4.1 Cities
We analyzed eight different cities located within the U.S. These vary substantially among each other – for example, in terms of size, population composition, and wealth, as described later in this section. We chose to focus on the U.S., as this country hosts a diverse range of cities, with mature Airbnb presence across many of them. Future studies may wish to explore to what extent the findings that hold within a country also span across different ones.
San Francisco. It is the city where Airbnb was founded in 2008 and is currently headquartered. As Airbnb’s hometown, it offers insights into the most developed Airbnb marketplace. Furthermore, it is the second most densely populated U.S. city and is home to many technology entrepreneurs who work in the nearby heart of the U.S. technology scene, the Silicon Valley. It is a very ethnically diverse city, has a very high average age and, despite having high median income, has a large disparity between the rich and poor.
Oakland. Unlike San Francisco, it serves as a center for trade and is the busiest port in California. Despite its close proximity to San Francisco, the characteristics of Oakland’s demographic makeup differ considerably and median pay is roughly two thirds that of San Francisco’s.
Manhattan. It is the most densely populated borough of New York City. It is also the city’s economic and administrative center, and it is often described as the cultural and financial capital of the world. Manhattan has the highest cost of living in the U.S., and also contains the country’s most profound level of income inequality. The majority of the population is white (65%), and approximately 27% are foreign born.
New Orleans. In stark contrast to Manhattan, New Orleans is the smallest of the chosen cities, with a population of 378,000, predominantly black (60.2% ). The city has seen a decline in population in recent times. As further proof of contrast to Manhattan, the median income of the city is $26,900 (2010 U.S. Census), to Manhattan’s $72,200, almost three times greater.
Austin. It differs vastly to both the metropolis of Manhattan and the quaint New Orleans. Austin is the fastest growing city of the top 50 largest U.S. cities and is not so ethnically diverse. The majority of Austin’s population is white (66.8%). It is also the youngest city in the dataset.
Seattle. The Pacific Northwest city of Seattle, in Washington, is an important center for technology, being home to Amazon, Microsoft, and Boeing. It is also a major gateway for trade with Asia. Like Austin, it is a predominantly white city. However, it is far older, has a much higher median income, and a greater cost of living.
San Diego. It is the third major city in our dataset (with a population greater than 1,000,000). The city, which has an immediate proximity to the Mexican border, is not a technology hub like Seattle or New York. Its main economic engines are the military and tourism. Due to its closeness to Mexico, it has a large Hispanic population and a low proportion of black people (6.7%).
Los Angeles. It is a global center of commerce and has a diverse economy in business, technology, culture and sport. It has the highest educational diversity in the country and ranks highly on the diversification of its economy business-sectors. Despite its size and economic power, it has a low median income and a disproportionately high cost of living.
Table
1 lists the eight cities chosen for this study (first column), and also summarizes their varying social and economic characteristics (next five columns), in terms of: population, median age, median income, percentage of white population, and cost of living – estimated from consumer prices of goods and services relative to the reference urban area of Manhattan [
23].
Table 1
Summary characteristics of the 8 chosen U.S. cities
4.4 City geography
Distance to Center. A previous Airbnb study [
13] of the city of London, UK, found that distance to the city center was one of the variables that most explained Airbnb presence in an area (i.e., the closer to the city center, the more Airbnb listings). We aim to explore whether the same holds for U.S. cities. Some of the analyzed cities (such as San Diego, Oakland and Seattle) are relatively small with a clear definition of city center. For other cities this may be not true and they may contain multiple urban hubs [
24]. For simplicity, we computed a single metric across all cities; specifically, we consider the ‘downtown district’ or CBD (central business district) as the center of the city. For each city, we compute distance to center as the shortest distance in meters between the CBD’s center, and the center of the tract under study.
Points of Interest. A point of interest (POI) is a geographic feature that might be useful or interesting. Examples of POIs include pubs, town halls and post offices. A study of the geography of Airbnb in London [
13] found that, together with ‘distance to center’, the ‘tourism factor’ of an area, as shown by the number of POIs within an area, had the greatest positive significance on the number of Airbnb offerings in that area. We expect that the relationship will hold for American cities too, such that areas of higher POI concentration, indicating greater tourist appeal, will also have increased Airbnb presence. To count the number of POIs within a given area, we used OpenStreetMap data; specifically, for each city, we extracted the latitude/longitude coordinates for all POIs that fell under the following OpenStreetMap categories: accommodation, attractions, eating and drinking, retail and sports, and entertainment.
Number of Hotels. Despite a previous analysis showing that in London there is little relationship between hotels and Airbnb adoption [
13], we do not know
a priori whether the same conclusion holds in U.S. cities as well. Airbnb’s economic blog, which reports and measures Airbnb’s effect on city economies, states that 72% of Airbnb properties in San Francisco are outside the central hotel district [
25]. However, little other evidence exists relating the spatial penetration of Airbnb listings to that of hotels. Intuitively, the number of hotels in an area should provide a reasonable proxy for the level of tourism of that area. Furthermore, results highlighting where Airbnb listings appear in a city relative to hotels will provide regulators with a source of quantitative information to make more informed decisions. We thus explore this variable in our analysis. Since there is no publicly available dataset for the number of hotels in all cities, hotel data was crawled from Google, searching for ‘city_name’ + ‘hotels’, and then retrieving their latitude-longitude pairs.
Bus Stops. The strength of an area’s infrastructure and transport links have historically been a key component in the performance of property prices, due to the ease of connection to major areas of that city. For tourists visiting a city, although they may spend time and money in tourist centers, their choice of where they stay is likely influenced by the connectivity of an area. Different cities may offer a variety of different public transport modalities. Since buses are present in all cities under study, we chose the number of bus stops in an area as proxy to the strength of said area’s transport links. Thus, we expect to see a relationship between Airbnb offerings and the number of bus stops. To compute this metric, we used a combination of OpenStreetMap data and city-specific datasets to obtain the latitude-longitude of bus stops; we then counted the number of stops within each area.
Population Density. This is a standard metric derived from the U.S. Census Bureau that provides information on how densely populated a specific area is. It is widely used as general statistical datum at the country as well as at the local level. It is calculated by dividing the number of people living in a certain area by the area’s total surface. Population density is an aspect considered crucial by many urbanists in explaining a number of urban aspects [
26‐
28]. Recent studies have found that this factor is linked to the spread of sharing economy services [
29]. We thus decided to include it as one of our geographic attributes.
4.5 Social indexes
Race Diversity Index. The Race Diversity Index is a metric derived from the U.S. Census Bureau; it provides a measure of how much racial diversity exists in an area. First coined by Meyer and Macintosh [
30], it is formulated as a Gini–Simpson Index [
31] and acts as a probability measure. It measures the likelihood that two people selected at random from a given area represent different types. In this case, it is a measure of whether the race of the chosen people is the same. We formulate the problem with seven distinct racial categories: white, black or African American, Hispanic or Latino, American Indian or Alaska native, Asian, native Hawaiian or Pacific Islander. The greater the race diversity index, the greater the probability that two people selected at random will be from different races.
Income Diversity Index. The income diversity index shows how diverse an area is in terms of average household income for the population of that area. It is derived from the U.S. Census Bureau and it is calculated using the Gini–Simpson index [
31] for three distinct wage bands: low income (annual incomes less than $35,000), middle band income (annual incomes between $35,000 and $100,000) and high income (annual incomes greater than $100,000).
Bohemian Index. A bohemian is a socially unconventional person with interests in art or literacy (
https://en.oxforddictionaries.com/definition/bohemian). Richard Florida’s paper “Bohemia and Economic Geography” [
6] examines the relationship between geographic concentrations of bohemia and a strong technology presence by directly measuring the bohemian population at an MSA (Metropolitan Statistical Area) level. Though there are other variations of the bohemian index [
32], we use Florida’s definition, which computes the proportion of the number of bohemians to the number of residents in an area, compared to the national proportion of bohemians to the number of the total population. We derived the Bohemian Index from the U.S. Census Bureau.
Talent Index. The talent index [
33] measures the education level of a populace, defined as the proportion of people with a bachelor’s degree or above. The index is normalized per thousand people and it is derived from the U.S. Census Bureau. Richard Florida hypothesizes that a high talent index is correlated with a larger concentration of bohemians. Given this, we may infer that areas with a strong technology presence, such as those areas with high Airbnb uptake, will have a higher index for talent.
Proportion of Young People. This was calculated as the proportion of people aged between 20 and 34 years old in a given area against the population of that area. Florida suggests that, as well as the bohemian index, a higher concentration of young people is often a driver of the technology uptake in that area [
6]. We derived this index from the U.S. Census Bureau.
4.6 Economic indexes
Unemployment Proportion. The unemployment proportion is calculated as the number of people aged 16 and over currently out of work (unemployed) against the total number of people in an area. This measure is provided by the U.S. Census Bureau. Unemployment rates often provide a strong indication of the economic health of an area. According to Florida’s work on the ‘creative class’ [
7], areas of lower unemployment (amongst other factors) are symbolic of a creative class, and transitively may lead to greater technology concentration. We would thus expect to see a negative correlation between Airbnb penetration and unemployment proportion. However, the Wall Street Journal [
34] found a large percentage of Airbnb renters were offering up living spaces due to unemployment. In Paris, only one third of Airbnb hosts were reported to have full time jobs [
35]. If the relationship holds across the U.S. too, then we may see a positive correlation between unemployment and Airbnb penetration instead.
Poverty by Income Percentage. Michael Zweig [
36] defines poverty as “a state of deprivation, or a lack of the usual or socially acceptable amount of money or material possessions”. In the U.S., the most common poverty metrics are the ‘poverty thresholds’, as defined by the U.S. Census Bureau [
37]. Our explanatory variable is then calculated, in a given tract area, as the percentage of households in poverty (as defined by their income) against the total number of households in that area. The underlying hypothesis is that Airbnb’s penetration will fall in areas of increased poverty.
Median Household Income. For each tract area, the U.S. Census Bureau measures the median household income for the local population. A temporal study on Airbnb in London [
13] showed that income became increasingly more negative correlated with Airbnb penetration over time, signaling that more people with low income were joining Airbnb as hosts, possibly using the extra income generated from Airbnb to support themselves.
Median Household Value. The U.S. Census Bureau also provides a measure of median household value for each area. Together with median household income, this variable should provide a strong indicator of socio-economic makeup of a city. This can also be used to identify clusters of cities with similar profiles.
Proportion of Owner Occupied Residences. Quattrone
et al. [
13] found that, in London, Airbnb hosts tend to rent rather than own the property. Therefore, we hypothesize that the proportion of owner occupied residences matter in the U.S. as well. We derived this metric from the U.S. Census Bureau.
Table
2 summarizes all the metrics introduced in this section, along with the sources from which they were taken.