The Identification of Response of Stock Returns to Monetary Policy Actions. Using Market-Based Measures of Monetary Policy Shocks.

Transcription

1 The Idenificaion of Response of Sock Rerns o Moneary Policy Acions Using Marke-Based Measres of Moneary Policy Shocks. by Dennis W. Jansen* Texas A&M Universiy and Chn-Li Tsai Naional Cheng Kng Universiy Absrac: We invesigae wo relaed approaches o dealing wih he possible join response bias in sing Kner s approach o idenifying moneary policy s impac on sock rerns he mehodology recenly sggesed by Thornon, and se of inraday daa. For all hree mehods, we find he esimaed impac of moneary policy acions on sock rerns is negaive and saisically significan, and we find ha his negaive impac is magnified dring bear markes and dring recessions. We find poin esimaes indicaing a posiive join response bias sing Thornon s mehodology, alhogh hese are no saisically significan. We find ha inraday daa provide he same qaliaive paern of resls, b he esimaed magnide of he impac of moneary policy on sock rerns is smaller compared o eiher approach sing daily daa. Dennis Jansen hanks he Privae Enerprise Research Cener a Texas A&M Universiy for research sppor. *Corresponding ahor. Dennis Jansen, Deparmen of Economics, 4228 TAMU, Texas A&M Universiy, College Saion, TX, , USA. dennisjansen@am.ed, Telephone: (1)

2 1. Inrodcion. Marke-based measres of moneary policy shocks have been sed a leas since Kner (2001). Kner sed changes in Federal Fnds Rae Fres prices o infer he srprise in he Federal Open Marke Commiee ineres rae annoncemen, and sed a regression of ineres raes on his annoncemen srprise measre in order o idenify he impac of moneary policy on ineres raes. One concern wih his widely sed procedre is he possibiliy ha an nobserved variable migh joinly impac boh he annoncemen srprise measre and he dependen variable here ineres raes. This wold lead o bias in or esimaed impac of moneary policy on ineres raes, a bias ha some have labeled he joinresponse bias in his lierare. Mch of he work sing marke measres of moneary policy have been a he daily level, wih he daily change in ineres raes compared o he daily change in he Federal Fnds Rae Fres price. One proposed solion o he join-response bias problem is o se higher freqency daa, in his case inraday daa, and o measre he policy srprise and he change in ineres raes in a narrow window arond he FOMC annoncemen. The idea is ha, by sing a mch smaller par of he day, he chance of some oher nobserved variable joinly impacing boh he Federal Fnds Rae Fres price and he marke ineres raes is mch redced, so ha se of higher freqency daa a leas aenaes his join response bias. More recenly, Thornon (2013) has sggesed an alernaive approach o miigae he join response bias. The above approaches, wheher sing daily daa or inraday daa, boh se daa only from he FOMC annoncemen day. Kner s original approach looked a he daily change in ineres raes and he daily srprise change in he Federal Fnds Rae Fres price. The se of inraday daa ses daa in a narrow window maybe 30 mines arond he annoncemen. Boh approaches are even 2

3 sdies, only considering daa a he FOMC evens, and ignoring all oher days in he sample. Thornon s alernaive is o se daily daa, b all of he daily daa, and o consrc measres of he change in ineres raes and he srprise change in he Federal Fnds Rae on each day of he sample, no js he FOMC even daes. Thornon hen rns a regression where he dependen variable he change in an ineres rae is regressed on he measre of he srprise change in he Federal Fnds Rae and on an indicaor for an FOMC even ineraced wih he measre of he srprise change in he Federal Fnds Rae. The coefficien on his laer erm is he measre of he impac of moneary policy on ineres raes, while he srprise change in he Federal Fnds Rae on all oher days serves as a conrol variable for he join-response bias. Thornon (2013) demonsraes he abiliy of his mehodology o redce he problemaic join-response bias. Here we se he hree approaches olined above o esimae he impac of moneary policy on sock rerns, and we compare or resls o see if here is evidence of he join-response bias in or esimaes of he impac of moneary policy on sock rerns. We also compare he wo approaches sggesed o miigae his problem. Imporanly, he join response bias ha Thornon discsses can be eiher negaive or posiive, depending on he correlaion of sock rerns and he federal fnds rae srprise wih he nobserved variable. Sock rerns and he federal fnds rae fres prices are likely o respond o similar nobserved news evens impacing financial markes, making he join response bias a poenially imporan isse. We will follow Thornon (2013) and esimae he following eqaion sing daa from every day in he sample and no js on FOMC even daes. We calclae he srprise change in he federal fnds rae fres each day of or sample, and esimae he following eqaion: 3

4 R i, 0 1 f 2 f FOMC i, (1) where f denoes Kner s (2001) marke-based measre of nexpeced federal fnds rae changes, and FOMC denoes a dmmy variable ha is 1 on days wih moneary policy evens and zero oherwise. The coefficien β 1 denoes he join response of sock rerns and marke-based measres of moneary policy shocks o ambien news, and β 2 denoes he join response of sock rerns and he marke-based measre o nexpeced policy evens. In oher words, he coefficien β 2 reflecs he marginal change in sock rerns associaed wih an nexpeced policy even. If β 2 is no significanly differen from zero, he marke s reacion o a srprise moneary policy even is no differen from is reacion o changes in he federal fnds rae de o ambien news. 2. Daa Or analysis is based on Federal Open Marke Commiee (FOMC) decisions, and or daa se conains informaion on FOMC decisions on he Federal Fnds Rae arge from 103 schedled FOMC meeings over he period March 1995 hrogh December We obain he exac dae and ime of he FOMC releases o se when measring moneary policy srprises wih inraday day. Or sock rern daa is calclaed from he S&P 500 index. Measre of Daily Srprise Kner shock We measre he srprise moneary policy shock --- and acally he srprise 1 The Federal Fnds Rae was argeed a a range of % by lae in 2008, and was engaged in he firs of an ongoing series of large asse prchases, so-called nonradiional moneary policy. Nonradiional moneary policy, inclding large scale Federal Reserve Sysem asse prchases, especially a he zero lower bond of he Federal Fnds Rae, implies ha he Federal Fnds Rae is no longer a saisfacory single indicaor of moneary policy acions. Ths we end or sample a he end of

5 change in he Federal Fnds rae each day -- sing Federal Fnds Rae Fres prices and he mehod proposed by Kner (2001). Kner sed marke daa on Federal Fnds fres conracs raded on he Chicago Board of Trade o exrac a measre of he srprise change in he Federal Fnds Rae on FOMC even daes. The idea is ha he Federal Fnds Rae Fres conrac price on he day prior o he FOMC annoncemen reflecs he marke s expecaion of he FOMC annoncemen on he scceeding day. The fres conrac price on he day of he FOMC annoncemen, and especially is change from he previos day, reflecs informaion in he annoncemen. The difference in he fres conrac prices a dae -1 and dae can be sed o calclae he change in he Federal Fnds Rae ha is a srprise o he marke. The acal calclaion ms be scaled o ake accon of he fac ha he fres conrac selemen price is based on he monhly average Federal Fnds Rae. For a change in he Federal Fnds Rae arge aking place on day of monh s, he srprise arge fnds rae change is calclaed as he 1-day change in he spo-monh fres rae. This change in he fres rae is scaled p by a facor relaed o he nmber of days in he monh affeced by he change, becase he conrac s selemen price is based on he monhly average Federal Fnds Rae. Tha is, he nanicipaed change in he Federal Fnds Rae arge is calclaed as: m 0 0 f ( f s, f s, 1 ) (2) m where f is he srprise Kner shock, 0 f s, is he spo-monh fres rae, and m is he nmber of days in monh s. In he case where he arge rae change occrs on he firs day of he monh, we replace f wih 0 s, 1 0 f s 1, m. Here f denoes he 1-0 s 1, m monh fres rae from he las day of he previos monh. In order o avoid he large noise a he end of he monh, if he arge rae change falls wihin he las seven days 5

6 of he monh we se he nscaled 1-monh fres rae change insead of he change in he spo-monh rae. Tradiionally his mehodology is applied o he FOMC even daes, so he srprise change in he Federal Fnds Rae is calclaed only on hose daes. In order o apply Thornon s mehodology we calclae he srprise change in he Federal Fnds Rae on each day of or sample no js FOMC even daes. Measre of Inraday Srprise Kner shock When sing inraday daa we consrc an inraday version of he Kner srprise by calclaing he change in he Federal Fnds Rae Fres prices in a narrow window arond he FOMC annoncemen ime, and hen inferring he srprise change in he Federal Fnds Rae in he annoncemen. We se varios alernaive windows, and here will repor only for a window of 40 mines widh. Oher window widhs (e.g. 25 mines wide, from 5 mines before o 20 mines afer he annoncemen) gave qie similar resls. The inraday Kner srprise o he federal fnds rae arge in a 40-mine window is calclaed as: m f, 40 ( f, 30 f, 10 ) (3) m where is he 40-mine window srprise arge rae change for he FOMC f, 40 annoncemen on day, f, 30 is he spo-monh fres rae a ime on day for ime 30 where τ is he annoncemen ime, m is he nmber of days in he monh, and f is he spo-monh fres rae a ime 10., 10 Measre of S&P 500 Index Sock Rerns Or daily sock rerns are calclaed for he S&P500 index sing daa from he CRSP (he Universiy of Chicago s Cener for Research in Secriy Prices) daa se. We se he closing price on he day prior o he FOMC annoncemen day and he 6

7 closing price on he day of he FOMC annoncemen. For inraday rerns, we again se he S&P 500 index, sing inraday daa from he Tick Inraday Fres & Indices daabase. The Tick Inraday Fres & Indices daabase is a collecion of inraday rades and qoes for many indices and commodiies. We calclae inraday rerns wihin he specified windows srronding FOMC annoncemen imes from index vales in he Tick Inraday Fres & Indices daabase. Rerns are calclaed sing log differences. Ths for he inraday 40 mine window we calclae he 40-mine sock rern as: R, *(log P, 30 log P, 10 ) (4) where R,40 is he 40-min inerval rern on S&P 500 index on day srronding an FOMC annoncemen a ime. Here P, 30 and P, 10 represen he sock prices for rades a ime 30 and 10 on day, respecively. Imporanly, hese rerns are rerns over a 40 mine period, so comparisons o daily rerns sed in he even-sdy mehods shold be made wih ha in mind. Measre of Bll and Bear Markes We look a differenial impacs of moneary policy in bll and bear markes, and also in recessions and expansions. While we se he NBER definiion of recessions, we need o define bll and bear markes. Or definiion follows Pagan and Sossonov (2003), a definiion also sed in Jansen and Tsai (2010). Pagan and Sossonov develop a classificaion of markes ino bll and bear ha is mally exclsive and exhasive. A peak will always follow a rogh and vice versa, and he even space is divided ino bll and bear marke periods. A bll marke is said o occr when he sock index is locaed beween he rogh poin and he peak poin, inclding he peak, 7

8 and a bear marke occrs oherwise. Figre 1 provides a graph of bll and bear markes over or sample. This graph clearly illsraes sock marke was in a bear marke from Sepember 2000 nil he rogh in Sepember 2002, and a bll marke occrs oherwise. Measre of Expansion and Recession Saes We se he decision of he Bsiness Cycle Daing Commiee of he Naional Brea of Economic Research o define Apr 2001 o Nov 2001 as a recession period dring or sample. Oherwise an expansionary sae occrs. 3. Descripive Saisics Table 1 provides descripive saisics for he daa sed in or hree esimaion mehods. Mehod 1 is he radiional daily even sdy mehodology based on calclaing Kner srprises. Mehod 2 is he even sdy mehodology based on calclaing Kner srprises on inraday daa, ideally redcing he join response bias. Mehod 3 is Thornon s mehod, a ime series mehodology sing daily daa, Kner shocks calclaed for all days, and an indicaor for he response of sock rerns o Kner shocks on FOMC even daes. Mehods 1 and 2 are based on 103 moneary policy evens. In he firs par of Table 1 we repor he nmber of evens 103 as well as he average Kner shock calclaed on hese 103 even daes, -.005, and he average sock index rern on hese 103 days, 0.24%. Mehod 2 looks a a narrow slice of ime on he same 103 even daes. In or 40- mine window, he average Kner shock is -.110, and he average sock rern is -.093%. We also provide saisics for he average sock rern from he close of day prior o he even dae hrogh he beginning of or even window on he even day, and he average sock rern from he end of he even window nil he marke close on he even day. We will say more abo hese saisics laer in he paper. 8

9 Mehod 3 looks a 3,231 days of daa spanning or sample from March December The average Kner shock over all hese days is mch lower han on he even days, -.001, and he average daily sock rern is 0.001%. Clearly mch more is happening on he 103 FOMC meeing days, on average, han on he rading day when he FOMC does no mee. We have 154 monhs of daa in or sample, wih 103 FOMC daes. Dring or sample we have 25 monhs when we classify he sock marke as being in a bear sae, js nder one-sixh of or sample. We have only 8 monhs when we classify he sae of he bsiness cycle as being a recession, or js over five percen of or sample. 4. Economeric Specificaions Or firs look is o se even-sdy mehodology wih daily and inraday daa o respecively invesigae he linear relaion beween sock rerns and marke-based measre of nexpeced arge changes by esimaing eqaion (5): R 1 f 0 (5) Again, his eqaion will be esimaed sing daily daa, and again sing or inraday daa, all for he 103 FOMC even days. Or second look is o employ s Thornon s mehodology o esimae he impac of moneary policy sing daily ime series daa o conrol for he possible join-response bias. For his we esimae eqaion (6): R 1 f 2 0 f FOMC (6) Here FOMC denoes a dmmy variable ha is 1 on days wih moneary policy evens and zero oherwise. Coefficien β1 measres he effec of he srprise change in he Federal Fnds rae (he Kner shock) on all days in or sample. Coefficien β2 measres he effec of hese srprise changes in he Federal Fnds rae when hey occr on an FOMC meeing day. This coefficien will be or esimae of he 9

10 moneary policy impac on sock rerns. We also se hese hree mehods o examine how he impac of moneary policy migh be sae dependen. We look separaely a wo sae variables. One is he sae of he sock marke, bll or bear. The oher is he sae of he economy, expansion or recession. For or wo even-sdy mehodologies, sing eiher daily daa or sing inraday daa, we esimae he following eqaion (7): R 0 1 f 2 f Sae (7) Here Sae is a dmmy variable for a bear marke, or a dmmy variable for a recession. Here when he sae variable is he sae of he sock marke, he impac of moneary policy in a bll marke is measred by β1, and he impac of moneary policy in a bear marke is measred by β1 + β2. Alernaive, when he sae variable is he sae of he bsiness cycle, β1 measres he impac of moneary policy dring an expansion, and β1 + β2 measres he impac of moneary policy dring a recession. Finally, we apply Thornon s mehodology by esimaing eqaion (8), which allows he sae of he sock marke, or alernaively he sae of he bsiness cycle, o aler he impac of moneary policy on sock rerns. R 1 f 2 f FOMC 3 f Sae 4 0 f FOMC Sae (8) Here β 2 + β 4 measres he effec of moneary policy on sock rerns dring a bear marke (alernaively a recession period), and β 2 measres he effec of moneary policy on sock rerns in a bll marke (alernaively an expansion period.) 5. Empirical Resls. Table 2 repors resls from esimaing eqaion (5) sing daily and inraday sock rerns. Esimaes of β1 are esimaes of he impac of he Kner shock on sock rerns, esimaed wih a daily even window and wih a 40 mine inraday window. 10

11 Given sandard idenificaion assmpions, hese are esimaes of he impac of moneary policy on sock rerns. Wih or daily daa or esimae of β1 is , indicaing ha a 1% increase in he Federal Fnds Rae resls in a % change in sock rerns. Wih inraday daa or esimae is , indicaing ha a 1% increase in he Federal Fnds Rae resls in a % change in sock rerns. Boh esimaes are saisically significan a he 1% significance level. Table 3 repors resls from applying Thornon s mehodology. The esimae of he coefficien on he conrol for he join response bias, β 1 is b no saisically significan. This coefficien indicaes ha sock rerns and Kner srprises are posiively correlaed on non-fomc days, sggesing ha he join response bias redces he poin esimae of he impac of moneary policy on sock rerns. The esimae of β 2 is and saisically significan. Ths we esimae ha a 1% increase in he Federal Fnds Rae redces sock rerns by 6.652%. The esimaed impac is abo one-hird larger han he esimaed impac of moneary policy sing daily daa and he ypical even-sdy framework, and abo wo-hirds larger han he esimaed impac of policy sing he inra-day daa in he even sdy framework. Ths he wo approaches sggesed o deal wih he join response bias resl in esimaes ha move in opposie direcions relaive o he radiional Kner approach. We also invesigae how he sae of he sock marke, or he sae of he bsiness cycle, impacs he response of sock rerns o moneary policy. In Tables 4 and 5 we repor resls for he impac of moneary policy in bll and bear markes. In Table 4 we esimae from daily daa ha he impac of moneary policy on sock rerns in a bll marke is b saisically insignifican, while he impac in a bear marke is and saisically significan a he 1% level. Frher, a es of he hypohesis ha hese wo esimaes are he same has a p-vale of abo 2%, indicaing we wold 11

12 rejec his hypohesis. I seems ha moneary policy has a mch larger impac on sock rerns in a bear marke. Table 4 also repors esimaes of he impac of moneary policy from inraday daa, and while he paern of resls is similar he esimaed magnide in a bear marke is mch lower sing inraday daa. We esimae ha he impac of moneary policy in a bll marke is and saisically significan a he 10% level, while he impac of moneary policy in a bear marke is and saisically significan a he 1% level. A es of he hypohesis ha hese impacs are idenical across saes has a p-vale of.173 and hence wold no be rejeced. Ths he inraday daa finds a mch smaller impac of moneary policy in a bear marke compared o he esimae from daily daa. Table 5 repors or esimaes of he bear marke/bll marke differenial sing he Thornon mehodology. Here we esimae ha he impac of moneary policy in a bll marke is and saisically insignifican, while he esimaed impac in a bear marke is and saisically significan a he 1% level. Frher, a es of he hypohesis ha hese wo impacs are idenical across sae has a p-vale of js nder 1%, and hence wold be rejeced. The coefficiens ha conrol for he join response bias are saisically insignifican. Or varios esimaes condiioned on he sae of he sock marke follow he general paern we fond wih or ncondiional esimaes. Tha is, esimaes from or daily daa for he impac in recessions find a mch sronger impac of moneary policy on sock rerns compared o eiher of or esimaes sing or inraday daa. Thornon s mehodology finds a sronger impac of moneary policy han he radiional Kner approach, while he inraday daa indicaes a mch weaker effec (hogh sill significan saisically and economically). Or final resls examine how he sae of he bsiness cycle changes he 12

13 magnide of he impac of moneary policy on sock rerns. Table 6 repors resls, sing daily daa, ha he esimaed impac of moneary policy on sock rerns in an expansion is and saisically significan a he 5% level, while he impac in a recession is and saisically significan a he 1% level. The p-vale for he hypohesis ha hese impacs are idenical across sae is.180, indicaing ha his hypohesis wold no be rejeced. For or inraday daa, we esimae he impac of moneary policy dring an expansion as , saisically significan a he 5% level, and he impac dring a recession as , saisically significan a he 1% level. The hypohesis ha he impac is he same across saes is no rejeced he p-vale is Finally, sing Thornon s mehodology, repored in Table 7, we esimae he impac of moneary policy in an expansion as and saisically insignifican, he impac dring recessions as and saisically significan a he 1% level, and he p-vale for he hypohesis ha he impac is consan across saes is.08, indicaing we wold no rejec a he 10% significance level. We find no saisically significan evidence of a join response bias. Again, we find ha wih daily daa Thornon s approach indicaes a sronger response o moneary policy acions han Kner s radiional approach, while sing inraday daa we find a weaker response o moneary policy acions. Again, sronger or weaker refer o magnide, as all resls are saisically significan. Overall, or resls ell a consisen sory. Moneary policy has a srong negaive impac on sock rerns. Or esimaes wih daily daa, wheher sed in he even sdy framework or in Thornon s approach, are larger in magnide han or esimaes sing inraday daa. We find ha moneary policy has a mch sronger negaive impac on sock rerns in a bear marke compared o a bll marke. Again, we find ha or esimaes of he magnide of he impac of moneary policy on sock rerns 13

14 in a bear marke is mch larger sing daily daa compared o inraday daa. Finally, we find ha moneary policy has sronger impacs on sock rerns when he economy is in a recession compared o when he economy is in an expansion. We also find he general paern ha he esimaed magnide of he impac in a recession is larger when we se daily daa compared o inraday daa. We also find no saisically significan join response bias, alhogh or ncondiional poin esimae indicaes an overall posiive bias, and or condiional poin esimaes indicae a posiive bias in he condiional esimaes when we are in a bear marke, or a recession. Wha can be made of he conflicing resls regarding he sign of he join response bias indicaed by he wo approaches, he Thornon mehodology applied o daily daa and he inraday daa approaches? In Table 1 we show ha sock rerns on even days average 0.240%, wih 0.325% occrring prior o he even window, % dring he even window, and almos zero occrring afer he even window. Frher, he average sock rern from all days (non-even days and even days) is %, again almos zero. Recenly Lcca and Moench (2013) have addressed his isse, labeling i pre- FOMC annoncemen drif, and afer an exensive invesigaion hey prononce his phenomenon a conining pzzle. For or prposes, his pzzle clods he isse of he relaive sccess of he wo mehods for dealing wih he join response bias, as he so-far nexplained sizable sock rerns on FOMC days pre-even window clearly impac esimaes of he impac of moneary policy in any daily daa approach inclding Thornon s mehodology, while his phenomenon does no (direcly) impac esimaes based on he narrow even window. Wheher we wan he policy impac o inclde he pre-annoncemen drif is an open isse, he resolion of which will depend on he resolion of he pre-fomc annoncemen drif pzzle. 14

15 6. Conclsion We invesigae he sing Thornon s mehodology o correc for a possible join response bias in he esimaed impac of moneary policy acions on sock rerns, as an alernaive o sing inraday daa. We find a negaive impac of moneary policy on sock rerns, an impac magnified dring bear markes or dring recessions. We find poin esimaes indicaing a posiive join response bias, b hese are no saisically significan. Thornon s approach and he even sdy approach sing daily daa provide similar esimaes of he impac of moneary policy on sock rerns. Inraday daa provide he same qaliaive paern of resls, b he esimaed magnide of he impac of moneary policy on sock rerns is smaller sing inraday daa compared o eiher approach sing daily daa. 15

16 References: Ammer, J., Vega, C., and Wongswan, J. (2010), Inernaional Transmission of U.S. Moneary Policy Shocks: Evidence from Sock Prices, Jornal of Money, Credi and Banking, 42(6), Basisha, A. and Krov, A. (2008), Macroeconomic Cycles and he Sock Marke s Reacion o Moneary Policy, Jornal of Banking and Finance, 32, P Bernanke, B. S. and Kner, K. N. (2005), Wha Explains he Sock Marke s Reacion o Federal Reserve Policy? Jornal of Finance, 6(3), Chlia, H., Marens, M., and van Dijk, D. (2010), Asymmeric Effecs of Federal Fnds Targe Rae Changes on S&P100 Sock Rerns, Volailiies and Correlaions, Jornal of Banking and Finance, 34, Ehrmann, M. and Frazscher, M. (2004), Taking Sock: Moneary Policy Transmission o Eqiy Markes, Jornal of Money, Credi, and Banking, 36(4), Farka, M. (2009), The Effec of Moneary Policy Shocks on Sock Prices Acconing for Endogeneiy and Omied Variable Biases, Review of Financial Economics, 18, Gerler, M. and Gilchris (1994), Moneary Policy, Bsiness Cycles, and he Behavior of Small Manfacring Firms Qarerly Jornal of Economics, Go, H. (2004), Sock Prices, Firm Size, and Changes in he Federal Fnds Rae Targe, The Qarerly Review of Economics and Finance 44, Grkaynak, R., Sack, B. and Swanson, E. (2007), Marke-Based Measres of Moneary Policy Expecaions, Jornal of Bsiness and Economics Saisics, 25(3), Hasman, J. and Wongswan, J. (2011), Global Asse Prices and FOMC Annoncemens, Jornal of Inernaional Money and Finance, 30, Jansen, D. and Tsai, C. (2010), Moneary Policy and Sock Rerns: Financing Consrains 16

17 and Asymmeries in Bll and Bear Markes Jornal of Empirical Finance, 17, Krov, A. (2010), Invesor Senimen and he Sock Marke s Reacion o Moneary Policy, Jornal of Banking and Finance, 34, Kner, K. N. (2001), Moneary Policy Srprises and Ineres Raes: Evidence from he Fed Fnds Fres Marke, Jornal of Moneary Economics, 47, Lcca, D. O., and Moench, E. (2013), The Pre-FOMC Annoncemen Drif, Federal Reserve Bank of New York Saff Repor, No Pagan, A. R. and Sossonov, K. A. (2003), A Simple Framework for Analyzing Bll and Bear Markes, Jornal of Applied Economerics, 18, Thornon, D. (2013), "The Idenificaion of he Response of Ineres Raes o Moneary Policy Acions Using Marke-Based Measres of Moneary Policy Shocks", Oxford Economics Paper, forhcoming. Thorbecke, W. (1997), On Sock Marke Rerns and Moneary Policy, Jornal of Finance 52,

18 Figre1: Bll and Bear Markes, Janary 1994 Jan December The shaded areas nderneah he sock marke index represen bear markes. The y-axis is he log of he S&P 500 index. Pagan and Sossonov s mehod idenifies marke peaks as Ags 2000 and Oc 2007; marke roghs are Jne 1994, and Sepember Bll markes are rogh o peak, bear markes are peak o rogh. 18

19 Table 1: Descripive Saisics (Sample Period: Mar Dec ) Mehod 1: Even Sdy Mehodology---Daily Daa The Nmber of FOMC Meeing Days 103 Mean Vale (Sd Deviaion): Srprise Federal Fnds Rae Change (0.051) S&P 500 Sock Rerns (%) 0.240(0.989) Mehod 2: Even Sdy Mehodology---Inraday Daa The Nmber of FOMC Meeing Days 103 Srprise Federal Fnds Rae Change wihin window (0.498) S&P 500 Sock Rerns (%) -- Previos day s close o 10 mines prior o he annoncemen 0.325(0.586) S&P 500 Sock Rerns (%) wihin he window (0.525) S&P 500 Sock Rerns (%) mines afer he annoncemen o marke close 0.000(0.008) Mehod 3:Join-response Bias Mehodology---Daily Daa The Nmber of Daily Observaions 3231 Srprise Federal Fnds Rae Change (all days) (0.032) S&P 500 index Sock Rerns (%) (all days) 0.001(0.029) 19

20 Table 2 : The Impac of Moneary Policy on Sock Rerns, Even Sdy Approaches. Daily Daa, Even Sdy Inraday Daa, Even Sdy β ***(1.825) ***(1.388) Noes: 1. Esimaes of eqaion (5). 2. Window size for inraday daa is 40 mines, 10 mines prior o 30 mines pos he FOMC annoncemen. 3. Parenheses conain robs sandard errors. 4. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 20

21 Table 3: The Impac of Moneary Policy on Sock Rerns, Thornon Mehodology. Thornon Mehodology β (1.917) β **(2.703) Noes: 1. Esimaes of eqaion (6). 2. Parenheses conain robs sandard errors. 3. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 21

22 Table 4 : The Impac of Moneary Policy on Sock Rerns in Bll and Bear Markes, Even Sdy Approaches Daily Daa, Even Sdy Inraday Daa, Even Sdy β 1 (The Impac in Bll ) (1.904) *(1.699) β 1 + β 2 (The Impac in Bear) ***(5.078) ***(1.155) Difference P vale 0.018** Noes: 1. Esimaes of eqaion (7). 2. Parenheses conain robs sandard errors. 3. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 22

23 Table 5 : The Impac of Moneary Policy on Sock Rerns in Bll and Bear Markes, Thornon Mehodology Thornon Mehodology β (2.015) β (2.856) β (2.970) β ***(5.808) β 2 (The Impac in Bll) (2.856) β 2 + β 4 (The Impac in Bear) ***(5.057) Difference P vale 0.009*** Noes: 1. Esimaes for eqaion (8). 2. Parenhesis conains robs sandard errors. 3. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 23

24 Table 6: The Impac of Moneary Policy on Sock Rerns in Expansion and Recession Cycles, Even Sdy Approaches Daily Daa, Even Sdy Inraday Daa, Even Sdy β 1 (The Impac in Expansion ) **(1.986) **(1.628) β 1 + β 2 (The Impac in Recession) ***(2.985) ***(0.637) Difference P vale Noes:1. Esimaes of eqaion (7). 2. Parenheses conain robs sandard errors. 3. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 24

25 Table 7: The Impac of Moneary Policy on Sock Rerns in Expansion and Recessions, Thornon Mehodology Thornon Mehodology β (2.479) β (3.223) β (3.197) β *(4.888) β 2 (The Impac in Expansion ) (3.223) β 2 + β 4 (The Impac in Recession) ***(3.676) Difference P vale 0.080* Noes: 1. Esimaes of eqaion (8). 2. Parenheses conain robs sandard errors. 3. *** indicaes significan a he 1% confidence level; ** significan a 5%; * significan a 10 %. 25