Estimating the Impact of Airbnb Activity on Housing Prices in New York City WILLIAMS COLLEGE

78 Pages·2016·9.22 MB·English

Estimating the Impact of Airbnb Activity on
Housing Prices in New York City
Author: Advisor:
Andrew Udell Professor Stephen Sheppard
A thesis submitted in partial fulfillment
of the requirements for the
Degree of Bachelor of Arts with Honors
in Economics
WILLIAMS COLLEGE
Williamstown, MA
May 2016
Contents
1 Introduction 7
1.1 Research Question and its Implications . . . . . . . . . . . . . . . . . 7
1.2 Contemporary Policy Debates . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Methodology and Analytical Framework . . . . . . . . . . . . . . . . 11
2 Literature Review 14
2.1 Literature Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 The Hedonic Model of Housing Markets . . . . . . . . . . . . . . . . . 15
2.3 The Application of Hedonic Models . . . . . . . . . . . . . . . . . . . . 17
2.4 Research on Peer-to-Peer Platforms . . . . . . . . . . . . . . . . . . . 21
2.5 Contributions to Existing Literature . . . . . . . . . . . . . . . . . . 23
3 Theoretical Model 25
3.1 The Perfectly Complementary City . . . . . . . . . . . . . . . . . . . . 25
3.2 Partial Equilibrium Model . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4 Data and Methodology 33
4.1 Data & Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Empirical Identification Strategy . . . . . . . . . . . . . . . . . . . . . 42
5 Results 44
5.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6 Conclusion 50
6.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.1.1 Summary of Results . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.1.2 Policy Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.1.3 Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7 Appendix 52
7.1 Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2 Meeting with Airbnb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.3 Stata Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.3.1 Generating Time Series Airbnb Data . . . . . . . . . . . . . . . 62
7.4 ArcGIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.4.1 Near Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.4.2 Spatial Join . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.5 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.5.1 Different Spatial Measures of Airbnb Activity . . . . . . . . 66
7.5.2 Linear Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.6 Year of Sale Hedonic Estimates . . . . . . . . . . . . . . . . . . . . . . 68
2
List of Figures
1.1 Transmission Mechanisms for the Impact of Airbnb Activity on
Housing Prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1 Monocentric Urban Model . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Perfect Complements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Land Rent Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4 Theoretical Impact of a Rise in Population. . . . . . . . . . . . . . . 32
3.5 Theoretical Impact of a Rise in Income . . . . . . . . . . . . . . . . . 32
3.6 Theoretical Impact of a Decrease in ↵ . . . . . . . . . . . . . . . . . . 32
4.1 Construction of Airbnb Dataset . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Airbnb Listings Over Time . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Sales & Buffer Zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.1 Housing Price Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.1 Airbnb Density: June 2015 . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.2 Average Sale Price: September 2014 - August 2015 . . . . . . . . . . 54
7.3 Bivariate Choropleth: Airbnb Activity and Sales Prices . . . . . . 55
7.4 VisualizationofTable7.8: DifferentSpatialMeasuresofAirbnb
Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3
List of Tables
4.1 Data Sources and Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Descriptive Statistics: Airbnb Activity Measures . . . . . . . . . . . 40
4.3 Descriptive Statistics: Sales and Controls . . . . . . . . . . . . . . . 41
5.1 Airbnb’s Positive Impact on New York City Housing Prices . . . . 47
5.2 Alternative Airbnb Activity Proxies . . . . . . . . . . . . . . . . . . . 48
7.1 Number of Listings by Type by Borough (as of 11/17/2015) . . . . 57
7.2 NumberofHostswithActiveEntireHomeListings(asof11/17/15) 58
7.3 Number of Listings by Type by Nights Hosted . . . . . . . . . . . . . 59
7.4 (Cont’d) Number of Listings by Type by Nights Hosted . . . . . . . 60
7.5 Median Listing Revenue by Type of Listing . . . . . . . . . . . . . . . 61
7.6 Median Listing Revenue by Borough . . . . . . . . . . . . . . . . . . . 61
7.7 Median Nights Booked per Listing by Borough . . . . . . . . . . . . 61
7.8 Different Spatial Measures of Airbnb Activity . . . . . . . . . . . . 67
7.9 Linear Hedonic Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.10 Year of Sale Estimates from Table 5.1 . . . . . . . . . . . . . . . . . . 69
7.11 Year of Sale Estimates from Table 5.2 . . . . . . . . . . . . . . . . . . 70
7.12 Year of Sale Estimates from Table 7.8 . . . . . . . . . . . . . . . . . . 71
4
Abstract
I evaluate the impact of Airbnb activity on housing prices in New York City. Using a
hedonic regression to estimate Airbnb’s impact, I find that Airbnb activity in New York
City has positively impacted housing sales prices. This impact is robust to changes in the
specification of my model and a 1% increase in the number of Airbnb listings within 300m
increasessalespricesbyapproximately0.08%,whichincludescontrolsforthepropertyitself
and local neighborhood fixed e↵ects to capture time-invariant neighborhood quality. The
impact varies slightly, in the full specifications, from 0.06% to 0.1%, depending on the
di↵erent proximate and descriptive measures of Airbnb activity; across all specifications,
the impact is positive and both economically and statistically significant. Additionally, I
estimatethetheoreticalimpactofAirbnbactivityonhousingpricesinamonocentricurban
partial equilibrium model.
5
Acknowledgments
I am filled with gratitude for the incredible amount of support I have received over the
course of the past year. Professor Stephen Sheppard truly went above and beyond as an
advisor, supporting me throughout each step of this process (even as he spent the second
halfofthisyearintheNetherlands!). Thisthesisundoubtedlywouldnothavebeenpossible
without his generous support.
ProfessorSaraLaLumiaandProfessorSarahJacobsonwereconstantsourcesofhelpdur-
ing the past year. Their feedback and suggestions were critical to the success of this thesis.
Ben Blackshear provided many hours of technical advice for using ArcGIS. A conversation
wehadinMarchof2014servedastheinspirationforthisresearch. Thoseseedsweresowed
as a result of an internet economics course I took with Dr. Greg Taylor, Fellow at the
Oxford Internet Institute. Professor David Love, my advisor in the economics department,
encouraged me to write a thesis and helped me iron out a strong proposal. He has served
as a mentor, formally and informally, throughout my entire academic career at Williams
College. Professor Matthew Gibson devoted many hours to helping me with this project.
Je↵rey Fossett, a Williams College alumnus and data scientist at Airbnb, provided incredi-
bly useful feedback. We had several conversations as well as a long-standing email thread.
My brother, Jacob Udell, provided me with comments and criticism that were immensely
helpful. He pushed and goaded me as only an older brother can. Matt Rock and Molly
Leonardprovidededitsonmanydrafts. MattMcNaughtonconsistentlyhelpedmetorefine
my ideas and stay positive. Ryann Noe, on more than one occasion, helped me debug my
Stata code, and served as a sounding board for ideas. I am thankful for Murray Cox open-
sourcing his InsideAirbnb dataset, which is the basis for this research. Airbnb generously
invited me to access their data and meet with their team. Ryan McCloskey drove to New
York City with me for my meeting with Airbnb. Of course, this thesis would not have been
possible without the support of Williams College and the Economics Department. Other
2015-2016 thesis writers in the Economics Department provided me with meaningful and
thoughtful feedback during presentations and conversations. I am thankful for the advice
I received from friends and family, and importantly, for dealing with me as I constantly
discussed this research. Thank you to you all.
All mistakes are my own.
6
1 Introduction
1.1 Research Question and its Implications
This paper estimates the impact of Airbnb activity on housing prices in New York City.
Airbnb is an internet-based peer-to-peer marketplace which allows individuals to “list, dis-
cover, and book” over 2,000,000 accommodations in over 34,000 cities across the world
(Airbnb, 2016). Airbnb acts as an intermediary between consumers and producers to re-
duce the risk and cost of o↵ering a home as a short-term rental, which enables suppliers
(homeowners) to flexibly participate in the commercial housing market. In New York City,
Airbnb activity tends to be heavily concentrated in the boroughs of Manhattan and Brook-
lyn. As of November 17th, 2015, there were a total of 35,743 active listings in New York
City.1 These listings constitute a sizable portion of the accommodations industry in New
YorkCity,asthereisatotalofapproximately102,000hotelroomsintheentirecity(Cuozzo,
2015).2 AirbnbhasasignificantpresenceinNewYorkCityandothercitiesacrosstheworld,
so much so that some local city governments are attempting to shut down the company.
Cities are concerned with Airbnb’s impact on a↵ordability as well as the way in which it
evades housing and tax regulations.
Airbnbispartofwhathasbecometobeknownasthe“SharingEconomy,”whichsimply
putreferstopeer-to-peerproducts, services, andcompanies. Alargepartofthemotivation
behind the Sharing Economy, according to the companies that self-define as part of the
sector, is to make use of otherwise under-utilized goods.3 In the case of housing, homes
mightnotbeutilizedtotheirfullextent(forexample, duringvacationsorduetoanunused
bedroom). This allows homeowners to “share” (e.g., rent) parts or the entirety of their
1ForfurtherinformationonthemarketdynamicsofAirbnbinNewYorkCity,seeSection7.2.
2There are 3,394,486 housing units in New York City measured in 2013 (Been et al., 2015), meaning
thatover1%ofhousingunitswerebeingactivelylistedonAirbnbonNovember17th,2015. Giventhatthe
distributionofAirbnbisnotnormallydistributedthroughoutthecity,weshouldexpectthatinsomeareas,
theratioofAirbnblistingstototalunitsissignificantlyhigher.
3See“TheSharingEconomy: FriendorFoe?”(Avitaletal.,2015)foraconcisesummaryofthedi↵erent
viewpointssurroundingthefutureoftheSharingEconomy.
7
homes during these times and earn revenue. The potential for and ease of these types of
transactions is greatly increased by better matching technologies, a trend which has been
driven by the Internet (Horton and Zeckhauser, 2016). Airbnb further reduces transaction
costsforbothconsumersandproducersbyprovidingafeedbackandreputationmechanism,
allowing for a safer and more streamlined transaction.
DespiteAirbnb’se�ciencyimprovementsandtheabilityitgiveshomeownerstogenerate
revenue, there are concerns about the economic and welfare e↵ects of Airbnb’s presence on
the residential housing market.4 The following research examines those economic e↵ects.
Put simply, the study is motivated by the following question: in a highly constrained and
regulated housing market, where residential homes are both in high demand and located in
dense neighborhoods, what is the impact of being able to transform residential properties
into revenue streams and partly commercial residences?
In New York City, this question has been raised explicitly, as the role of Airbnb has
been at the center of a number of policy discussions at the municipal level. In 2014, the
Attorney General of New York State, Eric Schneiderman, investigated Airbnb’s presence in
New York City (Schneiderman, 2014). The New York Attorney General O�ce’s findings
indicated that 72% of Airbnb listings in New York City violated property use and safety
laws and were therefore illegal.5 The Attorney General’s O�ce also found that over 4,600
units in New York City were booked for more than three months of the year, leaving the
Attorney General’s O�ce to question the impact that Airbnb has on the supply of housing
stock and New York City’s a↵ordability. As of the Attorney General’s investigation in
2014, Airbnb saw an increase of over 1000% in both listings and bookings from 2010 to
2014. To understand Airbnb’s scale of growth, or at least the way their investors value its
business, an oft cited statistic is that in its most recent funding round, Airbnb was valued
at approximately $25B. This suggests it is more valuable than Marriott International Inc.,
4There are several firms similar to Airbnb. As these types of companies become more prevalent and
continuetoexpand,thisareaofresearchbecomesincreasinglyimportant,assuchfirmsmostlyenterhighly
constrained and regulated markets, the dynamics of which often have welfare consequences. The analysis
here is not directly applicable to, for example, understanding the economic impact of Uber on a city, a
ride-sharingservice. However,theresearchpresentedinthispapersuggeststhatthesecompaniescanhave
asignificantimpact,oneworthyofstudy.
5This is largely due to New York State’s Multiple Dwelling Law, which imposes strict regulations on
safetyandhealthconditionsthatmustbemetaswellaslimitsonbusinessusesofhomes.
8
which has a market capitalization of $17.9B and which owns over 4,000 hotels. In 2014,
MarriottInternationalInc. had$13.8Binrevenue,overtentimesAirbnb’sprojected revenue
in 2015 (Kokalitcheva, 2015). That investors are still willing to purchase an equity stake in
Airbnb at its current valuation suggests an expectation of continued, extraordinary growth.
Their expected revenue for 2020 is $10B, implying an annual growth rate of approximately
+75% (Kokalitcheva, 2015).
Especiallygivenitsrapidgrowth,theAttorneyGeneral’sinvestigationrevealsthatthere
are real concerns about the presence of Airbnb in cities across the world. Central to this
consideration,accordingtoauthorDougHenwood,isthepotentialofAirbnb’s,“real,ifhard-
to-measure, impact on housing availability and a↵ordability in desirable cities,” (Henwood,
2015). Indeed, as this study will expand upon below, almost all of the welfare consequences
(both positive and negative) of Airbnb circle around the question of its impact on housing
prices. Assuch,thisresearchgrappleswiththequestionofAirbnb’simpactinNewYorkCity
by presenting both empirical and theoretical information on Airbnb’s impact on residential
housing prices – an issue that is brought up often, but rarely quantified. This paper seeks
not to make a judgment on whether or not Airbnb is good or bad for cities, but rather to
provide the first estimates on Airbnb’s impact on residential housing prices.
1.2 Contemporary Policy Debates
Residentsofcitiesandlocalgovernmentsacrosstheworld,bothinfavorandagainstAirbnb’s
presence, are growing increasingly vocal. The arguments against Airbnb focus primarily on
three areas:6 1) Airbnb’s impact on decreasing a↵ordability, 2) the negative externalities
causedbyAirbnbguests,7and3)theshadowhotelindustrythatallowscommercialoperators
6An article on the impact of Airbnb in Los Angeles articulates these concerns clearly: “Airbnb forces
neighborhoodsandcitiestobearthecostsofitsbusinessmodel. Residentsmustadapttoatighterhousing
market. Increased tourist tra�c alters neighborhood character while introducing new safety risks. Cities
lose out on revenue that could have been invested in improving the basic quality of life for its residents.
Jobsarelostandwagesareloweredinthehospitalityindustry”(Samaan,2015,p. 2).
7Hortondescribesthisphenomenonwell: “IfAirbnbhostsbringinginloudordisreputableguestsbut,
critically, still collect payment, then it would seem to create a classic case of un-internalized externalities:
thehostgetsthemoneyandherneighborsgetthenoise”(Horton,2014,p. 1).
9
to use Airbnb in order to evade important safety regulations and taxes.8 On the other side,
those who argue in favor of Airbnb’s presence tend to focus on its positive economic impact
on the city, including creating new income streams for residents as well as encouraging
tourism and its associated economic benefits for a city (Kaplan and Nadler, 2015).
The contemporary policy debates surrounding Airbnb can be summarized by the fol-
lowing question: should Airbnb be regulated and, if so, what is the appropriate type and
level of regulation? This has been debated in New York City Council Hearings, protests
haveformedinsupportofandagainstAirbnb, andthispastNovember(2015), Airbnbeven
made it onto the ballot in San Francisco through Proposition F.9 There is strong language
on both sides; some are scared of Airbnb’s impact on the a↵ordability of neighborhoods
and others suggest that its net welfare e↵ects are positive. Additionally, the policy debates
surroundingAirbnbandothersharingeconomycompaniesareconcernedthatthesecompa-
nies degrade important regulations. Arun Sundararajan argues that new regulations need
to be developed to protect individuals, both consumers and workers, as a result of these
companies: “Asthescaleofpeer-to-peerexpands,however,societyneedsnewwaysofkeep-
ing consumers safe and of protecting workers as it prepares for an era of population-scale
peer-to-peer exchange” (Sundararajan, 2014).
In the New York City Council hearings, as well as in protests and debates in the public
sphere, there is a consistent lack of data and analysis upon which people can rely. Because
of this void, arguments are, to put it bluntly, mostly rhetorical and ideological rather than
empirical. Thus, in addition to pursuing the analysis of Airbnb’s impact on housing prices
in New York City, the data collection work included in this paper will also hopefully begin
to fill that void so that individuals can better understand Airbnb’s impact in a way that is
mathematically rigorous and econometrically robust.
8MuchoftheuproarinNewYorkCityconcernsnon-uniformtaxationandregulation;hotelsandmotels
facetaxeswhichAirbnbisnotcurrentlysubjectto. InNewYorkCity,itisuptohoststopaytaxesonthe
revenuetheygeneratefromAirbnb. Insomeothercities,Airbnbhasa“collectandremit”featuretocollect
taxes.
9Proposition F was ultimately rejected but would have limited the number of nights an Airbnb could
beavailableeachyear.
10