Effects of Monetary Policy on Household Expectations: The Role of Homeownership

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Effects of Monetary Policy on Household Expectations: The Role of Homeownership Hie Joo Ahn, Shihan Xie, and Choongryul Yang 2022-065 Please cite this paper as: Ahn, Hie Joo, Shihan Xie, and Choongryul Yang (2022). “Effects of Monetary Policy on Household Expectations: The Role of Homeownership,” Finance and Economics Discussion Series 2022-065. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2022.065. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Effects of Monetary Policy on Household Expectations: The Role of Homeownership* HieJooAhn† ShihanXie‡ ChoongryulYang§ FederalReserveBoard UniversityofIllinois FederalReserveBoard FirstDraft: April,2022 ThisDraft: September,2022 Abstract Westudytheroleofhomeownershipintheeffectivenessofmonetarypolicyonhouseholds’ expectations.Empirically,wefindthathomeownersrevisedowntheirnear-terminflation expectationsandtheiroptimismaboutfuturelabormarketconditionsinresponsetoarise inmortgagerates,whilerentersarelesslikelytodoso.Wefurthershowthatthemonetarypolicycomponentofmortgage-ratechangescreatesthedifferenceinexpectationrevisions betweenhomeownersandrenters. Thisresultsuggeststhathomeownersareattentiveto newsoninterestratesandadjusttheirexpectationsaccordinglyinamannerconsistent withtheintendedeffectofmonetarypolicy.Wecharacterizethesefindingsusingarational inattentionmodelwithtwotypesofhouseholds—homeownersandrenters. JELclassification:D83,D84,E31,E52 Keywords:Inflationexpectations,homeownership,rationalinattention *We would like to thank Dan Vine, Wenting Song, Rupal Kamdar, Andrew Figura and numerous seminar participants at the Federal Reserve Board, Applied Time Series Econometrics Workshop at St. Louis Fed, 4th BehavioralMacroeconomicsWorkshop,2022KER-KAEAInternationalConference,SEM2022andSETA2022for helpfulcommentsandsuggestions. Theviewsexpressedhereinarethoseoftheauthors,anddonotreflectthe viewsoftheFederalReserveBoardoranypersonassociatedwiththeFederalReserveSystem. †FederalReserveBoardofGovernors,20thStreetandConstitutionAvenueNW,Washington,DC20551,U.S.A. Email:econ.hjahn@gmail.com ‡DepartmentofEconomics,UniversityofIllinoisatUrbana-Champaign,1407WGregoryDr,MC-707,Urbana, IL61801,U.S.A.Email:shihanx@illinois.edu §FederalReserveBoardofGovernors,20thStreetandConstitutionAvenueNW,Washington,DC20551,U.S.A. Email:choongryul.yang@frb.gov 1

1 Introduction Thesuccessofmonetarypolicyreliesonhoweffectivelythecentralbank’scommunicationand policyimplementationaffectstheexpectationsofeconomicagents(Blinder,2009). Forward guidance policy, for example, is designed to work through economic agents’ expectations at thezerolowerboundwhenstandardpolicyinstrumentsareconstrained. However,recentempiricalstudiesfindthattheFederalReserve’scommunicationaboutmonetarypolicyhaslittle effectontheinflationexpectationsofhouseholds(e.g.,LamlaandVinogradov2019;Coibion, GorodnichenkoandWeber2022;D’Acunto,HoangandWeber2022). Moreover,householdsin low-inflationcountriesevenreportthattheyarelargelyunawareofmonetarypolicyannouncementsandtheroleofthecentralbank(Coibion,GorodnichenkoandKumar2018b;D’Acunto, Hoang,PaloviitaandWeber2021). Thesefindingsonthelackofhouseholds’attentiontomonetarypolicyraiseimportantquestionsforpolicymakerswhoseektounderstandhoweconomic agents’expectationsrespondtopolicycommunicationandactions: Doesmonetarypolicyaffect economicagents’expectationsoffutureeconomicconditions? Ifso, whataretheimportant channelsformonetarypolicytoinfluenceexpectationformation? Whoseexpectationsaremore responsivetochangesineconomicconditionsandmonetarypolicy? Thispaperconcernstheeffectofmonetarypolicyonhouseholds’expectationsandtherole ofhousehold heterogeneity initseffectiveness.1 We pay attentionto homeownershipas the keyheterogeneity,becausehomeownersandrentersmayhavedifferentincentivestoacquire informationonchangesininterestrates. Homeownersmaymakeregularmortgagepayments orconsiderrefinancingtheirmortgages,andhencechangesininterestrateslikelyhavedirect effectsontheirhouseholdfinances. Inthisregard,homeownershavemoreincentivestopay closeattentiontonewsoninterestratesandtheeconomythanrentersdo. Relatedly,Grundl and Kim (2019) find that households who receive newsletters on home loans and mortgage refinancing make more informed decisions on their household finances than those who do not. Thisevidencesuggeststhathomeownershavemoreaccesstoinformationoninterestrates thanrentersdo. Alltold,whenitcomestoeconomicexpectations,homeownersmaybebetter informedofnewsoninterest-ratechangesandmonetarypolicythanrenters. Weempiricallyinvestigatehowresponsivetheexpectationsofhomeownersandrentersare 1Therecentliteraturehighlightstheimportanceofhousehold-andfirm-levelheterogeneityinexpectations formationandtheimplicationofthisheterogeneityonmonetarypolicy.SeeWeber,D’Acunto,Gorodnichenkoand Coibion(2022)forarecentreviewaboutinflationexpectationsformationofhouseholdsandfirms. 2

to interest-rate changes and monetary policy shocks with microdata from the University of MichiganSurveysofConsumers(MSC)andtheSurveyofConsumerExpectations(SCE).Wefind thathomeownerslowertheirinflationexpectationsone-tothree-yearsaheadinresponsetoa risein30-yearmortgagerates,whilerentersarelesslikelytodoso. Thisrelationship,however, isnotobservedinthelonger-runinflationexpectations. Additionally,inresponsetoarisein mortgagerates,homeownersreducetheiroptimismaboutlabormarketconditionsmore,and are more likely to expect interest-rate rises in the future than renters. We further show that monetarypolicyshockspassthroughto30-yearmortgagerateswithstatisticalsignificance,and homeownersrespondparticularlytotheportionofmortgageratechangesdrivenbymonetary policyshocks.2 Wealsoconsiderthe30-yearT-bondrateandfederalfundsrate,butthesensitivity ofexpectationstomortgage-ratechangesexhibitsthelargestdifferencebetweenhomeowners andrenters. Alltheempiricalevidenceindicatesthathomeownersareattentivetotheevolution ofmortgageratesandadjusttheireconomicoutlookinawaythatisconsistentwiththeintended consequencesofmonetarypolicy. Whatexplainstheheterogeneousresponseofinflationexpectationstointerestratechanges byhomeownershipstatus? Onepotentialexplanationistherefinancingmotive. Homeowners holdingmortgageslikelyhaveastrongincentivetopayattentiontomortgageratechangesto seekanopportunitytorefinancetheirmortgagewithalowerrate. Consideringthatmorethan 60% of U.S. homeowners carry a mortgage, the refinancing motive may be the key driver of thestrongresponsivenessofhomeowners’expectationstotheratechanges.3 Consistentwith this conjecture, we find that the sensitivity of homeowners’ revisions in short-term inflation expectationsislargertoadecreaseinmortgageratethantoanincreaseintherate. We offer corroborating evidence for our main claim based upon the American Time Use Survey.Timespentonfinance-relatedactivitiesisanaturalmeasureofhouseholds’attentiveness tofinancialmarketsandmacroeconomicdevelopments. Wefindthathomeownersaremore likelytoengageinfinance-relatedactivitiessuchascheckingfinancialmarketsandresearching investmentsthatlikelyexposethemtocurrentmacroeconomicconditionsandinterestrates, 2Noticethatourevidencesuggeststhatconditionalonademandshocklikemonetarypolicyshock,inflation expectationsandthelabormarketoutlookarepositivelycorrelated. Thisfinding,however,doesnotcontradict theobservationinKamdar(2019)thatunconditionalexpectationsofinflationandlabormarketsarenegatively correlated,asiftheassociationsreflectedtheconsequenceofasupplyshock. 3Recentstudieshighlighttherefinancingmotiveasanimportanttransmissionchannelofmonetarypolicy(e.g., Amromin,BhuttaandKeys,2020;Berger,Milbradt,TourreandVavra,2021;Wong,2021;Eichenbaum,Rebeloand Wong,2022). 3

andalsospendmoreontheseactivitiesthanrenters. Ourfindingsseemtobeinconsistentwiththerecentevidencebasedonsurveysandexperiments that points to little effect of monetary policy on economic agents’ expectations (e.g., Coibion,GorodnichenkoandKamdar2018a;LamlaandVinogradov2019;Coibionetal.2022; D’Acuntoetal.2022,2021). Thesestudiesshowthateconomicagents—householdsinparticular—donothaveagoodunderstandingofmonetarypolicyorthecentralbank’scommunication about the future policy path. Nonetheless, these findings do not necessarily contradict our empiricalresult. Thoughhouseholdsmaynotknowconceptslike“FederalReserve,”“monetary policy,”and“inflationtarget,”theymayhaveasolidunderstandingoftheeffectofinterest-rate changesontheirhouseholdfinancesandtheoveralleconomy. Householdsmayhavelearned aboutitfromtheirownexperiencesorconversationswithpeoplethattheyinteractwithsuchas loanofficers. Inotherwords,evenifhouseholdsrespondtoasurveythattheydonotknowabout monetarypolicy,thisresponsedoesnotnecessarilymeanthathouseholdsdonotpayattention toachangeininterestratesanddonotunderstandtheconsequencestofuturemacroeconomic conditions. Wecharacterizeourmainempiricalfindingswithasimpletwo-periodrationalinattention modelwithaparticularfocusontheassociationbetweeninflationexpectationsandinterestratechanges. Facinguncertaintyaboutfutureeconomicfundamentals,includinginflationand interestrates,householdsacquirecostlyinformationtooptimizetheirconsumptionandhousing choices. Householdsarecomposedofhomeownersandrenters. Ahomeowner,afterpurchasing ahousewithahomeloan,paysmortgageinteresteachperiod,whilearenterpaysafixedrent andtradesaone-periodnominalbond. Sincehomeownersvaluehousingservicesintheirutility, changes in goods’ prices affect the relative price of housing, which has a substitution effect ontheirconsumption. Thisadditionalsubstitutioneffectisabsentfromtherenters’problem. Therefore,homeowners’beliefsaboutinflationrespondmorestronglytointerestratevariations thanrenters’beliefs. Alltold, homeownershipcreatesdifferencesintheincentivetoacquire information,andhenceresultsinheterogeneityinthesensitivityofinflationexpectationsto interestratechangesbetweenhomeownersandrenters. This paper contributes to multiple strands of research. First, there is growing literature regardingtheeffectivenessofmonetarypolicyoneconomicagents’expectations(e.g.,Cavallo, CrucesandPerez-Truglia2017;Coibionetal.2022;D’Acuntoetal.2022;Coibion,Gorodnichenko 4

and Weber 2021).4 The majority of recent research finds little evidence for the effectiveness of the Fed’s communication or monetary policy in guiding economic agents’ expectations, thoughsomestudiesincludingHoffmann,Moench,PavlovaandSchultefrankenfeld(2021)and KryvtsovandPetersen(2021)reachadifferentconclusion.Thispaperalsospeakstotheliterature on determinants of expectation formation (e.g., Carroll 2003; Coibion and Gorodnichenko 2015;Bauer2015;Stillwagon2018;KilianandZhou2021). Recentstudiesfocusontheroleof economicdevelopmentsorindividualattributesintheexpectationofeconomicagents(e.g., Das,KuhnenandNagel2020;KuchlerandZafar2019;Armona,FusterandZafar2019;D’Acunto, Hoang,PaloviitaandWeber2019;D’Acunto,Malmendier,OspinaandWeber2021;D’Acunto, MalmendierandWeber2022). Ourpaperbridgesthesestrandsofliteraturebysheddinglighton theimportanceofhomeownershipintheeffectivenessofthemonetarypolicyonhouseholds’ expectation formation. In this regard, this paper is closest to Claus and Nguyen (2020), but isdifferentfromitintworespects.5 First,wefocusonhouseholdsintheUnitedStates,while Claus-Nguyen’sanalysisconcernsAustralianhouseholds. Second, ClausandNguyendonot considerhowhomeownershipdeterminesthesensitivityofinflationexpectationstomonetary policyshocks,whichisthemainfocusofourpaper. Therestofthispaperisorganizedasfollows. Section2introducesthedataadoptedforour empirical study. Section 3 presents the empirical analyses. Section 4 explores the potential mechanismbehindourmainfindingsandcheckstherobustnessofresults. Section5discusses amodelofrationalinattentivehouseholdsthatisdesignedtoexplainourempiricalfindings. Section6concludes. 2 Data Thissectiondescribesmeasuresofhouseholdexpectationsandmonetarypolicyshocksusedin thispaper. Asnotedearlier,ourdataonhouseholdexpectationscomefromtheMSCandSCE. Bothdatasetshavebeenpopularlyusedintheliteraturetostudyhouseholdexpectations. For monetarypolicyshocks,weadoptthemeasurefromBu,RogersandWu(2021).6 Thesample 4Forinstance,D’Acuntoetal.(2022)findthathouseholdexpectationsareaffectedbyunconventionalfiscalpolicy, notforwardguidance,andarguethatpoliciesaimedathouseholdsdirectlyarenoteffectivebecausehouseholdsdo notunderstandthem. 5Adelino,SchoarandSeverino(2018)investigatetheeffectofexpectationsabouthousingpricesonhomeownership,whichisdifferentfromourfocus—theeffectofhomeownershiponexpectations. 6WealsoconsidermonetarypolicyshocksinSwanson(2021)forrobustnesschecksintheappendix. 5

periodoftheempiricalanalysesrangesfrom1990:M1through2020:M12. 2.1 Measuresofhouseholdexpectations TheMSCquestionnairesaredesignedtotrackconsumerattitudesandexpectations. Thesurvey has been conducted by telephone monthly since 1978 and constitutes a sample of over 500 householdsrepresentativeoftheU.S.population. Itcontainsdemographicinformationsuchas respondents’educationlevel,age,andhouseholdincome. In1990,theMSCstartedcollecting informationaboutrespondents’homeownership,homevalue,andhomepriceexpectations. TheMSChasalongtimeseriesbutdoesnottrackindividualhouseholdsovertime. About40% ofthehouseholdswhowereinterviewedsixmonthsagoarere-contacted. Inourstudy,wefocus onthepost-1990sampletoexploittheinformationonhomeownershipandtherepeatedsample featureofthesurvey. TheSCEisamonthlysurveyconductedbytheFederalReserveBankofNewYorkthatfocuses onexpectationsaboutinflation,labormarketconditions,andhouseholdfinance. Withdata startinginJune2013,thisinternet-basedsurveyconstitutesasampleofabout1,300nationally representativehouseholds. Surveyrespondentsparticipateconsecutivelyforupto12months. Thesurveymeasureshouseholdexpectationsusingtwodifferentapproaches. First,respondents reporttheirexpectedgrowthrateofthevariableofinterestoveraspecifiedtimehorizon. Second, forcertainvariablessuchasinflation,incomegrowth,andunemployment,respondentsreport the probability they assign to a percentage change in the respective variable. Based on this information,ageneralizedbetadistributionisestimatedtocharacterizethedistributionthat eachsurveyparticipantperceivesaboutchangesineachvariable. Inourstudy,weusethemean ofthisdistributionasameasureofhouseholdexpectations. Inthissurvey, homeownersare thosewhoindicatethattheyowntheirprimaryresidenceandrentersarethosewhodonot. Only asmallfractionofrenters,about1.2%ofallrespondents,indicatethatownotherhomes.7 2.2 Measuresofmonetarypolicyshocks Weselectmeasuresofmonetarypolicyshocksthatareavailableforthesampleperiodincluding thepost-zerolowerbound(ZLB)period. DuringtheZLBperiod,monetarypolicyhasbecome 7AppendixTableB.1showsthesummarystatisticsforthehomeownershiprateandtheaverageinflationexpectationsbyhomeownershipstatus. 6

moremultidimensionalwiththeadoptionofunconventionalmonetarypolicytools. Acommon approachintheliteratureisthehigh-frequencyidentificationmethodthatfocusesonmovements in asset prices in a narrow window around Federal Open Market Committee (FOMC) meetings. TheseFOMCmeetingannouncementsusuallycontainbothcentralbankinformation effectandmonetarypolicyshocks. Forthisreason,weadopttheunifiedmeasureofmonetary policyshocksproposedbyBuetal.(2021)thatcontainsnosignificantinformationeffect(BRW shockhenceforth)8. 3 Empirical investigation Section3.1analyzestheeffectofinterestratechangesoninflationexpectationsofhomeowners andrenters. Section3.2conductssimilaranalysesforhouseholds’expectationsoflabormarket conditions. Section3.3examinestheresponsivenessofinterest-rateexpectationstochanges ininterestrates. Section3.4explorestheextenttowhichmonetarypolicyshocksaccountfor households’expectationsviatheinterest-ratechannel. 3.1 Theeffectofinterest-ratechangesonhouseholds’inflationexpectations Thissectioninvestigateshowmuchhomeownersandrentersrevisetheirinflationexpectations inresponsetointerestratechanges. Forthisempiricalanalysis,weemploythefollowingmodel specification: E h−yr −E h−yr =α+β homeowner ×∆R +β renter ×∆R +δX +ϵ , (3.1) i,t+6 i,t 1 i t 2 i t i,t i,t whereE h−yr isrespondenti’sh-year-aheadinflationexpectationforh=1,5attimet fromthe i,t MSC;homeowner andrenter aredummiesforhomeownerandrenter,respectively;∆R isa i i t changeininterestratesduringthepastsixmonths,and X arecontrolsfortherespondent’s i,t characteristicswhichincludegender,education,birthcohort,andincome. For∆R ,weconsider t changesinthe30-yearfixedmortgagerate,thefederalfundsrate,andthe30-yearT-bondrate. ThisspecificationisbasedonthemodelbyCoibionandGorodnichenko(2015)thatanalyzes theeffectofoilpricechangesoninflationexpectations. Thereareafewdifferencesbetweenour 8BauerandSwanson(2022)alsodistinguishmonetarypolicysurprisesfromFedinformationeffectbyorthogonalizingthesurpriseswithrespecttomacroeconomicandfinancialdatathatpre-datetheannouncement. 7

Table1: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO CHANGES IN INTEREST RATES Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.MichiganSurveyofConsumers(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.246 ∗∗∗ 0.118 ∗∗ -0.285 ∗∗∗ t 1 (0.062) (0.046) (0.061) Renter×∆R (β ) -0.151 0.00961 -0.289 ∗∗ t 2 (0.114) (0.087) (0.115) Numberofobs. 50,834 50,834 50,834 R2 0.0140 0.0138 0.0143 PanelB.MichiganSurveyofConsumers(5-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.0176 0.0468 -0.0585 t 1 (0.051) (0.034) (0.050) Renter×∆R (β ) 0.00269 -0.0837 -0.0474 t 2 (0.107) (0.077) (0.107) Numberofobs. 48,842 48,842 48,842 R2 0.0107 0.0107 0.0107 Notes:ThistablereportstheregressionresultsfromEquation(3.1).Dependentvariablesarethesix-monthchangeinthe MSC’s12-monthaheadinflationexpectations(PanelA)andthesix-monthchangeintheMSC’s5-yearaheadinflation expectations(PanelB)."Homeowner"and"Renter"indicatedummiesforhomeownerandrenterrespectively.∆Rtrefersto thesix-monthchangeininterestrate.Weusethe30-yearmortgagerateinColumn(1),theFederalFundsRateinColumn (2),andthe30-yearTreasuryBondrateinColumn(3). Wecontrolfortheobservedsurveyrespondents’characteristics, includinggender,education,birthcohort,andthelevelofincome.Robuststandarderrorsarereportedintheparenthesis. ∗∗∗ ∗∗ ∗ , , denotesstatisticalsignificanceat1%,5%,and10%levelsrespectively. Sources:Authors’calculation. modelandCoibionandGorodnichenko’s. First,ourmodelhastermsthatcapturethedifferent sensitivitiesofhomeownersandrenterstoachangeininterestratesasareplacementofthe argument capturing the effect of oil-price changes in Coibion and Gorodnichenko’s model. Second,weuseapastchangeininterestratestoreflectthedelayedeffectofmonetarypolicy due,forinstance,toinformationrigidity,whileCoibionandGorodnichenko(2015)considera changeinoilpricesinthecurrentperiod. Third,weexplicitlycontrolforadditionalobservable individualcharacteristics. Panel A of Table 1 reports the estimation results for inflation expectations in the next 12 monthsfromtheMSC.Asshowninthefirstcolumn,thecoefficientonhomeowner isnegative i and statistically significant, while that on renter is not statistically significant. This result i suggeststhathomeownerstakesignalsfromchangesinmortgagerateswhenprojectinginflation ayearahead,whilerentersarelesslikelytodoso. Homeownerslikelymakeregularmortgage paymentsandconsiderrefinancingtheirhomeloans. Therefore,homeownersmaypaycloser 8

attention to the evolution of mortgage rates than renters do, because a change in mortgage rates likely has a direct effect on their household finances. Meanwhile, both homeowners’ and renters’ coefficients on changes in the 30-year T-bond rate are negative and statistically significant. Consideringloaninterestratesarecloselyassociatedwiththe30-yearT-bondrate, thisresultmayreflectthathomeownersandrentershaveloanstopayoffandthuspayattention tonewsontheinterestrate. Thisobservationindicatesthathouseholdsdoadjusttheirinflation expectationstointerestratechangestowhichtheypayattention.9 Unliketheestimationresultsfromone-year-aheadinflationexpectations,households’fiveyear-aheadinflationexpectationsdonotseemtorespondtointerestratechanges,regardlessof homeownershipstatus. AsshownbyPanelBofTable1,thecoefficientsoninterestratechanges areclosetozeroandnotstatisticallysignificant. Overall,householdsarelesslikelytochange theirlong-runinflationexpectationsinresponsetoachangeininterestrates. WefurtherexaminetherobustnessofempiricalresultsbasedontheSCE.Themodelspecificationisasfollows: E h−yr −E h−yr =α +β homeowner ×∆R +β renter ×∆R +ϵ , (3.2) i,t+6 i,t i 1 i t 2 i t i,t whereE h−yr isrespondenti’sh-yearaheadinflationexpectationforh =1,3attime t andα i,t i capturesindividualfixed-effects. Ashouseholdsparticipateinthesurveyrepeatedlyforupto12 months,weemploytheindividualfixedeffectα toabsorbtheindividualheterogeneity. i TheestimationresultsaredocumentedinTable2. Homeownerslowertheirthree-year-ahead inflationexpectationstoarisein30-yearmortgageratewithstatisticalsignificance(PanelA).In addition,homeownersrevisedowntheirone-year-aheadinflationexpectationsinresponseto ariseinthe30-yearT-bondrateandthefederalfundsrate(PanelB).Noneofthecoefficients capturingrenters’responsivenessarestatisticallysignificant. Tosummarize,theevidencebased ontheSCEclearlyshowsthathomeownerspayattentiontointerest-ratechangesandrevise theirinflationexpectationsaccordingly,whilerentersarelesslikelytodoso. Alltold,homeownersareattentivetodevelopmentsinmortgageratesandT-bondrates,as if a rise in the interest rates is a contractionary macroeconomic factor. However, renters are 9Quitedifferently, wedonotobserveasimilarassociationbetweenachangeinthefederalfundsrateand households’one-year-aheadinflationexpectations.Rather,homeowners’coefficientispositiveandstatistically significant.Onepotentialexplanationforthedifferenceisthathomeownersmaytakeadifferentsignalfromthe federalfundsrate.Homeownersmayinferinformationaboutfutureinflationfromariseinthefederalfundsrate, whichisoftenreferredtoastheinformationeffect(NakamuraandSteinsson,2018). 9

Table2: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO CHANGES IN INTEREST RATES Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.NYFedSurveyofConsumerExpectations(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.212 -0.385 ∗ -0.279 ∗∗ t 1 (0.130) (0.226) (0.124) Renter×∆R (β ) -0.278 0.389 -0.408 t 2 (0.309) (0.431) (0.281) Numberofobs. 42,081 42,081 42,081 R2 0.0001 0.0002 0.0003 PanelB.NYFedSurveyofConsumerExpectations(3-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.257 ∗ -0.301 -0.197 t 1 (0.134) (0.214) (0.127) Renter×∆R (β ) -0.257 0.236 -0.283 t 2 (0.294) (0.469) (0.268) Numberofobs. 42,183 42,183 42,183 R2 0.0002 0.0001 0.0001 Notes: ThistablereportstheregressionresultsfromEquation(3.2). Dependentvariablesarethesix-monthchangein theSCE’s1-yearaheadinflationexpectations(PanelA)andthesix-monthchangeintheSCE’s3-yearaheadinflation expectations(PanelB)."Homeowner"and"Renter"indicatedummiesforhomeownerandrenterrespectively.∆Rtrefersto thesix-monthchangeininterestrate.Weusethe30-yearmortgagerateinColumn(1),theFederalFundsRateinColumn (2),andthe30-yearTreasuryBondrateinColumn(3). Weemploytheindividualfixedeffectstocontroltheindividual ∗∗∗ ∗∗ ∗ heterogeneity.Robuststandarderrors(clusteredattheindividual-level)arereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. Sources:Authors’calculation. lesslikelytodoso. Theempiricalfindingsimplythathouseholdspayattentiontointerestrate changesaboutwhichtheyhaveanincentivetoacquireinformation. 3.2 Howdoexpectationsonlabormarketconditionsreacttointerest-rate changes? Thissectioninvestigatestowhatextentinterestratechangesaffecthouseholds’expectationsof labormarketconditions. Ifaninterestrateincreasealsohasnegativeeffectsonhouseholds’job marketoutlook,wecaninterpretthattheinterestratechangeinfluenceshouseholds’expectationsinawaysimilartoacontractionarymonetarypolicyand,furthermore,mayactuallyreflect aconsequenceofmonetarypolicy. Themainchallenge,however,inthisanalysisisthatexpectationsoflabormarketconditions arecapturedbycategoricalresponses,unlikeinflationexpectations. IntheMSC,forexample, 10

thequestiononexpectationsofjoblessnessinthenext12monthsispostulatedasfollows: Howaboutpeopleoutofworkduringthecoming12months—doyouthinkthatthere willbemoreunemploymentthannow,aboutthesame,orless? 1. Moreunemployment 3. Aboutthesame 5. Lessunemployment Sincewearechieflyinterestedinchangesinexpectations,weconstructacategoricalvariable thatreflectsthedirectionofexpectationrevisions. Thisvariablehasthreeoutcomes—improved, unchanged,andworsened.Ifthenumericvalueoftheresponseintheoriginalquestionincreases, weregardtheexpectationtohave“improved.” Ifthenumericvaluedecreases,weinterpretthe expectationtohave“worsened.” Ifthenumericvaluestaysthesame,weassign“unchanged.” Withtheconstructedcategoricalvariablecapturinghouseholds’revisionofunemployment expectations,werunamultivariatelogit regressiontoexaminehowachangeintheinterestrate sixmonthsagoaffectstherevision. Themodelisspecifiedasfollows: (cid:181) (cid:182) p log ik,t =α +α renter +β homeowner ×∆R +β renter ×∆R +δX +ϵ , (3.3) 0 1 i 1 i t 2 i t i,t i,t p ij,t wherep istheprobabilitythathouseholdi’sresponseisk∈{“improved”,“worsened”}from ik,t periodt tot+6,andp istheprobabilitythathouseholdi’sresponseis j =“unchanged”from ij,t periodt tot+6.Theregressorshomeowner andrenter aredummiesforhomeownerandrenter, i i respectively;∆R isachangeintheinterestrateduringthepastsixmonths;X arecontrolsfor t i,t individualcharacteristics. Wetreattheresponse“unchanged”asthebasecategoryandestimate theprobabilityofhouseholdi torespond“improved”or“worsened”relativetothatofhousehold i torespond“unchanged.” ThecoefficientestimatesarereportedinTable3. Tomaketheresultsmoreinterpretable,wecomputethemarginalprobabilitiesofhouseholds to change their unemployment expectations, and display the probabilities in Figure 1. As depictedbythedownward-slopinglinesinthetop-leftpanel, householdsbecomelesslikely to expect that the labor market conditions will improve, when 30-year mortgage rates rise. Consistentwiththisobservation,householdsbecomemorelikelytoanticipatethatthelabor marketconditionswilldeterioratewithariseinthe30-yearmortgagerate,asindicatedbythe upward-slopinglines(toprightpanel).However,wefindstatisticallysignificantdifferencesinthe 11

Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate Figure1:MARGINALPROBABILITYOFCHANGESINHOUSEHOLDEXPECTATIONSOFUNEMPLOYMENT Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsofunemploymenttochangesin interestrates.Theinterestratesconsideredfromtoptobottomarethe30-yearmortgagerate,federalfundsrate,and30-year T-bondrate.TheresultsarecalculatedbasedontheestimatesofthelogitregressionresultsfromEquation(3.3)asreported inTable3.Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birthcohort,and thelevelofincome.Shadedareasrepresent95%confidenceintervals. Source:Authors’calculation. 12

Table3: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ UNEMPLOYMENT EXPECTA- TIONS TO CHANGES IN INTEREST RATES Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.Unemployment:Improve Renter(α ) 1.090 ∗∗∗ 1.086 ∗∗∗ 1.080 ∗∗∗ 1 (0.025) (0.025) (0.025) Homeowner×∆R (β ) 0.814 ∗∗∗ 0.874 ∗∗∗ 0.872 ∗∗∗ t 1 (0.017) (0.013) (0.019) Renter×∆R (β ) 0.863 ∗∗∗ 0.914 ∗∗∗ 0.877 ∗∗∗ t 2 (0.030) (0.022) (0.031) PanelB.Unemployment:Worsen Renter(α ) 1.043 ∗ 1.042 ∗ 1.043 ∗ 1 (0.025) (0.025) (0.025) Homeowner×∆R (β ) 1.032 1.119 ∗∗∗ 1.014 t 1 (0.022) (0.017) (0.022) Renter×∆R (β ) 1.028 1.159 ∗∗∗ 1.012 t 2 (0.036) (0.031) (0.037) Numberofobs. 59,917 59,917 59,917 R2 0.0022 0.0035 0.0016 Notes:ThistablereportsthelogitregressionresultsfromEquation(3.3).Dependentvariablesarethelogoftheprobability thatunemploymentratewillbeimproved(PanelA)orworsened(PanelB)inthenextsixmonthsrelativetotheprobability thatunemploymentratewillbeunchangedinthenextsixmonths. "Homeowner"and"Renter"indicatedummiesfor homeownerandrenterrespectively.∆Rtreferstothesix-monthchangeininterestrate.Weusethe30-yearmortgageratein Column(1),theFederalFundsRateinColumn(2),andthe30-yearTreasuryBondrateinColumn(3).Thecoefficientsare reportedintermsofrelativeriskratios.Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender, ∗∗∗ ∗∗ ∗ education,birthcohort,andthelevelofincome.Robuststandarderrorsarereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. Sources:Authors’calculation. optimisticrevisionsbetweenhomeownersandrenters(topleftpanel)butnotinthepessimistic revisions(toprightpanel). Homeownersrevisedowntheiroptimismmoreinresponsetoan increaseinthemortgageratethanrentersdo. Wefindsimilarresultswithchangesinthefederal fundsrateandthe30-yearT-bondrate(middleandbottompanels). Nonetheless,thesensitivity ofreducedoptimismbetweenhomeownersandrentersexhibitsthelargestdifference,when conditionedonachangein30-yearmortgagerates. Tosummarize,householdsbecomelessoptimisticaboutlabormarketoutlook,whenthereis ariseininterestrates.Thedecreasedoptimismisstrongeramonghomeownersthanrenters,and thedifferencebetweenthetwogroupsislargestwithanincreaseinmortgagerates. Combined withtheevidenceoninflationexpectations,theempiricalresultindicatesthatariseinmortgage ratesparticularlyinfluenceshomeowners’macroeconomicexpectationsinawaysimilartoa 13

monetarypolicyshock. 3.3 Howdoexpectationsonfutureinterestratesrespondtorecentchanges ininterestrates? Inthissection,wefurtherinvestigatethesensitivityofhouseholds’expectationsoffutureinterest ratestoapastinterest-ratechange. Weexaminethissensitivityasapotentialchannelthrough whichtheraterisehaspersistentcontractionaryeffectsonhouseholdexpectations. Responses tothequestiononinterest-rateexpectationarealsoacategoricalvariable. Therefore,weemploy specification(3.3)withoutconstructinganewcategoricalvariable.10 Thequestiononinterest rateexpectationsispostulatedasfollows: No one can say for sure, but what do you think will happen to interest rates for borrowingmoneyduringthenext12months—willtheygoup,staythesame,orgo down? We treat the response “stay the same” as the base category and estimate the probability of householdi responding“goup”or“godown”relativetothatofhouseholdi responding“staythe same”. Therefore,inthedependentvariable,p istheprobabilitythathouseholdi’sresponse ik,t isk=goup/godown,andp istheprobabilitythathouseholdi’sresponseis j =staythesame. ij,t Table4reportstheparameterestimates. Figure2displaysthemarginalprobabilityestimates. Whenthereisariseintheinterestrate,householdsbecomemorelikelytoexpectfutureinterest raterises,butbecomelesslikelytobelievethattheinterestratewouldeitherstaythesameor godown. Theupsiderevisionsarelargerthanthedownsiderevisionstoasameordeclining interestrate. Thisobservationimpliesapersistentcontractionaryeffectofinterestratehikeson householdexpectations. Notably, the responsiveness of homeowners is larger than that of renters with statistical significance. Similar to the earlier results, the difference in upward revision between homeownersandrentersisalsothelargestwhenconditionedonachangein30-yearmortgagerates. Theempiricalfindingsprovideapartialexplanationforcontractionaryeffectsofrisesininterestrates—mortgagerates,inparticular—onhouseholds’expectations,withalargereffecton homeownersthanonrenters. 10Noticethatweareinterestedinexpectationsonfutureinterestratenottherevisionsofexpectationsunlikethe previousanalyses.Forthisreason,weusethecategoricaldataastheyare. 14

Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate Figure2: MARGINAL PROBABILITY OF HOUSEHOLD EXPECTATIONS OF INTEREST RATES Notes:Thisfigurereportsthemarginalprobabilitiesofhouseholdexpectationsof1-yearaheadinterestratestopastchanges ininterestrates.Theinterestratesconsideredarethe30-yearfixedmortgagerates,federalfundsrates,and30-yearT-bond rates.TheresultsarecalculatedbasedontheestimatesofthelogitregressionresultsfromEquation(3.3)asreportedin Table4.Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birthcohort,andthe levelofincome.Shadedareasrepresent95%confidenceintervals.Theconfidencebandsaresonarrowthattheydonot clearlyshowthroughtothefigures. Source:Authors’calculation. 15

Table4: SENSITIVITY OF HOMEOWNERS AND RENTERS’ EXPECTATIONS ON FUTURE INTEREST RATES TO CHANGES IN INTEREST RATES Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.Interestrateincrease Renter(α ) 1.153 ∗∗∗ 1.177 ∗ 1.171 ∗∗∗ 1 (0.016) (0.016) (0.016) Homeowner×∆R (β ) 2.015 ∗∗∗ 1.757 ∗∗∗ 1.618 ∗∗∗ t 1 (0.025) (0.016) (0.019) Renter×∆R (β ) 1.481 ∗∗∗ 1.361 ∗∗∗ 1.348 ∗∗∗ t 2 (0.031) (0.021) (0.028) PanelB.Interestratedecrease Renter(α ) 1.398 ∗∗∗ 1.414 ∗∗∗ 1.355 ∗∗∗ 1 (0.030) (0.030) (0.028) Homeowner×∆R (β ) 0.590 ∗∗∗ 0.589 ∗∗∗ 0.676 ∗∗∗ t 1 (0.012) (0.007) (0.012) Renter×∆R (β ) 0.704 ∗∗∗ 0.638 ∗∗∗ 0.766 ∗∗∗ t 2 (0.022) (0.013) (0.024) Numberofobs. 168,201 168,201 168,201 R2 0.0243 0.0390 0.0014 Notes:ThistablereportsthelogitregressionresultsfromEquation(3.3).Dependentvariablesarethelogoftheprobability thatinterestratewillincrease(PanelA)ordecrease(PanelB)inthenextsixmonthsrelativetotheprobabilitythatinterest ratewillstaythesameinthenextsixmonths."Homeowner"and"Renter"indicatedummiesforhomeownerandrenter respectively.∆Rtreferstothesix-monthchangeininterestrate.Weusethe30-yearmortgagerateinColumn(1),theFederal FundsRateinColumn(2),andthe30-yearTreasuryBondrateinColumn(3).Thecoefficientsarereportedintermsofrelative riskratios.Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birthcohort,and ∗∗∗ ∗∗ ∗ thelevelofincome.Robuststandarderrorsarereportedintheparenthesis. , , denotesstatisticalsignificanceat1%, 5%,and10%levelsrespectively.ThistablereportstheestimatesofEquation(3.3)onoutcomesofunemployment. Sources:Authors’calculation. 3.4 Pass-throughofmonetarypolicyshockstohouseholds’expectations Theempiricalevidenceintheprevioussectionspointstothepossibilitythataninterestrate change influences households’ inflation expectations in a way similar to a monetary policy shock. Inthissection,weevaluatetheextenttowhichmonetarypolicyshocksaccountforthe empiricalevidence. Weparticularlyfocusonthe30-yearmortgagerateandthe30-yearT-bond rate,becausehouseholds’one-year-aheadinflationexpectationsfromtheMSChavestatistically significantnegativeassociationswiththeinterestrates. Notice that there are factors other than monetary policy that create changes in the 30yearT-bondrateandthe30-yearmortgagerate. Examplesincluderiskpremiums,thefuture economicoutlook,andhousingmarketconditions. However,thesefactorshaveoppositeeffects on households’ inflation expectations. To illustrate, a rise in the interest rate, if driven by 16

monetarytighteningorrisesindefaultrisks,lowersinflationexpectations. Quitedifferently,an increasedinterestratedrivenbyarosiereconomicoutlookraisesinflationexpectations.11 In thiscontext,thenegativeassociationsbetweeninterestratechangesandhouseholds’inflation expectationscapturethenetoutcomesofmultiplesources,withthecontractionary(former) effectsbeingstrongerthantheexpansionary(latter)effects. Therefore,itisex-anteunclearhow monetary policy shocks would influence households’ expectations through the interest-rate channels. Toinvestigatetowhatextentinformationonmonetarypolicycontainedintheinterestrate changesaffectshouseholds’inflationexpectations,wefirstanalyzethepass-throughofmonetary policy shocks to the 30-year mortgage rate and the 30-year T-bond rate.12 We then examine theresponsivenessofhouseholds’inflationexpectationstothemonetarypolicycomponentof interestratechanges. Considerthefollowinghigh-frequencyspecifications,withinterestratechangesataweekly frequency: 3 (cid:88) ∆R t+1 =α+β 1 UnifiedShock t + δ j ∆R t−j +ϵ t (3.4) j=1 wherethedependentvariable∆R isachangeintheinterestrateoverweekt,andUnifiedShock t referstothemonetarypolicyshockfromBuetal.(2021). Theweeklymonetarypolicyshocks aretheestimatesaroundanFOMCmeeting,ifthemeetingfallsinweekt,butaresettozero, otherwise. Thecoefficientsofinterestareβ’s,whichmeasuretheresponsivenessoftheinterest ratetoamonetarypolicyshock. Wecontrolforthreelagsofchangesintheinterestrate. Table5documentsthecoefficientontheunifiedshock. Changesinthemortgageratehavea statisticallysignificantpositivecorrelationwiththeunifiedshock(Column1),whilechangesin theT-bondratedonot(Column2). AsshowninColumns3and4,wearriveatasimilarresult, evenifwenarrowthesampleperiodtothepost-2013era—thesampleperiodoftheSCE.Insum, amonetarypolicyshockpassesthroughtothe30-yearmortgageratewithstatisticalsignificance, whilethepass-throughisunclearforthe30-yearT-bondrate. With the coefficient estimates in Table 5, we recover the portions of changes in the 30yearmortgagerateattributabletomonetarypolicyshocks. Let∆R˜Mort denotechangesinthe t,brw 11Increasedhousingdemandcanalsoraisemortgageratesand,atthesametime,households’expectationsof inflationviaanincreaseinthecostofowner-occupiedhousing,amongothers.Meanwhile,ariseinthemortgagerate duetocontractionarymonetarypolicyorincreaseddefaultriskinhousingmarketscanlowerinflationexpectations. 12We donotconsider thefederal funds rate, becausethe interestratehasanambiguous effectoninflation expectationsasshowninTables1and2. 17

Table5: RESPONSES OF INTEREST RATES TO MONETARY POLICY SHOCKS Dependentvariable (1)∆RMort (2)∆RT-bond (3)∆RMort (4)∆RT-bond t t t t Unifiedshock 0.532*** -0.001 0.870** 0.007 (0.183) (0.156) (0.396) (0.301) Sampleperiod 1991–2019 1991–2019 2013–2019 2013–2019 Numberofobs. 1,299 1,334 340 350 R2 0.0214 0.0059 0.0695 0.0149 Notes:ThistablereportstheregressionresultsfromEquation(3.4).Dependentvariablesaretheweeklychangeinthe30-year mortgagerateinColumns(1)and(3)andtheweeklychangeinthe30-yearTreasuryBondrateinColumns(2)and(4).The sampleperiodisfromApril1991toJuly2019fortheestimatesreportedinColumns(1)and(2).Thesampleperiodforthe estimatesinColumns(3)and(4)isfrom2013onward.Newey-Westrobuststandarderrorsarereportedintheparenthesis. ∗∗∗ ∗∗ ∗ , , denotesstatisticalsignificanceat1%,5%,and10%levelsrespectively. Sources:Authors’calculation. mortgage rate predicted by the unified shock.13 With the predicted values, we now analyze howresponsivehouseholds’expectationsofinflationandrealeconomicactivitiesaretothe informationonmonetarypolicycontainedinthemortgagerate. Forthisanalysis,wereplace ∆RMort inEquations(3.1)and(3.2)withthecorrespondingpredictedvalue. t Theestimationresultforhouseholds’inflationexpectationsisdocumentedinTable6. Column 1 shows that in the MSC, homeowners do strongly react to news on monetary policy reflectedintheinterestratechangeswhenrevisingshort-terminflationexpectations. Consistent withthebaselineresult,noneofthecoefficientsarestatisticallysignificantinpredictingfiveyear-aheadinflationexpectations(Column2),suggestingthathouseholds’long-runinflation expectationsarenotresponsivetonewsonmonetarypolicy. However,noneofthecoefficients arestatisticallysignificantbasedupontheSCE(Columns3and4). Next,weshowtheestimationresultforhouseholds’expectationsoflabormarketconditions. Forinterpretability,wereportthemarginalprobabilityofhouseholdschangingtheirexpectations offutureunemploymentinFigure3.14 Strikingly,rentersdonotrevisedowntheiroptimismon labormarketoutlookinresponsetothemonetarypolicycomponentofmortgage-ratechanges, while homeowners do. The different responsiveness between the two groups is statistically significant. To summarize, monetary policy shocks pass through to the 30-year mortgage rate with statisticalsignificance. Homeowners,takingsignalsfromthemonetarypolicycomponentor 13Wedonotreporttheestimationresultwith∆R˜T-bond,becausethecoefficientofshockfromBuetal.(2021)is t,brw smallandnotstatisticallysignificant. 14ThelogitregressionresultsarereportedinColumn(1)ofAppendixTableB.2. 18

Table6: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO CHANGES IN INTEREST RATES PREDICTED BY MONETARY POLICY SHOCK MichiganSurvey(MSC) NYFedSurvey(SCE) Dependentvariable (1)1-year (2)5-year (3)1-year (4)3-year Homeowner×∆R (β ) -2.749 ∗∗∗ -0.717 0.881 -0.631 t 1 (0.687) (0.489) (1.119) (1.121) Renter×∆R (β ) -0.193 0.009 1.386 1.304 t 2 (1.399) (1.144) (2.417) (2.259) Numberofobs. 43,434 42,240 38,775 38,861 R2 0.0137 0.0096 0.0000 0.0001 Notes: ThistablereportstheregressionresultsfromEquations(3.1)and(3.2). Dependentvariablesarethesix-month changesintheMSC’s1-yearaheadinflationexpectations(Column(1)),the5-yearaheadinflationexpectations(Column (2)),theSCE’s1-yearaheadinflationexpectations(Column(3)),andtheSCE’s3-yearaheadinflationexpectations(Column (4))."Homeowner"and"Renter"indicatedummiesforhomeownerandrenterrespectively.∆Rt referstothesix-month changeininterestrate.Weusethe30-yearmortgageratepredictedbytheunifiedshockfromBuetal.(2021).InColumns(1) and(2),wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birthcohort,andthe levelofincome.Robuststandarderrorsarereportedintheparenthesis.InColumns(3)and(4),weemploytheindividual fixedeffectstocontroltheindividualheterogeneity.Robuststandarderrors(clusteredattheindividual-level)arereportedin ∗∗∗ ∗∗ ∗ theparenthesis. , , denotesstatisticalsignificanceat1%,5%,and10%levelsrespectively. Sources:Authors’calculation. mortgage-ratechanges,adjusttheirmacroeconomicexpectationsaccordingly,whilerentersare lesslikelytodoso. Alltold,homeownersareattentivetonewsonmonetarypolicyreflectedon mortgagerates,buttheevidenceisweakforrenters. 4 Mechanism and Robustness Section4.1examinesthepotentialnonlinearityinthemainresult,anditsimplications. Section 4.2providescorroboratingevidencebasedupontheAmericanTimeUseSurvey. Section4.3 checkstherobustnessofourmainempiricalresults. 4.1 Asymmetric effects of mortgage-rate changes on households’ expectations Thissectionexaminestheasymmetriceffectsofmortgage-ratechangesonhouseholds’inflation expectations. Homeownersholdingmortgageswouldkeepseekinganopportunitytorefinance theirmortgageswithlowerrates. Therefore,homeownershavemoreincentivetopayattention tomortgage-ratedeclines,andhencearelikelytomakemoreinformedpredictionsaboutfuture inflation in times of declining mortgage rates. In other words, attention heterogeneity due 19

Improve Unchanged Worsen .4 .35 Homeowner Renter .3 -.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03 ytilibaborP Change in mortgage rate explained by BRW shocks Figure3:MARGINALPROBABILITYOFCHANGESINHOUSEHOLDEXPECTATIONSOFUNEMPLOYMENT Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsofunemploymenttochanges ininterestratesexplainedbymonetarypolicyshocks.Theexplanatoryvariableconsideredisthe30-yearmortgagerate predictedbytheunifiedshockfromBuetal.(2021).Theresultsarecalculatedbasedontheestimatesofthelogitregression resultsfromEquation(3.3)asreportedinColumn(1)ofAppendixTableB.2.Wecontrolfortheobservedsurveyrespondents’ characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareasrepresent95%confidence intervals. Source:Authors’calculation. to the refinancing motive could lead to an asymmetric response of inflation expectations to mortgage-ragechanges—specifically,largersensitivityintimesofdecliningmortgageratesthan intimesofrisingmortgagerates.15 We consider the following specification which is a variant of Equation (3.1) in order to separatetheeffectsofincreasesanddecreasesinmortgagerates: E h−yr −E h−yr =α+β homeowner ×∆R ×I ++β homeowner ×∆R ×I − i,t+6 i,t 1 i t t 2 i t t (4.1) +β renter ×∆R ×I ++β renter ×∆R ×I −+δX +ϵ , 3 i t t 4 i t t i,t i,t + − whereI (I )isadummyvariableindicatinganincrease(decrease)inthemortgagerate. For t t ∆R , we consider a changein30-year mortgage (∆RMort) and achange in themortgage rate t t attributedtomonetarypolicyshocks. Thelargernegativeandstatisticallysignificantcoefficient onhomeowner×I − thanthatonhomeowner×I + suggeststherefinancingmotiveofhomeowners t t isineffect. WeestimatethismodelwiththeMSCdataincludingthesamesetofhousehold-level controlsforourmainempiricalspecification. 15Severalrecentworkshighlightthatthisrefinancingmotivewouldbeimportantforthetransmissionofmonetary policy(e.g.,Amromin,BhuttaandKeys,2020;Berger,Milbradt,TourreandVavra,2021;Wong,2021;Eichenbaum, RebeloandWong,2022.Relatedly,BarnichonandMatthes(2016)arguethatcontractionarymonetarypolicyhasa significantlystrongereffectontheeconomythanexpansionarypolicy. 20

Table7: ASYMMETRIC EFFECTS OF MORTGAGE-RATE CHANGES 1-yearaheadinflationexpectations 5-yearaheadinflationexpectations Interactions (1)∆RMort (2)∆R˜Mort (3)∆RMort (4)∆R˜Mort t t,brw t t,brw Homeowner×I + -0.1072 -1.6605 ∗ 0.0653 -0.0392 t (0.1243) (0.9614) (0.1116) (0.7199) Renter×I + -0.5835 ∗∗∗ 1.5777 -0.4370 ∗∗ -0.0577 t (0.1870) (1.9134) (0.1816) (1.5933) Homeowner×I − -0.3293 ∗∗∗ -3.7813 ∗∗∗ -0.0564 -1.3554 ∗∗ t (0.1078) (0.9699) (0.8270) (0.6578) Renter×I − 0.1756 -1.8928 0.3443 ∗∗ 0.0668 t (0.1739) (2.0247) (0.1541) (1.6436) Numberofobs. 50,834 43,434 48,842 42,240 R2 0.0144 0.0138 0.0111 0.0096 Notes:ThistablereportstheregressionresultsfromEquation(4.1).Dependentvariablesarethesix-monthchangeinthe MSC’s12-monthaheadinflationexpectations(Columns(1)and(2))andthesix-monthchangeintheMSC’s5-yearahead inflationexpectations(Columns(3)and(4)).“Homeowner”and“Renter”indicatedummiesforhomeownerandrenter + − respectively.I andandI indicatedummiesforperiodsofincreaseanddecreasein30-yearmortgageratesrespectively. t t Columns(1)and(3)reportresponsestochangesin30-yearmortgagerate;Columns(2)and(4)reportresponsestothe changesin30-yearmortgageratepredictedtheunifiedshockfromBuetal.(2021). Wecontrolfortheobservedsurvey respondents’characteristics,includinggender,education,birthcohort,andthelevelofincome.Robuststandarderrorsare ∗∗∗ ∗∗ ∗ reportedintheparenthesis. , , denotesstatisticalsignificanceat1%,5%,and10%levelsrespectively. Source:Authors’calculation. Table7reportstheestimationresult. Overall,theestimationresultsupportstherefinancing motive as an important factor driving the sensitivity of homeowners’ inflation expectations to mortgage-rate changes. Homeowners’ inflation expectations—both short-term and longterm—respondtothemonetary-policycomponentofmortgage-ratedeclinesmorethantheydo tothatofmortgage-raterises(Columns2and4).Inlinewiththisresult,homeowners’short-term inflationexpectationsincreasetoadeclineinmortgagerateswithstatisticalsignificance,while thecoefficient’sstatisticalsignificancedisappearstoariseinmortgagerates(Column1). For long-term inflation expectations, homeowners’ asymmetric sensitivity to the mortgage-rate changeisunclear(Column3). However,strikingly,homeownersraisetheirlong-terminflation expectations with statistical significance in response to a drop in mortgage rates caused by monetarypolicyshocks. Insummary,adropinmortgageratedrivenbymonetarypolicyraises homeowners’inflationexpectationsinthelongrunaswellasintheshortrun. Quite differently, renters’ inflation expectations respond more strongly to an increase in mortgage rate, and are not driven by monetary policy shocks. While renters do not have an incentivetorefinance,theymightworryaboutahighermortgagerateaspotentialhome-buyers, 21

leadingtheirexpectationstorespondmorestronglytotheratehikesthantotheratecuts. Theobservedasymmetricsensitivityofhomeowners’inflationexpectationstoachangein mortgage rates and monetary policy offers new insight about the effectiveness of monetary policy. Thisanalysissuggeststhathomeownershipandhomeowners’refinancingmotivemaybe akeydriverofunequaleffectsofmonetarypolicyonhouseholds’expectations. 4.2 EvidencefromAmericanTimeUseSurvey Thissectionprovidescorroboratingevidenceforthegreaterattentivenessofhomeownerstothe macroeconomicdevelopmentsthanthatofrentersbasedupontheAmericanTimeUseSurvey (hereafter,ATUS). Here,weprovidebriefexplanationsonthesurveyandthevariablesusedfortheempirical analysis.16 TheATUScollectsdataonthetimethatanindividualspendsforvariousactivities duringaday. ThesampleofATUSisfromtheeighthoutgoingrotationgroupoftheCurrent Population Survey. Therefore, each individual in the ATUS is surveyed once. The ATUS is a monthlysurvey,whichbeganfrom2003. Hence,thesampleperiodofourempiricalanalysisis from2003:M1to2020:M12. The ATUS has information about an individual’s time spent on finance-related activities, whichisanaturalmeasureofhouseholds’attentivenesstofinancialmarketsandmacroeconomic developments. Specifically, the ATUS has data on time spent on financial management and purchasingfinancialandbankingservices. Activitiesonfinancialmanagementincludetrading andcheckingstocks,researchinginvestments,payingmortgages,checkingcryptocurrencyor bitcoin balance, and so on. Activities on purchasing financial and banking services include applying for a loan or mortgage, talking to/with a loan officer, meeting with a stockbroker, insuranceagent,bankmanager,etc. Inaddition,theATUShasrespondents’socio-economic characteristicsincludinghomeownershipaswellasdemographicattributes. Therefore,thedata allowustoanalyzetheassociationbetweenhomeownershipandattentivenesstodevelopments infinancialmarketsandthemacroeconomy.17 16ForinterestedreadersontheATUSusedforotherresearch,see,forexample,Ke(2021),Mukoyama,Patterson andS¸ahin(2018)andAhnandShao(2021). 17TheATUSmayunderstatetimespentonfinancialmanagementandpurchasingfinancialandbankingservices, becausetheATUSsurveystherespondent’sprimaryactivityonly.Ifanindividualchecksstockpriceswhileworking orwatchingTV,thisactivitymaybeclassifiedas“working”or“TVwatching.” 22

Table8: HOMEOWNERSHIP AND TIME SPENT ON FINANCE-RELATED ACTIVITIES Dependentvariable (1)Extensive(E ) (2)Intensive(N ) (3)Time(Time ) i i i PanelA.Financialmanagement ∗∗∗ ∗∗∗ ∗∗∗ Homeowner 0.0076 4.6085 0.5331 (0.0010) (1.7788) (0.0810) Numberofobs. 219,368 8,583 219,368 R2 0.0084 0.0366 0.0064 PanelB.Purchasingfinancialandbankingservices ∗∗∗ ∗∗∗ ∗∗∗ Homeowner 0.0029 0.6366 0.0636 (0.0010) (0.7869) (0.0228) Numberofobs. 219,368 5,618 219,368 R2 0.0020 0.0366 0.0008 Notes:Thistablereportstheestimatesofβ 1’sfromEquation(4.2).PanelAshowstheresultswhenweusetimespenton financialmanagementasdependentvariables.PanelBshowstheresultswhenweusetimespentforpurchasingfinancial bankingservicesasdependentvariables.InColumn(1),weusetheindicatorofrespondentsparticipatingintheactivity (extensivemargin). InColumn(2),weuseminutesspentfortheactivityconditionalonreportingnonzerominutesfor theactivity(intensivemargin).Lastly,inColumn(3),weusethetotaltimespentontheactivity.“Homeowner”indicates adummyforhomeowner. Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,age,race, education,andlaborforcestatus. Robuststandarderrorsarereportedintheparenthesis. ***,**,*denotesstatistical significanceat1%,5%,and10%levelsrespectively. Source:Authors’calculation. Weconsiderthefollowinglinearregressionmodel: Y =α+β homeowner +δX +ϵ (4.2) i 1 i i i where Y = {Time ; E ; N }. Time is individual i’s time spent on financial management, E i i i i i i denotestheindicatorofrespondenti’sparticipatingintheactivity,andN isminutesspentfor i financialmanagementconditionalonreportingnonzerominutesfortheactivity.NoticethatE is i theextensivemarginoffinancialmanagement. Ittakesvalue1,ifanindividualreportsanonzero minuteforfinancialmanagement,butiszero,otherwise.ThenotationN istheintensivemargin, i and takes always a positive value. Individual characteristics, denoted by X , include gender i (female),age(16-24,55andover),race(white),education(high-schoolgraduationorless,some collegeandassociatedegree),laborforcestatus(unemploymentandoutofthelaborforce).18 Table8reportsthecoefficientestimates. Beingahomeownerraisestheprobabilityofen- 18Weconsideralinearprobabilitymodelfortheextensivemarginasthebaseline.Wefurtherconsideralogitand probitmodelfortheextensivemargin,buttheoverallconclusionisthesameasthatfromthelinearprobability model.Ourresultsarerobustoncewecontrolforoccupation,regionand/ortime-fixedeffects. 23

gaginginfinancialmanagementandalsotimespentforfinancialmanagementamongthose whoengageintheactivitywithstatisticalsignificance(PanelA).Essentially,similarresultsare obtained,ifwereplacethedependentvariablewithtimespentforpurchasingfinancialandbankingservices(PanelB).Thisresultsuggeststhathomeownersaremorelikelytoengageinactivities thatexposethemtocurrentmacroeconomicconditionsandinterestrates,andalsotospend moretimeontheseactivities.Alltold,thisdirectevidencefromATUSconfirmsthathomeowners tendtopaymoreattentiontooverallmacroeconomicconditionsthanrenters,corroborating whyhomeowners’expectationsofthemacroeconomyaremoresensitivetointerest-ratechanges thanothers. 4.3 Robustnesschecks Thissectionoutlinestherobustnessofempiricalresults. Thedetailsarefoundintheappendix. First, we control for house-price expectations to examine the possibility that differences in households’inflationexpectationsjustreflecthouse-priceexpectations. Evenifwedoso,the mainresultisrobustasdocumentedinPanelAofAppendixTableB.3.Second,sinceinflationrate itselfrespondstointerestratechanges,ourmeasuredresponsivenessofhouseholds’inflation expectationscouldcomefromtheirresponsestotheactualinflationoverthe6-monthperiod. Toruleouthouseholds’responsestochangesininflationasthemainchannelthatdrivesour results, we include concurrent inflation rate as an additional explanatory variable. Panel B of Appendix Table B.3 shows that the responsiveness of homeowners’ and renters’ inflation expectationstointerestratechangesremainsrobust. Third,weconsiderdifferenthorizonsof interest-ratechanges. Inourbaselinespecification,weemploychangesininterestrateduring thepastsixmonths. Instead,weadoptinterest-ratechangesduringthepastthree-monthsor nine-months. AsshowninAppendixTablesB.4andB.5,ourmainresultontheheterogeneous responseofinflationexpectationstointerest-ratechangesstillsurvivesregardless.19 However, theempiricalrelationshipbecomesweaker,ifthehorizonofinterest-ratechangesbecomesthe pasttwelvemonthsorlonger. Fourth,wereplaceexpectationsonunemploymentconditions withthoseonfuturebusinessconditionstoexaminethecontractionaryeffectofinterest-rate changesonoveralleconomicconditions. AsshowninAppendixTableB.6andFigureB.5,our 19Similarly,wealsofindthatthedifferenthorizonsofinterestratechangesdonotchangeourconclusiononthe heterogeneousresponsesofunemploymentexpectations(AppendixFiguresB.1andB.2)andthoseof1-yearahead interestrateexpectations(AppendixFiguresB.3andB.4). 24

mainconclusionremainsrobust. We employ an alternative measure of monetary policy shocks constructed by Swanson (2021). ThreeorthogonalfactorsofFOMCannouncementscapturechangesinfederalfunds rate, forward guidance, and large scale asset purchases (LSAPs), respectively. Each of three measures from Swanson (2021) is correlated with the BRW estimate.20 Appendix Table B.7 documentsthecoefficientsonthefederalfundsrate,forwardguidance,andtheLSAPshocks. Werecovertheportionsofchangesinthe30-yearmortgagerateandT-bondrateattributableto forwardguidanceandLSAPshocks. Withthepredictedvalues,weestimatetheresponsivenessof households’expectationsofinflation(AppendixTableB.8)andrealeconomicactivities(Appendix Table B.2 and Appendix Figure B.6) to the information on monetary policy contained in the mortgagerateandT-bondrate. Insummary,ourresultsarerobusttothisalternativemeasureof monetarypolicyshocks. 5 Model Inthissection,weprovideastructuralmodelthatexplainsourempiricalfindingswithafocus oninflationexpectations. Thestructuralmodelisdesignedtocharacterizethetwofindings: First,homeownersrevisedowntheirinflationexpectationstoariseininterestratemorethan rentersdo;Second,homeownershaveamoreincentivetopayattentiontonewsoninterest-rate changesbutrentersdonot. Webuildatwo-periodpartialequilibriummodelofrationallyinattentiveconsumerswhoare eitherahomeownerorarenter. Thehomeownermakesaonce-and-for-allhousepurchaseand consumeshousingserviceswhiletherenterdoesnotmakeadecisiononhousingpurchase,but rentsit.21 20Thetwotypesofshockestimatescapturethedifferentcomponentsofmonetarypolicychanges.Duringthe pre-ZLBperiod,theBRWshockhasacorrelationof0.44withthefederal-funds-ratefactor,andacorrelationof0.45 withtheforward-guidancefactor.ThecorrelationbetweenBRWshockandforwardguidancefactorincreasesto 0.55inthepost-ZLBperiod,whilethecorrelationwithfederal-funds-rateshockdropsclosetozero.TheBRWshock haslowcorrelationswiththeLSAPfactoreitherpre-orpost-ZLBperiod. 21Thisassumptioniswithoutlossofgenerality.Thoughhomeownersmayrentandrentersmayownahouse,we donotconsideritarepresentativecase.Forinstance,intheMSC,homeownersarethosewhoindicatethatthey currentlyownorarebuyingtheirhomeandrentersarethosewhodonot.Similarly,intheSCE,homeownersare thosewhoindicatethattheyowntheirprimaryresidenceandrentersarethosewhodonot.Onlyasmallfractionof renters,about1.2%ofallrespondents,indicatethattheyownotherhomes.See,forinstance,Garriga,Kydlandand Šustek(2017)forasimilarsetuptoanalyzetheeffectofmonetarypolicywithhomeownersandrenters. 25

5.1 Homeowner Supposetherearetwoperiods,withtimedenotedbyt =0,1. Ineachperiod,thehomeowneris endowedwithafixedrealincome,w,andconsumesCo fort =0,1. Intime0,thehomeowner t makesaonce-and-for-allhousepurchase,H,financingafractionθofthepurchasewithaloan andafraction1−θwithhomeowner’sincome. Theloancanbeusedonlyforthehousingpurchase;thehouselastsfort =0,1,thenitfullydepreciates.Thelifetimeutilityofthehomeowneris u (cid:161) Co(cid:162)+βu (cid:161) Co(cid:162)+(cid:161) 1+β(cid:162) v(H). ThehomeownermaximizesitsutilitybychoosingCo,Co,andH, 0 1 0 1 subjecttobudgetconstraints:Co+H =w+ l andCo=w− m,wherel =θP H isthenominal 0 P0 1 P1 0 valueoftheloan,m=Rml isnominalmortgagepayments,Rm isthe(gross)mortgageinterest rate,P isaggregatepricelevel,andΠ= P1 istheinflationrate. t P0 Weassumethatthehomeownerdoesnotknowthenominalmortgagerateandaggregate pricesbutmayobservecostlyandnoisysignal(s)aboutthemgiventheinformationalcost,which wediscusslaterinfurtherdetail. Thistwo-periodproblemisbrokenintothreesequentialsteps: (1)obtainnoisysignal(s),(2)committotheamountofthehousingpurchaseandthemortgage payment,and(3)consumesothatthebudgetconstraintbinds.22 Bycombiningthebudgetconstraintsintothehomeowner’sutility,weobtaintheutilityasa functionofthechoicevariable,H,andthetwounknownfundamentals, (cid:169) Rm,Π(cid:170) : U (cid:161) H,Rm,Π(cid:162)=u(w−(1−θ)H)+βu (cid:181) w−θ (cid:181) Rm(cid:182) H (cid:182) +(cid:161) 1+β(cid:162) v(H). Π Wetakethesecond-orderTaylorapproximationtotheutilityfunctionaroundthenonstochastic steadystate. Atthissteadystate,theoptimalhousingpurchaseisdenotedbyH¯ andsolvesthe firstordercondition,U (cid:161) H¯,R¯m,Π¯(cid:162)=0.Denotelog-deviationswithlowercasevariables(e.g., 1 h=logH−logH¯,π=logΠ−logΠ¯,andim=logRm−logR¯m).Weassumethatπandimaredrawn fromindependentGaussiandistributionswithmeanzeroandvarianceσ2,implyingthatinflation andthemortgageratehavezerocorrelationunderthemodel’struedata-generatingprocess. Letuˆ(h,im,π)=U (cid:161) H¯eh,R¯meim ,Π¯e π(cid:162)=U(H,Rm,Π)denotethehomeowner’sutilityfunction expressedintermsoflogdeviationsandu˜ denotethesecond-orderTaylorapproximationofuˆ at 22Thetimingimpliesthebudgetconstraintwillholdintherealizationsoftheunknowns,notjustinexpectations. Thisassumptioniscommonlyusedinthemodelsofrationallyinattentivehouseholds.See,forinstance,Kamdar (2019). 26

thesteady-state: u˜ (cid:161) h,im,π(cid:162)=uˆ h+ 1 uˆ h2+uˆ him+uˆ hπ+termsindependentofhousing. 1 11 12 13 2 Notice that uˆ = 0 because the housing purchase is the choice variable, and uˆ < 0 with a 1 11 standard convexity property of the utility function. Let x =[im π] ′ denote the vector of the twounknowns. Then,wecanrewritethehomeowner’sutilityfunctionas−h ′ Dh+x ′ Bh where D= 1|uˆ |andB =[uˆ uˆ ] ′ . Giventhehomeowner’sinformationsetS ,theoptimalhousing 2 11 12 13 choiceish ∗= D −1B ′ x˜ wherex˜=E[x|S ]. Lastly,werewritetheutilityfunctionas 2 (cid:181) D −1B ′ D −1B ′ (cid:182)′ (cid:181) D −1B ′ D −1B ′ (cid:182) U =− E[x|S ]− x D E[x|S ]− x =−(x˜−x) ′Ωo(x˜−x), 2 2 2 2   uˆ2 uˆ uˆ whereΩo= BD 4 −1B ′ = 2|uˆ 1 11 | uˆ 1 uˆ 2 1 uˆ 2 2 13 . 12 13 13 Weassumethehomeownerisrationallyinattentiveandchoosesthehousingpurchasebased onimperfectinformationabouttwounknownsintheeconomy.Wemodelthecostofinformation processing as a linear function in Shannon’s mutual information function. Let H(X|Y) be a conditional entropy of a random variable of X given knowledge of Y. We define the flow costofinformationwiththeinformationsetS =S −1 ∪s asλI(x;S |S −1 ),whereI(x;S |S −1 )= H(x|S )−E[H(x|S )|S −1 ]isthereductioninuncertaintyaboutunknownfundamentalsthatthe homeownerexperiencesbyobservingasetofsignals,s,giventhehomeowner’spriorinformation set,S −1 ,andλisthemarginalcostofabitofinformation. Becausethehomeowner’spriorand posteriorbeliefsarebothGaussian,wecanrewritethehomeowner’sproblemaschoosingthe posteriorvariance-covariancematrixabouttheunknownfundamentals,Σo,tomaximizeutility giventhecostofinformationprocessing: λ max −(x˜−x) ′Ωo(x˜−x)− log (cid:175) (cid:175) Σo(cid:175) (cid:175), Σo≤Γ 2 whereΓ=σ2I isthepriorvariance-covariancematrix.23 23Anyn-dimensionalnormallydistributedvectorhasentropy n+nlog(2π)+1log|Σ|whereΣistheposterior 2 2 2 variance-covariancematrixabouttheunknownfundamentals. 27

5.2 Renter Ineachperiod,therenterisendowedwithafixedrealincome,w,andconsumesCr fort =0,1. t Unlikethehomeowner,therenterdoesnotbuyahouse,butrentsonewithaconstantrental rate γ =γ =γ. Let w˜ =w−γ denote the rent-adjusted real income for the renter. At time 0 1 0, the renter trades a one-period nominal bond, B, with an exogenous nominal rate R. The rentermaximizeslifetimeutility,u (cid:161) Cr(cid:162)+βu (cid:161) Cr(cid:162) ,bychoosingCr,Cr,andB,subjecttobudget 0 1 0 1 constraints:Cr+ B =w˜ andCr =w˜+R B .Weassumethattherenterdoesnotknowthenominal 0 P0 1 P1 rate and aggregate prices, but may observe costly and noisy signal(s) about them given the informationalcost.24 Bycombiningbudgetconstraintswiththerenter’sutility,weobtaintheutilityasafunction ofthechoicevariable,Cr,andthetwounknownfundamentals,{R,Π}: 0 (cid:181) (cid:182) U (cid:161) Cr,R,Π(cid:162)=u (cid:161) Cr(cid:162)+βu w˜ + R (cid:161) w˜ −Cr(cid:162) . 0 0 Π 0 Denotelog-deviationswithlowercasevariables(e.g.,cr =logCr −logC¯r andi =logR−logR¯). 0 0 0 Similar to the homeowner’s problem, we take the second-order Taylor approximation to the renter’s utility around the non-stochastic steady state. Let uˇ (cid:161) cr,i,π(cid:162)=U (cid:161) C¯rec 0 r ,R¯ei,Π¯e π(cid:162)= 0 U (cid:161) Cr,R,Π(cid:162) denotetheutilityfunctionexpressedintermsoflogdeviationsandletu˜ˇ denotethe 0 second-orderTaylorapproximationofuˇ atthesteady-state: u˜ˇ (cid:161) cr,i,π(cid:162)=uˇ h+ 1 uˇ (cr)2+uˇ cri+uˇ crπ+termsindependentofcr. 0 1 2 11 0 12 0 13 0 0 Therenterisalsorationallyinattentiveandchoosesconsumptionbasedonimperfectinformationabouttwounknownsintheeconomy. Let xr =[i π] ′ denotethevectorofthetwo unknowns for the renter. After incorporating the cost of information processing as a linear functioninShannon’smutualinformationfunction,therenter’sproblemcanbeexpressedas choosingtheposteriorvariance-covariancematrix,Σr,tosolvethefollowing: λ max −(cid:161) x˜r −xr(cid:162)′ Ωr(cid:161) x˜r −xr(cid:162)− log (cid:175) (cid:175) Σr(cid:175) (cid:175), Σr≤Γ 2 24Here,weassumeaggregatepricesasconsumerpricesexcludinghousingservices. 28

  uˇ2 uˇ uˇ whereΓ=σ2I isthepriorvariance-covariancematrixandΩr = 2|uˇ 1 11 | uˇ 1 uˇ 2 1 uˇ 2 2 13 . 12 13 13 5.3 Solution Toderiveananalyticalsolutionoftherationalinattentionproblems,weassumethatthehomeownerhasalogutilityinconsumptionandhousing,u(C)+v(H)=logC+logH,andtherenter alsohasalogutilityinconsumption: u(Cr)=logCr. Wealsoassumethatthenominalrateis identicaltothemortgagerate,i =im,implyingthatbothinterestratesperfectlyco-move,and havezerocorrelationwithinflationunderthemodel’struedata-generatingprocess. Wesolvetherationalinattentionproblemsusingarecentlydevelopedsolutionmethod(e.g., Kamdar 2019; Ko˝szegi and Mateˇjka 2020; Afrouzi and Yang 2021). For a symmetric matrix X withspectraldecompositionX=UDU ′ ,wedefineMax(X,λ)=Umax(D,λ)U ′ wheremax(D,λ) operates on every element on the diagonal. Then, as shown in Afrouzi and Yang (2021), the optimalposteriorcovariancematrix,Σi,ischaracterizedbythefollowingpolicyfunction: (cid:104) (cid:179) (cid:180)(cid:105)−1 Σi =λσ2 Max σ2Ωi,λ fori ∈{o,r}, Fori ∈{o,r},thespectraldecompositionofamatrixσ2Ωi showsonepositiveeigenvalue,Λi >0, 1 and another zero eigenvalue, Λi = 0. We set the marginal cost of information, λ, to satisfy 2 Λo=σ2 uˆ 1 2 2 >λandΛr =σ2 uˇ 1 2 2 >λ.25 Inthiscase,boththehomeownerandtherenterchoose 1 |uˆ11 | 1 |uˇ11 | toobserveonesignalalongthecorrespondingfirstdimension. Wecharacterizetheoptimalposteriormeansoftheagents’beliefusingBayesianupdating. Thefollowingpropositioncomparestheresponseoftheagents’inflationexpectationswithan exogenouschangeintheinterestrate. Proposition. Letπ˜o andπ˜r beinflationexpectationsofthehomeownerandtherenter,respectively. Then,ifβ> 1, 2 ∂π˜o ∂π˜r < <0. ∂i ∂i Proof. Thehomeowner’soptimalposteriormeanbeliefofinflationis (cid:195) (cid:33) π˜o= 1− λ|uˆ 11 | (cid:163)−i+π+σzo(cid:164)= (cid:181) 1− λ (cid:181) 1+β(cid:182)(cid:182) (cid:163)−i+π+σzo(cid:164) . σ2uˆ2 σ2 2β2 12 25InAppendixA.,wederivethesecondderivativesofthehomeowner’sandtherenter’sutilityfunction. 29

(cid:179) (cid:180) where zo ∼N 0, λ is the homeowner’s rational inattention error. Similarly, the renter’s Λo−λ 1 inflationexpectationcanbeexpressedasfollows: (cid:195) (cid:33) π˜r = 1− λ|uˇ 11 | (cid:163)−i+π+σzr(cid:164)= (cid:181) 1− λ (cid:181) 1+β(cid:182)(cid:182) (cid:163)−i+π+σzr(cid:164) σ2uˇ2 σ2 β 12 (cid:179) (cid:180) wherezr ∼N 0, λ istherenter’srationalinattentionerror. Then,ifβ> 1, ∂π˜o < ∂π˜r <0. Λr−λ 2 ∂i ∂i 1 Consistentwithourempiricalfindings,themodelimpliesthatthehomeowner’sinflation expectationsaremoresensitivetoachangeininterestratesthantherenter’sinflationexpectations. Because,attime0,bothhomeownerandrentermaketheirintertemporalsavingdecision onnominalassets(housingforthehomeownerandaone-periodnominalbondfortherenter), they have an incentive to acquire information about inflation.26 Notice that the changes in inflationaffecttherelativepriceofhousing,whichhasasubstitutioneffectonconsumptionfor thehomeowner,becausehomeownersvaluehousingservicesintheirutility. Thisadditional substitutioneffectisabsentfromtherenter’sproblem. Ifboththehomeownerandtherenterare sufficientlypatient(β> 1),thenthisadditionalsubstitutioneffectgivesastrongerincentivefor 2 thehomeownertopayattentiontotheunknownfundamentalscomparedwiththerenter.27 This findingimpliesthatanexogenousincreaseininterestratesreducesthehomeowner’sinflation expectationsmorethantherenter’sexpectations. 6 Conclusion Thispaperinvestigatestheroleofhouseholdheterogeneityintheeffectivenessofmonetary policy on households’ expectations with a particular focus on homeownership. The survey datasuggestthathomeownersrevisedowntheirexpectationsofinflationandthelabor-market outlookinthenearterminresponsetoariseinlong-termmortgageratesorT-bondrates,while rentersarelesslikelytodoso. Wefindthatthemonetary-policycomponentofmortgage-rate 26Whilewefocusonthedifferentincentivestoacquireinformationaboutinflationandassumethatthemarginal costofinformation,λ,isidenticalforbothtypesofhouseholds,onemightthinkofacasewherethehomeowner hasasmallercostthantherenterduetoeitherotherhouseholdcharacteristics(e.g.,educationandincome)orthe greateravailabilityofnewsrelatedwithinterestrates(e.g.,mortgagenewsletters).Thiscase,however,isonlylikely tostrengthenourresultsasthehomeownerwillacquiremoreinformationaboutinflationwithalowermarginal costofinformation. 27This relationship can be easily found when we compare the homeowner’s benefit matrix, Ωo = β (cid:179) β (cid:180)(cid:181) 1 −1 (cid:182) ,withtherenter’sbenefitmatrix,Ωr =1 (cid:179) β (cid:180)(cid:181) 1 −1 (cid:182) . 1+β −1 1 2 1+β −1 1 30

changesdrivesthecontractionaryeffectonhomeowners’expectations. Weinterpretthisresult tosuggestthathomeownersareattentivetonewsoninterestratesandrevisetheirexpectations accordinglyinamannerconsistentwiththeintendedeffectofmonetarypolicy. Wecharacterizethesefindingsusingarationalinattentionmodelwithaninformationacquisitioncostwherehomeownersmakingmortgagepaymentshaveanincentivetopayattentionto newsoninterestratesand,hence,adjusttheirexpectationsinresponsetoachangeininterest rates more than renters do. Our results imply that the macroeconomic effects of monetary policymaydependonthedistributionofhouseholdhomeownershipstatusintheeconomy. A fully-fledgedgeneralequilibriummodelmaybeabletoaddressthemacroeconomiceffectsof monetarypolicyanditswelfareimplicationsforhouseholds. Weleavetheextendedanalysesfor futureresearch. References Adelino,Manuel,AntoinetteSchoar,andFelipeSeverino,“PerceptionofHousePriceRiskand Homeownership,”WorkingPaper25090,NationalBureauofEconomicResearch2018. Afrouzi,HassanandChoongryulYang,“DynamicRationalInattentionandthePhillipsCurve,” CESifoWorkingPaperNo.88402021. Ahn,HieJooandLingShao,“TheCyclicalityofOn-the-JobSearchEffort,”TheB.E.Journalof Macroeconomics,January2021,21(1),185–220. Amromin,Gene,NeilBhutta,andBenjaminJ.Keys,“Refinancing,MonetaryPolicy,andthe CreditCycle,”AnnualReviewofFinancialEconomics,2020,12(1),67–93. Armona, Luis, Andreas Fuster, and Basit Zafar, “Home Price Expectations and Behaviour: EvidencefromaRandomizedInformationExperiment,”ReviewofEconomicStudies,2019,86 (4),1371–1410. Barnichon,RégisandChristianMatthes,“GaussianMixtureApproximationsofImpulseResponsesandTheNon-LinearEffectsofMonetaryShocks,”CEPRDiscussionPapers11374July 2016. Bauer, Michael D., “Inflation Expectations and the News,” International Journal of Central Banking,2015,11(2),1–40. andEricT.Swanson, “A Reassessment of Monetary Policy Surprises and High-Frequency Identification,”NBERMacroeconomicsAnnual,2022,37. Berger,David,KonstantinMilbradt,FabriceTourre,andJosephVavra,“Mortgageprepayment andpath-dependenteffectsofmonetarypolicy,”AmericanEconomicReview,2021,111(9), 2829–78. 31

Blinder,Alan,“TalkingAboutMonetaryPolicy: TheVirtues(andVice?) ofCentralBankCommunication,”BISWorkingPaperNo.274March2009. Bu,Chunya,JohnRogers,andWenbinWu,“AunifiedmeasureofFedmonetarypolicyshocks,” JournalofMonetaryEconomics,2021,118,331–349. Carroll,ChristopherD.,“MacroeconomicExpectationsofHouseholdsandProfessionalForecasters,”TheQuarterlyJournalofEconomics,2003,118(1),269–298. Cavallo,Alberto,GuillermoCruces,andRicardoPerez-Truglia,“InflationExpectations,Learning,andSupermarketPrices:EvidencefromSurveyExperiments,”AmericanEconomicJournal: Macroeconomics,2017,9(3),1–35. Claus,EddaandVietHoangNguyen,“Monetarypolicyshocksfromtheconsumerperspective,” JournalofMonetaryEconomics,2020,114(C),159–173. Coibion,OlivierandYuriyGorodnichenko,“IsthePhillipsCurveAliveandWellafterAll? InflationExpectationsandtheMissingDisinflation,”AmericanEconomicJournal: Macroeconomics, 2015,7(1),197–232. , ,andMichaelWeber,“FiscalPolicyandHouseholds’InflationExpectations: Evidencefrom aRandomizedControlTrial,”CESifoWorkingPaperSeries8905,CESifo2021. , ,and ,“Monetarypolicycommunicationsandtheireffectsonhouseholdinflationexpectations,”JournalofPoliticalEconomy,2022,130(6),1537–1584. , ,andRupalKamdar,“TheFormationofExpectations,Inflation,andthePhillipsCurve,” JournalofEconomicLiterature,2018,56(4),1447–91. , , and Saten Kumar, “How Do Firms Form Their Expectations? New Survey Evidence,” AmericanEconomicReview,2018,108(9),2671–2713. D’Acunto,Francesco,UlrikeMalmendier,andMichaelWeber,“WhatDotheDataTellUsAbout InflationExpectations?,”WorkingPaper29825,NationalBureauofEconomicResearch2022. , ,JuanOspina,andMichaelWeber,“ExposuretoGroceryPricesandInflationExpectations,” JournalofPoliticalEconomy,2021,129(5),1615–1639. Das,Sreyoshi,CameliaMKuhnen,andStefanNagel,“Socioeconomicstatusandmacroeconomicexpectations,”TheReviewofFinancialStudies,2020,33(1),395–432. D’Acunto,Francesco,DanielHoang,andMichaelWeber,“Managinghouseholds’expectations withunconventionalpolicies,”TheReviewofFinancialStudies,2022,35(4),1597–1642. , ,MarittaPaloviita,andMichaelWeber, “IQ,expectations, andchoice,” WorkingPaper 25496,NationalBureauofEconomicResearch2019. , , ,and ,“HumanFrictionsintheTransmissionofEconomicPolicies,”WorkingPaper 29279,NationalBureauofEconomicResearchSeptember2021. Eichenbaum,Martin,SergioRebelo,andArleneWong,“State-DependentEffectsofMonetary Policy: TheRefinancingChannel,”AmericanEconomicReview,March2022,112(3),721–61. Garriga,Carlos,FinnE.Kydland,andRomanŠustek,“MortgagesandMonetaryPolicy,”The ReviewofFinancialStudies,052017,30(10),3337–3375. 32

Grundl,SerafinandYouSukKim,“Consumermistakesandadvertising: Thecaseofmortgage refinancing,”QuantitativeMarketingandEconomics(QME),2019,17(2),161–213. Hoffmann,Mathias,EmanuelMoench,LoraPavlova,andGuidoSchultefrankenfeld,“TheeffectsoftheECB’snewinflationtargetonprivatehouseholds’inflationexpectations,”Technical Report43rdedition,DeutscheBundesbank2021. Kamdar,Rupal,“TheInattentiveConsumer: SentimentandExpectations,”workingpaper,2019. Ke,Da,“Whowearsthepants? Genderidentitynormsandintrahouseholdfinancialdecisionmaking,”TheJournalofFinance,2021,76(3),1389–1425. Kilian,LutzandXiaoqingZhou,“TheImpactofRisingOilPricesonU.S.InflationandInflation Expectationsin2020-23,”WorkingPapers2116,FederalReserveBankofDallas2021. Ko˝szegi,BotondandFilipMateˇjka,“ChoiceSimplification: ATheoryofMentalBudgetingand NaiveDiversification,”TheQuarterlyJournalofEconomics,2020,135(2),1153–1207. Kryvtsov,OleksiyandLubaPetersen,“Centralbankcommunicationthatworks: Lessonsfrom labexperiments,”JournalofMonetaryEconomics,2021,117,760–780. Kuchler, TheresaandBasitZafar, “Personal Experiences and Expectations about Aggregate Outcomes,”JournalofFinance,2019,74(5),2491–2542. Lamla,MichaelJ.andDmitriV.Vinogradov,“Centralbankannouncements: Bignewsforlittle people?,”JournalofMonetaryEconomics,2019,108,21–38. Mukoyama,Toshihiko,ChristinaPatterson,andAys¸egülS¸ahin,“JobSearchBehavioroverthe BusinessCycle,”AmericanEconomicJournal: Macroeconomics,January2018,10(1),190–215. Nakamura,EmiandJónSteinsson,“High-FrequencyIdentificationofMonetaryNon-Neutrality: TheInformationEffect,”TheQuarterlyJournalofEconomics,2018,133(3),1283–1330. Stillwagon,JoshR.,“TIPSandtheVIX:SpilloversfromFinancialPanictoBreakevenInflationin anAutomated,NonlinearModelingFramework,”OxfordBulletinofEconomicsandStatistics, 2018,80(2),218–235. Swanson,EricT,“MeasuringtheeffectsofFederalReserveforwardguidanceandassetpurchases onfinancialmarkets,”JournalofMonetaryEconomics,2021,118,32–53. Weber,Michael,FrancescoD’Acunto,YuriyGorodnichenko,andOlivierCoibion,“TheSubjectiveInflationExpectationsofHouseholdsandFirms: Measurement,Determinants,and Implications,”JournalofEconomicPerspectives,August2022,36(3),157–84. Wong,Arlene,“RefinancingandTheTransmissionofMonetaryPolicytoConsumption,”Working Papers2021-57,PrincetonUniversity.EconomicsDepartment.March2021. 33

Appendix For Online Publication Theappendixiscomposedofasectiononthetheory(SectionAppendixA.)andasectionon thesupplementaryempiricalresults(SectionAppendixB.). Appendix A. Second Derivatives of Consumer Utility Functions Inthisappendix,wederivesecondderivativesofconsumers’utilityfunction,whicharenecessary forourmaintheoreticalresults. Homeowner Recallthehomeowner’sutilityfunction: u (cid:161) Co(cid:162)+βu (cid:161) Co(cid:162)+(cid:161) 1+β(cid:162) v(H)=logCo+βlogCo+(cid:161) 1+β(cid:162) logH. 0 1 0 1 Bycombiningthebudgetconstraints,Co+H =w+ l andCo=w−mwithl =θP H andm=Rml, 0 P0 1 P1 0 wegettheutilityasafunctionofthechoicevariable,H,andthetwounknownfundamentals, (cid:169) Rm,Π(cid:170) : U (cid:161) H,Rm,Π(cid:162)=log(w−(1−θ)H)+βlog (cid:181) w−θ (cid:181) Rm(cid:182) H (cid:182) +(cid:161) 1+β(cid:162) logH. Π Theutilityfunctionwritteninlog-deviationsfromitssteady-stateis: (cid:179) (cid:180) uˆ (cid:161) h,im,π(cid:162)=U H¯eh,R¯meim ,Π¯e π =log (cid:179) w−(1−θ)H¯eh (cid:180) +βlog (cid:195) w−θ (cid:195) R¯meim(cid:33) H¯eh (cid:33) +(cid:161) 1+β(cid:162) logH¯eh. Π¯e π Theoptimalhousingchoice,h,satisfiesthefollowingfirst-ordercondition: (1−θ)H¯eh +β θ (cid:179) R¯ Π m ¯e e π im(cid:180) H¯eh =(cid:161) 1+β(cid:162) . w−(1−θ)H¯eh w−θ (cid:179) R¯meim(cid:180) H¯eh Π¯eπ Noticethatusingthebudgetconstraints,wegetthesteady-staterealmortgagerate, R¯m = 1−θ . Π¯ θ Then,thesteady-statehousingchoiceisH¯ = w . 2(1−θ) Thesecond-orderapproximationofuˆ withrespecttohousingevaluatedatthesteady-state 34

is: (1−θ)H¯ (cid:181) w (cid:182) θ (cid:179) R¯ Π¯ m (cid:180) H¯  w  uˆ 11 =− w−(1−θ)H¯ w−(1−θ)H¯ −β w−θ (cid:179) R¯m (cid:180) H¯  w−θ (cid:179) R¯m (cid:180) H¯  Π¯ Π¯ =−2 (cid:161) 1+β(cid:162) . Thesecond-orderapproximationofuˆ withrespecttohousingandmortgagerateevaluatedat thesteady-stateis: (cid:179) (cid:180) θ R¯m H¯ Π¯ uˆ =−β w =−2β 12 (cid:179) w−θ (cid:179) R¯m (cid:180) H¯ (cid:180)2 Π¯ Similarly,thesecond-orderapproximationofuˆ withrespecttohousingandinflationevaluated atthesteady-stateis: (cid:179) (cid:180) θ R¯m H¯w Π¯ uˆ =β =2β. 13 (cid:179) w−θ (cid:179) R¯m (cid:180) H¯ (cid:180)2 Π¯ Thisimpliesthatuˆ =−uˆ andtheproductofthetwowillbeanegativenumber. 12 13 Renter Recalltherenter’sutilityfunction: u (cid:161) Cr(cid:162)+βu (cid:161) Cr(cid:162)=logCr +βlogCr. 0 1 0 1 By combining the budget constraints, Cr + B = w˜ and Cr = w˜ +R B , we get the utility as a 0 P0 1 P1 functionofthechoicevariable,Cr,andthetwounknownfundamentals,{R,Π}: 0 (cid:181) (cid:182) U (cid:161) Cr,R,Π(cid:162)=log (cid:161) Cr(cid:162)+βlog w˜ + R (cid:161) w˜ −Cr(cid:162) . 0 0 Π 0 Theutilityfunctionwritteninlog-deviationsfromitssteady-stateis: (cid:179) (cid:180) uˇ (cid:161) cr,i,π(cid:162)=U C¯rec 0 r ,R¯ei,Π¯e π 0 0 (cid:179) (cid:180) (cid:181) R¯ei (cid:179) (cid:180) (cid:182) =log C¯rec 0 r +βlog w˜ + w˜ −C¯rec 0 r . 0 Π¯e π 0 Theoptimalhousingchoice,cr,satisfiesthefollowingfirst-ordercondition: 0 1=β Π R ¯ ¯ e e π i C¯ 0 rec 0 r w˜ + Π R ¯ ¯ e e π i (cid:161) w˜ −C¯ 0 rec 0 r(cid:162) 35

Noticethatusingthebudgetconstraints,wegetthesteady-stateconsumption,C¯r =w˜. Then, thesteady-statehousingchoiceis R¯ = 1. Π¯ β The2nd-orderapproximationofuˆ withrespecttohousingevaluatedatthesteady-stateis:   uˇ 11 =−β w˜ +R¯ Π R ¯ ¯ (cid:161) C w ¯ ˜ 0 r −C¯r (cid:162)  w˜ + w˜ R¯ + (cid:161) w˜ Π R ¯ ¯ − w˜ C¯r (cid:162)  =− 1+ β β Π¯ 0 Π¯ 0 The2nd-orderapproximationofuˆ withrespecttohousingandmortgagerateevaluatedatthe steady-stateis:   uˇ 12 =−β w˜ +R¯ Π R ¯ ¯ (cid:161) C w ¯ ˜ 0 r −C¯r (cid:162)  w˜ +R¯(cid:161) w w ˜ ˜ −C¯r (cid:162)  =−1 Π¯ 0 Π¯ 0 Similarly,thesecond-orderapproximationofuˆ withrespecttohousingandinflationevaluated atthesteady-stateis:   uˇ 13 =β w˜ +R¯ Π R ¯ ¯ (cid:161) C w ¯ ˜ 0 r −C¯r (cid:162)  w˜ +R¯(cid:161) w w ˜ ˜ −C¯r (cid:162)  =1. Π¯ 0 Π¯ 0 Thisimpliesthatuˆ =−uˆ andtheproductofthetwowillbeanegativenumber. 12 13 Appendix B. Supplementary Empirical Results AppendixTableB.1: HOMEONWERSHIP RATES AND DIFFERENCES IN INFLATION EXPECTATIONS ACROSS HOMEOWNERS AND RENTERS (1)MSC (2)SCE Homeownershiprate 75% 74% Differencesininflationexpectations:Ehomeowner −Erenter −1-yearaheadinflationexpectations -0.342 ∗∗∗ -0.282 ∗∗∗ (0.025) (0.033) −3-yearaheadinflationexpectations -0.249 ∗∗∗ (0.034) −5-yearaheadinflationexpectations -0.458 ∗∗∗ (0.021) Notes:Thistablereportshomeownershipratesandtheaveragedifferencesininflationexpectationsbetween homeownersandrentersfromtheMichiganSurveysofConsumers(Column(1))andtheNYFedSurveyof ∗∗∗ ∗∗ ∗ ConsumerExpectations(Column(2)).Robuststandarderrorsarereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. 36

AppendixTableB.2: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ UNEMPLOYMENT EXPECTATIONS TO CHANGES IN INTEREST RATES Interactions (1)∆R˜Mort (2)∆R˜Mort (3)∆R˜T-bond t,brw t,LF t,LF PanelA.Unemployment:Improve Renter(α ) 1.090 ∗∗∗ 1.076 ∗∗∗ 1.071 ∗∗∗ 1 (0.027) (0.025) (0.025) Homeowner×∆R (β ) 0.788 ∗∗∗ 0.820 ∗∗∗ 0.904 ∗∗∗ t 1 (0.021) (0.021) (0.022) Renter×∆R (β ) 0.788 ∗∗∗ 0.857 ∗∗∗ 0.912 ∗∗ t 2 (0.036) (0.035) (0.037) PanelB.Unemployment:Worsen Renter(α ) 1.057 ∗∗ 1.053 ∗∗ 1.054 ∗∗ 1 (0.027) (0.025) (0.026) Homeowner×∆R (β ) 1.039 1.049 ∗ 1.022 t 1 (0.028) (0.026) (0.025) Renter×∆R (β ) 1.023 1.022 0.902 t 2 (0.048) (0.044) (0.019) Numberofobs. 51764 55641 55641 R2 0.0024 0.0021 0.0014 Notes:ThistablereportsthelogitregressionresultsfromEquation(3.3).Dependentvariablesarethelogofthe probabilitythatunemploymentratewillbeimproved(PanelA)orworsened(PanelB)inthenextsixmonths relativetotheprobabilitythatunemploymentratewillbeunchangedinthenextsixmonths. "Homeowner" and"Renter"indicatedummiesforhomeownerandrenterrespectively.∆R referstothesix-monthchangein t interestrate.Weusethe30-yearmortgageratepredictedbytheunifiedshockfromBuetal.(2021)inColumn(1), the30-yearmortgageratepredictedbytheLSAPandforwardguidanceshocksfromSwanson(2021)inColumn (2),andthe30-yearTreasuryBondratepredictedbytheLSAPandforwardguidanceshocksfromSwanson (2021)inColumn(3).Thecoefficientsarereportedintermsofrelativeriskratios.Wecontrolfortheobserved surveyrespondents’characteristics,includinggender,education,birthcohort,andthelevelofincome.Robust ∗∗∗ ∗∗ ∗ standarderrorsarereportedintheparenthesis. , , denotesstatisticalsignificanceat1%,5%,and10% levelsrespectively. 37

AppendixTableB.3: ROBUSTNESS CHECK ON SENSITIVITY OF REVISIONS IN HOMEOWNERS’ INFLA- TION EXPECTATIONS TO INTEREST-RATE CHANGES BY CONTROLLING FOR HOUSE PRICE EXPECTA- TIONS OR OTHER MACRO VARIABLES Interactions Baseline (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.Controllingforhousepriceexpectations ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Housepriceexpectations -0.030 -0.030 -0.029 -0.030 (0.006) (0.006) (0.006) (0.006) ∆R -0.489 ∗∗∗ 0.431 ∗∗∗ -0.262 ∗∗∗ t (0.097) (0.071) (0.085) Numberofobs. 14071 14071 14071 14071 R2 0.0129 0.0146 0.0155 0.0135 PanelB.Controllingforconcurrentinflationrate ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Inflationrate 0.338 0.335 0.333 0.341 (0.037) (0.037) (0.039) (0.037) Homeowner×∆R (β ) -0.233 ∗∗∗ 0.038 -0.288 ∗∗∗ t 1 (0.062) (0.047) (0.062) Renter×∆R (β ) -0.163 -0.052 -0.316 ∗∗ t 2 (0.114) (0.088) (0.115) Numberofobs. 50834 50834 50834 50834 R2 0.0155 0.0158 0.0155 0.0157 Notes:Thistablereportstheestimatesofβ fromEquation(3.1)bycontrollingforhousepriceexpectationsor 1 othermacroeconomicvariables.Dependentvariablesarethesix-monthchangeintheMSC’s12-monthahead inflationexpectations.PanelAincludeshousepriceexpectationsoverthenextyearasanadditionalexplanatory variable.PanelBincludestheconcurrentinflationrateoverthesix-monthasanadditionalexplanatoryvariable. ∗∗∗ ∗∗ ∗ Robuststandarderrorsarereportedintheparenthesis. , , denotesstatisticalsignificanceat1%,5%,and 10%levelsrespectively. 38

AppendixTableB.4: ROBUSTNESS CHECK ON SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO INTEREST-RATE CHANGES OF THE PAST 3 MONTHS Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.MichiganSurveyofConsumers(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.339 ∗∗∗ 0.203 ∗∗∗ -0.316 ∗∗∗ t 1 (0.085) (0.042) (0.083) Renter×∆R (β ) -0.146 0.038 -0.367 ∗∗ t 2 (0.096) (0.125) (0.143) Numberofobs. 50834 50834 50834 R2 0.0140 0.0138 0.0141 PanelB.MichiganSurveyofConsumers(5-yearaheadinflationexpectations) Homeowner×∆R (β ) 0.013 0.071 -0.057 t 1 (0.071) (0.035) (0.070) Renter×∆R (β ) -0.08 -0.077 -0.172 t 2 (0.119) (0.105) (0.121) Numberofobs. 48842 48842 48842 R2 0.0107 0.0107 0.0107 Notes: ThistablereportstheregressionresultsfromEquation(3.1). Dependentvariablesarethesix-month changeintheMSC’s12-monthaheadinflationexpectations(PanelA)andthesix-monthchangeintheMSC’s 5-yearaheadinflationexpectations(PanelB)."Homeowner"and"Renter"indicatedummiesforhomeowner andrenterrespectively.∆R referstothechangeininterestrateoverthepastthreemonths.Weusethe30-year t mortgagerateinColumn(1),theFederalFundsRateinColumn(2),andthe30-yearTreasuryBondratein Column(3).Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birth ∗∗∗ ∗∗ ∗ cohort,andthelevelofincome. Robuststandarderrorsarereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. 39

AppendixTableB.5: ROBUSTNESS CHECK ON SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO INTEREST-RATE CHANGES OF THE PAST 9 MONTHS Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.MichiganSurveyofConsumers(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.202 ∗∗∗ 0.068 ∗∗ -0.202 ∗∗∗ t 1 (0.048) (0.030) (0.083) Renter×∆R (β ) -0.181 ∗∗ -0.045 -0.213 ∗∗ t 2 (0.078) (0.052) (0.143) Numberofobs. 50834 50834 50834 R2 0.0141 0.0138 0.0141 PanelB.MichiganSurveyofConsumers(5-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.053 0.026 -0.085 ∗∗ t 1 (0.041) (0.025) (0.042) Renter×∆R (β ) -0.051 -0.067 -0.019 t 2 (0.066) (0.044) (0.121) Numberofobs. 48842 48842 48842 R2 0.0107 0.0107 0.0107 Notes: ThistablereportstheregressionresultsfromEquation(3.1). Dependentvariablesarethesix-month changeintheMSC’s12-monthaheadinflationexpectations(PanelA)andthesix-monthchangeintheMSC’s 5-yearaheadinflationexpectations(PanelB)."Homeowner"and"Renter"indicatedummiesforhomeowner andrenterrespectively.∆R referstothechangeininterestrateoverthepastninemonths.Weusethe30-year t mortgagerateinColumn(1),theFederalFundsRateinColumn(2),andthe30-yearTreasuryBondratein Column(3).Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birth ∗∗∗ ∗∗ ∗ cohort,andthelevelofincome. Robuststandarderrorsarereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. 40

AppendixTableB.6: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ EXPECTATIONS ON FUTURE BUSINESS CONDITIONS TO CHANGES IN INTEREST RATES Interactions (1)∆RMort (2)∆RFFR (3)∆RT-bond t t t PanelA.FutureBusinessConditions:Improve Renter(α ) 1.040 ∗ 1.0045 ∗ 1.044 ∗ 1 (0.025) (0.025) (0.025) Homeowner×∆R (β ) 0.878 ∗∗∗ 0.881 ∗∗∗ 0.916 ∗∗∗ t 1 (0.019) (0.013) (0.020) Renter×∆R (β ) 0.875 ∗∗∗ 0.897 ∗∗∗ 0.946 t 2 (0.031) (0.022) (0.034) PanelB.FutureBusinessConditions:Worsen Renter(α ) 1.070 ∗∗∗ 1.070 ∗∗∗ 1.070 ∗∗∗ 1 (0.026) (0.025) (0.026) Homeowner×∆R (β ) 1.000 1.018 1.034 t 1 (0.022) (0.016) (0.023) Renter×∆R (β ) 1.010 1.034 1.036 t 2 (0.036) (0.027) (0.038) Numberofobs. 58791 58791 58791 R2 0.0017 0.0022 0.0014 Notes: ThistablereportsthelogitregressionresultsfromEquation(3.3). Dependentvariablesarethelogof theprobabilitythatfuturebusinessconditionswillbeimproved(PanelA)orworsened(PanelB)inthenext sixmonthsrelativetotheprobabilitythatfuturebusinessconditionswillbeunchangedinthenextsixmonths. "Homeowner"and"Renter"indicatedummiesforhomeownerandrenterrespectively. ∆R referstothesixt monthchangeininterestrate. Weusethe30-yearmortgagerateinColumn(1),theFederalFundsRatein Column(2),andthe30-yearTreasuryBondrateinColumn(3).Thecoefficientsarereportedintermsofrelative riskratios.Wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birth ∗∗∗ ∗∗ ∗ cohort,andthelevelofincome. Robuststandarderrorsarereportedintheparenthesis. , , denotes statisticalsignificanceat1%,5%,and10%levelsrespectively. 41

AppendixTableB.7: RESPONSES OF INTEREST RATES TO MONETARY POLICY SHOCKS Dependentvariable (1)∆RMort (2)∆RT-bond (3)∆RMort (4)∆RT-bond t t t t FederalFunds 0.009 -0.004 -0.069 0.048 (0.009) (0.007) (0.043) (0.043) ForwardGuidance 0.024*** 0.007 0.031*** 0.024** (0.008) (0.005) (0.011) (0.009) LSAP 0.027*** 0.026** 0.095** 0.028* (0.017) (0.012) (0.039) (0.015) Sampleperiod 1991–2019 1991–2019 2013–2019 2013–2019 Numberofobs. 1416 1464 330 342 R2 0.0273 0.0112 0.1492 0.0373 Notes:ThistablereportstheregressionresultswhenweusethemonetarypolicyshocksidentifiedinSwanson (2021).Dependentvariablesaretheweeklychangeinthe30-yearmortgagerateinColumns(1)and(3)andthe weeklychangeinthe30-yearTreasuryBondrateinColumns(2)and(4).ThesampleperiodisfromApril1991to July2019fortheestimatesreportedinColumns(1)and(2).ThesampleperiodfortheestimatesinColumns ∗∗∗ ∗∗ ∗ (3)and(4)isfrom2013onward.Newey-Westrobuststandarderrorsarereportedintheparenthesis. , , denotesstatisticalsignificanceat1%,5%,and10%levelsrespectively. 42

Appendix Table B.8: SENSITIVITY OF REVISIONS IN HOMEOWNERS AND RENTERS’ INFLATION EXPECTATIONS TO CHANGES IN INTEREST RATES PREDICTED BY MONETARY POLICY SHOCK Interactions (1)∆R˜Mort (2)∆R˜T-bond t,LF t,LF PanelA.MichiganSurveyofConsumers(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -4.523 ∗∗∗ -4.341 ∗∗∗ t 1 (0.606) (1.027) Renter×∆R (β ) -3.500 ∗∗∗ -2.988 t 2 (1.212) (2.156) Numberofobs. 47008 47008 R2 0.0143 0.0134 PanelB.MichiganSurveyofConsumers(5-yearaheadinflationexpectations) Homeowner×∆R (β ) -0.417 -0.455 t 1 (0.505) (0.723) Renter×∆R (β ) 0.273 2.249 t 2 (0.969) (1.768) Numberofobs. 45622 45622 R2 0.0105 0.0106 PanelC.NYFedSurveyofConsumerExpectations(1-yearaheadinflationexpectations) Homeowner×∆R (β ) -1.329 -2.641 t 1 (1.058) (1.756) Renter×∆R (β ) -0.730 2.056 t 2 (2.248) (3.421) Numberofobs. 37258 37258 R2 0.0000 0.0000 PanelD.NYFedSurveyofConsumerExpectations(3-yearaheadinflationexpectations) Homeowner×∆R (β ) -2.170 ∗∗ -4.743 ∗∗∗ t 1 (1.028) (1.672) Renter×∆R (β ) -0.889 -0.609 t 2 (2.235) (3.426) Numberofobs. 37337 37337 R2 0.0001 0.0003 Notes: ThistablereportstheregressionresultsfromEquations(3.1)and(3.2). Dependentvariablesarethe six-month changes in the MSC’s 1-year ahead inflation expectations (Panel A), the 5-year ahead inflation expectations(PanelB),theSCE’s1-yearaheadinflationexpectations(PanelC),andtheSCE’s3-yearahead inflationexpectations(PanelD)."Homeowner"and"Renter"indicatedummiesforhomeownerandrenter respectively.∆R referstothesix-monthchangeininterestrate.Weusethe30-yearmortgageratepredicted t bytheLSAPandforwardguidanceshocksfromSwanson(2021)inColumn(1),andthe30-yearTreasuryBond ratepredictedbytheLSAPandforwardguidanceshocksfromSwanson(2021)inColumn(2).InPanelAand B,wecontrolfortheobservedsurveyrespondents’characteristics,includinggender,education,birthcohort, andthelevelofincome.Robuststandarderrorsarereportedintheparenthesis.InPanelCandD,weemploy theindividualfixedeffectstocontroltheindividualheterogeneity. Robuststandarderrors(clusteredatthe ∗∗∗ ∗∗ ∗ individual-level)arereportedintheparenthesis. , , denotesstatisticalsignificanceat1%,5%,and10% levelsrespectively. 43

Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Improve Unchanged Worsen .45 .4 .35 Homeowner Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate AppendixFigureB.1: ROBUSTNESS CHECK ON MARGINAL PROBABILITY OF CHANGES IN HOUSE- HOLD EXPECTATIONS OF UNEMPLOYMENT (INTEREST-RATE CHANGES OF THE PAST 3 MONTHS) Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsofunemployment tochangesininterestratesoverthepastthreemonths.Theinterestratesconsideredfromtoptobottomsare 30-yearmortgagerates,federalfundsrate,and30-yearT-bondrates.Theresultsarecalculatedbasedonthe estimatesofthelogitregressionresultsfromEquation(3.3).Wecontrolfortheobservedsurveyrespondents’ characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareasrepresent95% confidenceintervals. Source:Authors’calculation. 44

Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate AppendixFigureB.2: ROBUSTNESS CHECK ON MARGINAL PROBABILITY OF CHANGES IN HOUSE- HOLD EXPECTATIONS OF UNEMPLOYMENT (INTEREST-RATE CHANGES OF THE PAST 9 MONTHS) Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsofunemployment tochangesininterestratesoverthepastninemonths. Theinterestratesconsideredfromtoptobottomare 30-yearmortgagerates,federalfundsrates,and30-yearT-bondrates.Theresultsarecalculatedbasedonthe estimatesofthelogitregressionresultsfromEquation(3.3).Wecontrolfortheobservedsurveyrespondents’ characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareasrepresent95% confidenceintervals. Source:Authors’calculation. 45

Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate AppendixFigureB.3: ROBUSTNESS CHECK ON MARGINAL PROBABILITY OF HOUSEHOLD EXPECTA- TIONS OF INTEREST RATES (INTEREST-RATE CHANGES OF THE PAST 3 MONTHS) Notes: Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsof1-year-ahead interestratestochangesininterestratesoverthepastthreemonths. Theinterestratesconsideredfromtop tobottomare30-yearmortgagerates,federalfundsrates,and30-yearT-bondrates.Theresultsarecalculated basedontheestimatesofthelogitregressionresultsfromEquation(3.3).Wecontrolfortheobservedsurvey respondents’characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareas represent95%confidenceintervals. Source:Authors’calculation. 46

Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Go up Same Go down .8 .6 Homeowner .4 Renter .2 0 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate AppendixFigureB.4: ROBUSTNESS CHECK ON MARGINAL PROBABILITY OF HOUSEHOLD EXPECTA- TIONS OF INTEREST RATES (INTEREST-RATE CHANGES OF THE PAST 9 MONTHS) Notes: Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsof1-year-ahead interestratestochangesininterestratesoverthepastninemonths.Theinterestratesconsideredfromtopto bottomare30-yearmortgagerates,federalfundsrates,and30-yearT-bondrates. Theresultsarecalculated basedontheestimatesofthelogitregressionresultsfromEquation(3.3).Wecontrolfortheobservedsurvey respondents’characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareas represent95%confidenceintervals. Source:Authors’calculation. 47

Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in mortgage rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in fed funds rate Improve Unchanged Worsen .45 .4 Homeowner .35 Renter .3 .25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 -1 -.75 -.5 -.25 0 .25 .5 .75 1 ytilibaborP Change in T-bond rate AppendixFigureB.5: MARGINAL PROBABILITY OF CHANGES IN HOUSEHOLD EXPECTATIONS OF FUTURE BUSINESS CONDITIONS Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsoffuturebusiness conditionstochangesininterestratesoverthepastsixmonths.Theinterestratesconsideredfromtoptobottom are30-yearmortgagerates,federalfundsrates,and30-yearT-bondrates.Theresultsarecalculatedbasedonthe estimatesofthelogitregressionresultsfromEquation(3.3).Wecontrolfortheobservedsurveyrespondents’ characteristics,includinggender,education,birthcohort,andthelevelofincome.Shadedareasrepresent95% confidenceintervals. Source:Authors’calculation. 48

Improve Unchanged Worsen .4 .35 Homeowner Renter .3 -.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03 ytilibaborP Change in mortgage rate explained by LSAP & FG Improve Unchanged Worsen .4 .35 Homeowner Renter .3 -.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03-.03 -.02 -.01 0 .01 .02 .03 ytilibaborP Change in T-bond rate explained by LSAP & FG AppendixFigureB.6: MARGINAL PROBABILITY OF CHANGES IN HOUSEHOLD EXPECTATIONS OF UNEMPLOYMENT Notes:Thisfigurereportsthemarginalprobabilitiesofchangesinhouseholdexpectationsofunemploymentto changesininterestratesexplainedbymonetarypolicyshocks.Explanatoryvariablesconsideredinthetopand bottomarethe30-yearmortgageratepredictedbytheLSAPandforwardguidanceshocksfromSwanson(2021), andthe30-yearTreasuryBondratepredictedbytheLSAPandforwardguidanceshocksfromSwanson(2021). TheresultsarecalculatedbasedontheestimatesofthelogitregressionresultsfromEquation(3.3)asreported inColumn(2)and(3)inAppendixTableB.2.Wecontrolfortheobservedsurveyrespondents’characteristics, includinggender,education,birthcohort,andthelevelofincome. Shadedareasrepresent95%confidence intervals. Source:Authors’calculation. 49