Montary Policy Shocks and Stock Rtrns: Evidnc from th British Markt A. Grgorio a, A. Kontonikas b*, R. MacDonald b, A. Montagnoli c a Norwich Bsinss School, Univrsity of East Anglia b Dpartmnt of Economics, Univrsity of Glasgow c Dpartmnt of Economics, Univrsity of Stirling May 2009 Abstract This papr xamins th impact of anticipatd and nanticipatd intrst rat changs on aggrgat and sctoral stock rtrns in th UK. Th montary policy shock is gnratd from th chang in th thr-month strling LIBOR ftrs contract. Rslts from tim-sris and panl analysis indicat an important strctral brak in th rlationship btwn stock rtrns and montary policy shifts. Particlarly, whil bfor th crdit crnch th stock markt rspons to both xpctd and nxpctd intrst rat changs is ngativ and significant, th rlationship bcoms positiv dring th crdit crisis. Th lattr finding highlights th inability, so far, of montary policy-makrs to rvrs, via intrst rat cts, th ngativ trnd obsrvd in stock prics sinc th onst of th crdit crisis. JEL classification: C33; E44; E52; G13. Kywords: Asst Prics; Montary Policy; Panl Data. Corrsponding athor: A. Kontonikas, Dpartmnt of Economics, Univrsity of Glasgow, Adam Smith Bilding, Glasgow, G12 8RT, UK. Email: A.Kontonikas@lbss.gla.ac.k 1
1. Introdction Whil montary policy objctivs ar xprssd in trms of macroconomic variabls sch as inflation and ral otpt, policy actions affct ths variabls indirctly and with a lag. On th othr hand, financial markts tnd to ract qickly to th rlas of nw information. Ths, th s of financial data shold hlp to idntify a mor dirct and immdiat ffct of changs in montary policy and improv or ndrstanding of th transmission mchanism sinc asst prics play a ky rol in svral channls. Among asst prics, stock prics ar typically closly monitord and ar commonly rgardd as bing highly snsitiv to conomic nws. As Brnank and Kttnr (2005) point ot, whn stablishing th link btwn montary policy changs and stock rtrns on shold accont for th possibility that anticipatd policy actions may hav alrady bn incorporatd by markt participants into thir invstmnt dcisions. Brnank and Kttnr (2005) s Kttnr s (2001) ftrs mthodology to dcompos th fdral fnds rat changs into xpctd and nxpctd and find that an nanticipatd montary policy tightning xrts a ngativ impact on th US stock markt. Brdin t al. (2007) adopt a similar mpirical approach sing UK data and find that nanticipatd policy changs hav a significant impact on both aggrgat and th majority of th sixtn sctoral stock rtrns that thy mploy. In this papr w xamin th impact of anticipatd and nanticipatd actions by th Bank of England Montary Policy Committ (MPC), on UK aggrgat and sctoral stock rtrns. Th montary policy shock is gnratd from th chang in th thr-month strling LIBOR ftrs contract at th tim of th MPC mting and th sampl priod rns from Jn 1999 to Novmbr 2009. Unlik Brdin t al. (2007), w tilis both tim-sris and panl rgrssion analysis. In th contxt of th lattr, w tiliz both th cross sctional and tim sris proprtis of th data as wll as ncapslat possibl ndognity and joint dtrmination by mploying a GMM panl stimator, in addition to th standard OLS panl stimator. Morovr, w incorporat a highr lvl of sctoral disaggrgation to stimat th impact of montary policy annoncmnts on th UK stock markt. 2
Finally, w ar th first stdy, to th bst of or knowldg, to invstigat th possibility of a strctral brak in th rlationship btwn intrst rat changs and stock rtrns arond th rcnt crdit crnch. Or rslts ar intrsting and novl indicating a significant chang in th natr of th stock markt rspons to montary policy shifts sinc Agst 2007. Th rst of th papr is strctrd as follows. Th nxt sction dscribs th stock markt data and th calclation of th montary policy shock. Sction 3 prsnts th mpirical modls and rslts. Sction 4 conclds. 2. Data Or sctoral stock rtrns datast covrs svnty FTSE Indstrial sb-sctors which form th tn basic UK indstris: oil & gas, basic matrials, indstrials, consmr goods, halthcar, consmr srvics, tlcommnications, tilitis, financials, and tchnology. W masr stock rtrns for sbsctor i at day t that th Montary Policy Committ (MPC) mts, y it,, as th first diffrnc of th natral log of th daily closing stock pric (S it ): yit = 100*(ln Sit ln Si, t 1). Aggrgat stock rtrns, agg y t, ar masrd as th first diffrnc of th natral log of th daily closing val of th FTSE 100 agg indx (FTSE t ): yt = 100*(ln FTSEt ln FTSEt 1). Following Kttnr (2001), w s data from intrst rat ftrs contracts in ordr to driv th montary policy shock. In th UK, thr is no ftrs contract tracking th MPC-controlld policy instrmnt (sch as th two-wk rpo rat sd in US-basd stdis). Th closst sbstitt that xists, and which w s, is a ftrs contract basd pon th thr-month LIBOR rat, and this rat is widly accptd as a vry good indicator of markt xpctations of ftr policy changs. 1 Ths, th montary policy shock, Δ i, is proxid by th chang in th rat of th thr-month strling 1 Th ftrs contract is basd on th British Bankrs Association London Intrbank Offrd Rat (BBA LIBOR) for thr month strling dposits at 11:00 on th last trading day. Th sttlmnt pric is 100 mins th BBA LIBOR rondd to thr dcimal placs. Lindholdt and Wthrilt (2004) mploy LIBOR rats at varios matritis and find that thr has bn a clar improvmnt in th ability of th markt to forcast policy rat changs by th Bank of England. 3
LIBOR ftrs contract as tradd on th Eronxt/LIFFE markt, rlativ to th day bfor th (monthly) MPC mting: Δ it = fm, t fm, t 1 (1) whr f mt, is th implid ftrs rat (100 mins th ftrs contract pric) associatd with th contract that xpirs on th month that th MPC mts. 2 Finally, w masr th xpctd chang in intrst rats, Δ i, as th actal chang in th thr-month LIBOR rat mins th srpris: Δ i =Δi Δ i (2) t t t Th sampl priod ndr invstigation is Jn 1999 March 2009, providing s with 119 MPC mtings, which will b th tim-sris dimnsion in th panl analysis. 3 3. Economtric modls and rslts 3.1 Tim-sris modls: Aggrgat rtrns W start or mpirical invstigation by rgrssing FTSE 100 rtrns on xpctd and nxpctd intrst rat changs: y = α + β Δ i + β Δ i + (3.1) agg t t t t whr th rror trm, t, rprsnts factors othr than montary policy that affct th stock markt on vnt days. [TABLE 1 HERE] 2 For xampl, to calclat th montary policy shock associatd with th 9/06/2005 MPC mting, w s th rat implid by th thr-month LIBOR ftrs contract that xpirs on 13/06/2005. Kttnr (2001) and Brnank and Kttnr (2005), in thir analysis of US montary annoncmnts adjst th fdral fnds rat for th nmbr of days rmaining in th month. This is ncssary in thir cas bcas thy s th fdral fnds ftrs rat, th sttlmnt of which is basd pon th avrag fd fnds rat of th last month in th ftrs lif. W do not adjst or LIBOR rats for th nmbr of rmaining days in month bcas in th UK th 3 month LIBOR ftrs sttls at th 3m-libor of th last trading day. 3 Th particlar start dat was chosn to tak into accont two constraints. First, th Montary Policy Committ mtings commncd on Jn 1997, and scond, prior to Jn 1999 th Libor ftr contracts did not hav dlivry dat for ach month of th yar. Brdin t al. (2007) s an arlir start dat (1993-2004), which maks th pr-jn 1997 analysis hard to intrprt givn th lack of schdld monthly mtings. 4
Th OLS rslts in colmn 2 of Tabl 1 indicat that th stimatd stock markt rspons to both th xpctd and srpris componnt of montary policy changs is statistically insignificant. Modl 3.1 dos not appar to b wll spcifid sinc th rsidals sffr from both htroscdasticity and srial corrlation. Visal inspction of th rsidal sris from this modl indicats th prsnc of two vry larg ngativ otlirs on Agst 2002 and Octobr-Novmbr 2008. Ths xtrm obsrvations ar associatd with th stock markt downtrns of 2002 and 2008. Particlarly, lat in th smmr of 2002 th dot-com bbbl brst startd raching its final stag, whil atmn 2008 was charactrizd by th collaps of th Lhman Brothrs at th pak of th crdit crnch. To accont for ths otlirs, w agmnt Modl 3.1 by inclding two additiv dmmy variabls: agg y = α + γ1dlhman + γ2d2002 + β Δ i + β Δ i + (3.2) t t t t t t whr DLhman t is qal to 1 dring Octobr-Novmbr 2008 and 0 othrwis, and D2002 t is qal to 1 on Agst 2002 and 0 othrwis. From Tabl 1 colmn 3 w notic that th fit of th modl improvs considrably with th adjstd R 2 incrasing by 31%, whil th rsidals ar now fr of htroscdasticity and srial corrlation. Apart from th two additiv dmmy variabls, which ar highly significant, only on of th montary policy chang variabls is significant, th xpctd componnt, and that at th 10% lvl only. Th srpris componnt ffct, whil xhibiting a ngativ sign in lin with US vidnc by Brnank and Kttnr (2005), is not significantly diffrnt from zro. Ths findings may b th rslt of strctral instability in th stock markt rspons to montary policy changs. Indd, from th onst of th crdit crisis on Agst 2007 onwards, qity markt participants hav witnssd dcrasing valation in tandm with sharp cts in intrst rats. This sggsts a positiv, as opposd to th typically ngativ, association btwn stock markt rtrns and intrst rat changs. In ordr to accont for possibl strctral chang dring th financial crisis w intract th xpctd and nxpctd componnts of montary policy changs with a (mltiplicativ) dmmy variabl: 5
y = α + γ DLhman + γ D2002 + ( β + δ DCrisis ) Δ i + ( β + δ DCrisis ) Δ i + (3.3) agg t 1 t 2 t 1 t t 2 t t t whr DCrisis t is qal to 1 from Agst 2007 onwards and 0 othrwis. Th rslts in Tabl 1 colmn 4 ar intrsting and novl. Th adjstd R 2 incrass to 43% and, xclding th constant, all variabls ar statistically significant at th 5% lvl or lss. Th stock markt rspons to both xpctd and nxpctd intrst rat changs is ngativ. Hnc, or rslts ar similar to Brnank and Kttnr (2005) rgarding th lattr, bt diffr from thm rgarding th formr. Th pr-crdit crnch rslts imply a 3.26% on-day rtrn in rspons to a 50 basis points srpris rat incras. 4 Howvr, dring th crdit crnch th stimatd cofficint of both srpris and xpctd changs is positiv. This sggsts that montary policy makrs hav not bn abl so far to boost stock markts by intrst rat cts highlighting th svrity of th crrnt financial trmoil. 3.2 Tim-sris modls: Sctoral rtrns Or analysis has so far considrd th rspons of a broad stock indx to montary policy. Nvrthlss, it is also important to xamin th stock markt raction across diffrnt sctors in ordr to xamin if any intrsting pattrns aris. Ths, w stimatd Modl 3.3 for ach of th 70 sctors in or datast. Ovrall, th sctoral rslts ar similar to th broad indx basd rslts in trms of th signs of th stimatd cofficints, indicating a ngativ (positiv) raction on avrag to intrst rat changs bfor (dring) th crdit crnch. [FIGURE 1 HERE] As w s in Figr 1, which plots th histogram of th stimatd sctoral cofficints, arond 80% of th xpctd and srpris-rlatd cofficints ar ngativ bfor th crdit crnch. Dring th financial crisis, howvr, mor than thr qartrs of stimatd cofficints ar positiv indicating 4 W rpatd th analysis sing th FTSE All Shar rtrns and th rslts (availabl pon rqst) wr vry similar. Morovr, w agmntd Modl 3.3 by inclding DCrisis in an additiv mannr. Rslts (availabl pon rqst) wr vry similar. 6
an important chang in th natr of th stock markt rspons to policy shifts. Som intrsting diffrncs across th sctors incld th cas of nin indstris whr thir rspons to (xpctd and nxpctd) policy shifts bfor th crdit crisis is opposit to avrag bhavior. 5 Also, thr ar twlv sctors which xhibit ngativ rspons to (xpctd and nxpctd) policy shifts dring th crisis. 6 Ths findings sggst that portfolio divrsification opportnitis, with rspct to montary policy changs, ar not vry common bt nvrthlss do xist. 3.3 Panl modls For compltnss, w stimat th rlationship btwn montary policy and stock rtrns at a sctoral lvl sing panl stimators. W first considr th OLS fixd ffcts modl: y = α + X θ + (4) ' it i it it whr yit is dfind as abov, α i is th tim-invariant nobsrvd sctor-spcific fixd ffct, X is th vctor of th xplanatory variabls and θ is a vctor of cofficints. W xamin two altrnativ cass rgarding th xplanatory variabls vctor in Eq. (4). In Modl 4.1 it inclds th variabls of Modl 3.3, whil in Modl 4.2 w add th laggd dividnd yild (dy it-12 ) in ordr to xamin th robstnss of or rslts with rspct to th prsnc of an additional control variabl rlatd to th financial position of ach sb-sctor. 7 it On potntial drawback with th OLS modl is that it dos not dal with th likly prsnc of ndognity in or data. Althogh an Instrmntal Variabl (IV) stimator wold addrss sch 5 Ths sctors ar tility companis, nvironmntal srvics, gambling and casino facilitis, indstrial machinry and factory qipmnt, mploymnt srvics, prodcrs and distribtors of pns, papr goods, light blbs, battris and claning prodcts, insranc brokrs, gas distribtors to nd srs and tobacco. 6 Ths sctors ar nvironmntal srvics, prodcrs and distribtors of pns, papr goods, light blbs, battris and claning prodcts, manfactrrs and distribtors of varios drabl hoshold goods, soft drinks, insranc brokrs, proprty and casalty insranc, tobacco, hom improvmnt rtailrs, apparl rtailrs, mdical spplis, tility companis and food prodcts. 7 Brdin t al. (2007) also agmnt thir baslin tim-sris rgrssion modls sing, howvr, not financial position variabls bt othr variabls inclding aggrgat stock rtrns in slctd Eropan contris and th xchang rat. 7
ndognity it, in trn, fails to captr th cross-sctoral htrognity in or data st. Thrfor, in ordr to tackl both th ndognity and cross-sctoral htrognity in th data w s a GMM stimator. Spcifically w s th GMM stimator dvlopd by Arllano and Bond (1991) which maks s of intrnal instrmnts for ach tim priod to tackl ndognity. conditions: ( it z it) E y E = holds for z t across all th sctors thn it rprsnts th following momnt If ( ) 0 it, 0 iz Δ = for z 2; t = 3,..., T. (5) If X it ar wakly xognos thn w also hav th following additional momnt conditions: ( it z it) E X Δ = for z 2; t = 3,..., T. (6), 0 Th singl qation GMM panl stimator gnrally spcifis a dynamic panl modl in first diffrncs and xploits th abov momnt conditions. 8 Thrfor, th laggd (two tim priods or mor) lvls of ndognos and wakly ndognos variabls of th modl bcom appropriat instrmnts for addrssing ndognity. Th GMM panl stimator provids consistnt cofficint stimats. [INSERT TABLE 2 HERE] Th panl stimation rslts ar prsntd in Tabl 2. Using OLS, th fixd ffcts of th panl ar significant at th 10% only whn th laggd dividnd yild is addd (Modl 4.2), whil with GMM fixd ffcts ar significant across both spcifications. Th Sargan tst rslts confirm th validity of th instrmnts in th GMM modl. OLS and GMM stimats of Modls 4.1 and 4.2 provid similar rslts which ar in lin with th broad indx findings. Particlarly, th stock markt rspons to both xpctd and nxpctd intrst rat changs is significantly ngativ bfor th 8 Th modl is transformd into first diffrncs in ordr to liminat th fixd ffcts. 8
crdit crnch and positiv throghot it. Th additiv dmmy variabls associatd with th stock markt downtrns of 2002 and 2008 ar statistically significant. Finally, th laggd dividnd yild cofficint is positiv and significant which is in conformity with th litratr on th prdictability on stock rtrns whr highr dividnd yild is associatd with highr ftr rtrns (s.g. Campbll t al., 1997). 4. Conclsions This papr invstigats th impact of anticipatd and nanticipatd montary policy annoncmnts of th Bank of England s Montary Policy Committ (MPC) on UK aggrgat and sctoral stock rtrns. Th montary policy shock is gnratd from th chang in th thr-month strling LIBOR ftrs contract for a sampl priod from Jn 1999 to March 2009. Using tim-sris and panl rgrssion modls w show that both th xpctd and nxpctd componnts of montary changs impact significantly on stock rtrns. Or rslts docmnt an important strctral brak in th stock markt raction to montary policy changs sinc th onst of th crdit crisis. Whil bfor th crisis th stock markt rsponds ngativly to highr intrst rats, th stock rtrns-intrst rat chang rlationship bcoms positiv dring th crdit crnch. Th lattr finding indicats that, so far, highly xpansionary montary policy has not bn abl to rvrs th ngativ trnd in stock prics highlighting th svrity of th ongoing crisis. 9
Rfrncs Arllano, M. and S. Bond, 1991. Som Tsts of Spcification for Panl Data: Mont Carlo Evidnc and an Application to Employmnt Eqations, Rviw of Economic Stdis, 58, 277-297. Brnank, B. and K. Kttnr, 2005. What Explains th Stock Markt's Raction to Fdral Rsrv Policy?, Jornal of Financ, 60, 1221-1257. Brdin, D., Hyd, S, Nitzsch, D., and G. O Rilly, 2007. UK Stock Rtrns and th Impact of Domstic Montary Policy Shocks, Jornal of Bsinss Financ and Acconting, 34, 872-888. Campbll, J.Y., Lo, A. and C. MacKinlay, 1997. Th Economtrics of Financial Markts, Princton Univrsity Prss. Kttnr, K., 2001. Montary Policy Srpriss and Intrst Rats: Evidnc from th Fd Fnds Ftrs Markt, Jornal of Montary Economics, 47, 523-544. Lildholdt, P. and A. Wthrilt, 2004. Anticipation of Montary Policy in UK Financial Markts, Bank of England Working Papr 241. 10
TABLES Tabl 1: OLS stimats, FTSE 100 rtrns. Variabl Modl 3.1 Modl 3.2 Modl 3.3 constant -0.22 (1.9) * -0.13 (1.34) -0.1 (1.00) DLhman t - -5.5 (6.48) *** -5.34 (4.94) *** D20002 t - -4.83 (4.67) *** -4.96 (4.93) *** Δ i t -2.87 (1.01) -4.08 (1.75) * -8.17 (3.1) *** DCrisis t * Δ i t - - 14.74 (2.98) *** Δ i t 2.97 (1.02) -3.3 (1.3) -6.52 (2.27) ** DCrisis t * Δ i t - - 12.18 (2.22) ** Diagnostics Adjstd R 2 0.08 0.39 0.43 Q(1) 4.837 ** 0.098 0.186 Q 2 (1) 10.604 *** 0.314 0.219 Ht 4.475 ** 0.507 0.303 NOTES: This tabl rports OLS stimats of Modls 3.1-3.3 ovr th tim priod Jn 1999-March 2009 (119 MPC mtings). Figrs in parnthss rprsnt th absolt val of th t-statistic associatd with th nll hypothsis that th cofficint is qal to zro. Q, Q 2 dnot th Ljng-Box Q, Q 2 tst statistics for srial corrlation and ARCH-typ volatility clstring. Ht is th Brsch-Pagan-Godfry F tst statistic for htroscdasticity. ***, **, * indicat statistical significanc at th 1, 5, 10 % lvl, rspctivly. 11
Tabl 2: Panl stimats, sctoral stock rtrns. Variabl OLS GMM Modl 4.1 Modl 4.2 Modl 4.1 Modl 4.2 constant -0.14 (5.91) *** -0.39 (5.6) *** -0.13 (4.66) *** -0.11 (4.03) *** DLhman t -3.61 (13.35) *** -3.5 (13.18) ** -3.79 (11.62) *** -3.62 (11.56) *** D20002 t -2.63 (10.47) *** -2.58 (10.48) *** -3.13 (10.13) *** -2.78 (9.46) *** Δ i t -4.66 (7.07) *** -4.71 (7.09) *** -4.11 (5.56) *** -4.01 (5.67) *** DCrisis t * Δ 9.81 (7.93) *** 9.6 (7.85) *** 8.12 (5.61) *** 7.65 (5.01) *** i t Δ i t -4.11 (5.73) *** -2.39 (3.04) *** -3.58 (4.08) *** -3.32 (4.16) *** DCrisis t * Δ i t 8.21 (5.98) *** 6.49 (4.68) *** 5.43 (3.32) *** 5.12 (3.05) *** dy it-12-0.07 (3.55) *** - 0.06 (2.82) *** Diagnostics α i [0.43] [0.06] [0.00] [0.00] Sargan χ 2 () r - - [0.49] [0.55] NOTES: This tabl rports panl stimats of Modls 4.1 and 4.2 ovr th tim priod Jn 1999-March 2009 (119 MPC mtings) across 70 sctors. Figrs in parnthss rprsnt th absolt val of th t-statistic associatd with th nll hypothsis that th cofficint is qal to zro. Sargan tsts follow th X 2 distribtion with r dgrs of frdom ndr th nll hypothsis of valid instrmnts. Th ndognos xplanatory variabls in th panl ar GMM instrmntd stting z 3. α i ar fixd ffcts. [.] ar p vals***, **, * indicat statistical significanc at th 1, 5, 10 % lvl, rspctivly. 12
Figr 1: Histograms of OLS sctoral stimats. Δ i DCrisis* Δ i 9 8 7 6 5 4 3 2 1 0-18 -16-14 -12-10 -8-6 -4-2 0 2 4 20 16 12 8 4 0-30 -20-10 0 10 20 30 40 50 Δ i DCrisis* Δ i 14 16 12 14 10 12 8 6 10 8 6 4 4 2 2 0-15 -10-5 0 5 0-40 -30-20 -10 0 10 20 30 40 50 NOTES: Ths figrs show th histograms of th stimatd sctoral cofficints from Modl 3.3. Th hadr abov ach figr dnots th variabl corrsponding to th aformntiond cofficint. 13