Uncertainty about Federal Reserve policy and Its transmission to emerging economies: Evidence from Twitter

Tillman, Pe e Wo king Pape Unce ain y abou Fede al Rese e policy and I s ansmission o eme ging economies: E idence om Twi e ADBI Wo king Pape , No. 592 P o ided in Coope a ion wi h: Asian De elopmen Bank Ins i u e (ADBI), Tokyo Sugges ed Ci a ion: Tillman, Pe e (2016) : Unce ain y abou Fede al Rese e policy and I s ansmission o eme ging economies: E idence om Twi e , ADBI Wo king Pape , No. 592, Asian De elopmen Bank Ins i u e (ADBI), Tokyo This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/161468 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h ps://c ea i ecommons.o g/licenses/by-nc-nd/3.0/igo/ ADBI Wo king Pape Se ies Unce ain y abou Fede al Rese e Policy and I s T ansmission o Eme ging Economies: E idence om Twi e Pe e Tillman No. 592 Sep embe 2016 Asian De elopmen Bank Ins i u e The Wo king Pape se ies is a con inua ion o he o me ly named Discussion Pape se ies; he numbe ing o he pape s con inued wi hou in e up ion o change. ADBI’s wo king pape s e lec ini ial ideas on a opic and a e pos ed online o discussion. ADBI encou ages eade s o pos hei commen s on he main page o each wo king pape (gi en in he ci a ion below). Some wo king pape s may de elop in o o he o ms o publica ion. Sugges ed ci a ion: Tillman, P. 2016. Unce ain y abou Fede al Rese e Policy and I s T ansmission o Eme ging Economies: E idence om Twi e . ADBI Wo king Pape 592. Tokyo: Asian De elopmen Bank Ins i u e. A ailable: h p://www.adb.o g/publica ions/unce ain y-abou - ede al- ese e-policy-and-i s- ansmission-eme ging-economies/ Please con ac he au ho s o in o ma ion abou his pape . Email: pe e . ill[email p o ec ed]-giessen.de Pe e Tillmann is p o esso o Mone a y Economics a Jus us Liebig Uni e si y Giessen, Ge many. The iews exp essed in his pape a e he iews o he au ho and do no necessa ily e lec he iews o policies o ADBI, ADB, i s Boa d o Di ec o s, o he go e nmen s hey ep esen . ADBI does no gua an ee he accu acy o he da a included in his pape and accep s no esponsibili y o any consequences o hei use. Te minology used may no necessa ily be consis en wi h ADB o icial e ms. Wo king pape s a e subjec o o mal e ision and co ec ion be o e hey a e inalized and conside ed published. Asian De elopmen Bank Ins i u e Kasumigaseki Building 8F 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: +81-3-3593-5500 Fax: +81-3-3593-5571 URL: www.adbi.o g E-mail: [email p o ec ed] © 2016 Asian De elopmen Bank Ins i u e ADBI Wo king Pape 592 P. Tillman Abs ac I is well known ha a igh ening o easing o he Uni ed S a es’ mone a y policy a ec s inancial ma ke s in eme ging economies. This pape a gues ha unce ain y abou u u e mone a y policy is a sepa a e ansmission channel. We ocus on he ape an um episode in 2013, a pe iod wi h an ele a ed unce ain y abou mone a y policy, and use a da a se ha con ains 90,000 Twi e messages (“ wee s”) on Fede al Rese e ape ing. Based on his da a se , we cons uc a new index abou mone a y policy unce ain y using a lis o unce ain y keywo ds. An ad an age o his index is ha i e lec s unce ain y abou a speci ic policy decision. An es ima ed ec o au o eg ession (VAR) shows ha unce ain y shocks lead o a all in asse p ices and a dep ecia ion o local cu encies. We also discuss he policy implica ions o his unce ain y channel o mone a y policy ansmission. JEL Classi ica ion: E32, E44, E52 ADBI Wo king Pape 592 P. Tillman Con en s 1. In oduc ion .................................................................................................................. 3 2. Measu ing Policy Unce ain y wi h Twi e Da a .......................................................... 4 3. The Empi ical Model .................................................................................................... 8 4. Resul s ........................................................................................................................ 9 4.1 The Responses o Unce ain y Shocks ......................................................... 10 4.2 The Con ibu ion o Unce ain y Shocks ........................................................ 11 4.3 Robus ness ................................................................................................... 11 5. Policy Implica ions ..................................................................................................... 13 6. Conclusions ............................................................................................................... 15 Re e ences ............................................................................................................................ 16 Appendix ............................................................................................................................... 18 ADBI Wo king Pape 592 P. Tillman 1. INTRODUCTION Tigh ening and easing policy s eps o he Uni ed S a es’ (US) Fede al Rese e (Fed) a e known o a ec eme ging ma ke economies h ough a ious channels. Typically, a Fed igh ening leads o a wi hd awal o capi al om eme ging economies, a d op in asse p ices, and a eal dep ecia ion o local cu encies. These spillo e e ec s a e o en he subjec o majo conce ns o policymake s in eme ging ma ke economies. A e he 2007–2009 global inancial c isis, mone a y policy s eps ha e led o pa icula ly s ong spillo e e ec s. The a ious ounds o Quan i a i e Easing (QE1, QE2, QE3), he subsequen exi om QE3 (“ ape ing”) and he e en ual li -o o he Fede al Funds a e om he ze o lowe bound in Decembe 2015 ha e led o sha p adjus men s in eme ging ma ke s. Announcemen s o u he mone a y easing led o massi e capi al in lows in o eme ging economies. The e en ual unwinding o asse pu chases and he p ospec o a li -o e e sed hese lows—wi h all he pain ul consequences. These spillo e s o igina e om speci ic mone a y policy decisions. O e he yea s since he global inancial c isis, a second ype o spillo e s has eme ged: spillo e s s emming om he unce ain y abou u u e mone a y policy. Unce ain y spillo e s exis abo e and beyond hose s emming om speci ic policy s eps. I is he unce ain y abou he likely cou se o mone a y policy ha led in es o s o adjus hei po olios, hus leading o side e ec s on eme ging ma ke s. This unce ain y channel o in e na ional policy ansmission has been highligh ed by he inancial p ess. The Financial Times no only ecen ly i led he abo e-men ioned s a emen , bu also s a ed ha “Eme ging ma ke s call on Fed o li a es and end unce ain y” (9 Sep embe 2015) and “Fed cla i y on a es is he ca alys ha unce ain ma ke s need” (12 Oc obe 2015). In his pape we quan i y unce ain y spillo e s, wi h a special ocus on Asia, based on a no el da a se . We ocus on he “ ape an um” episode in 2013, which was igge ed by Fed chai man Ben Be nanke’s es imony on 22 May 2013. When he e lec ed abou he possibili y o exi ing Quan i a i e Easing (QE) and s a ing he g adual no maliza ion o mone a y policy, ma ke s eac ed wi h high unce ain y, a ise in ola ili y, and a sha p inc ease in US bond yields. I is he s ong ma ke eac ion ha ga e his pe iod i s name he “ an um” episode. These spillo e e ec s du ing he ape an um we e pa icula ly se e e o eme ging ma ke economies, which expe ienced s ong dep ecia ion o domes ic cu encies and a sudden igh ening o mone a y condi ions. The ape an um pe iod is esembling he ecen deba e abou he i s inc ease o he Fede al Funds a e a ge a e 9 yea s o nea -ze o policy a es: he ape ing decision was expec ed o he June o July Fede al Open Ma ke Commi ee (FOMC) mee ing, hen o Sep embe 2013, and had inally been delayed o ea ly 2014. Bo h he iming and he delays a e simila o he “li o ” in 2015, causing obse e s o e e o a eappea ance o he ape an um. The e is some quan i a i e e idence on ape ing spillo e s. The li e a u e includes Eicheng een and Gup a (2013); Aizenman, Binici, and Hu chison (2014); Rai and Suchanek (2014); Sahay e al. (2014); Mish a e al. (2014); and Lim, Mohapa a, and S ocke (2014). These pape s assess he magni ude o spillo e s and he impo ance o domes ic mac oeconomic undamen als o he sensi i i y o ape ing- ela ed spillo e s. They do no , howe e , shed ligh on he unce ain y channel o in e na ional policy ansmission. 3 ADBI Wo king Pape 592 P. Tillman The key con ibu ion o his pape is a no el da a se : ou da a se consis s o 90,000 Twi e messages (“ wee s”), which e lec he en i e discussion on he Fed ape ing be ween Ap il and Oc obe 2013 on Twi e .com. F om his we cons uc an indica o o unce ain y abou u u e mone a y policy, which we hen eed in o a s anda d ec o au o eg essi e (VAR) model. We use a comp ehensi e wo d lis om he li e a u e ha con ains keywo ds ha desc ibe unce ain y and apply his o all ou wee s. This gi es us an index o unce ain y ha is di ec ed o a e y speci ic mone a y policy decision. This is an impo an ad an age o e unce ain y indica o s used in he li e a u e, e.g., Bake , Bloom, and Da is (2013), which o en e lec a b oad, unspeci ic no ion o unce ain y. 1 A second ad an age is ha we measu e unce ain y di ec ly om messages sen , ead, and o wa ded by ma ke pa icipan s. A hi d ad an age is ha his index is a ailable on a daily equency.2 We ind ha an inc ease o unce ain y abou he iming o he exi om QE leads o signi ican e ec s on eme ging ma ke economies. Unce ain y leads o a all in bond and s ock p ices, a dep ecia ion o local cu encies, and an inc ease in c edi de aul swap (CDS) sp eads. Se e al obus ness checks such as con olling o he mac oeconomic en i onmen , he in o ma ion coming om FOMC mee ings, and he numbe o e wee s con i m ou indings. Thus, his pape p o ides e idence o an unce ain y channel o in e na ional mone a y policy ansmission. The esul s highligh he need o clea and consis en communica ion o majo policy decisions o cen al banks in ad anced economies. Majo policy u na ounds such as he exi om QE should be designed wi h special a en ion o he ulne abili ies o eme ging economies. This is pa icula ly ele an o , among o he s, he Eu opean Cen al Bank and he Bank o Japan, which will ha e o communica e he exi om uncon en ional mone a y policy a some poin in he u u e. Ou esul s also sugges ha main aining s ong mac oeconomic undamen als in eme ging ma ke s is bes sui ed o wi hs and unce ain y spillo e s. The emaining pa s o he pape a e o ganized as ollows. Sec ion 2 in oduces ou da a se and explains he cons uc ion o he new measu e o unce ain y abou US mone a y policy. Sec ion 3 explains he empi ical model used o es ima e he e ec s o unce ain y shocks. Sec ion 4 discusses he de ails and he obus ness o he indings. A pe spec i e on policy implica ions in Sec ion 5 and a concluding sec ion comple e he pape . 2. MEASURING POLICY UNCERTAINTY WITH TWITTER DATA We use a sample ha cap u es he 2013 ape an um episode and, hence, a pe iod in which ma ke unce ain y abou u u e US mone a y policy was high and ola ile. I is also a pe iod wi h a majo shi in expec a ions abou he u u e cou se o US mone a y policy. To es ima e he e ec s o unce ain y, we need da a (i) ha is a ailable on a high equency in o de o e lec he high-speed ansmission o in o ma ion on inancial ma ke s, (ii) ha e lec s unce ain y o ac ual ma ke pa icipan s, no o he o e all economy, and (iii) ha allow us o isola e he speci ic unce ain y abou 1 Is e i and Piloiu (2013) use he Bake , Bloom, and Da is (2013) index o policy unce ain y o he US, he Uni ed Kingdom, Ge many, and he eu o a ea o quan i y he esponse o in la ion expec a ions o a b oadly de ined no ion o unce ain y. 2 The same da a is used in Meinusch and Tillmann (2015) and Tillmann (2015). None o hese pape s, howe e , is in es iga ing he e ec s o policy unce ain y in he US on eme ging ma ke s‘ asse p ices and inancial condi ions, espec i ely. 4 ADBI Wo king Pape 592 P. Tillman mone a y policy om he unce ain y abou o he ields o economic policy. Fo he la e eason, he commonly used indica o o policy unce ain y om Bake , Bloom, and Da is (2013) canno be used as i summa izes unce ain y on a wide ange o policy a eas. The co e con ibu ion o his pape is a new da a se ha allows he esea che o ex ac a measu e o ma ke pa icipan s’ unce ain y on a high equency. Ins ead o measu ing expec a ions by coun ing newspape a icles like in Bake , Bloom, and Da is (2013) o in e ing in es o unce ain y om inancial p oduc s, we u n o social media. Due o i s ubiqui y and esponsi eness, social media is a p omising da a sou ce o all esea ch ques ions o which unce ain y o ma ke pa icipan s plays an impo an ole. We ha e a da a se ha includes he en i e discussion abou he ape an um episode on Twi e .com. Twi e is a social media ne wo k on which use s can exchange sho ex messages ( wee s) o a maximum 140 cha ac e s. Twi e is popula among inancial ma ke p o essionals and Fed wa che s. Impo an ly, Twi e use s can no only send messages o ollowe s, bu also o wa d wee s o commen on o he messages. Thus, we ha e a da a se a hand ha ul ills he c i e ia ou lined be o e. To be speci ic, we ha e he en i e Twi e olume on Fed ape ing be ween 15 Ap il and 30 Oc obe 2013. This is he pe iod o which mos o he ma ke eac ions could be obse ed. The sample pe iod is ela i ely sho , which is dic a ed by da a a ailabili y, bu no oo sho o es ima e a pa simonious model. Wi h his in mind, we use a model below ha lea es su icien ly many deg ees o eedom. The aw da a has been p o ided by Gnip.com, which is now pa o Twi e .com. The da a was deli e ed in 10-minu e in e als o ma ed in JASON.3 The da a se includes 87,621 wee s om 27,276 use s loca ed in 136 economies and he exac ime hey we e sen . We ans o m he da a in o a sp eadshee o ma and no malize he ime o each message o New Yo k ime. We emo e public holidays and weekends in he US om hese se ies because he e is no inancial da a a ailable on hese days. In addi ion, e y ew wee s in languages o he han English had o be emo ed. Panel (a) in Figu e 1 plo s he se ies o wee s agg ega ed o a daily se ies. The ho izon al axis gi es he days in he sample pe iod. We see ha he numbe o wee s explodes a he Sep embe 2013 FOMC mee ing, o which he ape ing decision o he Fed had been expec ed. Howe e , he Fed decided o pos pone ape ing o a subsequen mee ing. The numbe o wee s sys ema ically inc eases no only a ound he Sep embe FOMC mee ing, bu also a ound o he FOMC mee ings. Fo ha eason we will p esen below an es ima ion speci ica ion in which we con ol o FOMC mee ing days. The scaling o he g aph unde s a es he magni ude o luc ua ions in he numbe o wee s, which a y s ongly o e ime. Based on hese ex messages we wan o de i e a quan i a i e measu e o unce ain y. Fo his pu pose, we eso o ex mining and use a dic iona y wi h keywo ds ha signal unce ain y o he sende o a gi en Twi e message. This dic iona y o wo ds is aken om he widely ci ed pape o Lough an and McDonald (2011).4 This wo d lis is pa icula ly use ul o ou pu poses because i is ela i ely comp ehensi e and pa icula ly calib a ed o e lec unce ain y in inancial communica ion. A dic iona y app oach is p e e able in ou con ex because Twi e 3 The da a is p op ie a y such ha he aw da a canno be made a ailable. 4 Thus, we c ea e an indica o o unce ain y which is based on he wisdom o he c owd in inancial ma ke s. This is simila in spi i o he index c ea ed by Bachmann, Els ne , and Sims (2013) om i m su eys, al hough we measu e a mo e speci ic o m o in es o unce ain y. 5 ADBI Wo king Pape 592 P. Tillman messages a e ypically highly poin ed, o en cynical s a emen s wi h equen use o ma ke ja gon. O he ex mining ools such as opic modeling migh ha e di icul ies g asping he ull no ion o each message. Table 1: Unce ain y Keywo ds Abeyance Abeyances Almos Al e a ion Al e a ions Ambigui ies Ambigui y Ambiguous Anomalies Anomalous Anomalously Anomaly An icipa e An icipa ed An icipa es An icipa ing An icipa ion An icipa ions Appa en Appa en ly Appea Appea ed Appea ing Appea s App op ia e App oxima ed App oxima ely App oxima es App oxima ing App oxima ion App oxima ions A bi a ily A bi a iness A bi a y Assume Assumed Assumes Assuming Assump ion Assump ions Belie e Belie ed Belie es Belie ing Cau ious Cau iously Cau iousness Cla i ica ion Cla i ica ions Concei able Concei ably Condi ional Condi ionally Con uses Con using Con usingly Con usion Con ingencies Con ingency Con ingen Con ingen ly Con ingen s Could C oss oad C oss oads Depend Depended Dependence Dependencies Dependency Dependen Depending Depends Des abilizing De ia e De ia ed De ia es De ia ing De ia ion De ia ions Di e Di e ed Di e ing Di e s Doub Doub ed Doub ul Doub s Exposu e Exposu es Fluc ua e Fluc ua ed Fluc ua es Fluc ua ing Fluc ua ion Fluc ua ions Hidden Hinges Imp ecise Imp ecision Imp ecisions Imp obabili y Imp obable Incomple eness Inde ini e Inde ini ely Inde ini eness Inde e minable Inde e mina e Inexac Inexac ness Ins abili ies Ins abili y In angible In angibles Likelihood May Maybe Migh Nea ly Nonassessable Occasionally O dina ily Pending Pe haps Possibili ies Possibili y Possible Possibly P ecau ion P ecau iona y P ecau ions P edic P edic abili y P edic ed P edic ing P edic ion P edic ions P edic i e P edic o P edic o s P edic s P elimina ily P elimina y P esumably P esume P esumed P esumes P esuming P esump ion P esump ions P obabilis ic P obabili ies P obabili y P obable P obably Random Randomize Randomized Randomizes Randomizing Randomly Randomness Reassess Reassessed Reassesses Reassessing Reassessmen Reassessmen s Recalcula e Recalcula ed Recalcula es Recalcula ing Recalcula ion Recalcula ions Reconside Reconside ed Reconside ing Reconside s Reexamina ion Reexamine Reexamining Rein e p e Rein e p e a ion Rein e p e a ions Rein e p e ed Rein e p e ing Rein e p e s Re ise Re ised Risk Risked Riskie Riskies Riskiness Risking Risks Risky Roughly Rumo s Seems Seldom Seldomly Some ime Some imes Somewha Somewhe e Specula e Specula ed Specula es Specula ing Specula ion Specula ions Specula i e Specula i ely Spo adic Spo adically Sudden Suddenly Sugges Sugges ed Sugges ing Sugges s Suscep ibili y Tending Ten a i e Ten a i ely Tu bulence Unce ain Unce ainly Unce ain ies Unce ain y Unclea Uncon i med Undecided Unde ined Undesigna ed Unde ec able Unde e minable Unde e mined Undocumen ed Unexpec ed Unexpec edly Un amilia Un amilia i y Un o ecas ed Un o seen Ungua an eed Unhedged Uniden i iable Uniden i ied Unknown Unknowns Unobse able Unplanned Unp edic abili y Unp edic able Unp edic ably Unp edic ed Unp o ed Unp o en Unquan i iable Unquan i ied Un econciled Unseasonable Unseasonably Unse led Unspeci ic Unspeci ied Un es ed Unusual Unusually Unw i en Vaga ies Vague Vaguely Vagueness Vaguenesses Vague Vagues Va iabili y Va iable Va iables Va iably Va iance Va iances Va ian Va ian s Va ia ion Va ia ions Va ied Va ies Va y Va ying Vola ile Vola ili ies Vola ili y Sou ce: Lough an and McDonald (2011). h p://www3.nd.edu/~mcdonald/Wo d_Lis s.h ml 6 ADBI Wo king Pape 592 P. Tillman We see ha ou baseline indings emain unchanged. An unce ain y shock leads again o a d op in asse p ices. The EMBI alls by 0.2% on impac . Hence, con olling o he mac oeconomic en i onmen does no change ou o e all indings. Tha also means ha he wee s in ou sample do no simply e lec he in e p e a ion o news eleases such as non- a m pay oll employmen and in la ion. Taken oge he , we ind obus e idence o an unce ain y channel o in e na ional policy ansmission. The unce ain y abou u u e policy i sel has an e ec on eme ging ma ke s’ asse p ices. This e ec comes on op o he egula spillo e s o policy changed in he US on eme ging inancial ma ke s. The unce ain y channel gi es ise o impo an policy implica ions, which a e discussed in he ollowing sec ion. 5. POLICY IMPLICATIONS The esul s p esen ed in his pape unde line wo impo an policy implica ions, which a e discussed in u n. The i s implica ion pe ains o he design o mone a y policy in ad anced economies. Ou esul s sugges a need o be e communica ion o policy changes. In gene al, mone a y policy should be as p edic able and anspa en as possible in o de o allow ma ke s o p ice in u u e policy mo emen s. This is impo an om he pe spec i e o he domes ic economy, bu also ele an om he pe spec i e o eme ging economies. The changing unce ain y s udied in his pape has been he esul o an ill-designed communica ion o a majo policy u na ound. When Fed chai man Be nanke ga e his es imony on 22 May 2013, he did men ion he possibili y o a educ ion in bond pu chases unde QE3—an ob ious consequence o a pe sis en economic eco e y. Al hough his ema ks we e ca e ully c a ed, ma ke s ne e heless o e eac ed. While pa o he o e eac ion migh ha e been due o an i a ional panic, he ape an um sugges s ha ca e ully communica ing policy s eps and, in pa icula , policy e e sals, is c ucial. This is an impo an lesson o cen al banks such as he Bank o England, he Bank o Japan, o he Eu opean Cen al Bank, o which he exi om uncon en ional measu es and he e u n o policy no maliza ion is ye o come. These cen al banks could lea n om he ape an um case s udy abou how o con ain ma ke ji e s du ing he p ocess o policy no maliza ion. The Fed i sel has lea ned om he communica ion mis akes in 2013. The li -o o he Fede al Funds a e a ge by 0.25 pe cen age poin s in Decembe 2015, al hough delayed se e al imes like in 2013, has been e y well an icipa ed by inancial ma ke s. This was he consequence o an imp o ed communica ion s a egy o he Fede al Rese e. Ma ke s eac ed in ad ance o he ac ual in e es a e ise, such ha he e en ual decision was ecei ed wi hou an inc ease in ola ili y. The u u e pa h o in e es a e no maliza ion will again equi e clea and p edic able policy communica ion. Whe he he Fed will manage subsequen in e es a e changes as well as he ecen one emains o be seen. Some obse e s a e skep ical: Bloombe g (20 No embe 2015) i les ha [Fede al Rese e Bank o S . Louis P esiden James] “Bulla d Says In es o s Should P epa e o Unce ain y E a” and o esees ha each FOMC mee ing in 2016 will be “su ounded by ele a ed deg ees o unce ain y.” The ele a ed ola ili y a he beginning o 2016, i.e., a e he i s Fed in e es a e hike, lends impo ance o a c edibly policy s a egy. Some a gue ha because o he apidly changing global en i onmen he i s in e es a e hike was a mis ake and should be e e sed. The Fed co ec ly esponded o his by s essing how ca e ul i will be as ega ds he iming o he nex a e inc ease. 13 ADBI Wo king Pape 592 P. Tillman The second implica ion add esses mac oeconomic policymaking in eme ging economies. I is widely accep ed ha domes ic mac oeconomic weaknesses make eme ging ma ke economies mo e p one o policy spillo e s om ad anced economies. Recen wo k by Ahmed, Coulibaly, and Zla e (2015), Geo giadis ( o hcoming), and o he s shows ha he exposu e o spillo e s depends on domes ic mac oeconomic undamen als, wi h he s eng h be ween undamen als and spillo e s a ying s ongly ac oss economies and o e ime. Fu he mo e, he spillo e s o Fed ape ing ecei ed special a en ion. Eicheng een and Gup a (2013) and Aizenman, Binici, and Hu chison (2014) ound ha s onge undamen als do no shield eme ging economies om spillo e e ec s, while he pape by Ahmed, Coulibaly, and Zla e (2015) a gued ha economies wi h s onge undamen als a e less a ec ed by ape ing shocks. This line o esea ch ypically es ima es whe he a policy change in he US leads o asse p ice changes in eme ging economies and whe he his esponse is a ying ac oss economies oge he wi h mac oeconomic undamen als. Impo an ly, hese pape s ocus on spillo e s o igina ing om igh ening o easing s eps, espec i ely, o he Fede al Rese e—no om he unce ain y su ounding he u u e policy pa h. A small amoun o li e a u e s udies he ole o undamen als o he suscep ibili y o eme ging economies o mone a y policy unce ain y. Gau in, McLoughlin, and Reinha d (2014) p o ided e idence o gene al policy unce ain y in ma u e economies as a de e minan o in e na ional capi al lows o eme ging economies. They used he policy unce ain y indica o s o Bake , Bloom, and Da is (2013), bu no a ocused measu e o speci ic mone a y policy unce ain y such as he one used in his pape . They also showed ha undamen als play a ole in de e mining he magni ude o spillo e e ec s. Addi ional e idence is p o ided by Ca iè e-Swallow and Céspedes (2013). These au ho s showed ha global unce ain y shocks, again measu ed by iden i ying episodes o excep ionally high ealiza ions o he VIX index in he US economy, d i e capi al lows. Mos impo an ly, domes ic mac oeconomic ulne abili y inc eases he e ec s o unce ain y shocks on eme ging economies. Based on he esul s p esen ed in his pape we can see ha he exchange a e esponse is s onge in B azil, India, Indonesia, and Tu key compa ed wi h he Republic o Ko ea and Thailand. While he o me coun ies belong o wha has been called he “ agile i e” economies wi h pa icula ly weak undamen als, he Republic o Ko ea and Thailand had much s onge undamen als. Hence, we ind p ima acie e idence ha he sensi i i y o unce ain y shocks does indeed a y acco ding o domes ic undamen als, al hough a mo e sys ema ic analysis is needed he e o unde line his inding. Taken oge he , s eng hening he esilience o eme ging economies should educe he sensi i i y o policy spillo e s and also unce ain y spillo e s. A policy ha ackles imbalances in go e nmen budge s and cu en accoun balances, a oids excessi e indeb edness, and s eng hens domes ic inancial ins i u ions is bes sui ed o cope wi h spillo e s o any kind. In addi ion, mac op uden ial policy ins umen s a e o en seen as a p omising oolki o a oiding inancial imbalances in he ligh o ola ile capi al in lows. 14 ADBI Wo king Pape 592 P. Tillman 6. CONCLUSIONS This pape p o ided e idence o an unce ain y channel o in e na ional mone a y policy ansmission. We showed ha asse p ices in eme ging ma ke economies esponded o heigh ened unce ain y abou u u e US mone a y policy. The key con ibu ion o his pape is he use o social media da a om Twi e .com, om which an indica o o unce ain y is de i ed. An ad an age o his da a se is ha we de i e unce ain y di ec ly om ma ke pa icipan s’ Twi e commen s and ha unce ain y ocuses on a pa icula policy decision. Based on his da a and an es ima ed VAR model we ound ha unce ain y a ec s eme ging ma ke s abo e and beyond he ac ual igh ening and easing policy s eps. Hence, mi iga ing unce ain y i sel is bene icial om an eme ging ma ke ’s poin o iew. Ou esul s no only ha e implica ions o economic policymaking, as elabo a ed in Sec ion 5, bu also o u u e esea ch on mone a y policy in small open economies. Da a om social media sou ces has no ye been widely used o unde s anding mone a y policy. In many ields o mone a y policy, expec a ions o ma ke pa icipan s ma e —whe he i is abou u u e policy o u u e in la ion. In hese cases, social media migh be pa icula ly use ul as i allows esea che s o measu e ma ke iews di ec ly om he ele an people and on a e y high equency. E en a e he Fed has exi ed uncon en ional mone a y policies, many cen al banks such as he Eu opean Cen al Bank and he Bank o Japan a e cu en ly ex ending hei uncon en ional se o policies. F om he pe spec i e o eme ging ma ke economies, in pa icula hose in Asia, he ime o s ong policy spillo e s is a om o e . This pape p o ided some insigh s in o he likely consequences o u u e policy unce ain y. 15 ADBI Wo king Pape 592 P. Tillman REFERENCES Ahmed, S., B. Coulibaly, and A. Zla e. 2015. In e na ional Financial Spillo e s o Eme ging Ma ke Economies: How Impo an a e Economic Fundamen als? In e na ional Finance Discussion Pape 1135. Washing on, DC: Boa d o Go e no s o he Fede al Rese e Sys em. Aizenman, J., M. Binici, and M. H. Hu chison. 2014. The T ansmission o Fede al Rese e Tape ing News o Eme ging Financial Ma ke s. NBER Wo king Pape 19980. Camb idge, MA: Na ional Bu eau o Economic Resea ch. A uoba, S. B., F. X. Diebold, and C. Sco i. 2009. Real- ime Measu emen o Business Condi ions. Jou nal o Business and Economic S a is ics 27(4): 417–427. Bachmann, R., S. Els ne , and E. Sims. 2013. Unce ain y and Economic Ac i i y: E idence om Business Su ey Da a. Ame ican Economic Jou nal: Mac oeconomics 5(2): 217–249. Bake , S., R. N. Bloom, and S. J. Da is. 2013. Measu ing Economic Policy Unce ain y. Unpublished. S an o d Uni e si y. Caggione, G., E. Cas elnuo o, and N. G oshenny. 2015. Unce ain y Shocks and Unemploymen Dynamics in US Recessions. Unpublished. Uni e si y o Melbou ne. Calda a, D., C. Fuen es-Albe o, S. Gilch is , and E. Zak ajsek. 2014. The Mac oeconomic Impac o Financial and Unce ain y Shocks. Unpublished. Boa d o Go e no s o he Fede al Rese e Sys em. Ca iè e–Swallow, Y., and L. F. Céspedes. 2013. The Impac o Unce ain y Shocks in Eme ging Economies. Jou nal o In e na ional Economics 90(2): 316–325. Eicheng een, B., and P. Gup a. 2013. Tape ing Talk: The Impac o Expec a ions o Reduced Fede al Rese e Secu i y Pu chases on Eme ging Ma ke s. Unpublished. Uni e si y o Be keley. F y, R., and A. Pagan. 2011. Sign Res ic ions in S uc u al Vec o Au o eg essions. Jou nal o Economic Li e a u e 49(4): 938–960. Gau in, L., C. McLoughlin, and D. Reinha d . 2014. Policy Unce ain y Spillo e s o Eme ging Ma ke s: E idence om Capi al Flows. Wo king Pape 512. London: Bank o England. Geo giadis, G. Fo hcoming. De e minan s o Global Spillo e s om US Mone a y Policy. Jou nal o In e na ional Money and Finance. Is e i, K., and A. Piloiu. 2013. Economic Policy Unce ain y and In la ion Expec a ions. Unpublished. Goe he Uni e si y F ank u . Lim, J. J., S. Mohapa a, and M. S ocke . 2014. Tinke , Tape , QE, Bye? The E ec o Quan i a i e Easing on Financial Flows o De eloping Coun ies. Policy Resea ch Wo king Pape No. 6820, Washing on, DC: Wo ld Bank. Lough an, T., and B. McDonald. 2011. When is a Liabili y no a Liabili y? Tex ual Analysis, Dic iona ies, and 10-Ks. Jou nal o Finance 66(1): 35–65. Meinusch, A., and P. Tillmann. 2015. Uncon en ional Mone a y Policy and Tape ing Unce ain y: E idence om Twi e . Unpublished. Uni e si y o Giessen. 16 ADBI Wo king Pape 592 P. Tillman Mish a, P., K. Mo iyama, P. N’Diaye, and L. Nguyen. 2014. Impac o Fed Tape ing Announcemen s on Eme ging Ma ke s. IMF Wo king Pape 14/109. Washing on, DC: In e na ional Mone a y Fund. Rai, V., and L. Suchanek. 2014. The E ec o he Fede al Rese e’s Tape ing Announcemen on Eme ging Ma ke s. Wo king Pape 2014–50. O awa: Bank o Canada. Sahay, R., V. A o a, T. A ani is, H. Fa uquee, P. N’Diaye, and T. Mancini-G i oli. 2014. Eme ging Ma ke Vola ili y: Lessons om he Tape Tan um. IMF S a Discussion No e SDN/14/09. Washing on, DC: In e na ional Mone a y Fund. Tillmann, P. 2015. Tape ing Talk on Twi e and he T ansmission o Eme ging Economies. Unpublished. Uni e si y o Giessen. Uhlig, H., 2005. Wha a e he E ec s o Mone a y Policy on Ou pu ? Resul s om an Agnos ic Iden i ica ion P ocedu e. Jou nal o Mone a y Economics 52: 381–419. 17 ADBI Wo king Pape 592 P. Tillman APPENDIX Figu e A1: Response o EMBI o Unce ain y Shock EMBI = eme ging ma ke bond p ice index; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . Figu e A2: Response o MSCI o Unce ain y Shock MSCI = eme ging ma ke s ock p ice index; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 18 ADBI Wo king Pape 592 P. Tillman Figu e A3: Response o FX Ra e o B azil o Unce ain y Shock FX = o eign exchange; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . Figu e A4: Response o FX Ra e o Tu key o Unce ain y Shock FX = o eign exchange; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 19 ADBI Wo king Pape 592 P. Tillman Figu e A5: Response o FX Ra e o India o Unce ain y Shock FX = o eign exchange; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . Figu e A6: Response o FX Ra e o Indonesia o Unce ain y Shock FX = o eign exchange; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 20 ADBI Wo king Pape 592 P. Tillman Figu e A7: Response o FX Ra e o he Republic o Ko ea o Unce ain y Shock FX = o eign exchange; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . Figu e A8: Response o FX Ra e o Thailand o Unce ain y Shock FX = o eign exchange; VIS = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 21 ADBI Wo king Pape 592 P. Tillman Figu e A9: Response o CDS Sp eads o Indonesia o Unce ain y Shock CDS = c edi de aul swap; VIS = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . Figu e A10: Response o CDS Sp eads o Tu key o Unce ain y Shock CDS = c edi de aul swap; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 22 ADBI Wo king Pape 592 P. Tillman Figu e A23: Response o EMBI o Unce ain y Shock wi h FOMC Dummies wi h Mac oeconomic Fundamen als EMBI = eme ging ma ke bond p ice index; FOMC = Fede al Open Ma ke Commi ee; VIX = ola ili y index. No es: The shaded a ea e lec s 68% con idence bands. The ed line is he median o all d aws; he do ed line is he d aw ha is closes o he median. Sou ce: P epa ed by au ho . 29