0.509***	1

*p < 0.10.

**p < 0.05.

***p < 0.01.

And, in fact, once all variables are integrated into a general model (Table 3), only a few variables result significant: In Barcelona, it is the logarithm of price (notably, positively; thus unlikely capturing the price elasticity of demand, but rather, the higher valuation of tourists for certain neighborhoods) and the distance (negatively, as expected). In Madrid, also the distance is significantly negative, but one more variable shows such sign and significance: The recency of buildings, indicating possible higher tourist demand for lodging in historical parts of the city. Regardless, the explanatory capacity is high: R² reaches 0.752 in Madrid and 0.823 in Barcelona.

Table 3.

OLS regression of log-demand on log-quantity and neighborhood-level environmental variables.

	Barcelona	Madrid
	q	q
p	0.858** (0.289)	−0.185 (0.271)
Aging	−0.000516 (0.00254)	−0.00232 (0.00209)
Foreigners	0.0269 (0.0194)	0.0516 (0.0275)
Education	0.00553 (0.0174)	0.0378 (0.0194)
House density	−0.00103 (0.000870)	0.000469 (0.00143)
Recency	0.0115 (0.00934)	−0.0427 (0.0127)
House price	−0.000205 (0.000159)	0.0000207 (0.000284)
Unemployment	0.117 (0.0735)	0.0243 (0.0966)
Distance	−0.704*** (0.100)	−0.165** (0.0527)
Constant	−28.53 (18.64)	78.17** (26.34)
N	66	115
Adjusted R ²	0.823	0.752

Note: Standard errors in parentheses. OLS: ordinary least squares.

*p < 0.05.

**p < 0.01.

***p < 0.001.

In light of the previous results, we can attest the importance of distance in the market for peer-to-peer tourist accommodation. Indeed, for traditional lodgings, abundant empirical evidence is concerned with the impact of location (most of the time, relative to relevant attractions and/or the city center) on price. Certain studies found that proximity to the center increases hotel prices (Bull, 1994; Chen and Rothschild, 2010; Monty and Skidmore, 2003; Schamel, 2012; Thrane, 2007), while others did not find a significant relationship (Arbel and Pizam, 1977; Carvell and Herrin, 1990; Hung et al., 2010; Lee and Jang, 2012). The divergence may be due to methodological differences and the idiosyncrasy of particular cities.

As for peer-to-peer accommodation, Wang and Nicolau (2017) utilized data on 33 cities around the world to find that accommodations’ prices fall 0.59% per kilometer away from the city center. And, for the case of Barcelona, Benítez-Aurioles (2017) finds a sharper fall, of 2.21%, for the equivalent distance. The fall of prices with distance to the city center can be explained by both supply- and demand-side factors. q falls as we move away from the center: Each kilometer of distance to it supposes, on average, a 0.704% decrease (in Barcelona) and 0.165% (in Madrid) of visited accommodations per neighborhood resident. If we assume that tourists prefer accommodations that allow them to walk to places of interest (which tend to be located in the city center), then, we can explain greater density of tourist accommodation by greater demand. Since supply is unlikely to be either perfectly elastic or inelastic, greater demand may reflect not only in increments of prices but also of quantities.

The theoretical framework

In this section, we aim to model why tourist accommodation happens to be concentrated in the center of the city. A model explaining land use by residents, built from the seminal contributions of Alonso (1964), Mills (1967), and Muth (1969), is ought to Duranton and Puga (2014), which is based on the monocentric urban model. Some of the conclusions are that houses are smaller and more concentrated in the center, bigger in the periphery, and, as anyone can move freely, everyone enjoys the same utility no matter where they live. Housing rents are the mechanism to compensate for larger commuting costs; peripherical residents have to pay lower housing rents than central ones, which means that they can enjoy larger houses for the same income.

Tourists are not residents. We also observe more concentration of tourist accommodation in the center of the city, but accommodations do not seem to be “better” in the periphery. In fact, in our Barcelona data for which we have the share of surface area devoted to tourism and hospitality and the one dedicated to residential use, we observe that the former decreases more sharply with distance than the latter (Table 4).

Table 4.

Regression of distance on share of surface area devoted to tourism and hospitality and share of surface area devoted to residential use in Barcelona.

	β
Tourist surface	−0.1164573 (0.0317587)
Resident surface	−0.0422807 (0.0119317)
Constant	6.174766 (0.5265389)
Adjusted R ²	0.2273
N	72

Be it proximity to the center or any other generally desirable attribute (e.g. space and privacy), these attributes are enjoyed as long as one lives in that property. That is to say, unlike a meal or a haircut, consumption is not made at one point of time but continuously. Thus, home attributes are, in some way, durables.

But, unlike in the resident land use model, we do observe fewer accommodations in more eccentric locations, but they are not significantly cheaper nor better equipped, if not slightly worse and more expensive on average. If resident citizens are willing to move to the outskirts of a city, paying more commuting costs in exchange for lower housing rents, why are tourists not?

One explanation that may seem intuitive, but never formalized to our knowledge, is that the shorter one spends living in a place, the less home attributes matter; and distance might also matter less, but its importance decreases less with the brevity of the stay than does that of home attributes. A theoretical model will be presented in what follows to theorize this notion.

To frame this problem, we will consider closeness rather than distance, so that we can use the standard logic of a good rather than a bad. We denote closeness by c, such that c = R − d, where R is the radius of the city. This variable takes maximum value of R when the home is located in the center and 0 when it is in the circumference. Agents see homes as bundles of two goods: Proximity to the center, c, and, for simplicity, a single home attribute x.³ Agents only differ in their t, $t \in (0; + \infty)$ , which represents their time of stay, and is fixed exogenously by nature according to some probability distribution. Key to our model is that the fraction of the day that a person stays home is a function of t. We denote this function α(t) and we postulate that

\frac{\partial α (t)}{\partial t} > 0

That is, as agents stay longer in a city, the average daily time spent inside their home increases. Moreover

\frac{\partial^{2} α (t)}{\partial t^{2}} < 0

That is, α(t) is concave. It makes sense as we want a continuously derivable but bounded function for the entire support of t, because a “fraction of the day” cannot increase indefinitely. Also

lim_{t \to \infty} α (t) \leq 1

The market is competitive, and closeness and the home attribute have prices ρ and β, respectively. ρ may be the price that makes residents indifferent between living anywhere in the city; but, for tourists, this compensation for distance may not be large enough for them to be willing to settle on the outsides. We write the daily utility function as

U (α (t) x, c)

and the budget constraint is

β x + ρ c \equiv P \leq I

where I is an exogenously given level of daily income and P is our notation for the total price of the accommodation chosen. Naturally, marginal utility of either good is positive, $\frac{\partial U}{\partial x (t)} > 0 and \frac{\partial U}{\partial c} > 0$ , and decreasing, $\frac{\partial^{2} U}{\partial x {(t)}^{2}} < 0 and \frac{\partial^{2} U}{\partial c^{2}} < 0$ (alternatively, U is concave on x and c).

We will impose that the utility function is separable on x and c, so that U can be written as

U (α (t) x, c) = f (α (t)) x + g (c)

Note, thus, that α(t) is not factoring c. There are reasons why we choose not to weight closeness increasingly with the time of stay. First, residents may be more tolerant to distance than tourists, such as the existence of monthly and yearly transportation cards that discount the price, and the ease created by habit that spreads the fixed cost of learning the routes. The disadvantage of being distant to the center as a long-time inhabitant is the higher total amount of commuting to take over time. Yet, we do not have reasons to think that the daily frequency of trips to the center of residents and tourists (one two-way trip to the center) may be different—if anything, residents may travel to the center less on average per day because they may not feel the need to go the center on nonworking days.

The Lagrangian of the problem in equation (4) with constraint (5) is

L = U (α (t) x, c) - λ (β x + ρ c - I)

and the first order conditions are

\frac{\partial L}{δ x} = α (t) \frac{\partial U}{\partial x} - λ β = 0

\frac{\partial L}{δ c} = \frac{\partial U}{\partial c} - λ ρ = 0

\frac{\partial L}{δ λ} = β x + ρ c - I = 0

Taking equations (8) and (9)

α (t) \frac{\partial U}{\partial x} \frac{1}{β} = \frac{\partial U}{\partial c} \frac{1}{ρ}

which is the same as, calling u_x and u_c the marginal utilities of x and c, respectively, and because, we assumed that U is separable on x and c

u_{x} (x) \frac{ρ}{β} = \frac{u_{c} (c)}{α (t)}

Here, it is easy to see that the higher α(t), either closeness is lower or attributes are higher. Then, so far we cannot justify our claim that tourists will be more concerned with closeness than residents do. Isolating c

c = u_{c}^{- 1} (α (t) u_{x} (x) \frac{ρ}{β})

c is a function of a ratio of constants $(\frac{ρ}{β})$ , times a product, $α (t) u_{x} (x)$ . We established that α(t) is increasing with t, and in turn, x may be increasing with t (let us write x(t) to indicate this possible dependency). We established as well that α(t) is concave (2), and that U is concave on x(t). We require the further condition that

\frac{\partial u_{x} (x (t))}{\partial x (t)} > \frac{\partial^{2} α (t)}{\partial t^{2}}

That is, α to be more concave on t than U is on x(t). Then, α(t) dominantly drives down the product with t. Because u_c is decreasing too, $u_{c}^{- 1}$ is decreasing, and thus closeness is a decreasing function of time spent.

Empirical strategy

We have already determined why distance is notably more relevant to tourists than to locals. The fact that we included population in the denominator of our accommodation consumption variable Q, combined with the exposed theory, justifies why we should expect a negative estimated effect of distance on demand for accommodation: Even if local population were significantly concentrated in the center and if locals and tourist did not have significantly different tastes for distance versus other home attributes, the fact that we divide weighed accommodates by population should neutralize any correlation between tourist demand and distance. If, however, our model is accurate, then we would expect the coefficient on distance in the demanded quantity for accommodates to be negative, because that would mean that tourists have a marked taste for central locations, even relative to natives.

We assume the following demand function

q_{d} = β_{0} + β_{1} p + β_{2} d + u

Where q_d is the (log) demand for accommodation, p is the (log) price of an overnight, d is distance to the center, and u are unobservable consumer taste shifters.

q_{s} = γ_{0} + γ_{1} p + γ_{2} z + ∊

where q_s is the (log) supply for accommodation, and z and ∊ are factors that influence production that are, respectively, observable and unobservable.

Equilibrium imposes

q_{d} = q_{s} = q

Unlike p and q, the values for z, u, and ∊ are not determined by the model; they follow some externally given probability distribution. And given their realization, we can find p and q from the model

p = \frac{1}{γ_{1} - β_{1}} (β_{0} + β_{2} d + u - γ_{0} - γ_{2} z - ∊)

q = β_{0} + \frac{β_{1}}{γ_{1} - β_{1}} (β_{0} + β_{2} d + u - γ_{0} - γ_{2} z - ∊) + β_{2} d + u

We cannot obtain the demand-price elasticity β₁ from simply regressing p on q, because the OLS estimation formula for the slope coefficient on p will not produce an unbiased estimate of the true β₁

\frac{Cov (p, q)}{Var (p)} = β_{1} + \frac{Cov (p, u)}{Var (p)}

To overcome this problem, we require some variable z to instrument p . An instrumental variable needs to satisfy two conditions. The first is relevance, that is, it must be correlated with the variable to be instrumented

Cov (z, p) \neq 0

The second is exogeneity; that is, it must only be correlated with p through q and be orthogonal to all other determinants of demand

Cov (z, u) = 0

From equation (21), we can find β₁ as

β_{1} = \frac{Cov (z, q)}{Cov (z, p)}

From equation (16), if γ₂ ≠ 0, we have essentially one instrument that is valid for our purposes. Here, z corresponds to factors that affect the cost of providing accommodation that matter to the hosts but not to the tourists.

We will use the method of 2SLS. Thus, we will run a first stage in which we regress z on p (with some controls). From it, we obtain $\hat{p}$ , and we run a second stage in which we regress $\hat{p}$ on q (with some controls). Naturally, the second stage’s coefficient on $\hat{p}$ should be negative, as it corresponds to the price elasticity of demand.

Implementation and results

The instrument that we will use is the share of foreigners in a neighborhood. Theoretically, we can easily think that foreign population has a productivity advantage in providing Airbnb accommodation, if they are more familiar with the cultures and languages of the visitors. On the other hand, the origin of the host may only be of minor interest to the guest. Empirically, we find in both cities that the share of foreign population is positively correlated with demand (0.3272 in Madrid and 0.6612 in Barcelona) but negatively correlated with price (−0.1553 in Madrid and −0.0826 in Barcelona). It is important to control for distance in both stages of 2SLS, because both foreigners and tourists happen to be located in a distinctive manner. Moreover, we add one extra relevant control in the first stage that is specific to each city. Because, in Madrid, aging population happens to be concentrated with foreigners (correlation 0.0342) and older people may also have differential productivity in providing Airbnb hosting (presumably, below average as opposed to foreigners), we need to control for the aging ratio. This phenomenon does not happen in Barcelona (correlation −0.2381), so we instead include unemployment because coincidence of high unemployment and high foreign share is prevalent in both cities, and the economic conditions of residents may as well affect Airbnb hosting productivity (although we are uncertain in which direction).

2SLS results are presented in Table 5 for Barcelona and in Table 6 for Madrid. As expected, the second stage coefficient on $\hat{p}$ is negative in both cases and of similar magnitude: −2.2 in Barcelona and −2.4 in Madrid. In turn, the coefficient on distance in Barcelona is around the triple that of Madrid’s. A likely explanation is that tourists may care about relative rather than absolute distance, and indeed Madrid spans much more than Barcelona, where in fact the maximum accommodation-center distance found in our data (21.1 km) in Madrid is almost the triple than that of Barcelona (8.9 km).

Table 5.

2SLS regression (Barcelona).

	Stage 1 (log price)		Stage 2 (log demand)
Foreigners	−0.0130 (0.0194)	log $\hat{p}$	−2.200* (0.896)
Unemployment	−0.000650 (0.0885)
Distance	−0.169 (0.113)	Distance	−0.969*** (0.128)
Constant	5.120*** (0.147)	Constant	8.658* (4.266)
Adjusted R ²	0.211		0.743

Note: N = 72. Standard errors in parentheses. 2SLS: two-stage least squares.

*p < 0.05.

**p < 0.01.

***p < 0.001.

Table 6.

2SLS regression (Madrid).

	Stage 1 (log price)		Stage 2 (log demand)
Foreigners	−0.0210* (0.00883)	log $\hat{p}$	−2.401* (1.171)
Aging	−0.000312 (0.000819)
Distance	−0.0303* (0.0152)	Distance	−0.334*** (0.0618)
Constant	4.420*** (0.256)	Constant	5.680 (4.770)
Adjusted R ²	0.25		0.399

Note: N = 72. Standard errors in parentheses. 2SLS: two-stage least squares.

*p < 0.05.

**p < 0.01.

***p < 0.001.

Conclusions

Stating that distance to the places to visit matters for tourists’ accommodation choices is trivial to some extent. Nonetheless, explaining why such high concentration exists around the city centers requires analytical thinking. Traditional models based on residents’ preferences are insightful for urban development but fall short for the unequal spatial distribution of tourist accommodation. The model that we propose in which the “time of stay” of home users has a nontrivial effect in their utility function helps us understand both long-term residents’ and short-term tourists’ decisions regarding where to live.

The proposed methodology permits an empirical approximation whose results are compatible with the theorized framework of reference. Moreover, the methodology allows the estimation of the distance and price elasticity of demand for the cities under study. The latter is closely similar in both Madrid and Barcelona, being slightly above 2. On the other hand, the distance elasticity of demand in Madrid (0.969) is almost the triple of Barcelona’s (0.334), suggesting a possible influence of the city’s configuration on the importance of location of accommodations in the peer-to-peer market.

From our perspective, a new line of research to explore would be to replicate this analysis on other cities, to verify to which extent the sensibility of demand to the price of accommodations is similar across them and perhaps deepen into the explanation of the importance of distance to the decisions of tourists. From the technical angle, we acknowledge two limitations. The first one concerns the instrument used, which has different weaknesses, although also different strengths, in each of the cities: While the share of foreigners is significant for the accommodation prices in Madrid, unlike in Barcelona, the first stage performed for the state capital does not explain as much variation as it does in the Catalan city. Nonetheless, there is hardly ever a single correct instrument in such contexts and we hope that this research may inspire other neighborhood-level proxied cost instrumental strategies to advance the research on informal accommodation markets. The second limitation concerns our measure of demand, or rather, consumption of peer-to-peer tourist accommodation. As Airbnb does not disclose information on the length of stays, nor on the actual number of visitors—had they left a review or not, our measure of consumption may contain a bias. In particular, the second bias (an underestimation) would not be as problematic as the first one, whose direction to us is unknown.

Footnotes

Declaration of conflicting interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Airbnb (2017) About us. Available at: https://www.airbnb.com/about/about-us (accessed 13 July 2017).

Alonso

(1964) Location and Land Use: Toward a General Theory of Land Rent. Cambridge: Harvard University Press.

Arbel

Pizam

(1977) Some determinants of urban hotel location: The tourists inclinations. Journal of Travel Research 15(3): 18–22.

Benítez-Aurioles

(2017) Price and Booking Determinants in the Peer-To-Peer Market of Tourist Accommodation. The case for Airbnb in Barcelona. Master Thesis, CEMFI, Spain.

Bull

(1994) Pricing a motels location. International Journal of Contemporary Hospitality Management 6(6): 10–15.

Carvell

Herrin

(1990) Pricing in the hospitality industry: An implicit markets approach. FlU Hospitality Review 8(2): 27–37.

Chen

Rothschild

(2010) An application of hedonic pricing analysis to the case of hotel rooms in Taipei. Tourism Economics 16(3): 685–694.

Dowling

(2016) A tale of two cities: Madrid and Barcelona in Spain. In Cole

Payre

(eds) Cities as Political Objects: Historical Evolution, Analytical Categorisations and Institutional Challenges of Metropolitanisation. Cheltenham: Edward Elgar, pp. 79–95.

Duranton

Puga

(2014) The growth of cities. In: Aghion

Durlauf

(eds) Handbook of Economic Growth, 2(vol). Amsterdam: North-Holland, pp. 751–843.

10.

EY (2015) Impactos sociales y económicos sobre los destinos españoles derivados del exponencial aumento del alquiler de viviendas turísticas de corta duración, impulsado por los nuevos modelos y canales de comercialización P2P. Available at: http://www.exceltur.org/wp-content/uploads/2015/06/Alojamiento-tur%C3%ADstico-en-viviendas-de-alquiler-Impactos-y-retos-asociados.-Informe-completo.-Exceltur.pdf (accessed 7 August 2017).

11.

Gutierrez

Garcia-Palomares

Romanillos

. (2017) Airbnb in tourist cities: Comparing spatial patterns of hotels and peer-to-peer accommodation. Tourism Management 62: 278–291.

12.

Hung

Shang

Wang

(2010) Pricing determinants in the hotel industry: Quantile regression analysis. International Journal of Hospitality Management 29(3): 378–384.

13.

Lee

Jang

(2012) Premium or discount in hotel room rates? The dual effects of a central downtown location. Cornell Hospitality Quarterly 53(2): 165–173.

14.

Mills

(1967) An aggregative model of resource allocation in a metropolitan area. American Economic Review Papers and Proceedings 57(2): 197–210.

15.

Monty

Skidmore

(2003) Hedonic pricing and willingness to pay for bed and breakfast amenities in southeast Wisconsin. Journal of Travel Research 42(2): 195–199.

16.

Muth

(1969) Cities and Housing. Chicago: University of Chicago Press.

17.

Schamel

(2012) Weekend vs. midweek stays: Modeling hotel room rates in a small market. International Journal of Hospitality Management 31(4): 1113–1118.

18.

Thrane

(2007) Examining the determinants of room rates for hotels in capital cities: The Oslo experience. Journal of Revenue and Pricing Management 5(4): 315–323.

19.

Wang

Nicolau

(2017) Price determinants of sharing economy based accommodation rental: A study of listings from 33 cities on Airbnb.com. International Journal of Hospitality Management 62: 120–131.