资源描述
NBERWORKINGPAPERSERIES SMART(PHONE)INVESTING?AWITHININVESTORTIMEANALYSISOFNEW TECHNOLOGIESANDTRADINGBEHAVIOR. AnkitKalda BenjaminLoos AlessandroPrevitero AndreasHackethal WorkingPaper28363 nber/papers/w28363 NATIONALBUREAUOFECONOMICRESEARCH 1050MassachusettsAvenue Cambridge,MA02138 January2021 WethankShlomoBenartzi,JuhaniLinnainmaa,UlrikeMalmendier,BrianMelzer,andseminar participants atIndianaUniversityandBINorwegianBusinessSchoolforhelpfulcommentsand discussions.The viewsexpressedhereinarethoseoftheauthorsanddonotnecessarilyreflectthe viewsoftheNational BureauofEconomicResearch. NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeen peerreviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompanies official NBERpublications. 2021byAnkitKalda,BenjaminLoos,AlessandroPrevitero,andAndreasHackethal.Allrights reserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicit permission providedthatfullcredit,includingnotice,isgiventothesourceSmart(Phone)Investing?AwithinInvestortimeAnalysisofNewTechnologiesandTrading Behavior. AnkitKalda,BenjaminLoos,AlessandroPrevitero,andAndreasHackethal NBERWorkingPaperNo.28363 January2021 JELNo.G11,G40,G50 ABSTRACT Using transactionlevel data from two German banks, we study the effects of smartphones on investor behavior. Comparing trades by the same investor in the same month across different platforms, we find that smartphones increase purchasing of riskier and lotterytype assets and chasingpastreturns. Aftertheadoptionofsmartphones,investorsdonotsubstitutetradesacross platforms and buy also riskier, lotterytype, and hot investments on other platforms. Using smartphonestotradespecificassets orduringspecifichourscontributestoexplainourresults. Digital nudges and the device screen size do not mechanically drive our results. Smartphone effectsarenottransitory. AnkitKalda KelleySchoolofBusiness 1309E10thSt IndianaUniversity Bloomington,IN47405 akaldaiu.edu BenjaminLoos TUMSchoolofManagement Arcissstrasse21 Munich80333 Germany AlessandroPrevitero KelleySchoolofBusiness IndianaUniversity 1309E.10thStreet Bloomington,IN47405 andNBER alepreviindiana.edu AndreasHackethal GoetheUniversityFrankfurt Grneburgplatz1HouseofFinance 60323Frankfurta.M.Germany hackethalgbs.unifrankfurt.de1 Introduction Technology has dramatically changed how retail investors trade, from placing orders using direct dial-up connections in the 1980s or Internet-based trading in the 1990s to the more recent rise of robo-advisers. With few exceptions, the introduction of these new technologies is generally associated with a decline in investor portfolio eciency.1 Whethergoodorbadforinvestors,itisacceptedthatnewtechnologiesinuenceinvestor behavior. The empirical evidence in these studies comes from some comparisons of in- vestorbehaviorbeforeandaftertheadoptionofthenewtechnology,potentiallycontrasted with the behavior over time of another group that did not adopt the technology. Under the assumption that, absent the innovation, investors would have behaved in the exact same way, a common interpretation of this evidence is that new technologies inuence investorsand change their behavior. An alternativeexplanation isthat investors,instead, adoptthenewtechnologybecausetheyarewillingtochangetheirtradingbehaviorinthe rst place. Even if we could randomly assign the new technology to investors,2 it would still not be straightforward to conclude that the new technology changes the overall in- vestor portfolio. If investors manage investments across dierent accounts or platforms, they could decide to substitute across technologies. Therefore, observing trades on one platformmight not be informativeoftheoverallinvestortradingbehavior. Whilepreviousstudieslackthedatatodistinguishbetweenthesealternativeinterpre- tations,theirimplicationsare,however,starklydierent. Ifthenewtechnologyinuences investorpreferencesandbeliefs,absentthetechnologyinvestorswouldhavenotchanged theirtradingbehavior. If,instead,itfulllsuntappedinvestordemand,thenthenewtech- 1For example, when moving to online trading, investors increased turnover and reduced performance (Barber and Odean, 2002). More recent studies document, instead, that robo-advisers could reduce invest- ment mistakes (seeDAcunto,Prabhalaand Rossi, 2019;Loos etal., 2020). 2DAcunto, Prabhala and Rossi (2019) use the randomness in investors answering their phone to the marketing enrollmentcalls asa plausiblyexogenous shocktotheprobability ofjoining the robo-advisor. 1nology at best accelerates or makes less costly a change in investor behavior that would have happened anyway. Analogously, the new technology could just fulll substitute demand, if investors substitute trades across dierent platforms. Therefore, simple com- parisons of investor behavior pre- and post-adoption or analyses of trades on one single platform could vastly overestimate the eects of the new technology. Furthermore, the policy implications could not be any more dierent. Is the technology helping investors to achieve their goals by facilitating their trades? Or is technology inuencing adopters inprofound ways that could strayinvestorsawayfromtheiroriginalgoals?3 Inthispaper,weuseuniquedataonGermanhouseholdstoovercometheseempirical challenges and to weigh in on the question if technology drives changes or just fullls untapped or substitute investor demand. Our data comes from two large German retail banks that have introduced trading applications for mobile devices. For over 15,000 bank clients that have used these mobile apps in the years 2010-2017, we can observe all holdingsandtransactions,and,moreimportant,thespecicplatformusedforeachtrade (e.g.,personalcomputervs. smartphone). Theseuniquefeaturesofthedataprovefruitful forouranalyses. Thatis,wecanconductallourmaintestscomparingtradesdonebythe same investor in the same month across dierent platforms. Moreover, we can directly testforsubstitution eects. We present four set of results. First, we study if the use of smartphones induces dierences in the riskiness of trades. Comparing trades by the same investor in the same year-month, we nd that the probability of purchasing risky assets increases in smartphone trades compared to non-smartphone ones. Analogously, smartphone trades involve assets with higher volatility and more positive skewness. This evidence is best summarizedbyouranalysesoflotterytypestocks.4 Smartphonesincreasetheprobability 3Ina2020articletitled“RobinhoodHasLuredYoungTraders,SometimesWithDevastatingResults”,theNew York Times features a series of stories of investors that have lost substantial amount of money trading o their mobilephones. 4Following Kumar (2009), we dene as lottery-type stocks those assets with below median prices and 2ofbuyinglottery-type stocks by 67%oftheunconditionalmeanforsmartphoneusers. Second, we examine the eects of smartphones on the tendency to chase past returns. We nd that smartphones increase the probability of buying assets in the top decile of thepastperformancedistribution. Smartphonesincreasetheprobabilityofbuyingassets in the top 10 percent of past performance by 12.0 percentage points (or 70.6% of the unconditional mean). Third, we investigate if investors selectively use smartphone to execute their risky, lottery-type, and trend-chasing trades. In this scenario, investors could simply substitute their trades from one device to another, without any real consequences for their over- all portfolio eciency. Using a dierence-in-dierences design that compares iOS and Android users, we nd that, following the launch of smartphone apps, investors areif anythingmore likely to purchase risky and lottery-type assets and to chase hot invest- mentsalsoonnon-smartphoneplatforms. Whileinconsistentwithsubstitutioneects,this evidence potentially suggests that investors are learning to become overall more biased aftertheirinitial use of smartphonestotrade. Last, we evaluate the mechanisms that may drive these smartphone eects. We begin byexaminingwhethertheabilitytotradeanytimeandeverywhereviasmartphonesdrives ourresults. Toevaluatetheimportanceofthischannel,werepeatouranalyses,including year-by-time-of-the-day xed eects. In this specication, our estimates become smaller but remain economically and statistically signicant. This nding suggests that time of trade is important in our setting, but it does not fully explain our ndings. Consistent withthisinterpretation,heterogeneityanalysesshowthatsmartphoneeectsarestronger during after-hours (i.e. following exchange closure). Institutional dierences between trading on ocial exchanges and in after-hours markets do not drive this heterogeneity. Given that individuals are more likely to rely on the more intuitive system 1 later in the above medianskewnessand volatility. 3day(Kahneman,2011),strongereectsduringafter-hoursareconsistentwithsmartphones facilitating trades based more on system1thinking. Alternatively, investors may use smartphones to trade dierent investments and this selection of riskier asset classes may drive our results. We re-estimate our main anal- yses, including year-by-asset-class xed eects. We nd again smaller but still strong smartphone eects, suggesting that the choice of asset classes doesnt fully explain our ndings. Anotherpossibilityisthatdigitalnudgesmightcontributetoourresults. Smartphone trading apps in fact prominently feature stocks that have experienced dramatic positive (and negative) performance in the recent past. If these stocks are riskier and with higher skewness, digital nudges could mechanically inuence investor behavior. To test for this hypothesis, we re-run our main specications separately for dierent asset classes: indi- vidual stocks, mutual funds, and other investments (options, certicates, and warrants). Wendthatourresultsarestrongacrossallassetclassesandnotjustforindividualstocks that can be more prominently featured in the smartphone trading app. Additionally, we test if a physical attribute of smartphonestheir smaller screencontributes to our ndings. To explore this mechanism, we separately investigate the eects of trading via deviceswithdierentscreensizes(iPhonesvs. iPads). Giventhatwedonotndstronger resultsfortradesviaiPhones,weconcludethatthisphysicalattributeisnotlikelytodrive our ndings. Last, our results do not appear to be short-lived and driven by the initial enthusiasm or the learning curve of the new technology. Our estimates do not change signicantly between the rst quarter up to the tenth quarter after the initial use of the smartphone app. Ourndingscontributetoliteratureontheeectsoftechnologyoninvestorbehavior. Barber and Odean (2002) document that investors who switched from phone-based to onlinetradingstarttradingmorefrequently,butlessprotablythanbefore. Choi,Laibson, 4and Metrick (2002) document similar results in 401(k) plans. Our evidence complements these studies by documenting that smartphones increase the purchases of lottery-type stock and trend-chasing. More importantly, we document dierent behaviors within the same investor and same month, but across platforms. This identication strategy enables ustomoreconvincinglyaddressselectioneectswhenexamininghowanewtechnology impactsinvestor behavior. Given the large diusion of robo-advisers in the past decade, DAcunto, Prabhala and Rossi (2019) and Loos et al. (2020) have investigated the eects of this innovation on investor behavior. Both studies highlight that robo-advice has the potential to reduce investment biases and improve portfolio performance. Our evidence provide a more nuanced picture of the eects of new technologies on investor behavior. Smartphones appear to foster investment biases such as investing in lottery-type and hot stocks. Our paper contributes also to the recent literature on the eect of mobile apps on nancial behaviors. Levi and Benartzi (2020) and DAcunto, Rossi, and Weber (2020) study the eects of mobile applications on spending behaviors. We contribute to these studies by investigating investment decisions. Our setting provides a nice laboratory to understand the consequences of providing constant feedback and ease of execution of trades to retail investors. More recently, a series of studies have investigated the eects of trading smartphone apps on aggregate markets. Using data from the US retail brokerage company Robin Hood,Welch(2020)ndsthataportfoliomimickingtheaggregateholdingsofRobinHood investors did not underperform standard academic benchmarks.5 Using the same data, Barberetal. (2020)documentthatepisodesofintensebuyingactivitybyRobinhoodusers arefollowedbynegativereturns. UsingdatafromaleadinginvestmentadviserinChina, 5Robin Hood operates entirely online via a website and mobile apps. The vast majority of its trades are made usingthesmartphoneapps. 5Cen (2019) shows that, after the mobile app introduction, investors ows into mutual funds become more volatile and more sensitive to short-term fund returns and market sentiment. Our results nicely dovetail with the ndings in these studies and make three distinctive contributions. First, we focus on the consequences of smartphones on retail investors, and not aggregate markets. Aggregate eects might mask substantial investor heterogeneity, making it dicult to understand potential redistributive eects of this technology. Second, our investor trading data allow to sharpen the causal interpretation of smartphone eects and to investigate the mechanisms driving them. Third, while Robinhood investors are Millennials with little or no trading experience, the German investors that adopt smartphone trading are, on average, 45 years old with nine years of experienceinvestingwiththebanks. Therefore,wecancapturetheeectsofsmartphone tradingonmore experienced usersandamorerepresentativesampleoftraders. 2 Hypotheses Development New technologies can change the way households make economic decisions, from labor supply, to borrowing, to investor behavior.6 Broadly speaking, we investigate if smartphones inuence nancial risk-taking and investment biases. The eects of smart- phonesinbothsettingsarenotobviousex-ante. Byfacilitatingsearchingandmonitoring eorts, smartphones can reduce the participation costs in the stock market and promote nancial risk-taking. If investors are, instead, sensitive to short-term losses, the more frequent feedback via smartphones could reduce risk-taking, as predicted in the frame- work of myopic loss aversion by Benartzi and Thaler (1995). Consistent with myopic loss aversion, Haigh and List (2005) document that professional option traders take less risk 6Forexample,Fosetal(2019),Jackson(2019)andKoustas(2018)documenttheeectofride-sharingapps onlabormarketdecisions;DiMaggioandYao(2019),Buchak,Matvos,PiskorskiandSeru(2018)andFuster et al. (2018) document the eect of Fintech lending on borrowing decisions; and DAcunto, Prahabala, and Rossi (2019) documentthe eectof robo-advisingon investment decisions. 6whenrandomly assigned to th
展开阅读全文