HOTEL TAX INCIDENCE WITH HETEROGENEOUS FIRM EVASION: EVIDENCE FROM AIRBNB REMITTANCE AGREEMENTS Eleanor Wilking University of Michigan [email protected] February 10, 2016 ABSTRACT How does assignment of the remittance obligation affect consumption tax incidence? In classical tax theory, the responsibility of transferring tax revenue has no effect on which party bears the economic burden of a consumption tax. I explore this prediction in the context of agreements between city governments and a large digital platform firm that shifted the obligation to remit hotel taxes from independent rentors to the platform firm itself. Using variation in the location and timing of such agreements, I estimate their effect on rental prices. My results indicate that shifting the remittance obligation to the platform increases after-tax prices, suggesting that consumers bear a greater share of the tax burden when the remittance obligation is shifted to a party with fewer evasion opportunities. JEL classification: H22, H26, H32, K2. This is a preliminary draft. Please do not cite or distribute. 1 1. Introduction The proliferation of mobile internet-connected devices has given rise to a number of firms that rely on “digital platforms” to match consumers and service providers. These firms, such as Uber, Lyft, and Airbnb, are economically significant, expanding rapidly, and often valued in the tens of billions of dollars. Policymakers and investors wonder whether and how existing taxes and regulations should apply to platform firms. Opponents of enforcing existing rules claim they do not apply to services created out of excess capacity, and their application would stifle innovation and growth in this industry. Proponents argue instead that existing rules should be enforced on transactions of the same economic substance (e.g., regulations applying to taxicabs should apply to Uber rides). Given how differently these new firms are structured relative to their traditional counterparts, their response to such enforcement is theoretically ambiguous and thus an empirical question. This paper examines the effect of one recent policy to enforce greater tax compliance on platform firms: bilateral tax remittance agreements between Airbnb and a number of U.S. cities. Specifically, I examine the effect that shifting responsibility for remitting hotel taxes from hosts to Airbnb has on listing prices. 2. Background and Policy 2.1 What is Airbnb? Airbnb is the largest of several firms facilitating short-term, peer-to-peer residential space rentals through an online platform. Originally conceived as an online marketplace to connect couch surfers, Airbnb has experienced remarkable growth in recent years, expanding 2 exponentially in popular tourism cities around the globe.1 Hosts on Airbnb create listings for each of their properties. Each listing includes information about the space’s characteristics, such as the number of beds, kitchen availability, and whether it is a private apartment or a shared space. Hosts can designate a listing’s availability and set its price for each calendar day. 2.2 How are Airbnb transactions taxed? In addition to consumer safety concerns, local governments expressed frustration with Airbnb hosts’ avoidance of short-term rental taxes. In cities with significant tourism, the estimated loss of occupancy tax revenue is significant. Initially, Airbnb’s position was that its rentals were not subject to occupancy taxes because transactions were “peer-to-peer” rather than commercial in nature. In May 2014, the company officially retracted this view and announced that it believed its hosts were responsible for paying occupancy taxes to local governments. It also amended its “Terms of Service” agreement to inform hosts of their obligation to research and comply with applicable local taxes and regulations.2 On June 28, 2014, Airbnb announced that it had reached an agreement with the city of Portland, OR to collect an 11.5% occupancy tax on all reservations booked on its site, and to pay these taxes to the city at the end of each quarter. Crucially, the agreement explicitly prohibited Portland’s city government from requiring Airbnb to disclose information related to taxable 1 Paris is thought to have nearly 40,000 active Airbnb listings, the most of any city in the world. 2 Beginning May 1, 2014, Airbnb’s Terms of Service includes the following paragraph: YOU AS A HOST UNDERSTAND AND AGREE THAT YOU ARE SOLELY RESPONSIBLE FOR 2 Beginning May 1, 2014, Airbnb’s Terms of Service includes the following paragraph: YOU AS A HOST UNDERSTAND AND AGREE THAT YOU ARE SOLELY RESPONSIBLE FOR DETERMINING (I) YOUR APPLICABLE TAX REPORTING REQUIREMENTS, AND (II) THE TAXES THAT SHOULD BE INCLUDED, AND FOR INCLUDING TAXES TO BE COLLECTED OR OBLIGATIONS RELATING TO APPLICABLE TAXES IN LISTINGS. YOU ARE ALSO SOLELY RESPONSIBLE FOR REMITTING TO THE RELEVANT AUTHORITY ANY TAXES INCLUDED OR RECEIVED BY YOU. AIRBNB CANNOT AND DOES NOT OFFER TAX-RELATED ADVICE TO ANY MEMBERS. 3 transaction that could individually identify hosts. As part of the exchange, the Portland City Council agreed to pass a code revision that would legalize short-term home rentals if residents obtained a $180 permit and installed fire alarms. Between August 2014 and August 2015, similar agreements to collect and remit hotel sales taxes were signed with San Francisco, CA (14.5%), San Jose, CA (10%), Chicago, IL (4.5%), Washington, DC (14.5%), Philadelphia, PA (8.5%), Durham, NC (6%), San Diego, CA (10.5%), and Phoenix, AZ (3%), as well as several smaller municipalities. Typically, an agreement is announced two weeks before the date when Airbnb begins collecting taxes on all bookings in that jurisdiction. Airbnb notifies affected hosts of the policy change via email shortly after the announcement. 2.3 How do taxes appear to Airbnb guests? When a guest searches for a rental on Airbnb, she is presented with a set of search results that includes an image, location, and tax-exclusive estimate of the nightly fee for each listing (Figure 1).3 After a guest clicks on a listing, she is shown a more detailed accounting of the rental cost, including Airbnb’s service fee and any occupancy tax. Figure 2 shows examples of listings from two jurisdictions: one that has a bilateral agreement with Airbnb (Chicago, IL), and one that does not (Evanston, IL). Notice that both listings appear among same set of search results. Without clicking on a listing, it is not evident whether an occupancy tax applies to it. 3. Motivation 3 The price shown in the search results is the average cost per night of the room, excluding taxes and Airbnb’s service fee. For example, if a listing’s rental prices for Friday, Saturday, and Sunday are $90, $100, and $110, respectively, and the listing has a $30 cleaning fee, then the price displayed in the search results will be $110 (90+100+110+30 / 3 = 110). 4 In the traditional view, the economic incidence of a tax is independent of its statutory incidence. In the absence of differences in compliance and monitoring costs or evasion opportunities between the supply and demand sides of a market, shifting the remittance obligation from one to the other has no impact on which party bears the economic burden of the tax. However, recent empirical work suggests that the assignment of the remittance obligation can affect incidence in practice. I consider two potential reasons why the traditional view may not hold in the context of Airbnb’s market for short-term rentals. 3.1 Evasion technology and the point of collection Statutory irrelevance may not hold if economic agents differ in their ability to evade taxes (Kopczuk, Marion, Muehlegger, and Slemrod forthcoming, hereafter KMMS). KMMS find that the economic incidence of a quantity tax on diesel fuel depends on the point of collection within the supply chain. As the remittance obligation moves “up” the chain from retailers to distributors and prime suppliers, the pass-through rate of diesel taxes to the retail price increases, as do tax revenues. This suggests that differential evasion opportunities afforded to these agents explain the relevance of a tax’s collection point. The context I study differs from that of KMMS in several potentially consequential respects. First, Airbnb’s role in the provision of short-term rentals cannot be neatly confined to that of a supplier. Airbnb considers itself a “platform” or “market-maker,” rather than an upstream firm in a supply chain that provides paid lodging. Second, Airbnb hosts provide a heterogeneous product, unlike diesel fuel. Third, pre-policy remittance behavior by hosts was likely binary.4 Additional differences between the Airbnb and diesel fuel contexts—such as the 4 Although I am unable to observe pre-policy remittance behavior, based on online discussions between (anonymous) hosts and public comments by municipal tax authorities, I infer that evasion took place mainly on the 5 occupancy tax being ad valorem rather than specific, and the different costs of compliance and monitoring within each market—may also affect how changing the point of remittance affects the tax’s incidence, and will need to be carefully considered. 3.2 Host sophistication in price-setting Classically, tax incidence is exclusively determined by market-level factors such as the level of competition and supply and demand elasticities (see, e.g., Myles 1989; Weyl and Fabinger 2013). However, there is some empirical evidence that differences in firms’ characteristics, such as managerial resources that affect price-setting strategies, can lead to variation in tax incidence within a market. For example, small, independent firms are more likely to rely on simplified pricing rules, such as round number heuristics, and may not fully incorporate tax changes into price setting behavior (Harju, Kosonen, and Skans 2015). There is likely significant heterogeneity in host experience with, and attention to, optimal price setting. Some hosts are actually hotels, with substantial managerial resources devoted to setting prices. On the other end of the spectrum, many hosts list once or twice a year when they vacate their residences to travel, or in response to anticipated spikes in demand from local events (e.g., football games). Like consumers, hosts in the latter category may be inattentive to changes in Airbnb’s occupancy tax policies, or exhibit bounded rationality that constrains them from setting prices optimally (Chetty 2009; Finkelstein 2009). 4. Model extensive margin. That is, hosts either voluntarily remitted the occupancy taxes they owed, or they paid nothing at all. 6 This section sets forth a simplified model of Airbnb host behavior. Hosts are divided into two types: those who remit their tax obligations, and those who do not. Prior to the enactment of any agreement, both types of host choose listing prices to maximize profit. Between two otherwise identical hosts, the one who remits sets a higher price to offset the cost of remitting the tax. Once an agreement is signed, hosts who previously set high, tax-inclusive listing prices now set low, tax-exclusive listing prices. This, combined with reduced demand by guests now faced with a tax, drives down listing prices for all hosts in equilibrium. 4.1 Host price-setting under monopolistic competition Each Airbnb host supplies a product that is differentiated—by location, amenities, and so on—from those of her competitors. The market for Airbnb rentals is therefore monopolistically competitive, with each host facing a downward-sloping demand curve: 𝑞 = 𝑞 (𝑝 ,𝛾 ,𝑝 ,𝑁). ! ! ! ! !! This demand curve is a function of the host’s own price !!! < 0 , the desirability of the listing !! ! !!! > 0 , the prices set by all other hosts !!! > 0 , and the number of other hosts !!! < 0 . !! !! !" ! !! Assume that the fixed cost of entry into the Airbnb market is zero, and that the variable !! cost of hosting is heterogeneous across hosts and represented by 𝐶(𝑞 ,𝛿 ), where > 0. The 𝛿 ! ! ! !! ! parameter can be thought to represent a host’s idiosyncratic level of inconvenience from hosting someone in their property. Prior to the enactment of Airbnb’s remittance agreements, hosts were responsible for remitting any hotel taxes owed to the municipal government. While some hosts complied, it is likely that many did not. Prior to the policy, assume hosts were one of two types: remitting or 7 non-remitting. The former complied fully by remitting all taxes owed, while the latter remitted nothing.5 All hosts must choose a listing price 𝑝 to maximize profit: ! Remitting hosts: 𝜋! = 𝑝 𝑞 1−𝜏 −𝐶 𝑞 ,𝛿 (1) ! ! ! ! ! Non-remitting hosts: 𝜋!" = 𝑝 𝑞 −𝐶 𝑞 ,𝛿 (2) ! ! ! ! ! where 𝜏 denotes the hotel tax rate in the city in which the property is located. 4.2 Equilibrium Remitting and non-remitting hosts both choose listing prices that satisfy their respective first order conditions. If two listings are equally desirable and hosts have the same cost index, then a remitting host will optimally set a higher listing price to account for the additional cost imposed by the tax remittance obligation: 𝑝!∗ 𝑝 ,𝑁,𝛾 ,𝛿 ,𝜏 > 𝑝!"∗ 𝑝 ,𝑁,𝛾 ,𝛿 ! !! ! ! ! !! ! ! !!!!,!!!! Free entry implies that the marginal host will earn zero profit in equilibrium. As a result, there exists a threshold value of 𝛿 that characterizes the variable cost of the marginal host. For ! example, conditional on listing desirability, the variable cost of the marginal remitting host is lower than the variable cost of the marginal non-remitting host: 𝛿!∗ 𝑝 ,𝑁,𝛾 ,𝜏 < 𝛿!"∗ 𝑝 ,𝑁,𝛾 !! ! !! ! !!!! This reflects the fact that, due to the additional cost imposed by the tax remittance obligation, remitting hosts will tend to have lower idiosyncratic costs in order to remain in the market. 5 This model rules out evasion on the intensive margin. Remitting hosts can be thought of as those facing very large costs of evasion, such as having a neighbor who is a municipal tax officer. 8 In Nash equilibrium, no host can increase her profit given the prices set by other hosts. Assuming symmetric price-setting within host type and a large number of hosts, the prices charged by all other hosts, 𝑝 , can be expressed as the weighted average of the expected price !! set by each type of host: 𝑝 = 𝑝 = 𝛼!𝔼 𝑝! 𝛿 < 𝛿!∗]+(1−𝛼!)𝔼 𝑝!" 𝛿 < 𝛿!"∗] !! ! ! ! ! where 𝛼! represents the share of hosts who are remitters. 4.3 Shifting the remittance obligation to Airbnb Under the remittance agreements signed by Airbnb, hosts are relieved of their obligation to remit hotel taxes to municipal governments. Instead, any tax is added to the listing price set by hosts and remitted on behalf of guests by Airbnb. In the context of the model, this means that the fraction of hosts who are remitters falls to zero, and all hosts face the profit maximization in equation (2). This change will affect a number of variables in equilibrium, starting with prices. Hosts who previously remitted, and set tax-inclusive listing prices, will now charge lower, tax- exclusive listing prices. In addition, even hosts who previously did not remit will find it optimal to lower their prices: first, because the reduced prices of previously remitting hosts will lower average prices, and second, because guests now face a tax upon check-out and will reduce their demand accordingly. The combination of these effects will lead to lower listing prices after an agreement has been signed. In addition to affecting the average listing price, an agreement may also affect the composition of Airbnb hosts. For example, non-remitting hosts with high costs may find that they earn negative profits after the policy change, and will thus be compelled to exit the market. 9 4.4 Model limitations This simplified model of Airbnb host behavior leaves out a number of factors that will likely govern real-world behavior. First, the decisions of hotels, direct competitors to Airbnb hosts, are left out of the model. The hotel industry, which lobbied heavily for increased tax enforcement of Airbnb hosts, will face a reduction in downward price pressure once an agreement is signed. Second, the guest’s decision problem is omitted from the model. While it may be the case that demand for temporary lodging within a city is relatively inelastic in the short run, it is almost certainly more elastic in the long run. If this is the case, then the incidence of the hotel tax will be increasingly borne by hosts in the long run, as guests have their choice of visiting cities with and without remittance agreements. Finally, evasion is modeled here as a corner solution: a host either evades fully, or not at all. In other contexts where evasion is modeled, it is common to allow entities to choose a level of evasion (e.g., KMMS forthcoming). 5. Data 5.1 Collection To measure the response of hosts to Airbnb’s remittance agreements, I collect information on listings for selected U.S. cities and their surrounding areas between February 13 and August 30, 2015.6 I focus on cities that had announced, but not yet implemented, occupancy 6 These data are collected using an automated script or “crawler” that systematically browses Airbnb.com and collects information on listings associated with a particular geographic search 10