支付方式可得性对市场均衡的影响：基于Uber数据的研究（英文版）.pdf

NBERWORKINGPAPERSERIES ONTHEEFFECTSOFTHEAVAILABILITYOFMEANSOFPAYMENTS: THECASEOFUBER FernandoE.Alvarez DavidO.Argente WorkingPaper28145 nber/papers/w28145 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 November 2020 We want to thank Andy Abel, Daron Acemoglu, Manual Amador, Marios Angeletos, George Alessandria, Andy Atkeson, Gadi Barlevy, Ben Bernanke, Mark Bills, Anmol Bhandari, Stephane Bonhomme, Sara Castellanos, Gabriel Chodorow-Reich, Doireann Fitzgerald, Greg Kaplan, Narayana Kocherlakota, John List, Ellen Mc Grattan, Juanpa Nicolini, Francesco Lippi, Enrique Seira, Gaby Silva-Bavio, Rob Shimer, Harald Uhlig, Venky Venkateswaran, and Ivan Werning for their comments andsuggestions. Wealsowanttothanktheparticipantsin theseminarsattheFederalReserveBankofKansas,the FederalReserveBankofMinneapolis, theFederalReserveBankofChicago,theWhartonSchool attheUniversityofPennsylvania,the University of Rochester, the Applications Workshop at the University of Chicago, Columbia Business School, the Banco de Mexico, the 2019 SED in St. Louis, the MM group in the 2019NBERSummerInstitute,the2019ChicagoNBER EFG, the Bank of Canada, the Bank of International Settlements, the Federal ReserveBank of Richmond, Penn State, Dartmouth, UCLA, Arizona State, UC San Diego, the Swiss Macro Workshop, the University of Pennsylvania, theVirtualFinanceTheorySeminar,andMIT.We further want to thank Libby Mishkin and other members of the San Francisco Uber policygroup, and especially the Uber Mexico team for general support and assistance, programming, implementing the field experiments, and for countless queries of observational data andbackgroundinformation, in particular to Federico Ranero, Daniel Salgado, and Hector Argente. We thank Basil Halperin for his many contributions during the initial phase of the project. WealsothankDan Ehrlich, Francisca SaraZaror, Rafael Jimenez, Yan Luo, and Kevin DuBartell for excellent researchassistance.Noneofthetwoauthorsareemployeesorconsultants forUberandhave not received payments of any kind from Uber. David Argente is the brother of Hector Argente, who was the Research and Analytics Manager for Uber Latin America. First draft: March 2019. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompaniesofficial NBERpublications. 2020byFernandoE.AlvarezandDavidO.Argente.Allrightsreserved.Shortsectionsoftext, not to exceed two paragraphs, may be quoted without explicit permission provided that fullcredit, includingnotice,isgiventothesourceOntheEffectsoftheAvailabilityofMeansofPayments:TheCaseofUber FernandoE.AlvarezandDavidO.Argente NBERWorkingPaperNo.28145 November2020 JELNo.E41 ABSTRACT WeusethreequasinaturalexperimentsinMexicoandoneinPanamatoestimatetheeffectsof having theoptiontopaywithcashonUberrides.Theabilitytopayincashaffectsthedemand forrides,which isreflectedinlargechangesinthetotalnumberoftrips,fares,miles,andnumber ofusersafterUber introducedcashpayments,particularlyinlowerincomecityblocks.Onthe otherhand,theeffects onprices,estimatedtimesofarrival,andcompetitorpricingarenegligible, consistentwiththesupply oftripsbeingveryelastic.Althoughcashpaymentsnaturallyincrease thefractionofusersthatpay exclusivelywithcash,morethanhalfoftheusershaveaccessto bothcardsandcash,andalternate betweenpaymentmethods.Wefindevidenceconsistentwith cash and card payments being imperfectly substitutable at both the intensive and extensive margins,whichmagnifiestheimpactofpoliciesthat restricttheavailabilityofpaymentmethods. FernandoE.Alvarez UniversityofChicago DepartmentofEconomics 1126East59thStreet Chicago,IL60637 andNBER falvarez1uchicago.edu DavidO.Argente PennsylvaniaStateUniversity DepartmentofEconomics 403KernBuilding UniversityPark StateCollege,PA16801 dargentepsu.edu Adataappendixisavailableatnber/dataappendix/w281451 Summary and Introduction For a number of economists and policymakers, the persistence of cash as a form of payment is potentially problematic. Some have called for the elimination of large-denomination bills, in part because such currency is often the primary transaction method for organized crime and tax evasion, see e.g. Rogos (2017) book The curse of cash, and the ensuing scholarly debate about whether to stage a war on cash (e.g. Bundesbank (2017). Indias demonetization eort in 2016 was a concrete policy that expressed this line of thinking. Nonetheless, for millions of people who have no credit or debit cards or who are disinclined to use them, cash is essential for facilitating economic activity. Chodorow-Reich, Gopinath, Mishra and Narayanan (2018), for instance, estimate that a contraction in employment and economic output as measured by night-lights data following Indias demonetization translated into a 2% decline in the countrys quarterly growth rate. Economically disadvantaged households tend to use cash much more than others, so policies that restrict the use of cash limit economic access for the poor and can have important distributional consequences. For this reason, several cities in the US have discussed or implemented a ban on cashless stores. 1 Uber accepts cash payments in more than 400 cities worldwide; however, some governments have restricted cash payment for ride-hailing services. In Mexico, cash was originally not allowed in several cities (for example in Mexico City or Quer etaro) and was temporarily banned in the cities of Puebla and San Luis Potos . Recently, the Mexican Supreme Court ruled local jurisdictions prohibitions on cash payments for select services as unconstitutional. 2 Cash payments have also been restricted in other countries, such as Panama and Uruguay. In this paper, we estimate the eect of the availability of cash as a payment option on the intensive and extensive margins of Uber trips in Mexico. We use three quasi-natural experiments in Mexico and one in Panama to estimate how cash payments aect rides, prices, and the use of other payment methods. First, we take advantage of the asynchronous entries of cash payments across cities in Mexico. 3 We consider the introduction of cash payments in the Uber app as a demand shock to Uber trips. Since Uber is merely connecting riders with drivers, we analyze the entry of cash as a change in an industry equilibrium. The entry of cash leads to large increases in quantities (i.e. it doubles the number of trips, fares, riders, drivers) but does not increase 1 Cities And States Are Saying No To Cashless Shops, NPR, February 6, 2020. 2 See the decision of the Suprema Corte de Justicia de la Nacion in the case of Ley de Movilidad Sustentable pare el Estado de Colima in October of 2018. 3 Currently, there are more than 40 cities in Mexico where cash is available as means of payment for Uber trips. Greater Mexico City, which is composed by Mexico City and its adjacent municipalities in the State of Mexico, is one of the ten largest metropolitan areas in the world in terms of the gross number of Uber trips. 1prices (i.e. surge multiplier, estimated time of arrival (ETA), prices of taxis). This evidence is consistent with an elastic supply of drivers (in terms of number of active drivers as well as hours worked per driver), which implies that the entry (or ban) of cash has small eects on riders that pay for their trips exclusively with cards or on the producer surplus. Importantly, despite the fact that prices do not change, we observe a small decrease in the number of trips paid in card, which is consistent with a certain degree of substitutability across the two means of payment. Second, we use the dierences in the availability of payment methods across contiguous city blocks in Greater Mexico City to validate our ndings about cash payments under a dierent set of identication assumptions. Using geolocalized trip information, we show that the entry of cash substantially increases the fraction of users that pay for rides exclusively with cash and disproportionately increases the number of rides that begin in lower-income city blocks. We again nd no eect on the prices of Uber rides or those of regular taxis. Using data from the application EC Taximeter, we document that the wait times for regular taxis were also unaected by Ubers introduction of cash payments. Consistent with the results of our event study, we observe a decrease in the number of trips paid for with a card in the city blocks in which Uber was active before it accepted cash payments. Lastly, we study bans on cash payments for ride-hailing services that took place in two cities: Puebla and Panama City. Consistent with our evidence about the introduction of cash payments, we do not nd any evidence of changes in prices. The ban on cash in Puebla immediately reduced the number of trips. We distinguish between the eect on riders that use both payment methods (mixed riders), and the eect on riders that do not register a payment card in the app (pure cash riders). Approximately half of Uber users in Mexico pay with both cash and card. Consistent with cash and credit being far from perfect substitutes, we nd that mixed users that paid for more trips in cash before the ban took fewer trips on Uber after the ban. Cash and credit are also imperfect substitutes at the extensive margin; only about a third of pure-cash users registered a card with Uber after the ban, in excess of the normal rate of migration from cash to credit. Data about Panama Citys ban on cash payments show that, as happened in Mexico, the prices of competing ride-hailing companies and public-transport options were unaected by the change in payment options for Uber rides. Although the data from Panama is relatively limited in scope and granularity, it oers the advantage of observing both the ban on cash payments and the reentry of cash payments months later. Our focus on Uber rides oers two advantages. First, we are able to exploit several quasi- natural experiments to study the changes in the supply and demand of the same good that can be paid for with varying means. An Uber user can, in principle, alternate between paying 2with cash or card, and Uber tracks which was used. Second, Uber measured specics about how prices and quantities of rides changed with changes in payment options. The richness of the data allows us to follow users decisions with ne geographic and spatial resolution. Our results from city-level, block-level, and individual-level data all point to the same conclusions qualitatively (if not quantitatively) and are robust to dierent methodologies and identication strategies (i.e. event study, coarsened exact matching, regression discontinuity, and synthetic control methods). In summary, Uber users pay using cash very often when the option is available, and the availability of cash payment has no signicant impact on prices, either monetary or non- monetary (i.e. wait times), or on the prices of Ubers competitors (i.e. prices and wait times for taxis). We nd evidence that cash and credit are imperfectly substitutable at both the extensive and intensive margins of a change in the availability of cash payments. Our results indicate that policies restricting the use of cash have a negative impact on the fares of pure- cash riders and on the fares paid by riders that use both payment methods, which are the majority of users in Mexico. The imperfect substitutability of cash as means of payment also indicates that the recent increase in contactless payments due to the health risks associated with COVID-19 is not without cost. Our work contributes to the literature about the continued prevalence of consumers who mix their use of cash and card payments in the broader marketplace. One possibility is that households use multiple payment methods in order to diversify the source and timing of funding among dierent means for payment (Shy, 2019). Another alternative is that the high use of cash payments for other goods makes the use of cash complementary, even for those users that own cards. Briglevics and Schuh (2014) nd that consumers with very large amounts of cash in their wallets are more likely to use cash and that consumers try to postpone withdrawals until a favorable opportunity is available. Similarly, Arango, Hogg and Lee (2015) use shopping diaries from Canada with information on consumers payment choices and nd that cash burns, meaning that the more cash individuals hold at the beginning of a 3-day shopping period, the more likely they are to use cash even when they have access to debit/credit cards. They also nd that consumers dislike the possibility of running out of cash, since they face costs in terms of time, eort, and fees to get more. Alvarez and Lippi (2017) construct a decision-making model in which cash and credit are used simultaneously in a way that is consistent with the evidence from developed countries in Arango, Hogg and Lee (2015). Deviatov and Wallace (2014) develop a model in which some fraction of the population is unbanked and uses only cash; for this reason, in equilibrium, even those who have access to banking services nd it convenient to hold and use cash. Another possibility is that cash and debit/credit cards are imperfect substitutes from the 3consumers perspective. Koulayev, Rysman, Schuh and Stavins (2016) shows evidence for substitution across payment methods, particularly between cash and debit cards. Consistent with their evidence, we nd that a fraction of users switch to cards after the use of cash is made less attractive, but the degree of substitution we observe is far from perfect. This evidence is complementary to evidence reported in Alvarez and Argente (2020), who nd similar results using eld experiments in Mexico. We believe that understanding both the reasons for the prevalence of mixed users and their adjustments following changes to the availability of payment methods is not only relevant for the theoretical literature in cash- credit but also to evaluate the impact of policies that restrict or enable various means of payment. In the next section we briey summarize the four quasi-natural experiments we study. Entry of Cash Across Mexican Cities For the entry of cash, we use two dierent strategies. First is an event study of the asynchronous entry of cash to 15 dierent cities where Uber had previously only accepted payment via credit or debit card. This part of the analysis is described in Section 4. Our understanding of Ubers decision to introduce cash in these cities is that after the successful introduction of cash in May of 2015 in Hyderabad (India), Uber decided that cash could be introduced to all cities in developing countries where it was allowed. Thus, we assume that the entry is quasi-random since the dierence in the tim