U.S. Monetary Policy Surprises and Mortgage Rates

Transcription

1 U.S. Moneary Policy Surprises and Morgage Raes Pisun Xu Reiman School of Finance, Universiy of enver, enver, CO or Yufeng Han The Business School, Universiy of Colorado enver, enver, CO or Jian Yang The Business School, Universiy of Colorado enver, enver, CO or Absrac This paper examines how he U.S. moneary policy surprises impac he morgage raes in he naion and across five regions from 1990 o Regression analysis based on boosrapping shows ha surprises in he arge federal funds rae (he arge facor) have a significanly posiive impac on he 1-year adusable-rae morgage (ARM) rae wihin he week of he FOMC announcemens and he posiive impac lass up o one week afer he announcemens. Surprises in he fuure direcion of he Federal Reserve moneary policy (he pah facor) have significanly posiive impacs on boh he 1-year ARM rae and he 30-year fixed morgage raes in he firs week afer he announcemen. Furhermore, he responses of morgage raes are asymmeric and affeced by he size of moneary policy surprises, he sage of he business cycle and wheher he moneary policy is ighening or loosening. There also exiss heerogeneiy in he morgage rae pass-hrough process across regions and moneary policy surprises have differenial impacs on he regional morgage raes. The cross-region variaions are mainly correlaed wih he regional housing marke condiions, such as home vacancy and renal vacancy raes. Keywords: FOMC saemens, asymmery, morgage raes, wo-facor empirical specificaion JEL Classificaions: G14, G21, E44, E52 *Corresponding auhor: Jian Yang, The Business School, PO Box , Universiy of Colorado enver, enver, CO Jian.Yang@ucdenver.edu. Tel: (303) ; Fax: (303)

2 U.S. Moneary Policy Surprises and Morgage Raes The ransmission mechanism of moneary policy has been an imporan opic wih significan policy implicaions. The main channel for such ransmission is he ineres rae channel, in which he pass-hrough of he key policy rae o reail bank ineres raes is a fundamenal elemen. I is generally believed ha changes in he policy rae and money marke raes can be ransmied ino he reail bank ineres rae (e.g., deposi rae and loan rae), which can ulimaely influence he opporuniy coss of consumpion and invesmen and hence he aggregae demand and oupu. Therefore, an efficien pass-hrough process is criical o he successful implemenaion of moneary policy. In his paper, we invesigae he mechanism of pass-hrough o a key reail ineres rae, he morgage rae. This issue has become paricularly imporan in he conex of recen financial crisis. The U.S. Federal Reserve Bank aggressively cu he ineres rae in 2007 and 2008, which aimed a dramaically reducing morgage raes hrough an assumed raher complee pass-hrough and hus a rescue for he roubled U.S. housing marke. Bu an observed widening spread beween he policy rae and reail ineres raes suggess ha changes in he policy rae migh no be fully ransmied ino reail loan raes. Thus, a horough undersanding of he morgage rae pass-hrough mechanism is crucial for policy making. The pass-hrough process of morgage raes is closely relaed o he expecaions heory of he erm srucure of ineres raes. In his heory, long-erm raes are averages of curren shorerm raes and expeced fuure shor-erm raes. Moneary policy affecs long-erm raes o he exen ha i influences curren and expeced shor-erm raes in he fuure. Changes in longerm raes (e.g., long-erm bond raes) affec he banks marginal coss of loan funds. Thus, as poined ou in Sellon (2002, p.7), changes in moneary policy affec he supply of bank reserves 1

3 and he cos of bank funds, and banks end o pass on hese cos changes o loan raes. Goukasian and Mabouri (2010) also find ha morgage and banking indusries have he sronges responses o moneary policy acions. Furhermore, moneary policy acion migh affec he morgage raes hrough is impac on secondary morgage marke. I is a common pracice o resell he morgage loans in he secondary marke, which are hen securiized and issued o invesors as morgage-backed securiies. Sirmans and Benamin (1990) find ha he loan raes on fixed-rae morgages in he primary marke move in a one-o-one relaionship wih secondary marke yield. In his conex, changes in moneary policy may affec long-erm bond raes and hus he secondary morgage marke yield, which can be ransmied ino he morgage raes. This paper explores he pass-hrough mechanism of morgage raes by examining he impacs of unanicipaed moneary policy changes around he Federal Open Marke Commiee (FOMC) announcemens on he 1-year adusable-rae morgage (ARM) raes and he 30-year fixed morgage raes in he naion and across five regions. Following Gürkaynak, Sack and Swanson (2005), we use wo facors o measure moneary policy surprises. The arge facor is designed o measure he marke s surprise regarding changes in he curren federal funds arge rae while he pah facor is designed o measure he surprises in he FOMC s possible fuure policy as refleced in he wording of he FOMC saemens. The wo-facor empirical specificaion explores he possibiliy ha boh surprises in he federal funds arge rae and in he fuure direcions of ineres rae changes affec morgage raes. We find new evidence ha boh he arge and pah facors have significan impacs on he morgage raes. The arge facor has a significanly posiive impac on he 1-year ARM rae wihin he week of he FOMC announcemens and such impac lass up o one week afer he announcemens. The pah facor has significanly posiive impacs on boh he 1-year ARM rae and he 30-year fixed morgage raes in he firs week afer he announcemens. The resuls 2

4 sugges ha he shor-erm adusable morgage raes experience boh he higher degree of passhrough and he faser adusmen speed han long-erm fixed morgage raes. The resuls also indicae ha FOMC saemens likely exer heir effecs on morgage markes hrough heir influence on financial marke expecaions of fuure policy acions. We provide addiional new evidence ha morgage raes have asymmeric responses o moneary policy surprises. The speed of he adusmen of morgage raes in response o moneary policy surprises is affeced by he size of moneary policy surprises, he sage of business cycle and wheher he moneary policy is ighening or loosening. Furhermore, we find ha here exiss heerogeneiy in how he morgage raes reac o moneary policy surprises across five Freddie Mac regions. The resuls also show ha he crossregion variaions are mainly relaed o he regional housing marke condiions, such as he home vacancy and renal vacancy raes. Our sudy makes several conribuions o he exising lieraure. Firs, his paper provides new insighs ino he morgage rae pass-hrough mechanism in he U.S. marke. A number of sudies explore he reail ineres rae pass-hrough mechanism in differen counries (e.g., Scholnick, 1996; Kleimeier and Sander, 2006; Liu, Margariis and Tourani-Rad, 2008). They find ha he key policy rae changes may no be fully and insananeously passed hrough o reail raes. However, he pass-hrough of moneary policy acions is sill relaively underexplored in he U.S. marke. Mos previous lieraure on U.S. moneary ransmission mechanism assumes immediae and complee pass-hrough of changes in he inended federal funds rae o reail ineres raes (e.g., Bernanke and Gerler, 1995; Kashyap and Sein, 2000). There are few sudies ha have explored he reail ineres rae pass-hrough mechanism. Hannan and Berger (1991) and Scholnick (1999) have examined he shor-erm ineres rae pass-hrough process. They find ha he shor-erm consumer ineres raes (e.g., deposi rae) respond asymmerically 3

5 o changes in he money marke rae (e.g., U.S. Treasury bills rae). There are even fewer sudies ha invesigae he morgage rae pass-hrough mechanism, wih he noable excepion of Hamilon (2008). While Hamilon (2008) mainly focuses on he 30-year fixed morgage rae, his paper examines he adusmen process of boh he ARM raes and he fixed morgage raes and hus provides new evidence on he pass-hrough mechanism of differen ypes of morgage raes. Furhermore, his paper examines no only he magniude of pass-hrough bu also he persisence of pass-hrough, which sheds new ligh on he moneary ransmission mechanism. Second, his paper invesigaes how morgage raes reac no only o surprises in he federal funds arge rae (he arge facor) bu also o surprises in he fed s fuure moneary policy direcions (he pah facor). Over he pas decade, Federal Reserve moneary policy decisionmaking has moved significanly in he direcion of greaer ransparency. As a consequence, Federal Open Marke Commiee (FOMC) saemens disseminae addiional informaion on he fuure pah of moneary policy beyond he informaion on he arge federal funds rae. 1 This paper exends he recen cenral bank communicaion lieraure on oher U.S. financial markes (Bernanke, Reinhar and Sack, 2004; Gürkaynak, Sack and Swanson, 2005; Wang, Yang and Wu, 2006; Wang, Yang and Simpson, 2008; Wongswan, 2009; Hausman and Wongswan, 2011; Xu and Yang, 2011) and uses wo facors o beer measure U.S. moneary policy surprises. We find evidence ha morgage raes reac o boh he arge and pah facors in his sudy. Therefore, by using he wo-facor empirical specificaion, we would no underesimae he magniude of morgage rae pass-hrough. Our findings indicae ha FOMC saemens likely exer heir effecs on morgage markes hrough heir influence on financial marke expecaions of fuure policy acions, which has imporan implicaions on he conduc of moneary policy. Third, he paper is among he firs o examine he asymmery of morgage rae passhrough process in response o U.S. moneary policy surprises. As in he previous lieraure (e.g., 4

6 Hannan and Berger, 1991; Scholnick, 1999; Fueres, Heffernan and Kaloychou, 2010), we allow for he asymmery associaed wih he direcion of rae movemens. In addiion, we explore hree oher ypes of asymmeries associaed wih he sign of he surprises, he size of he surprises and he sage of business cycle, which has no ye been much explored in his line of he lieraure. 2 Finally, exending Hamilon (2008), his paper explores wheher he morgage rae passhrough mechanism differs across regions in he U.S. marke. We provide new evidence on how moneary policy surprises impac he regional morgage raes. Furhermore, we examine which regional economic and housing facors have significan impacs on he heerogeneiy. The res of he paper is as follows. Secion 2 discusses he mehodology. Secion 3 describes he daa used. Secion 4 presens he analysis based on he naional morgage raes. Secion 5 presens he analysis based on he morgage raes in five Freddie Mac regions. Secion 6 concludes. Mehodology The Baseline Model Following he previous lieraure (e.g., Wang, Yang and Wu, 2006; Wang, Yang and Simpson, 2008; Wongswan, 2009; Hausman and Wongswan, 2011; Xu and Yang, 2011), we use he even sudy mehodology o examine he change in morgage raes around he FOMC announcemens. We furher refine he mehodology in he previous lieraure in wo aspecs. Firs, we employ he Seemingly Unrelaed Regressions (SUR) which, compared o he Ordinary Leas Squares (OLS) esimaor commonly used in earlier sudies, can improve efficiency by accouning for he cross-region correlaion and is necessary for cross-equaion hypohesis esing. Second, we use he boosrapping approach in he SUR framework o address he poenial errors-in-variables 5

7 problem due o he fac ha independen variables use in his sudy are all esimaed raher han observed variables. Specifically, for all model esimaions below, we draw random samples wih replacemen repeaedly for 1000 imes from he sample daa, and hus obain he robus sandard errors. The basic wo-facor model is se up as follows: M i i, i, i, S & PS & Pne, i, RS & P500, i, 1,2,,6 (1) i, where M i, is he change in he morgage rae i, of he naion or he five Freddie Mac regions, over one-week window around he FOMC announcemen. Closely following Gürkaynak, Sack and Swanson (2005), we use wo facors o measure moneary policy surprises. is he arge facor and is he pah facor on day. The arge facor is designed o measure he marke s surprise regarding a change in he curren federal funds arge rae, while he pah facor is designed o represen he surprises in he FOMC s possible fuure policy (and is inerpreaion of economic oulook) as refleced in he wording of he FOMC saemens. The wo-facor empirical specificaion explores he possibiliy ha surprises in boh he federal funds arge rae and he fuure direcions of ineres rae changes affec morgage raes. A significan coefficien of or i, would sugges a direc impac of i, U.S. moneary policy changes on U.S. morgage raes. We include several addiional variables o conrol for he impac of non-fomc news evens. S&P ne, is he esimaed residual aken from running he regression of he S&P500 fuures reurn over one-day window around he FOMC announcemen on he arge and pah facors, which conrols for he impac of non-fomc news even on he day of he announcemens, following Wongswan (2006, 2009) and Hausman and Wongswan (2011). R S&P500,- is he lagged 6

8 S&P500 fuures reurn for he las six days prior o he FOMC announcemen o conrol for news of prior days lasing for more han a week, following Hamilon (2008). We furher examine how persisen morgage raes reac o he moneary policy surprises. We use he same specificaion as above bu replace he dependen variable wih weekly changes in morgage raes over he window of one-week and wo-weeks afer he FOMC announcemens. Asymmeric Effec Regressions The previous lieraure (e.g., Wang, Yang and Simpson, 2008) demonsraes he economeric imporance of allowing for asymmery in dealing wih he model misspecificaion problem. We herefore also explore he poenial asymmeric effecs of he moneary policy surprises on morgage raes relaed o he sign of surprises, he size of surprises, he direcion of rae movemens and he sage of he business cycle. Sign response. uring he sample period, here are 122 posiive surprises and 33 negaive surprises in he federal funds arge rae changes. To invesigae he asymmeric responses of morgage raes o he sign of moneary policy surprises, we esimae he following regression using SUR: M i, i i, i, i, S & P S i, i, & Pne, i, RS & P500, i, (2) where,,, and are he sign dummy variables. equals one when he arge facor is posiive; equals one when he arge facor is negaive; equals one wih posiive surprises in he pah facor, and equals one wih negaive surprises in he pah facor. Noe ha he model is free from he problem of dummy variable rap. 7

9 Size response. Anoher possibiliy of asymmery is ha he magniude of he responses depends on he size of he moneary policy surprises. We esimae he following SUR regression: M i, i b i, b s i, i, S& P s S b i, b s i, s & Pne, i, RS& P500, i, (3) where b s b s b,,, and are size dummy variables. is ermed as a big surprise dummy which equals one if he surprise in he arge facor is among he upper 50% in erms of absolue values of all surprises. s is ermed as a small surprise dummy which equals one if he surprise is among he lower 50%. b s and are defined similarly. irecion of rae movemens. The hird ype of asymmery we examine is he direcion of rae movemen. The responses of morgage raes can differ during expansionary or ighening periods of he moneary policy. uring he sample period, here are 19 rae increases and 34 rae decreases. We creae wo dummy variables. funds rae, and is esimaed: i ff equals one if here is an increase in he federal d ff equals one if here is a decline in he federal funds rae. The following model M i, i i i, i ff d i, d ff i, S & P S i i, i ff d i, d ff & Pne, i, RS & P500, i, (4) Noe ha in equaion (4), he same dummy variables are used for boh he arge and pah facors, as we canno meaningfully define separae dummy variables for he pah facor. Sage of he business cycle. Finally, we examine wheher he responses o moneary policy surprises depend on he sage of he business cycle. Following he previous lieraure (e.g., 8

10 Wang, Yang and Simpson, 2008), we divide he sample ino periods of economic expansion (from February 1990 o June 1990, from April 1991 o February 2001, and from ecember 2001 o Augus 2007) and periods of economic conracion (from July 1990 o March 1991, from March 2001 o November 2001, and from Sepember 2007 o Sepember 2008) based on he NBER business cycle daes. uring our sample period, here are 137 observaions in he economic expansion and 28 observaions in he economic recession. We creae wo dummy variables o capure he sage of he business cycle. e (or r ) equals one when he economy is in an expansion (or a recession) and zero oherwise. The following SUR model is esimaed: M i, i e i, e r i, r i, S& P S e i, e r i, & Pne, i, RS& P500, i, r (5) aa U.S. Moneary Policy Surprises aa The decision-making of he FOMC has moved significanly in he direcion of greaer ransparency in las fifeen years. The FOMC relied on he subsequen open-marke operaions o signal shifs in moneary policy before Since February 1994, he FOMC began releasing saemens ha accompanied changes in he federal funds arge rae, which were issued occasionally and explained he raionale for he policy acion. Afer May 1999, FOMC saemens were issued afer every meeing, and he FOMC sared o communicae o he public he fuure direcion of moneary policy by releasing is views abou he balance of risks faced by he economy going forward. The balance of risk assessmen would indicae wheher he risk for he economy over he foreseeable fuure was weighed oward economic weakness or heighened inflaion pressures, or if he risk was balanced. 9

11 As a consequence, he poenial impac of U.S. moneary policy has drawn much aenion. Kuner (2001) is perhaps he firs one o examine he impac of moneary policy surprises on he financial markes. He measures he moneary policy surprises based on changes in he federal funds fuures rae and finds ha unanicipaed changes in moneary policy affec he financial markes while anicipaed changes have lile effecs. Following Kuner (2001), a large number of sudies use a one-facor empirical specificaion, which focuses on he impac of he federal funds arge rae surprises. Gürkaynak, Sack and Swanson (2005) is among he firs o horoughly examine he informaional conens of he FOMC saemens and finds evidence ha wo facors are required o beer capure he unanicipaed changes in U.S. moneary policy. Specifically, he arge facor is designed o measure he marke s surprises regarding a change in he curren federal funds arge rae, while he pah facor is designed o represen he surprises in he FOMC s possible fuure policy (and is inerpreaion of economic oulook) as refleced in he wording of he FOMC saemens. The pah facor is shown o have a significan effec on financial markes in several earlier sudies (e.g., Wang, Yang and Wu, 2006; Wang, Yang and Simpson, 2008; Wongswan, 2009; Hausman and Wongswan, 2011; Xu and Yang, 2011). Similar o Gürkaynak, Sack and Swanson (2005), we use a wo-facor empirical specificaion in he analysis. The wo facors are consruced using principal componens of price changes in he federal funds fuures and Eurodollar fuure conracs around he FOMC announcemens. 3 The arge surprise is defined as he difference beween he announced federal funds arge rae and expecaions derived from he federal funds fuures conrac. I is compued from he changes in he curren-monh federal funds fuures conrac rae in a 30-min window around he FOMC announcemens (10 minues prior o and 20 minues afer he announcemen). Because federal funds fuures have a payou ha is based on he average effecive federal funds rae ha 10

12 prevails over he calendar monh specified in he conrac, he change in he federal funds fuures rae is adused for a facor ha is calculaed based on he number of days in he monh affeced by he change in he federal funds arge rae. For an FOMC announcemen a ime on day d of a monh wih days, he federal funds fuures rae en minues before he announcemen, ff -10, is basically a weighed average of he federal funds rae ha has prevailed so far in he monh (r 0 ) and he rae ha is expeced o prevail for he remainder of he monh (r 1 ), plus a erm of risk premium (ρ) ha may be presened in he conrac. d d ff 10 r0 E 10( r1 ) 10 By leading he above equaion o weny minues afer he FOMC announcemen (+20), differencing, and assuming a consan risk premium, he arge facor ( ) is calculaed as follows: d ( ff 20 ff 10) To consruc he pah facor, we ake he changes in five federal funds fuures conracs in he hiry-minue window around he FOMC announcemens (10 minues prior o and 20 minues afer he announcemens) o creae he marix X. The firs wo columns in X are he changes in he curren-monh and hree-monh-ahead federal funds fuures. The nex hree columns in X are he changes in prices of he second, hird, and fourh Eurodollar fuures conracs, which have 1.5, 2.5 and 3.5 quarers o expiraion on average. Afer normalizing each column of X o have zero mean and uni variance, we can exrac he wo principal componens, where he wo componens Z 1 and Z 2 are made orhogonal o each oher. Z 1 is furher normalized so ha i mirrors he above-menioned federal funds fuures surprise ff ff ) d (

13 one-for-one, and Z 2 is also scaled so ha i has he same magniude effec on he year-ahead Eurodollar fuures rae as Z 1 has on ha rae (abou 53 basis poins). Such decomposiion appears o work very well, and he correlaion coefficien beween he arge facor and he surprises in he federal funds arge rae is over 95%. 4 U.S. Morgage Raes Weekly U.S. morgage rae daa are based on he Freddie Mac's Primary Morgage Marke Survey (PMMS). Freddie Mac surveys lenders across he naion on he raes and poins for heir mos popular morgage producs. The survey is based on firs-lien prime convenional conforming morgages wih a loan-o-value of 80 percen. In addiion, he adusable-rae morgage (ARM) producs are indexed o he U.S. Treasury yields. The average raes and poins for each produc are repored weekly for he naion and five regions specified below. Freddie Mac has grouped he saes ino five regions based primarily on heir locaions and similariies in socioeconomic characerisics. The five regions are Norh Cenral (OH, IN, IL, MI, WI, MN, IA, N, S), Norheas (NY, NJ, PA, E, M, C, VA, WV, ME, NH, VT, MA, RI, CT), Souheas (NC, SC, TN, KY, GA, AL, FL, MS, PR, VI), Souhwes (TX, LA, NM, OK, AR, MO, KS, CO, NE, WY) and Wes (CA, AZ, NV, OR, WA, UT, I, MT, HI, AK, GU). In his paper, we use he 1-year reasury-indexed adusable-rae morgage and 30-year fixed-rae morgage series of he naion and he five Freddie Mac regions. We obain he Freddie Mac morgage rae daa from aasream. The weekly changes in morgage raes are calculaed by aking he difference in he morgage raes from he previous week o he week of he FOMC announcemen. 12

14 Regional Facors Following he previous lieraure on regional economy and business cycles (e.g., Carlino and efina, 1998), we consruc various proxies for he poenial facors ha may impac how morgage raes in differen regions respond o U.S. moneary policy changes differenly. We use seven proxies o measure he economic and housing marke condiions in each sae. The growh in personal income (Growh in Income) is he annual growh rae of per capia personal income. The per capia personal income is he oal personal income of he sae s residens divided by he populaion of he sae. The unemploymen rae (Unemploymen rae) is he number of unemployed as a percenage of he labor force. The home ownership rae (Home Ownership) is he number of owner-occupied housing unis divided by he number of occupied housing unis or households. The home vacancy rae (Home Vacancy) is he number of vacan unis for sale divided by he sum of he owner-occupied unis and he number of vacan unis ha are for sale only. The renal vacancy rae (Renal Vacancy) is he number of vacan unis for ren divided by he sum of he rener-occupied unis and he number of vacan unis for ren. The Change in Building Permis is he monhly change of he sae s housing unis auhorized building permis. The Change in Housing Sars is he monhly change of he sae s privaely owned housing sars. The building permis and housing sars daa can be used o creae nearerm forecass of new home sales and gauge he overall srengh of he housing indusry. We collec he economic and housing marke daa for each sae from he FRE (Federal Reserve Economic aa) daa of he Federal Reserve Bank of S. Louis and consruc proxies for each sae. We hen furher calculae he regional facors for each of he five Freddie Mac regions by aking he weighed average across he saes in ha region wih he weigh being he populaion of he sae divided by he oal populaion in he region. 13

15 Conrol Variables In addiion, we conrol for oher non-fomc news or shocks ha migh affec how morgage raes reac o U.S. moneary policy surprises. Following (Wongswan, 2006, 2009; Hausman and Wongswan, 2011), we include he conrol variable, S&P ne, in he empirical analysis. S&P ne is he esimaed residual aken from running he regression of he S&P500 fuures reurn over one-day window around he FOMC announcemen on he arge and pah facors. Following Hamilon (2008), we also include he lagged S&P500 fuures reurn of previous six days before he FOMC announcemen dae o adequaely conrol for news of prior days lasing for more han a week. Our sample spans from February 8, 1990 hrough Augus 22, Noe ha he effecive federal funds rae and federal funds arge rae were close o zero (0-0.25%) saring from ecember 2008 due o he Federal Reserve s aggressive ineres rae cu, which makes i difficul o accuraely esimae he arge and pah facors using he federal funds fuures daa. Table 1 repors he summary saisics. Resuls Based on Naional Morgage Raes Baseline Resuls Table 2 Panel A presens he boosrapping resuls of esimaing he wo-facor model described in equaion (1), using he naional morgage raes. In Panel A, we use he change in average 1- year ARM rae and 30-year morgage rae across he naion as he dependen variables. The resuls show ha wihin he week of he FOMC announcemens, surprises in he arge federal funds raes (he arge facor) have significanly posiive impacs on he naional 1-year ARM rae. On average, an unanicipaed 100 basis-poin increase in he arge federal funds rae (or he arge facor), in he absence of any surprises in he accompanying FOMC saemens, would on 14

16 average lead o an increase of 0.22% or 22 basis poins in he average 1-year ARM rae across he naion (a he 1% significance level). 6 We furher examine he impac of he moneary policy surprises on he naional morgage raes over he firs and second weeks, respecively, afer he FOMC announcemen. Table 2 Panel B repors he persisence regression resuls. We find ha in he firs week afer he FOMC announcemens, he naional 1-year ARM rae has significanly posiive responses o boh he arge and pah facors (a he 1% and 10% level, respecively). On average, an unanicipaed 100 basis-poin increase in he arge federal funds rae (he arge facor) would lead o an increase of 24 basis poins in he naional 1-year ARM rae while an unanicipaed 100 basis-poin increase in he four-quarer-ahead Eurodollar fuures rae (he pah facor) would lead o an increase of 6 basis poins in he naional 1-year ARM rae. We also find ha in he firs week afer he announcemen, he naional 30-year fixed morgage rae has a significanly posiive response o he pah facor (a he 10% level). On average, an unanicipaed 100 basis-poin increase in he four-quarer-ahead Eurodollar fuures rae (he pah facor) would lead o an increase of 10 basis poins in he naional 30-year fixed morgage rae during he firs week afer he announcemens. We find no significan responses in he second week afer he FOMC announcemens. The adused R-squared also suggess ha moneary policy surprises have larger impac on he 1-year ARM rae naionwide han on he 30-year fixed rae. Obviously, he shor-erm adusable morgage raes experience boh he higher degree of pass-hrough and he faser adusmen speed han he long-erm fixed morgage raes. In summary, we find ha he arge facor has significanly posiive impacs on he naional 1-year ARM rae wihin he week of he announcemens and he impacs las up o one week afer he announcemens. However, he arge facor has no significan impacs on he 30-year 15

17 fixed morgage rae. The resul suggess ha surprises in he arge federal funds rae mainly influence he shor-erm floaing rae. To he bes of our knowledge, our paper is among he firs o provide evidence on how he naional morgage raes reac o he pah facor. We find ha he pah facor has significanly posiive impacs on boh he naional 1-year ARM and 30-year morgage raes in he firs week afer he FOMC announcemens. The evidence is well in line wih earlier sudies which find he significan impacs of he pah facor on oher financial markes (e.g., Wang, Yang and Wu, 2006; Wang, Yang and Simpson, 2008; Wongswan, 2009; Hausman and Wongswan, 2011). The resuls indicae ha he FOMC saemens likely exer heir effecs on morgage markes hrough heir influence on marke expecaions of fuure moneary policy acions. The resuls also sugges ha previous sudies (e.g., Hamilon, 2008) could underesimae he impacs of moneary policy surprises on morgage raes by focusing only on he surprises in he arge federal funds rae. In addiion, we provide new evidence ha he response of he long-erm fixed ineres rae (i.e., he 30-year morgage rae) o moneary policy surprises is lagged behind ha of he floaing ineres rae (i.e., he 1-year ARM). The reason why such lagged responses exis is no clear. We conecure ha moneary policy arge raes are he benchmark for shor-erm floaing ineres raes while here are many oher facors poenially affecing he formaion of he longerm fixed ineres rae and hus he marke migh ake longer ime o diges he implicaions of moneary policy surprises for he long-erm fixed ineres rae. Anoher possibiliy is he delayed updae of he 30-year morgage rae. As poined ou by Hamilon (2008, p. 1176), in he curren sysem in which he weekly morgage daa are released on Thursday, Freddie Mac officials sop collecing morgage raes on Wednesday, and hus mos of he repors from individual banks come in on Monday or Tuesday. On he oher hand, he scheduled FOMC meeings ypically 16

18 release he news on Wednesdays, and hus he impac of moneary policy news migh be refleced on nex week s 30-year morgage raes. Asymmeric Responses We furher examine he asymmeric responses of morgage markes o moneary policy surprises in he following dimensions: sign, size, direcion of rae movemen and sage of business cycle. As he resuls of baseline regression and persisence regression show ha moneary policy surprises have significan influences on he naional 1-year ARM rae boh wihin he week of FOMC announcemen and during he firs week afer he announcemen, we examine he asymmeric responses of he naional 1-year ARM rae in boh windows. As he naional 30-year morgage rae has significanly one-week lagged responses o moneary policy surprises, we examine he asymmeric responses of he naional 30-year morgage rae over he window of he firs week afer he FOMC announcemen. Table 3 repors he asymmeric resuls. Sign response. Table 3 panel A presens he resuls of he sign regression. We find evidence of he asymmeric influences of he moneary policy surprises on boh he naional 1-year ARM rae and he naional 30-year morgage rae. We find ha he negaive surprises in he arge facor have significan impacs on he 1-year ARM raes (a he 1% level) wihin he week of he FOMC announcemens while he posiive surprises in he arge facor have significan impacs on he 1-year ARM raes (a he 5% level) in he firs week afer he FOMC announcemens. The resuls sugges ha impacs of he arge surprises come mosly from he unanicipaed cus in he arge federal funds rae wihin he week of he announcemens bu mosly from he unanicipaed increases in he arge federal funds rae in he firs week afer he announcemens. In addiion, we find ha he posiive surprises in he pah facor have significan impacs on he naional 30-17

19 year morgage rae a he 10% level, which sugges ha he impacs of he pah facor come mosly from he unanicipaed posiive change in he fuure moneary policy direcions. In general, he srong sign-relaed asymmeric impac on morgage raes is in line wih Wang, Yang and Simpson (2008) on currency markes and Hausman and Wongswan (2011) on global equiy markes. Size response. Table 3 panel B repors he resuls on size responses. We find ha he naional 1- year ARM rae responds only o he big surprises in eiher he arge or pah facor while he size of moneary policy surprises has no impac on he naional 30-year morgage rae. The 1-year ARM reacs o he big surprises in he arge facor wihin he week of FOMC announcemen while reacs o he big surprises in boh he arge and pah facor during he firs week afer he announcemen. We provide new evidence on how he sizes of moneary policy surprises impac he morgage raes. The evidence sugges ha he big surprise in eiher he arge federal funds rae or he fuure direcion of he moneary policy have significanly larger impac on he shor-erm floaing morgage rae while he sizes of moneary policy surprises have no impac on he longerm fixed morgage rae. Response o direcion of rae movemens. Table 3 panel C presens he resuls of he direcion of rae movemens. We find ha he naional 1-year ARM rae reacs o surprises in he arge facor during he periods of boh ighening and loosening wihin he week of he announcemen and during he firs week afer he announcemen. The naional 1-year ARM rae reacs o he surprises in he pah facor mainly during he period of ighening (a he 10% level). The resuls 18

20 sugges ha he shor-erm floaing morgage rae mainly reacs o surprises in he fuure moneary policy direcions during he periods of moneary ighening. We find no evidence of asymmery in how 30-year morgage raes reac o moneary policy surprises during he period of ighening and loosening. The resuls are consisen wih hose in Scholnick (1999) which finds no evidence of asymmery in how 30-year morgage raes reac o changes in he 30-year U.S. governmen consan-mauriy Treasury bond rae. Response o sages of he business cycle. Table 3 Panel presens he resuls when he sample is subdivided ino he period of economic expansion and he period of economic recession. We find ha he arge facor has significan impacs on he naional 1-year ARM rae during he periods of boh economic expansion and economic recession wihin he week of he announcemens bu only during he periods of economic expansion in he firs week afer he announcemens. We also find ha he arge and pah facor have significan impacs on he naional 30-year morgage rae only during he period of economic expansion, where ineres rae ends o increase. In conclusion, here is srong evidence of asymmeric adusmens of morgage raes in response o moneary policy surprises, especially o large versus small surprises and in he expansionary periods versus he recessionary periods. The resuls show ha i seems inadequae o eiher ignore he asymmery or only focus on asymmery relaed o he direcion of rae movemens, as applied in many earlier sudies (e.g., Hannan and Berger, 1991; Scholnick, 1999). Robusness Checks We conduc several robusness checks o ensure ha he findings repored above are no driven by oher facors. 7 19

21 Firs, since lenders ie he changes in he ARM rae o an underlying index rae, we conduc addiional analysis o compare how he ARM raes and he morgage index raes may reac o moneary policy surprises differenly. Since he maoriy of ARM producs are indexed o eiher he 1-year consan-mauriy reasury yields or he 1-year London Inerbank Offered Rae (LIBOR), we examine he responses of hese wo morgage indexes o moneary policy surprises around FOMC announcemen days (i.e., on even days and up o wo days afer he evens). We find ha moneary policy surprises have no significan impacs on he daily 1-year reasury yield while he pah facor has a significanly posiive impac on he daily 1-year LIBOR one day afer he announcemen. Furhermore, since he 1-year ARM raes in he Freddie Mac Primary Morgage Marke Survey (PMMS) are indexed o U.S. reasury yields, we examine how moneary policy surprises impac he morgage spread beween he 1-year ARM rae and he 1- year reasury yield. We find ha he arge facor has a significanly negaive impac on he morgage spread wihin he week of he FOMC announcemens. The resuls sugges ha he impac of moneary policy surprises on he ARM raes is differen from ha on morgage indexes. The differences could be due o he underlying differences in he risk characerisics of he morgage producs and reasury securiies. Second, he convenional wisdom suggess ha he 30-year fixed morgage raes move closely wih he 10-year reasury bonds. Hence, we also invesigae how he 10-year reasury bonds reac o he arge and pah facors. We find no significan responses of he 10-year reasury yields o moneary policy surprises during he sample period. Thus, surprises in moneary policy acions apparenly have differenial impacs on he 30-year fixed morgage raes and he 10-year reasury yields, respecively. 8 20

22 Resuls Based on Regional Morgage Raes In his secion, we examine he impac of moneary policy surprises on he morgage raes in he five Freddie Mac regions (Norh Cenral, Norheas, Souheas, Souhwes and Wes). The idea ha moneary policy surprises have differenial impacs on morgage raes in differen regions can be deduced from radiional credi-based heories (e.g., Kashyap, Sein and Wilcox, 1993) of he moneary policy ransmission mechanism. We invesigae no only wheher here exis he cross-region variaions in he morgage markes bu also wha economic or housing facors migh cause he differen responses o moneary policy surprises in he regional morgage markes. Baseline Resuls Table 4 Panel A repors he boosrapping resuls of esimaing he wo-facor model describes in equaion (1). We use he morgage raes in he five Freddie Mac regions as he dependen variables. We find evidence ha here exis significan variaions in how he 1-year ARM raes respond o moneary policy surprises across he five regions. The 1-year ARM rae in he Souhwes region has no significan response o he arge facor while he 1-year ARM raes in oher four regions have significanly posiive responses o he arge facor wihin he week of he FOMC announcemens. The magniude of he responses of he regional 1-year ARM raes o he moneary policy surprises is similar o ha of he average morgage raes naionwide. On average, an unanicipaed 100 basis-poin increase in he arge federal funds rae (or he arge facor) would on average lead o an increase of 22 basis poins in he 1-year ARM raes of he Norheas and Souheas regions (a he 10% and 1% level, respecively), an increase of 20 basis poins in he 1- year ARM raes of he Norh Cenral region (a he 10% level) and an increase of 28 basis poins in he 1-year ARM raes of he Wes region (a he 1% level). We do no find significan 21

23 responses of he regional 30-year morgage raes o moneary policy surprises. The resuls are consisen wih hose of he naionwide 30-year morgage rae. We also examine he impac of he moneary policy surprises on he regional morgage raes over he firs and second week afer he FOMC announcemen, respecively. Table 4 Panel B repors he persisence regression resuls. We find ha he regional morgage raes reac o surprises in boh he arge and pah facors in he firs week afer he announcemen. The arge surprise has significanly posiive impacs on he 1-year ARM raes in he Norh Cenral, Souhwes and Wes regions (a he 10%, 5% and 1% level, respecively). The pah surprise has significanly posiive impacs on he 1-year ARM raes in he Norheas region (a he 5% level) and on he 30-year morgage raes in he Norh Cenral and Souheas regions (a he 10% level). The impacs of he arge and pah facors are subsanially weakened in he second week afer he announcemen excep in he Norheas region where he arge facor gains a significanly posiive impac on he 1-year ARM rae (a he 5% level). The adused R-squared suggess ha he arge facor has he larges impac on he 1-year ARM rae in he Souhwes region and he pah facor has he larges impac on he 30-year morgage rae in he Norh Cenral region. Overall, he resuls sugges ha here exis significan cross-region variaions in how he morgage raes reac o moneary policy surprises. Asymmeric Responses In his secion, we examine he asymmeric responses of morgage markes in five regions o moneary policy surprises in he following dimensions: sign, size, direcion of rae movemen and sage of business cycle. Table 5 repors he asymmeric regression resuls. 22

24 Sign Response. Table 5 panel A presens he resuls of he sign regression. We find evidence of he asymmeric influences of he moneary policy surprises on he regional 1-year ARM raes. Wihin he week of he FOMC announcemen, in Norheas, Souheas and Wes regions, he 1- year ARM raes only reac o negaive surprises in he arge facor (a he 1% level). In he firs week afer he announcemen, while he 1-year ARM rae has significan response only o negaive surprises in he arge facor in he Norh Cenral (a he 10% level), he 1-year ARM rae has significan responses only o posiive surprises in he arge facor in he Souheas and Souhwes regions (a he 5% levels). In Wes region, he 1-year ARM rae has significan responses o boh posiive and negaive surprises in he arge facor (a he 10% and 5% level). In he Norheas region, he 1-year ARM rae has significan responses o posiive surprises in boh he arge and pah facors (a he 10% and 5% level, respecively). We also find he asymmeric influences of he moneary policy surprises on he regional 30-year morgage raes. In he firs week afer he announcemens, in Norh Cenral and Souheas regions, he 30-year morgage rae has significan responses only o posiive surprises in boh he arge and pah facors. In Souhwes region, he 30-year morgage rae has significan responses only o he posiive surprises in he pah facor (a he 10% level). In summary, we find ha only he arge facor has asymmeric impacs on he regional 1- year ARM raes wihin he week of he announcemen. Bu boh he arge and pah facors have significanly asymmeric influences on he regional 1-year ARM raes in he firs week afer he announcemen. In addiion, boh he arge and pah facors have significanly asymmeric influences on he regional 30-year morgage raes in he firs week afer he announcemen. Size response. Table 5 panel B repors he resuls on size responses. Wihin he week of he announcemens, he regional 1-year ARM raes respond mainly o he big surprises in he arge 23

25 facor; wih one excepion in he Wes region, he 1-year ARM rae also has significan responses o small surprises in he arge facor a he 5% level. In he firs week afer he announcemens, we find sronger evidence on he cross-region variaions. In he Norh Cenral region, he 1-year ARM rae reacs posiively o he big surprises in he arge facor (a he 10% level) and negaively o he small surprises in he pah facor (a he 10% level). In he Norheas region, he 1-year ARM rae reacs o boh big and small surprises in he pah facor (a he 5% level). In he Souhwes and Wes regions, he 1-year ARM raes mainly reac o big surprises in he arge facor (a he 5% and 1% level, respecively). We also find ha in he Norh Cenral region, he 30-year morgage rae has significan responses o big surprises in he pah facor during he firs week afer he announcemen. The evidence suggess ha he big surprises in eiher he arge federal funds rae or he fuure direcion of he moneary policy have larger impac on he regional morgage markes han he small surprises. irecion of rae movemens. Table 5 panel C presens he resuls of he direcion of rae movemens. The regional 1-year ARM raes reac o he direcion of rae movemens differenly. Wihin he week of he announcemens, he 1-year ARM rae in he Norh Cenral region reacs o he surprises in boh he arge and pah facors during he period of moneary ighening (a he 1% level). The 1-year ARM rae in he Norheas region reacs o he surprises in he arge facors during he period of moneary loosening (a he 10% level). The 1-year ARM rae in he Souheas region reacs o he surprises in he arge facor during he period of moneary loosening (a he 1% level) and o he surprises in he pah facors during he period of moneary ighening (a he 1% level). The 1-year ARM rae in he Souhwes region reacs o he surprises in he pah facor during he period of moneary loosening (a he 10% level). The 1-year ARM 24

26 rae in he Wes region reacs o he surprises in he arge facor during boh he periods of moneary ighening and loosening (a he 10% and 5% level, respecively). uring he firs week afer he announcemen, he 1-year ARM raes in he Norheas and Norh Cenral regions reac more srongly o he pah facor during he period of moneary ighening (a he 10% level) while he 1-year ARM raes in he Souheas and Souhwes regions reac more srongly o he arge facor during he period of U.S. moneary ighening (a he 1% level). The 1-year ARM rae in he Wes region has significan responses o he arge facor during boh he period of moneary ighening and moneary loosening (a he 5% and 1% level, respecively). In he firs week afer he announcemens, he 30-year morgage rae in he Norh Cenral region reacs only o he pah facor during he periods of U.S. moneary ighening (a he 5% level). Overall, he resuls show ha moneary policy surprises have asymmeric impacs on he regional morgage raes during boh he periods of moneary ighening and moneary loosening. Sage of he business cycle. Table 5 panel presens he resuls when he sample is subdivided ino he period of economic expansion and he period of economic recession. Wihin he week of he FOMC announcemen, he 1-year ARM rae in he Norheas region reacs o he surprises in he arge facor during he recession periods (a he 10% level) while he 1-year ARM raes in he Souheas, Souhwes and Wes regions reac o surprises in he arge facor during he expansion period. uring he firs week afer he announcemen, we find uniform resuls in he five regions ha surprises in he arge facor has significanly posiive impacs on boh he 1-year ARM raes and he 30-year morgage raes mainly during he expansionary periods of he economy, when ineres raes end o increase. 25

27 The resuls sugges ha he regional morgage markes reac o moneary policy surprises quie differenly during differen saes of he economy. Why is There Heerogeneiy in Responses o he FOMC Saemens across Regions? Boh baseline and asymmeric regression resuls show ha here exiss heerogeneiy in how he moneary policy surprises are passed ono regional morgage raes. Morgage raes end o be suscepible o economic aciviy. For example, ob loss could lead o paymen defaul or early repaymen. Hence, a region s income level, ob marke and housing marke condiions can significanly affec he morgage raes. In his secion, we aim o empirically idenify he economic facors ha would conribue o he cross-region differences. Following he previous lieraure (e.g., Carlino and efina, 1998; Fraanoni and Schuh, 2003; Owyang, Rapach and Wall, 2009), we consruc seven facors o capure he differences in each region s macroeconomic and housing marke condiions and examine he poenial influences on he cross-region variaions in he morgage rae pass-hrough. The facors under consideraion are as follows: growh in personal income, he unemploymen rae, he home ownership rae, he home vacancy rae, he renal vacancy rae, he change in building permis, and he change in housing sars. The unemploymen rae, he change in building permis, and he change in housing sars are monhly daa, while he oher proxies are annual daa. Table 6 presens he summary saisics of he seven proxies for each region. Table 7 presens he Spearman correlaion marix of he proxies. We find ha he renal vacancy rae is highly correlaed wih unemploymen rae (negaive), home ownership rae and home vacancy rae, and unemploymen rae is highly negaively correlaed wih homeownership, renal vacancy rae and growh in income. 26

28 Pooled regression resuls. To explore facors ha migh affec he cross-region variaions in he morgage rae pass-hrough process, we run he pooled regressions and inerac he measures of moneary policy surprises wih he regional facors. Specifically, we esimae he following regression: M i, R Ri, S& PS & Pne, RS& P500, (6) M i, R Ri, S& PS & Pne, RS& P500, (7) Where R i, is a regional facor. The coefficiens on he ineracion erms ( R and R ) capures he impac of he regional facors on he cross-region variaions in he response o moneary policy surprises. 9 Table 8 presens he resuls for he role of each regional facor proxy in explaining he cross-region variaions in responses o he arge facor. Table 8 Panel A repors he resuls for he 1-year ARM rae. We find ha wihin he week of he announcemens, he coefficien on he ineracion erm of home vacancy is significanly posiive a he 5% level, which suggess ha a region wih a higher (lower) home vacancy rae migh have larger (smaller) responses o he surprises in he arge federal funds raes. uring he firs week afer he announcemen, he coefficien on he ineracion erm of unemploymen rae is significanly posiive a he 10% level, which suggess ha a region wih higher (lower) unemploymen rae migh have larger (smaller) responses o unexpeced changes in he federal funds arge rae. Table 8 Panel B repors he resuls for he 30-year fixed-rae morgage. The coefficien on he ineracion erm of unemploymen rae is significanly posiive a he 1% level, while he coefficien on he ineracion erm of home vacancy is significanly negaive a he 1% level. The resuls sugges ha he 30-year morgage rae in a region wih higher unemploymen rae and lower home vacancy rae responds more o he arge facor. 27

29 Table 9 presens he resuls for he role of each proxy in explaining he cross-region variaions in response o he pah facor. Table 9 Panel A shows he resuls for he 1-year ARM rae. Wihin he week of he FOMC announcemen, responses o he pah facor are significanly correlaed wih growh in income (significanly posiive a he 1% level), he home vacancy rae (significanly negaive a he 1% level), and he renal vacancy rae (significanly negaive a he 1% level). The 1-year ARM rae in a region wih higher growh in income, lower home vacancy rae, and lower renal vacancy rae responds more o he pah facor. uring he firs week afer he announcemen, we do no find significan impacs of regional facors on how he 1-year ARM rae reacs o he pah facor. Table 9 Panel B shows ha for he 30-year morgage rae, he coefficien on he ineracion erm of growh in income is significanly posiive a he 1% level. The 30-year morgage rae in a region wih higher (lower) growh in income responds more (less) o he pah facor. To furher disinguish among differen regional facors, we re-esimae equaion (6) and (7) using he muliple proxies oinly. Table 10 Panel A repors he resuls for responses o he arge facor. We find ha a region s home vacancy and renal vacancy raes have significan impacs on how he regional 1-year ARM raes reac o he arge facor wihin he week of he announcemens. We also find ha a region s unemploymen rae, home vacancy and renal vacancy raes have significan impacs on how he regional 1-year ARM ad 30-year morgage raes reac o he arge facor during he firs week afer he announcemens. Table 10 Panel B repors he resuls for responses o he pah facor. We find ha a region s growh in income and home vacancy rae have significan impacs on how he regional 1-year ARM raes reac o he pah facor wihin he week of he announcemens. We also find 28