Content uploaded by Zhenpeng Zou
Author content
All content in this area was uploaded by Zhenpeng Zou on Sep 02, 2020
Content may be subject to copyright.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
1
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
2
ABSTRACT
As on-demand short-term rentals (STRs) grow popular at the rise of sharing platforms like
Airbnb, regulations on the STR market have become center of the debate among policymakers,
housing interest groups, hotel & lodging industry, and STR platforms. Washington D.C., the
nation’s capital and one of the most popular tourist destinations in the U.S., is on the frontline
of legalizing and regulating the STR business. With the heated policy debate over whether
STRs disrupt the rental housing market in D.C., a concrete discussion about what STRs impose
on the owner housing market is left out. Using web-scraped data from Airbnb and property
level data from the city, I investigated the net impact of STRs on single-family property prices
through a series of hedonic analyses. The results suggest that having Airbnb establishments in
the neighborhood can significantly inflate property prices. Due to the uneven spatial market
penetration of STRs, such price impact could inequitably affect low-income homebuyers and
add another hurdle to resolve housing affordability issue faced by policymakers in Washington
D.C.
KEY WORDS:
Short-term rentals (STRs), Airbnb, Housing price, Hedonic Analysis
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
3
1. From niche to mainstream: the global and local rise of STRs
Charles Dickens would probably reckon, had he lived in the 21st century, that “It is the
best of times; It is the worst of times – for sharing”: We hail a ride with strangers in an Uber;
we sit in a cubicle next to an entrepreneur at WeWork; we, certainly, dare an adventure to stay
with other travelers in an Airbnb rental. The ideology behind “sharing” is collaborative
consumption – a concept built upon a set of principles, such as a critical mass of idling
capacities, belief in the commons, and trust in strangers (Sundararajan, 2016). The sharing
economy is a utopia painted by some as a solution to the underutilized resources in our society
and a dystopia suspected by many as a road to digital elitism (Kenny and Zysman, 2016).
The global success of on-demand short-term rental (STR) platforms like Airbnb highlights
the phenomenal sharing economy. Thanks to the advancements of information and
communication technologies (ICTs) and the advent of an integrated (matching, booking,
payment, etc.) peer-to-peer marketplace, searching cost for STRs has notably decreased for
both the demand and the supply side (Einav et al., 2016). Contrary to a centralized economy,
where transactional cost is lowered through economies of scale, the sharing economy creates a
decentralized market that facilitates heterogeneous product choices (Einav et al., 2016). In
addition, crowd-based networks and access without ownership remove the hurdle for ordinary
people to participate in the sharing economy, blurring the boundary between a personal
property and a professional establishment (Sundararajan, 2016).
In the global context, the soaring sharing economy translates into a rapid STR market
expansion: Since its first booking in 2008, Airbnb has accumulated more than five million
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
4
listings in 191 countries around the world and accommodated more than 300 million guests in
the past decade (Airbnb, 2018). In the local context – the study area of this paper, Airbnb
entered Washington D.C. in 2009 and other platforms like HomeAway and VRBO followed
suit. A typical STR host accommodates guests 32 days out of a calendar year and makes an
average income of $3,400, according to a survey conducted by Airbnb in 2016 (Airbnb, 2016).
As of August 2017, the number of Airbnb listings in Washington D.C. exceeded 8,000 based
on web-scraped Airbnb data
1
. The number of listings peaked around the inauguration of the
Trump Administration and the following Women’s March in the middle of January 2017, when
hundred-thousands of visitors gathered in the nation’s capital to witness those historical
moments (New York Times, 2017). When filtered by whether a listing has a review, an
indicator of STR business activities (Barron et al., 2017), active Airbnb listings grew steadily
in number. Figure 1 shows time trends for the total number of listings accessible through
Airbnb.com and the number of listings with at least one review from August 2015 to July 2017.
According to an Airbnb’s report (2017), 88% of the hosts in Washington D.C. share space in
their permanent home. In 2016, 7,100 entire home listings hosted at least one stay. In another
report (Airbnb, 2016), the platform claimed that 76% of its hosts rent out their primary dwelling
for STR activities. Cross-referencing different data sources, I come up with the following first
impressions of STRs in D.C.: (1) Washington D.C. is an emerging STR market, owing to its
unique status as the nation’s capital and its numerous tourist attractions; (2) The majority of
1
The main data source for this paper is from the Inside Airbnb website supported by Tom Slee
(http://insideairbnb.com/about.html). I appreciate his data collection efforts, both in terms of frequency and quality.
However, the data collection process stopped by the mid of 2017. According to another source, Airdna, current
number of Airbnb listings in Washington D.C. fluctuates around 7,000. This could be a result of market saturation,
policy uncertainty, or a combination of both.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
5
Airbnb’s thousands of listings were “registered” under a primary residential dwelling, though
Airbnb (or other STR platforms) never revealed the number of additional listings registered by
a single host or whether all hosts complied with local zoning codes, which may strictly prohibit
STRs at certain locations; And (3) there is a sizeable commercial STR market, in which the
primary function of a property is STR business instead of long-term rental or residency.
Spatially, STR listings tend to cluster at tourist hot spots and mixed-use residential areas.
I plotted two kernel density maps of Airbnb listings at two points in time (February 2015 and
February 2017) based on web-scraped, geocoded Airbnb listing data
2
(See Figure 2a and
Figure 2b). In addition to clusters in the densely populated historical and commercial
neighborhoods, STRs also expanded to residential neighborhoods in the Northwest, the
Northeast, and across the Anacostia River (Southeast) within a two-year span. Such a market
expansion can be intriguing as the east side of D.C. is traditionally a less heated housing market
with a noticeable growth in recent years (the Washington Post, 2018a).
Innovations in business and technology oftentimes outpace legislation that confines the
boundaries of their practice. Once a niche market product, STRs are no exception. While
triumphed by many who profited in the sharing economy, STR platforms increasingly clash
with cities as issues, such as illegal listings and unmannerly guest behavior, start to make
headlines. The central research question in this paper asks whether the thriving STR business
in Washington D.C. is a significant factor that drives up single-family property prices in the
2
According to the declaimers on Inside Airbnb, the locational information of an Airbnb listing that is publicly
available on airbnb.com is typically within a 450-feet distance from its actual address to protect anonymity of a
host’s information. This is not problem for the purpose of this study because Airbnb listing is characterized as a
“density” attribute within a certain buffer distance.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
6
owner housing market. In addition, it is vital to understand which neighborhoods are most
impacted by STRs, especially the neighborhoods with high shares of racial minorities.
Many issues and discussions about STRs are described in literature. In the following
section, I thoroughly review broader literature on this novel yet controversial topic with a focus
on the welfare impacts STRs have imposed on different communities.
2. STR Literature Review
2.1. Virtues and Vices of STRs
STRs only became a popular research subject recently because of its novelty. Early
research focused on descriptive analyses of successes and setbacks of the STR business model:
By adopting a trust and reputation system, STR platforms managed to minimize the potential
risks of sharing with strangers (Frenken and Schor, 2017; Abrahao et al., 2017). On the other
hand, a rating system could introduce unintended statistical and social biases due to information
asymmetry. For instance, Zervas et al. (2015) found that ratings on Airbnb were
overwhelmingly positive, disguising variations in service quality. In addition, STR platforms
introduced a two-sided feedback system for guests and hosts, where ratings were usually
inflated out of fear of retaliation (Tadelis, 2016). Fradkins et al. (2017) conducted two field
experiments to improve the effectiveness of the rating system for Airbnb. They found that both
financial rewards and simultaneous reviews could readily eliminate strategic reciprocity in the
STR rating process.
While addressing the importance of designing a robust rating system for STR platforms,
researchers also found worrisome evidence where social biases were held against STR
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
7
participants of color. Edelman et al. (2017) implemented an audit experiment on Airbnb and
found a significantly higher number of booking request rejections against African American
guests as compared to Caucasian guests. In addition, black hosts were found to earn
significantly less rent from STRs than their white counterparts after controlling for housing
conditions and location factors (Edelman and Luca, 2014). STR platforms claimed that they
were not liable for such social biases as a result of their ambiguous policies on user profile
photos and listing descriptions (Edelman and Luca, 2014). As allegations against
discriminatory cases accumulate, public appeals for regulatory measures to hold STR platforms
accountable for nondiscriminatory business conducts also increase.
Having observed the global success of STRs, researchers in tourism and hospitality tried
to assess how this emerging market would impact the traditional lodging industry. Zervas et al.
(2017) suggested that Airbnb could be responsible for 8% - 10% revenue loss for traditional
hotel chains in Austin, TX. The new competition from STR platforms, however, can
substantially benefit consumers as lodging cost is brought down (Guttentag, 2015). It is no
surprise that the incumbent hotels and lodging establishments will defend their business
interests by pursuing legislation/regulation against the disruptive STRs. A major argument
against the platforms is that they essentially created a deregulated market without enforcing
regulation, such as business registration, on their participating hosts (Guttentag, 2015).
Unlicensed accommodation providers could impose safety and public health risks on guests
(Gurran, 2018). Furthermore, unlicensed STR listings could escape tax liabilities, providing an
unfair advantage against traditional lodging establishments that obey tax rules (Gurran, 2018;
Guttentag, 2015). This tax issue is typically resolved through tax agreements between a city
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
8
government and STR platforms, allowing a city to collect hotel-like taxes on each booking
(Bibler et al., 2018). Yet, it is not commonly practiced at all levels of city governments in the
U.S., especially in small cities (DiNatale, et al., 2018).
2.2. STRs’ Externalities
In addition to affecting subscribers and the hospitality industry, STRs also impact the
welfare level of the broader community through externalities. Externalities exist naturally as
the market is imperfect. While subscribers (hosts and guests) and STR platforms are tied to a
legally binding contract, non-subscribers cannot hold platforms accountable for their behavior.
Neither can non-subscribers invoke market incentives, such as withholding their patronage, to
change platforms’ behavior (Edelman and Geradin, 2016).
In the context of STR, the most obvious externality comes from changes to quality of life.
Neighborhood quality, unbounded by ownership, could fall victim to a “tragedy of the
commons”, such as constant interruptions to the neighbors from STR guests, over-consumption
of rivalrous public goods (e.g. parking space), and reckless guest behavior (e.g. hosting loud
parties) (Edelman and Geradin, 2016). Filippas and Horton (2017) theoretically articulate that
negative home-sharing externalities cannot be entirely internalized in a “tenant decide” regime.
The externalities associated with STRs are complicated in that they are both “technological”
(i.e. spillovers) and “pecuniary” (Scitovsky, 1954). Technological externalities of STRs refer
to the social cost incurred by STR guests and borne by the public. Pecuniary externalities of
STRs, on the other hand, refer to the overall housing price and value changes as a result of the
advent of STRs in a city (Filippas and Horton, 2017). Empirically, quantifying externalities is
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
9
a difficult task due to its non-market nature. Hedonic pricing is a popular empirical approach
for non-market goods valuation, which implicitly embeds non-market locational characteristics
into determinants of property prices/values (Rosen, 1974).
While policymaking towards eliminating technological externalities is straightforward,
such as restrictions against the use of STRs for events and zoning compliance (e.g. Office of
Short-Term Rentals San Francisco, 2018), policymaking towards remedying pecuniary
externalities involves a complicated planning issue. Specifically, STR platforms are
condemned for exploiting the affordable rental housing stock that could have been rented by
long-term renters and for inflating rent and property value (Gurran, 2018; Gurran and Phibbs,
2017; Edelman and Geradin, 2016). Pecuniary externalities are a product of interdependence
among members of the economy. They cannot be resolved by simply applying policy tools to
move the economic equilibrium from the private optimum to the social optimum (Scitovsky,
1954). A change in policy to address pecuniary externalities, such as restricting the number of
listings per host, is likely to change the dynamics of the entire STR market. A summary of STR
externalities is provided in Figure 3 (modified from Sheppard and Udell, 2016).
Unlike green space or air pollution, which can be unambiguously categorized as an
amenity or a dis-amenity to quality of life, having STRs in a neighborhood can be considered
both an amenity and a dis-amenity. What’s revealed through differences in property
prices/values is the net effect of STR externalities. Recent empirical results suggest that STRs
seem to boost property values or rent (Wachsmuth and Weisler, 2018; Horn and Merante, 2017;
Wachsmuth et al., 2017; Sheppard and Udell, 2016), indicating that the positive externalities
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
10
associated with STRs dominate the negative ones.
Previous literature theorizes potential mechanisms of STRs’ positive impact on property
prices. STRs offers an extra income that can help property owners hold onto the ownership for
longer as the cost of ownership is reduced (Sheppard and Udell, 2016). This extra income
stream is capitalized into property prices (Barron et al., 2017). This is a plausible mechanism
in particular for those who would have been evicted from their property due to financial
struggles. In addition, STRs could generate new interests in real estate investment: Urban space
becomes more valuable as tourists and residents take advantage of STRs (Sheppard and Udell,
2016). With limited urban land supply for new development, investors will seek to convert the
existing housing stock into STRs, bidding up property prices and making life more difficult for
first-time homebuyers and long-term renters. This is exactly what Wachsmuth and Weisler
(2018) described as “gentrification without redevelopment”: A rent (price) gap emerged as a
result of a strong tourist demand for STRs. A strong economic incentive followed for real estate
investors to evict existing long-term tenants or to cash out existing homeowners. They then
converted properties into STRs without building anything new.
2.3. STRs’ Housing Market Implications
Empirically, existing research reached an early consensus that the advent of STR
platforms, such as Airbnb, resulted in net increases in either property prices or rent (Barron et
al., 2017; Horn and Merante, 2017; Sheppard and Udell, 2016; Wachsmuth et al., 2017). As
new evidence emerged, the debate intensified over whether STRs exacerbated the housing
affordability crisis in major U.S. and international cities. Nevertheless, a lack of robust rental
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
11
housing transaction data made it difficult for housing policy researchers to produce fruitful
results to stir up a conversation. Previous analyses on rental data are aggregated either at the
census tract level (e.g. Horn and Merante, 2017) or the metropolitan area level (e.g. Barron et
al., 2017). No property/parcel level rental housing analysis exists at this point to my knowledge.
Many STR proponents found the argument of a direct substitution between STRs and long
term rentals unconvincing. A report on the impact of Airbnb on the Portland housing market
suggested that “somewhere between 83 and 377 units (or 0.03% of the total housing stock in
Portland) would be considered full-time Airbnb rentals (ECONorthwest, 2016).” It is unclear
whether restraining the number of full-time STR listings in a city could significantly shrink the
rental housing shortage. Opponents of unregulated STRs focused on the issue in regard to
commercial STR hosts, who rented multiple listings for an extended number of days in a year
(from three months to all year round). According to a local nonprofit organization, more than
1/3 of all listings in D.C. could be categorized as “commercial listings” (D.C. Working Families,
2017). In Canada, researchers found that 13,700 “entire homes” out of 81,000 Airbnb listings
were rented more than 60 days a year in Montreal, Toronto, and Vancouver (Wachsmuth et al.,
2017). The definition of “an entire home” is tricky, since it does not necessarily mean that the
property owner lives elsewhere. In the ECONorthwest’s report (2016), a fair observation was
made that the definition of “entire home” from Airbnb also includes (a) accessory dwellings
attached to a property, (b) parts of a property with a separate entrance and private rooms, or (c)
a basement unit without a separate entrance. In addition, a property owner can list multiple
bedrooms as multiple listings on the platform, contrary to the D.C. report’s argument that a
commercial host must have rented out more than one property. As Wachsmuth et al. (2017)
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
12
point out, current observations about Airbnb are based on third-party information and data
sources (e.g. web-scraped data). Any statement with a high level of confidence would require
data from STR platforms directly with accurate details.
The rest of the paper is organized as follow: In Section 3, I highlight controversies around
STRs in Washington D.C. and ongoing efforts towards regulating the STR market. In Section
4, I summarize Airbnb and property data used in the analysis. Empirical frameworks and results
are presented in Section 5. Robustness checks are provided in Section 6. Lastly, I discuss the
policy implications and conclude the paper in Section 7.
3. STR Controversies and Regulations in the District of Columbia
3.1. Growing STR Business amid Controversies
There are no doubts that STR platforms like Airbnb provide economic benefit to D.C.
residents. However, the relationship between STR platforms and the city is not always cheerful.
A major concern about STRs is that commercial hosts occupy precious housing resources that
could have housed long-term renters in the city. In a defense from Airbnb (2017), the platform
argued that only 0.22% of the “entire home” listings were booked for more than half a year in
2016. In addition, the average monthly income for an STR host ($680) is only a fraction of the
average monthly rental income in D.C. ($2,299)
3
. Therefore, from an economic perspective,
part-time STRs, which consist of 60% of all entire home units, can hardly substitute long-term
rentals.
3
According to Insider Airbnb, the estimated full-time STR monthly income is about $986
(http://insideairbnb.com/washington-dc/), still much lower than the average rental price (even for a studio).
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
13
Another concern regarding STRs’ housing market impact has to do with its spatial
concentration around tourist hot spots. Areas like downtown and Capitol Hill are real estate
heavens and attract heavy tourist traffic. It is, to state the least, worrisome that STRs may
significantly change the housing market dynamics in these areas. If a property price premium
is transmitted to the rental housing market in such areas, then long-term renters will have to
endure inflating rent as a spillover from increasing housing prices.
Other stories unfolded that STR platforms barely regulated their hosts on business
registrations or compliance with local zoning ordinances, such as the strict condominium rules
that prohibit short-term sublets (the Washington Post, 2017a). In one case, several apartment
buildings were converted illegally into full-time STRs as opposed to being leased to long-term
renters (Greater Greater Washington, 2017). STR platforms were not well-received by all.
Therefore, the city government decided to step up and intervene in the unregulated STR market.
3.2. D.C.’s STR Regulatory Framework
In January 2017, Kenyan McDuffie, city councilmember representing Ward 5, first
introduced the Home/Short-term Rental Regulation and Affordable Housing Protection Act of
2017 (B22-92), heralding the first official attempt to legalize and regulate STRs. Both
proponents and opponents fiercely exchanged their stands during the first public hearing in
April 2017 over the current state and practice of STRs in D.C. and to what extent the STR
business should be regulated. (Council of the District of Columbia, 2018).
The initial proposal was not well-received as STR platforms and subscribers described the
bill as “goes too far and is too restrictive” by capping the number of days in a year for STR
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
14
operation to 15 days (the Washington Post, 2017b). After inaction for more than a year, the city
council moved forward the legislation in October 2018 with significant amendments to the
original bill: An STR listing is capped to have 90 business days per calendar year; The monetary
penalty on violations is reduced; Any STR listing located outside of a host’s primary residence
requires a license for operation (Council of the District of Columbia, 2018). The city council
passed the bill unanimously in November 2018, marking the end of an era of unregulated STRs
in Washington D.C.
Table 1 highlights the legislative contexts of B22-92. It also describes approved STR bills
and ordinances from the neighboring counties, including Arlington County, VA, Prince
George’s County, MD, and Montgomery County, MD. Through parallel comparisons, we can
observe many similarities amongst these legislations: A STR is defined as the transient
occupancy of a residential dwelling (owned or rented); A business license is acquired
conditional upon inspections from the regulatory body; Only the primary residence is allowed
for STRs, where physical presence of the residents is required for at least 180 days in a calendar
year; The maximum number of STR days in a calendar year and the maximum number of
guests are specified. On the other hand, these bills and ordinances also differ from each other:
While both two counties in Maryland and the D.C. government passed jurisdictional bills,
Arlington County (VA) only revised its zoning ordinances. Having the zoning commission
enforcing the ordinances with the power to suspend or revoke a permit may yield better
enforcement outcomes, but it could also cause an administrative burden. D.C.’s STR bill
remains the most restrictive in terms of the 90-day cap for STR (as opposed to 120 days or 180
days) with special exemptions. In addition, B22-92 is the only bill that specifies the penalty on
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
15
each violation. In response to the legislative approval, STR platforms quickly denounced the
council’s action and warned to bring the case directly to a 2020 ballot initiative (the Washington
Post, 2018b).
If passing a legislation on STRs requires year-long efforts, then enforcing STR regulations
entails administrative readiness and coordination. Underprepared implementation of STR
regulations results in unintended consequences. One such consequence is a cumbersome
registration process. As one of the first cities to pass an STR legislation in 2016, San Francisco
only registered 2,168 Airbnb hosts as of early 2018, leaving the majority of its 8,000 STR
listings with no legal status (San Francisco Chronicle, 2018). Similarly, eight months after the
legislation took effect, Arlington County government only issued 101 transient rental permits
on an estimated 1,600 STR owners base (INSIDENOVA, 2017). If the low registration rate is
a mixed outcome of uncooperative STR owners and inefficient administrative procedures, then
the existence of unregistered/commercial listings heightens a lack of regulatory enforcement.
Airdna’s (2018) data suggest that 5,778 Airbnb listings in San Francisco remain active, despite
the fact that the municipal STR bill has been in effect for two years. Should the platforms be
fined for listing unregistered STRs? Should the city go after each unregistered STR owner?
These unresolved issues are common to municipal lawmakers and governing bodies
everywhere, including the District of Columbia.
In the housing policy debate over STRs’ impact on D.C.’s housing market, a missing piece
of the puzzle is how STRs could impact property owners and homebuyers. In the following
sections, I will empirically investigate this issue using unique open source data.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
16
4. Empirical Data
4.1. Data Sources
Airbnb data: While data from STR platforms are almost impossible to acquire, third-party
web-scraped data have become popular for research purposes (e.g. Wegmann and Jiao, 2017).
Web-scraped STR data are subject to some limitations, such as the use of location proxies. Yet,
such data provide a comprehensive set of information about an available listing, including
listing amenities and reviews. Through real-time data scraping, researchers can describe STR
activities subject to a degree of discretion. Researchers either design their own scraper (e.g.
Barron et al., 2017) or rely on third-party scrapers, such as Inside Airbnb (e.g. Gurran and
Phibbs, 2017; Horn and Merante, 2017) and Airdna (e.g. Wachsmuth and Weisler, 2018). In
this study, I used data collected by Tom Slee from September 2014 to July 2017
4
.
Six web-scraped Airbnb datasets at half-year intervals were combined to represent Airbnb
listings in D.C. from early 2015 to mid-2017. The half-year intervals deliberately take into
account seasonal fluctuations in tourism (March – August are typically the popular months for
D.C.). While the data don’t cover the initial entry of Airbnb into the D.C. market, they cover
the period when the STR business took off in D.C. (recall Figure 1).
Housing data: Housing information came from the Open D.C. data portal with
periodically updated property sales records and city-wide housing appraisal records. The
appraisal data provide underlying housing attributes, such as number of rooms, bathrooms,
4
The scraper operator, Tom Slee, stopped Airbnb data collection after the summer 2017 due to an overwhelming
number of requests. He directed requestors to other open data sources like Inside Airbnb.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
17
stories, square-footage, and the estimated building year. Property sales records from the
Integrated Tax System Public Extract (ITSPE) and appraisal data from the Computer Assisted
Mass Appraisal (CAMA) database were extracted and combined using a unique identifier,
Square Suffix Lot (SSL). After trimming the dataset by matching criteria, completeness of
attributes, and exclusions of extreme values, I derived the final dataset of property sales records
during September 2014 – July 2017.
Neighborhood data: Aside from housing attributes, neighborhood characteristics are also
deterministic in hedonic prices. I included the most important attributes in the final dataset,
such as access to Metrorail stations, public schools, and historical landmarks. In addition,
underlying population attributes at the census tract level were extracted from the American
Community Survey database and were incorporated into the final dataset.
4.2. Data Processing
Due to the size of the housing datasets, neither sales records nor appraisal data were
geocoded. I applied the Master Address Repository (MAR) geocoder to geolocate each SSL
within the ITSPE database by a 92 percent matching criterion. Only 7,334 out of the 110,883
records were dropped due to low matching rates. The ITSPE data were then merged with the
CAMA residential data based on a unique identifier, Square Suffix Lot (SSL). 52,577 single-
family property sales records were successfully matched
5
. 12,680 records between September
2014 and July 2017 were kept in the final single-family housing dataset.
5
Another 39,886 records were matched for condominium and multifamily sales records. Condominium data were
excluded from this study due to unobserved attributes (such as condominium management quality) that could be
crucial in determining their prices.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
18
I measured “Airbnb density” by counting the number of listings within a certain buffer
distance of a property sales point at a given period of time. Four buffer sizes were included in
the analyses: 100 feet, 200 feet, 500 feet, and 1,000 feet. The choice of buffer size is a state of
art: While a smaller buffer captures a STR’s most direct impact on a property’s price, a larger
buffer allows for more variations in “Airbnb density” and captures the broader economic
impact of Airbnb activities on the neighborhood. As a comparison, Sheppard and Udell (2016)
tested different buffer sizes from 200 meters (656 feet) to 2,000 meters (6,560 feet). Some
studies also calculated “Airbnb density” at the aggregated level, such as census tracts (Horn
and Merante, 2017). I did not include a buffer size smaller than 100 feet or a buffer size larger
than 1,000 feet because (a) the variation in Airbnb density was insignificant for a smaller buffer
and (b) the neighborhood impact of a single listing was too weak for a much larger buffer. With
an increasing buffer size, more listings will be included, but the listings farther away from the
centroid will have a smaller impact on property prices. Figure 4 illustrates the “Airbnb density”
at different buffers in the ArcGIS environment.
4.3. Summary Statistics
Summary statistics of the final dataset are presented in Table 2. The average number of
Airbnb listings within 100 feet of a single-family property sales point is 0.21. The variation is
small for this search radius that it may affect the precision of the point estimate in the hedonic
regression model. The “Airbnb density” increases to 0.85, 5.06, and 18.63 for the 200 feet, 500
feet, and 1,000 feet search radiuses from a property sales point, respectively. In theory, the
marginal effect of each Airbnb listing on a property’s price will decay as the buffer size
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
19
increases. Therefore, I anticipate a declining magnitude in hedonic point estimates for the
Airbnb density variable for a larger buffer.
The sample average single-family property price is $762,000 and the median price is
$630,000, higher than the median home value in D.C. of $544,000 in 2017
6
. The sample
average property land area is 3,000 square feet (sqft) and the average structure area is about
1,700 sqft, with 7.5 rooms, 2.2 bathrooms, 0.6 half-bathrooms, and 1.2 kitchens. In addition,
basic amenities are usually equipped, such as a fireplace, an air-conditioner, and a heating
system.
As for neighborhood attributes, a typical property resides in a populated urban area with
heavy traffic (as indicated by the number of crash incidents within a half-mile buffer) and some
crime incidents. A property usually has a good access to public schools within walking distance
(0.5 miles). A property also has an easy access to a Metrorail station and commercial areas. In
Washington D.C., it is especially common to have historical landmarks in the neighborhood.
Such amenities can have significant impacts on property prices.
In terms of neighborhood demographics, a typical D.C. property is located in a
neighborhood of an employed, educated, middle-class population. However, the population
demographics differ significantly by zip code. I carefully controlled for such “zip codes” fixed
effects and STR clustering effects in the models specified in the next section.
I conducted a Pearson’s correlation test
7
to examine the preliminary bivariate relationship
6
The median value for condominium is $440k, but the condominium sample was excluded due to lack of detailed
condominium attributes from the appraisal database.
7
Due to the size of the Pearson’s correlation matrix, I decided not to include it in the final paper.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
20
between “Airbnb density” and property prices and to detect the unusual signs of different
housing and neighborhood attributes in explaining property prices. All “Airbnb density”
variables were positively correlated with property prices, suggesting a net positive externality
from STRs. Most signs of the correlation statistics made sense. No perfect collinearity was
found except for income and education at the Census tract level.
5. Hedonic Analyses of Airbnb on Property Prices
Empirically, the hedonic pricing model is one of the most widely adopted approaches to
study consumers’ willingness to pay a non-market goods. In this study, Airbnb density, defined
by the number of Airbnb listings within a distance from a property sales point, runs into the
regression analyses as a hedonic attribute. I constructed three models to fully investigate
Airbnb’s impact on property prices: a pooled cross-sectional model, a fixed effects model at
the census block level, and a first-difference model.
5.1. Model Specifications
The full-sample cross-sectional model considers the most comprehensive set of
explanatory variables, including housing attributes, neighborhood factors, sociodemographic
attributes at the census tract level, and series of time and location fixed effects. The model is
specified as follows:
Housing price takes a logarithm form to account for the right-skewedness in distribution;
represent housing and neighborhood attributes; represent demographic attributes that
are common to each property i in census tract n.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
21
The census block level fixed effects model controls for unobserved time-invariant
characteristics that may jointly affect housing prices and Airbnb activities, such as commercial
activities, infrastructure, and public facilities. In addition, a time trend is added in the model to
control for common housing market fluctuations over different periods. The model is specified
as follows:
The unit of observation is a representative property at the census block b during period t.
Both block level fixed effects and common time trends are included.
The nation’s capital experienced a historical influx of visitors in January 2017. Both
supporters and protesters congested the city during the Trump administration’s inauguration
and the Women’s March a day after – the latter attracted much heavier traffic. Having sensed
the unprecedented demand for lodging, the local STR community expanded dramatically
between November 2016 and January 2017, from 5,975 listings to 9,097 listings according to
the web-scraped data. This exogenous demand shock created a unique opportunity for me to
conduct a “before/after” type of analysis on how new Airbnb listings/activities affected
property prices.
I selected block-level data between March 2016 and November 2016 for the “before”
period and data between February 2017 and July 2017 for the “after” period. The final dataset
consists of 2,047 observations for 1,027 blocks. I then applied a first-difference model to
understand how changes in Airbnb density fluctuated property prices:
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
22
5.2. Empirical Results
The main estimation results are presented in Table 3. Panel A reports the regression
coefficients and standard errors for the most important variables in the pooled cross-sectional
model. It is evident that (a) having Airbnb listings in the neighborhood mildly raises a single-
family property’s price and (b) the average effect of a listing decays as the search buffer
broadens. Other significant variables also help explain property prices, such as good property
appraisal grades and conditions, having public schools and historical landmarks within walking
distance, as well as dwelling in a wealthy community. The model’s goodness of fit is high with
R2 > 0.80.
Panel B shows regression coefficients and standard errors of the Airbnb listing density
variables for the fixed effects model. The coefficients on the Airbnb densities at the 200-foot,
500-foot, and 1000-foot buffers hold their statistical significance and they are slightly larger in
magnitude than those in Panel A. While the fixed effects model controls for unobserved time-
invariant characteristics at the census tract level, the model’s goodness of fit drops due to
aggregation. Nevertheless, the results from both models suggest a price premium on properties
due to the presence of Airbnb listings in the neighborhood.
Panel C shows hedonic regression results for the first difference model. Again, the
coefficients on Airbnb densities at the 200-foot, 500-foot, and 1000-foot buffers remain
statistically significant. The magnitudes are much larger due to the dramatic increase in Airbnb
density between November 2016 and January 2017. One possible explanation is that the
transition to a new administration led to a temporary spike in housing demand to accommodate
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
23
new residents. Airbnb (and STRs in general) fulfilled the transitional housing need.
5.3. STRs’ Inequitable Impact on Property Prices
To quantify the impact of Airbnb listings on property prices, I calculated the aggregate
impact by multiplying the point estimates from the fixed effects model and the average density
of Airbnb listings for each buffer size. The impacts were then summarized by zip code to
account for the unbalanced spatial distribution of Airbnb listings. The results are presented in
Table 4. In particular, the underlying demographic composition varies significantly across zip
codes in D.C. due to historical redlining (Lloyd, 2016). Certain zip code areas have a much
higher concentration of Hispanic/Latino and/or African American population. Historically,
displacement of the black population was prominent in D.C. (Jackson, 2014). It is vital to
understand whether STRs have significantly impacted people of color in the city.
For the entire city, Airbnb alone could account for an increase in single-family property
price by 0.66% to 2.24%. The impact was mild yet non-trivial. Alarmingly, Airbnb was
accountable for a significant leap (> 5%) in property prices at tourist hot spots, such as
downtown (zip code: 20005), Shaw (20001), Adams Morgan (20009), Dupont Circle (20036),
and Foggy Bottom – George Washington University (20037). These neighborhoods were
already overheated in housing demand due to their advantageous locations. STR-related
housing investment will only aggravate the housing affordability issue.
What’s more unsettling is the fact that Shaw (20001), NOMA – Trinidad (20002), Capitol
Hill (20003), and Columbia Heights (20010) also experienced a noticeable price inflation (>3%)
because of STRs. These zip code areas are populated with Hispanic and African Americans as
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
24
shown in the last two columns of Table 4. While the increasing price is good news to current
homeowners, it puts a potential hurdle to new homebuyers to move into these neighborhoods.
Moreover, it is reasonable to worry that the price premium will be eventually borne by long-
term renters, jeopardizing low-income minority renters who could be displaced from the city.
This is the missing piece previously ignored in the debates over STRs’ housing market
consequences in Washington D.C.: Not only could STR platforms occupy valuable housing
stock, but their business could significantly drive up housing cost in neighborhoods with a
concentrated minority population.
6. Robustness Checks
6.1. Robustness Checks on Active Airbnb Listings
As mentioned in Section 3, housing advocacy groups and other STR opponents were most
concerned about the “entire home” STR listings that might have consumed the existing housing
stock. To inquire into this issue, I further subset the Airbnb listing data by two additional criteria:
(a) A listing was categorized as “entire home”; and (b) A listing had at least one review to signal
its active status. About 70% of the observations were preserved after the additional screening.
After rerunning all three models, I present robustness check results in Table 5. Surprisingly,
while the statistical significance of the regression coefficients and the goodness of fit resemble
the results from those in Table 3, the magnitudes of coefficients are larger for the 100-foot and
200-foot buffers and smaller for the 500-foot and 1000-foot buffers as compared to the results
in Table 3. Such interesting results can be explained by a perfectly reasonable rationale: Active
STR listings have a stronger localized impact on property prices as their activeness indicates
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
25
business success and attractiveness to new investors. On the other hand, the broader economic
benefit usually requires a cluster of listings in a larger buffer area. With fewer listings in a large
buffer, the magnitude of the “Airbnb density” impact declines.
6.2. A Robustness Check on the Rental Housing Market
While the focus of this paper is the single-family owner housing market, it will enrich the
discussion by looking into STRs’ impacts on the rental housing market. I could not access
disaggregated rental transaction data, so the robustness check was done at the aggregate zip-
code level. I used Zillow Rent Index (ZRI), a smoothed measure of the median estimated
market rate rent, across zip codes in Washington D.C. over time for this exercise
8
. When
applied to the same empirical models, the rental data yielded statistically insignificant results
(See Table 6). The most plausible estimate is the coefficient on the Airbnb density at the 200-
foot buffer. The estimate is positive yet statistically insignificant. In addition, Washington D.C.
adopted a strict Rent Control Act, in which any rent hike falls under rent control except for a
few exemptions (such as rental units built after 1975 and Federally/District-subsidized rental
units)
9
. From the housing dataset, 74% of the single-family units and 60% of the
multifamily/condominium units were built prior to 1975, suggesting that the majority of the
older housing units in D.C. fall under the rent control umbrella. This is somewhat reassuring
to the most vulnerable renters in the city. Nevertheless, I acknowledge that thorough and robust
8
See the methodology to calculate the Zillow Rent Index here:
https://www.zillow.com/research/zillow-rent-index-methodology-2393/
9
See the Rent Control Fact Sheet here:
https://dhcd.dc.gov/sites/default/files/dc/sites/dhcd/service_content/attachments/Rent%20Control%20Fact%20S
heet%202018.pdf
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
26
research using high-quality disaggregated rental housing data must be conducted to solve the
rental housing puzzle of STRs’ housing market consequences.
7. Discussion
7.1. Policy Implications
This paper provides empirical evidence on STRs’ impacts on property prices. The topic
has pivotal welfare implications that should not be neglected. Previous attempts to understand
STRs’ housing market impacts in D.C. were descriptive and lacked in rigor. In this paper, I
took advantage of innovative web-scraped Airbnb data and demonstrated the indirect impact
(externalities) of Airbnb listings on single-family property prices through hedonic analyses.
The results suggest that unregulated growth in STR business created an inequitable property
price premium that could distress first-time homebuyers and negatively affect long-term renters
if the price premium results in higher rent.
This study comes out in a particularly meaningful time in the wake of new STR
regulations in the District of Columbia. The lengthy legislative process took almost two years
to finish, with yet another 11 months of transition period before the regulations come into effect.
While stories about how STR business helped struggling families afford their homes in one of
the nation’s most expensive cities (the Washington Post, 2018a) should not be neglected, cities
ought to realize that anxious STR investors can make life much harder for people who are still
seeking a home.
STR regulation should by no means deprive a resident’s right to earn an extra income
through home-sharing. The unanimous criticisms of the stiff cap on STR days in the original
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
27
bill proposal is a living proof. Strict as it still is, the final version allows for a primary dwelling
to be rented 90 days a year. While it is yet to be tested how effectively the regulation will be
enforced, the bill can hopefully cool down STR-related housing investment by prohibiting
commercial listings outside of a host’s primary dwelling. It remains challenging as the city
must get STR platforms on board to make considerable efforts to remove illegal listings. Any
attempt to resolve the conflict between pro-STR and anti-STR communities without a
collaborative approach has no chance to succeed.
From a planner’s perspective, functional zoning ordinances and an effective zoning board
play critical roles in regulating STRs. Table 1 shows that all passed STR legislations revise
zoning ordinances to unambiguously confine a residential property’s STR usage. In the case of
Arlington County, VA, the zoning commission is also the issuer of STR licenses, empowering
the county’s planning body to oversee STR operation and law compliance.
In addition to revising zoning codes, planning and housing authorities should also keep a
keen eye on the affordable housing stock and ensure that the valuable rental housing resources
for voucher holders and other affordable housing program participants are not jeopardized by
illegal or irrational STR investments. On the other hand, there is a silver lining to foster
collaboration between the housing authority and STR platform in home sharing programs (e.g.
HUD, 2016). Rather than treating STRs as a threat to affordable housing, cities could
potentially benefit from the crowd-sourcing technology supported by STR platforms to match
voucher holders and rental housing owners. Cities should embark on the smart city concept by
thinking and acting innovatively to address the existing conundrums. A new type of home
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
28
sharing program through STRs can be a great experiment to produce a social good through the
private-public partnership between a city and STR platforms.
7.2. Limitations and Beyond the Study
I acknowledge that this study cannot directly answer the question: How do STRs gentrify
a city? Gentrification is a complicated issue that goes beyond the scope of partial equilibrium
analyses presented in this paper. We will have to reflect on the money-chasing real estate
development that is by no means affordable to low-income households and racial minorities.
We will also have to ask homeowners why they prefer to invest in the STR business.
Instead, this study confirms a hypothesis that STRs do make it more expensive to own a
property in a tourist paradise like Washington D.C. Moreover, and perhaps more alarmingly,
they have made the historically minority-concentrated neighborhoods more expensive. Due to
the short observation time, the data did not support a parcel level repeated sales model, which
would have been a more robust empirical approach. Nevertheless, all three hedonic models
confirmed that STRs indeed inflated single-family property prices. To put this paper into
perspective, I compared the empirical results to the findings from previous studies: In this paper,
I find a 0.78% increase in property prices with respect to one additional Airbnb listing within
the 200-foot buffer; Barron et al. (2017) find a 0.64% increase in property prices with respect
to a 10% increase in Airbnb listings; And Sheppard and Udell (2016) find a 6% - 9% increase
in property prices when the number of Airbnb listings doubles within the 300-meter buffer,
which translates into a 1.30% - 1.96% increase in property prices with respect to one additional
listing in New York City. Different as methodologies, data, and studies areas are, we come to a
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
29
similar conclusion.
Although I included a robustness check on Airbnb’s price effect on aggregated median
rent at the zip code level, the results are rather inconclusive. Unsurprisingly, the level of
geographic aggregation and the length of the time series both limited the interpretability of the
results. Referring to Barron et al. (2017) and Horn and Merante (2017), I believe that the story
for Washington D.C. is probably not so different; i.e., STRs also drive up rent. Recent studies
using web-scraped Craigslist data (e.g. Boeing and Waddell, 2017) inspire a new research
agenda on STRs’ rental housing market consequences.
Last but not least, hedonic models were only able to allow me to derive the net impact of
“Airbnb density” on property prices. It is unclear what the driving factor is in determining the
positive net externality. Judging from the literature (Wachsmuth and Weisler, 2018), investors
bidding up prices due to the extra income from STR is the more plausible mechanism than the
other two (increasing quality of life and more space demanded by existing property owners).
As a new wave of jurisdictions start to legalize and regulate STRs, it will be interesting to
compare the STR market before and after regulations take effect. One of the greatest debates
of all time is whether innovation and technology improve quality of life. In the case of STR, it
is a housing policy debate centering on an innovation in technology that redefines how we live
and how we travel.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
30
References
Abrahao, B., Parigi, P., Gupta, A., & Cook, K. (2017). Reputation offsets trust judgments
based on social biases among Airbnb users. Proceedings of the National Academy of
Sciences, 114(37), 9848-9853. doi:10.1073/pnas.1604234114
Airbnb (2016). Overview of the Airbnb Community in Washington, DC.
https://www.airbnbaction.com/wp-
content/uploads/2016/10/AirbnbWashingtonDCEconomicImpactReportOctober2201.pdf
Airbnb (2017). Airbnb, Housing, and The District of Columbia.
https://www.airbnbcitizen.com/wp-content/uploads/sites/24/2017/04/2017-04-Airbnb-D.C.-
Housing-Report.pdf
Airbnb (2018). Fast Facts. https://press.airbnb.com/fast-facts/
Airdna (2018) San Francisco Overview. https://www.airdna.co/vacation-rental-
data/app/us/california/san-francisco/overview
Barron, K., Kung, E., & Proserpio, D. (2017). The Sharing Economy and Housing
Affordability: Evidence from Airbnb. Available at SSRN: https://ssrn.com/abstract=3006832
Bibler, A., Teltser, K., & Tremblay, M. (2018). Inferring Tax Compliance from Pass-Through:
Evidence from Airbnb Tax Enforcement Agreements. Available at SSRN:
https://ssrn.com/abstract=3117021
Boeing, G., & Waddell, P. (2017). New insights into rental housing markets across the united
states: Web scraping and analyzing Craigslist rental listings. Journal of Planning Education
and Research, 37(4), 457-476.
Council of the District of Columbia. (2018). B22-0092 – Short-term Rental Regulation and
Affordable Housing Protection Act of 2017. http://lims.dccouncil.us/Legislation/B22-0092
D.C. Working Families. (2017). Airbnb’s Impact in Washington D.C. – Selling the District
Short. http://dcsharebetter.org/wp-content/uploads/2017/03/D.C.-Housing-Report_Web.pdf
DiNatale, S., Lewis, R., & Parker, R. (2018). Short-term rentals in small cities in Oregon:
Impacts and regulations. Land Use Policy, 79, 407-423.
ECONorthwest. (2016). Housing affordability impacts of Airbnb in Portland. Portland, OR:
ECONorthwest.
Edelman, B., & Geradin, D. (2016). Efficiency and Regulatory Shortcuts: How Should We
Regulate Companies like Airbnb and Uber? Stanford Technology Law Review 19, No. 2
(2016): 293 – 328.
Edelman, B., & Luca, M. (2014). Digital Discrimination: The Case of Airbnb.com. Working
paper. https://www.hbs.edu/faculty/Publication%20Files/Airbnb_92dd6086-6e46-4eaf-9cea-
60fe5ba3c596.pdf
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
31
Edelman, B., Luca, M., & Svirsky, D. (2017). Racial discrimination in the sharing economy:
Evidence from a field experiment. American Economic Journal: Applied Economics, 9(2), 1-
22. doi:10.1257/app.20160213
Einav, L., Farronato, C., & Levin, J. (2016). Peer-to-peer markets. Annual Review of
Economics, 8, 615-635. doi:10.1146/annurev-economics-080315-015334
Filippas, A., & Horton, J.J. (2017) The Tragedy of your Upstairs Neighbor: When is the
Home-Sharing Externality Internalized? Working Paper.
Frenken, K., & Schor, J. (2017). Putting the sharing economy into perspective.
Environmental Innovation and Societal Transitions, 23.
Fradkin, A., Grewal, E., Holtz, D. (2017). The Determinants of Online Review
Informativeness: Evidence from Field Experiments on Airbnb. Working Paper.
Greater Greater Washington. (2017). DC wins a big victory against landlords who convert
housing to Airbnb-like hotels. https://ggwash.org/view/65046/dc-wins-a-big-victory-against-
landlords-who-convert-housing-to-airbnb-like-hotels
Gurran, N., & Phibbs, P. (2017) When Tourists Move In: How Should Urban Planners
Respond to Airbnb? Journal of the American Planning Association, 83:1, 80-92.
Gurran, N. (2018). Global Home-Sharing, Local Communities and the Airbnb Debate: A
Planning Research Agenda. Planning Theory & Practice, 19:2, 298-304, DOI:
10.1080/14649357.2017.1383731
Guttentag, D. (2015). Airbnb: Disruptive innovation and the rise of an informal tourism
accommodation sector (vol 18, pg 1192, 2015). Current Issues in Tourism, 18 (12).
Horn, K., & Merante, M. (2017). Is home sharing driving up rents? evidence from Airbnb in
Boston. Journal of Housing Economics, 38, 14-24. doi:10.1016/j.jhe.2017.08.002.
Department of Housing and Urban Development (HUD). (2016). Home Sharing.
https://www.huduser.gov/portal/casestudies/study-09282016-1.html
INSIDENOVA. (2017). Arlington seeing new tax revenue from Airbnb-style rentals.
http://www.insidenova.com/news/arlington/arlington-seeing-new-tax-revenue-from-airbnb-
style-rentals/article_1e191132-8e45-11e7-b384-175da4d36dd0.html
Jackson, J. (2015). The consequences of gentrification for racial change in Washington, DC.
Housing Policy Debate, 25(2), 353-373. doi:10.1080/10511482.2014.921221
Kenny, M., & Zysman, J. (2016). The rise of the platform economy. Issues in Science and
Technology, 32(3), 61-69.
Lloyd, J. (2016). Fighting redlining and gentrification in Washington, D.C.: The Adams-
Morgan organization and tenant right to purchase. Journal of Urban History, 42(6), 1091-
1109. doi:10.1177/0096144214566975
The New York Times. (2017). Crowd Scientists Say Women’s March in Washington Had 3
Times as Many People as Trump’s Inauguration.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
32
https://www.nytimes.com/interactive/2017/01/22/us/politics/womens-march-trump-crowd-
estimates.html
Office of Short-Term Rentals, San Francisco. (2018). About Short-Term Rentals.
https://shorttermrentals.sfgov.org/about
Rosen, S. (1974). Hedonic prices and implicit markets: Product differentiation in pure
competition. Journal of Political Economy, 82(1), 34-55. doi:10.1086/260169
San Francisco Chronicle (2018). Airbnb loses thousands of hosts in San Francisco.
https://www.sfchronicle.com/business/article/Airbnb-loses-thousands-of-hosts-in-SF-as-
12496624.php
Scitovsky, T. (1954). Two concepts of external economies. Journal of Political Economy,
62(2), 143-151. doi:10.1086/257498
Sheppard, S., Udell, A. (2016). Do Airbnb properties affect house prices? Working paper.
Sundararajan, A. (2016). The Sharing Economy: The End of Employment and the Rise of
Crowd-Based Capitalism. ISBN: 9780262034579.
Tadelis, S. (2016). Reputation and feedback systems in online platform markets. Annual
Review of Economics, 8, 321-340. doi:10.1146/annurev-economics-080315-015325
Wachsmuth, D., & Weisler, A. (2018). Airbnb and the rent gap: Gentrification through the
sharing economy. Environment and Planning A: Economy and Space, 50(6), 1147-1170.
doi:10.1177/0308518X18778038
Wachsmuth, D., Kerrigan, D., Chaney, D., Shillolo, A. (2017). Short-term cities: Airbnb’s
impact on Canadian housing markets. http://upgo.lab.mcgill.ca/airbnb/
The Washington Post. (2017a). D.C. sues company for allegedly treating rent-control
apartments like ‘hotel rooms’. https://www.washingtonpost.com/business/capitalbusiness/dc-
sues-company-for-allegedly-treating-rent-control-apartments-like-hotel-
rooms/2017/04/25/e42ebb50-291a-11e7-a616-
d7c8a68c1a66_story.html?utm_term=.06833c8f83ed
The Washington Post. (2017b). Money and liability are among issues driving the D.C. Airbnb
bill. https://www.washingtonpost.com/news/where-we-live/wp/2017/09/18/money-and-
liability-are-among-issues-driving-the-d-c-airbnb-bill/
The Washington Post. (2018a). East of the Anacostia, where there are no hotels, Airbnb has
been a boon for property owners and guests. https://www.washingtonpost.com/local/dc-
politics/east-of-the-anacostia-where-there-are-no-hotels-airbnb-has-been-a-been-a-boon-for-
property-owners-and-guests/2018/10/28/35f6bb14-d7d8-11e8-83a2-
d1c3da28d6b6_story.html
The Washington Post. (2018b). D.C. government to propose zoning changes to permit
Airbnbs in residential neighborhoods. https://www.washingtonpost.com/local/dc-politics/dc-
government-to-propose-zoning-changes-to-permit-airbnbs-in-residential-
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
33
neighborhoods/2018/10/30/c2e160ae-dbbd-11e8-b3f0-
62607289efee_story.html?utm_term=.d4dbb10ed53b
Wegmann, J., & Jiao, J. (2017). Taming airbnb: Toward guiding principles for local
regulation of urban vacation rentals based on empirical results from five us cities. Land Use
Policy, 69, 494-501.
Zervas, G., Proserpio, D., & Byers, J. (2015). A First Look at Online Reputation on Airbnb,
Where Every Stay is Above Average. https://ssrn.com/abstract=2554500.
Zervas, G., Proserpio, D., & Byers, J. (2017). The rise of the sharing economy: Estimating
the impact of Airbnb on the hotel industry. Jmr, Journal of Marketing Research, 54(5), 687-
705.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
34
Appendix
Table 1: Comparison among different STR legislative frameworks in the D.C. Metropolitan Area
Jurisdiction
Washington D.C.
Arlington County, VA
Montgomery County, MD
Prince George’s County, MD
Legislative
Framework
B 22-92 (Proposed bill &
amendments)
Zoning Code 12.9.11 & 12.9.12
Zoning Text Amendment 17-03 &
Senate Bill 2-16
CB-10-2018 & CB-11-2018
Definition
A STR means paid lodging for
transient guests with the host
presence, unless it is a vacation
rental. A STR is not a hotel, inn,
motel, boarding house, or b&b.
An accessory homestay is a special
type of home occupation that allows
the occupant of a residential dwelling
unit to host short-term overnight
guests.
A STR means the residential
occupancy of a dwelling unit for a fee
for less than 30 consecutive days. A
STR is not a Bed and Breakfast.
A STR means a residential dwelling
unit occupied by a STR guest, other
than a permanent occupant, for fewer
than 31 consecutive days and no more
than 90 days per calendar year.
Business license
A license issued by the Department
of Consumer and Regulatory
Affairs. Valid for a period of 2 years.
Accessory homestay permit from the
Zoning Administrator. Renew annually.
A license issued by the director of the
Department of Health and Human
Services is required. Renew annually.
Annual issuance of a license by the
Department of Permitting, Inspections,
and Enforcement.
Zoning ordinance
D.C. Zoning Commission will revise
zoning codes to permit STRs.
Arlington County Zoning Code 12.9.11
and 12.9.12
Montgomery County Zoning Text
Amendment 17-03
CB-10-2018 (Sec.27-464.09 “Tourist
Home as an ‘accessory use’.”)
Days of STRs in a
calendar year
90 days (unless the host has received
an exemption.)
180 days
120 days (no cap for rental days with
physical presence of the owner.)
90 days if not occupied by the owner
or 180 days if occupied by the owner.
Primary dwelling
requirement
Primary residence only, which
means the property is eligible for the
homestead deduction pursuant.
Primary residence only. The dwelling
unit must be occupied for at least 185
days per year.
Primary residence only (farm tenant
dwelling or on-site accessory dwelling
prohibited.)
Primary residence to get the license.
However, no stated restriction on rental
dwellings once license is obtained.
Maximum number of
dwellings per host
1
1 (single family. Multi-family is
subject to the same rule as
condo/apartment.)
1 (owner’s property or owner-
authorized resident’s primary
residence.)
Multiple. However, the combined
allowable time frames shall not exceed
the permissible calendar days.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
35
Maximum number of
rooms per dwelling
No cap, as long as all rooms/suites
within the property.
No cap. All rented bedrooms must be
in the main building. Accessory
dwelling allowed with a permit.
No cap. Only habitable rooms can be
used by guests.
No cap. Only habitable rooms can be
used by guests.
Maximum number of
guests per dwelling
8 (or 2 per bedroom, whichever is
greater.)
6 (or 2 per bedroom, whichever is
greater.)
6 (only counting guests 18 years or
older and maximum 2 per bedroom.)
8 (and no more than 3 guests per
bedroom.)
Requirement on
safety codes
Smoke detectors and carbon
monoxide detectors.
Fire extinguishers, smoke detectors,
and carbon monoxide detectors.
Smoke detectors and carbon monoxide
detectors. Sanitation facilities.
Smoke detectors and carbon monoxide
detectors. Fire extinguishers. A posting
of emergency contact and a floor plan.
Other requirements
Insurance of liability required. No
visitor parking permit for STR
guests.
Forbidden for commercial meetings, or
other gathering for direct or indirect
compensation.
HOA, Condo, and co-op associations
will be notified when an application is
filed. An application is not prohibited
by HOA, condominium document, or a
rental lease.
Insurance of liability required.
Compliance with the requirements of
HOA, condo association, etc. One
parking space for every three guests.
Tax
14.50%
7.25% transient occupancy tax (TOT)
7%
7%
Penalty
Any host who violates regulations is
subject to a civil penalty of $500,
$2,000, and $6,000 for the first,
second, and third violation,
respectively. Suspension and
revocation of the license.
The permit may be revoked with no
new permit for one year in the event of
three or more violations, failure to
comply with the zoning ordinance, or
refusal to cooperate in a complaint
investigation.
The license is suspended for an
applicant receiving at least three
complaints that are verified as
violations within a 12-month period.
No new issuance within 3 years after a
license is revoked.
A STR license may be suspended or
revoked at any time due to non-
compliance with the requirements,
citation, violations of the building,
electrical, plumbing or zoning code. In
addition, subject to a civil fine up to
$1,000.
Legislative outcome
Adopted on Nov. 13, 2018 and
effective in Oct. 2019.
Adopted in Nov. 2017 and effective
since Jan. 2018.
Senate Bill 2-16 and ZTA 17-03
became effective on July 1, 2018.
Adopted on Oct. 23, 2018 and
effective Oct. 1, 2019.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
36
Appendix
Table 2: Summary Statistics
Variable (name)
Mean
S.d.
Variable
Mean
S.d.
Airbnb attributes
Neighborhood attributes
Airbnb listings in 100 ft
(Airbnb100ft)
0.21
0.56
Annual N of traffic incidents
in .5 mile (numCrash)
152.3
114.7
Airbnb listings in 200 ft
(Airbnb200ft)
0.85
1.52
Annual N of crime incidents
in .5 mile (numCrime)
326.0
265.3
Airbnb listings in 500 ft
(Airbnb500ft)
5.06
7.35
Number of public schools
in .5 mile (pubschool)
2.38
1.82
Airbnb listings in 1,000 ft
(Airbnb1000ft)
18.63
26.12
Number of charter schools
in .5 mile (chaschool)
2.60
2.59
Housing attributes
Number of Metrorail stations
in .5 mile (metro)
0.43
0.66
Property Prices in
$ (last_sale_price)
762,842
754,505
Number of historical sites
in .5 mile (landmark)
9.53
15.12
Land area in 1,000 sqft
(landarea)
3.087
2.835
Demographic attributes (Census tract level)
Estimated year built (eyb)
1970
17.44
Total population in a tract
(totalpop)
3,904
1,458
Number of rooms (rooms)
7.44
2.51
Population density per acre
(popden)
15.20
9.69
Number of bathrooms (bathrm)
2.24
1.06
Percentage adult (pct_adult)
0.18
0.06
Number of half-bathrooms
(hf_bathrm)
0.65
0.60
Percentage Hispanic/Latino
(pct_hisp)
0.09
0.08
Number of kitchens (kitchens)
1.24
0.63
Percentage highly educated –
post-bachelor (pct_educated)
0.29
0.19
Number of fireplaces
(fireplaces)
0.60
0.89
Percentage high income
– >$20,000 (pct_highinc)
0.15
0.14
Square footage (sqft)
1,693
818.6
Unemployment (pct_unemp)
0.11
0.07
Air-conditioning – dummy
variable, 1 is yes (ac)
0.73
0.45
Poverty rate (pct_poverty)
0.15
0.10
Number of stories (stories)
2.19
0.80
Number of observations
12,680
Grade – 1 is low, 12 is
exceptional (grade)
4.25
1.38
Other housing attributes: exterior wall type (extwall),
roof type (roof), interior wall type (intwall), heating
type (heat), building structure (structure), land use
code (usecode)
Condition – 1 is poor, 6 is
excellent (condition)
3.81
0.80
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
37
Appendix
Table 3: Empirical results of three models
Panel A: Pooled Cross-Sectional Model
(dependent variable: logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0065
(0.006)
0.0051*
(0.003)
0.0026**
(0.001)
0.0011**
(0.000)
Landarea
0.0107***
(0.002)
0.0107***
(0.002)
0.0108***
(0.002)
0.0109***
(0.002)
Eyb
0.0002
(0.000)
0.0002
(0.000)
0.0002
(0.000)
0.0002
(0.000)
Ac
0.0723***
(0.017)
0.0725***
(0.017)
0.0724***
(0.017)
0.0723***
(0.017)
Fireplaces
0.0227***
(0.006)
0.0228***
(0.006)
0.0226***
(0.006)
0.0224***
(0.006)
Rooms
0.0051*
(0.003)
0.0051*
(0.003)
0.0051*
(0.003)
0.0051*
(0.003)
bathroom
0.0648***
(0.004)
0.0648***
(0.004)
0.0649***
(0.004)
0.0652***
(0.004)
hf_bthroom
0.0278***
(0.005)
0.0278***
(0.005)
0.0278***
(0.005)
0.0278***
(0.005)
Sqft
0.0002***
(0.000)
0.0002***
(0.000)
0.0002***
(0.000)
0.0002***
(0.000)
Stories
0.0002***
(0.000)
0.0002***
(0.000)
0.0002***
(0.000)
0.0002***
(0.000)
Grade
0.0397***
(0.010)
0.0397***
(0.010)
0.0397***
(0.010)
0.0397***
(0.010)
Condition
0.1233***
(0.007)
0.1233***
(0.007)
0.1231***
(0.007)
0.1232***
(0.007)
Kitchens
-0.0291
(0.018)
-0.0288
(0.018)
-0.0282
(0.018)
-0.0281
(0.018)
Pubschool
0.0072**
(0.003)
0.0071**
(0.003)
0.0068**
(0.003)
0.0064**
(0.003)
Metro
0.0229
(0.013)
0.0227
(0.013)
0.0220
(0.013)
0.0222
(0.013)
Landmark
0.0029***
(0.001)
0.0029***
(0.001)
0.0028***
(0.001)
0.0028***
(0.001)
pct_adult
0.3242**
(0.131)
0.3214**
(0.131)
0.3144**
(0.130)
0.3078**
(0.129)
pct_educated
0.5133***
(0.113)
0.5046***
(0.114)
0.4819***
(0.114)
0.4650***
(0.116)
pct_unemp
-0.4268***
(0.136)
-0.4306***
(0.136)
-0.4365***
(0.135)
-0.4387***
(0.137)
Constant
11.4146***
(0.882)
11.4154***
(0.879)
11.4058***
(0.862)
11.4174***
(0.846)
Other controlled
variables
heat type, land use type, structure type, interior & exterior wall type, roof type, # of traffic
& crime incidences, # of charter school, population density, % Hispanic population, % high
income household, poverty rate
Zip-code
dummies
✓
✓
✓
✓
Period dummies
✓
✓
✓
✓
Cluster s.e.
✓
✓
✓
✓
N
12,680
12,680
12,680
12,680
R2
0.8095
0.8095
0.8097
0.8099
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
38
Appendix
Panel B: Fixed Effects Model at Census Tract Level
(dependent variable: average logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0060
(0.008)
0.0078*
(0.003)
0.0037**
(0.001)
0.0012***
(0.000)
Other controlled
variables
land area, estimated year built, air-conditioning, fireplaces, rooms, bedrooms, bathrooms,
half-bathrooms, sqft., stories, grade, condition, heat type, land use type, structure type,
interior & exterior wall type, roof type, # of traffic & crime incidences, constant
Period dummies
✓
✓
✓
✓
N
7,624
7,624
7,624
7,624
N blocks
2,378
2,378
2,378
2,378
R2
0.3905
0.3910
0.3923
0.3925
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
Panel C: First Difference Model at Census Tract Level
(dependent variable: average logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0212
(0.016)
0.0136*
(0.008)
0.0103***
(0.002)
0.0031***
(0.001)
Other controlled
variables
land area, estimated year built, air-conditioning, fireplaces, rooms, bedrooms, bathrooms,
half-bathrooms, sqft., stories, grade, condition, heat type, land use type, structure type,
interior & exterior wall type, roof type, # of traffic & crime incidences, constant
After
0.0283
(0.017)
0.0275
(0.017)
0.0249
(0.017)
0.0240
(0.017)
N
2,047
2,047
2,047
2,047
N blocks
1,027
1,027
1,027
1,027
R2
0.3704
0.3712
0.3804
0.3792
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
39
Appendix
Table 4: Aggregate Impact of Airbnb on Property Price by Zip Code
Zip code
200-ft
density
200-ft
impact
500-ft
density
500-ft
impact
1000-ft
density
1000-ft
impact
%
Hispanic
% Black
20001
2.61
2.04%
15.33
5.67%
55.49
6.66%
9.22%
50.75%
20002
1.51
1.18%
8.78
3.25%
32.45
3.89%
4.39%
61.33%
20003
1.45
1.13%
8.64
3.20%
31.37
3.76%
5.12%
36.41%
20005
2.68
2.09%
24.23
8.97%
102.09
12.25%
16.77%
15.17%
20007
0.79
0.62%
4.92
1.82%
17.72
2.13%
7.12%
3.12%
20008
0.41
0.32%
2.53
0.94%
9.74
1.17%
7.67%
5.10%
20009
3.43
2.68%
20.63
7.63%
79.21
9.51%
15.13%
20.69%
20010
1.89
1.47%
11.41
4.22%
43.68
5.24%
30.11%
31.07%
20011
0.44
0.34%
2.7
1.00%
10.11
1.21%
21.18%
65.31%
20012
0.21
0.16%
1.4
0.52%
4.71
0.57%
11.22%
64.29%
20015
0.17
0.13%
1.09
0.40%
3.54
0.42%
6.52%
9.00%
20016
0.15
0.12%
0.98
0.36%
3.55
0.43%
7.30%
4.25%
20017
0.35
0.27%
2.15
0.80%
7.54
0.90%
6.49%
71.42%
20018
0.16
0.12%
1.05
0.39%
4.21
0.51%
5.87%
85.08%
20019
0.15
0.12%
0.7
0.26%
2.16
0.26%
2.41%
94.98%
20020
0.21
0.16%
1.32
0.49%
4.44
0.53%
1.41%
95.00%
20024
0.7
0.55%
5.62
2.08%
18.11
2.17%
5.16%
54.50%
20032
0.07
0.05%
0.38
0.14%
1.12
0.13%
2.33%
90.00%
20036
4.36
3.40%
29.64
10.97%
86.79
10.41%
7.62%
7.78%
20037
3.2
2.50%
19.97
7.39%
65.79
7.89%
5.77%
6.32%
D.C.
0.85
0.66%
5.06
1.87%
18.63
2.24%
9.10%
50.03%
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
40
Appendix
Table 5: Robustness Check with Entire-Unit Airbnb Listings with Reviews
Panel A: Pooled Cross-Sectional Model
(dependent variable: average logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0140
(0.008)
0.0071*
(0.003)
0.0028**
(0.001)
0.0011***
(0.000)
N
12,680
12,680
12,680
12,680
R2
0.8091
0.8091
0.8092
0.8092
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
Panel B: Fixed Effects Model at Census Tract Level
(dependent variable: average logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0096
(0.011)
0.0086*
(0.005)
0.0033**
(0.001)
0.0008***
(0.000)
N
7,624
7,624
7,624
7,624
N blocks
2,378
2,378
2,378
2,378
R2
0.3906
0.3909
0.3913
0.3910
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
Panel C: First Difference Model at Census Tract Level
(dependent variable: average logarithm of property price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
0.0258
(0.022)
0.0050
(0.011)
0.0057*
(0.003)
0.0017*
(0.001)
N
2,047
2,047
2,047
2,047
N blocks
1,027
1,027
1,027
1,027
R2
0.3705
0.3698
0.3718
0.3715
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
41
Appendix
Table 6: Empirical Results for Median Rent Price at the Zip Code Level
Panel B: Fixed Effects Model at the Zip Code Level
(dependent variable: logarithm of median rent price)
100 ft buffer
200 ft buffer
500 ft buffer
1000 ft buffer
Variable Name
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Coef.
s.e.
Airbnb density
-0.0002
(0.016)
0.0065
(0.005)
0.0011
(0.001)
0.0004
(0.000)
Other Controls
land area, estimated year built, air-conditioning, fireplaces, rooms, bedrooms, bathrooms, half-
bathrooms, sqft., stories, grade, condition, # of traffic & crime incidences, constant
N
119
119
119
119
N Zip Codes
20
20
20
20
R2
0.4202
0.4310
0.4283
0.4373
Robust s.e. in parentheses, *** p<.01, ** p<.05, * p<.1
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
42
Appendix
Figure 1. Number of Airbnb listings in Washington, DC, January 2015–July 2017.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
43
Appendix
Figure 2. (a) Airbnb listing locations, February 2015. (b) Airbnb listing locations,
February 2017.
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
44
Appendix
Figure 3. Short term rentals' (STRs') welfare impact and mechanism. Modified from
Sheppard and Udell (2016, p. 9).
THIS IS A PREPRINT. PLEASE SEE THE PUBLISHED MANUSCRIPT HERE:
https://doi.org/10.1080/10511482.2019.1681016
45
Appendix
Figure 4. Example of Airbnb density buffers around a property sales point.
