Short Term Capital Flows and Pressure on the Exchange Rate in Kenya 1

Transcription

1 Shor Term Cpil Flows nd Pressure on he Exchnge Re in Keny 1 Benjmin Ongwe Muru Absrc: Using he Byesin vecor uo-regression mehodology, we find, rher surprisingly, h posiive 1 sndrd deviion shock o ne shor erm cpil inflows induces significn depreciing effec, on impc, on he shilling-usa dollr exchnge re. We would like o hink h hese rher prdoxicl resuls owe o d mesuremen errors. The vrince decomposiion resuls re consisen wih he impulse response resuls becuse ne shor erm cpil inflows ply, pprenly, ply limied direc role in explining ou-of-smple exchnge re forecs errors. In conrs, we find h posiive shock o he risk djused ineres re differenils hs sisiclly significn nd immedie incresing effec on ne shor erm cpil inflows. However, he incresing effec swiches o reducing one wihin yer beyond which he reducing effec is sisiclly significn. These resuls re suggesive of he volile nure of shor erm cpil flows in response o sudden chnges in cross-border risk djused ineres re differenils nd h here is poenil for sudden shor erm cpil reversls in Keny. The vrince decomposiion resuls show h he risk djused ineres re differenils ccoun for 71.% of he one period hed forecs error in he exchnge re. Since he risk djused ineres re differenils is crucil fcor h consisenly drives he ne shor erm cpil inflows in Keny, brring d mesuremen errors, we infer h ne shor erm cpil inflows will in urn induce consisen influence on he exchnge re. 1 This pper is prepred for presenion he Cenre for Sudies on Africn Economies (CSAE) 1 Annul Inernionl Conference scheduled for 3-5 h Mrch 1. An erly drf of he pper ws presened he Cenrl Bnk of Keny s Keny School of Monery Sudies Seminr on Keny Shilling Exchnge Re Dynmics; nd Ocober 13. Assisn Direcor nd Hed of he Finncil Secor Anlysis Division of he Reserch Cenre. Cn be reched wih commens nd suggesions on his sudy on: murubo@ksms.or.ke. Pge 1 of 53

2 1. Inroducion There is no doub h empiricl nlysis of he impc of cpil flows on exchnge res coninues o be n exciing nd relevn heme considering he incresed economic inerdependence mong counries. As relive price of foreign exchnge in erms of he domesic currency, he level nd chnges in he exchnge re hve he poenil o significnly influence he volume of inernionl rde nd overll mcroeconomic performnce s whole. For insnce, when evlued s relive price of counry s bundle of impored goods nd services in erms of bundles of he counry s expored goods nd services, he rel exchnge re is criicl elemen in deermining he counry s inernionl rde compeiiveness. Also, excessive exchnge re voliliy, which is he sudden chnge in he exchnge re, is kin o n implici x on inernionl business rnscions hs he poenil of reducing he volume of inernionl rde nd by exension consrining susinble economic nd humn developmen. I is herefore no surprising h compeiive exchnge re is one of he mos imporn policy objecives pursued by open economies. So imporn is his objecive h wihou sufficien exchnge re policy coordinion nd surveillnce mong rding prner counries, he sbiliy of he globl finncil sysem cn be jeoprdized by compeiive exchnge re policies mong counries. In recogniion of he economic impornce of sysem of sble exchnge res, boh in erms of exchnge re vlues h re consisen wih economic fundmenls s well s orderly exchnge re chnges h re devoid of excessive voliliy, counries re greed on he need for ensuring orderly policy induced exchnge re chnges under he Inernionl Monery Fund (IMF) Aricles of Associion. Under he uspices of he IMF, member counries cn pricipe, eiher singly or hrough he coordined effor of ffeced member counries monery uhoriies, o sbilize he foreign exchnge mrke by ridding he mrke of observed excessive exchnge re voliliy. Pge of 53

3 Anlyiclly, virully ll heories of exchnge re deerminion poin o he imporn role counry s ne foreign sses ply in he deerminion of he curren nd fuure expeced vlues of counry s exchnge re. The sse price view of he exchnge re is priculrly illumining s o how exchnge res re deermined nd Muss (19), for insnce, presens such n elegn model which inegres he sock nd flow equilibri condiions s proxime bses for he deerminion of exchnge res. Even he microsrucure pproch o exchnge re deerminion, which ccords foreign exchnge order flows, suggess h cpil flows ply he single imporn role in deerminion of exchnge res. Secondly, he microsrucure pproch o exchnge re deerminion predics h ne purchses of counry s rel sses, which re essenilly ccompnied wih corresponding ne cpil inflows, induces n ppreciion of he domesic currency. In conrs, nd o he exen h moneized ne cpil inflows ne shor erm cpil inflows in priculr, re essenilly nominl shocks, he Dornbusch (197) ype of exchnge re overshooing models such s Muss (19) predics h ne cpil inflows will, on impc, deprecie fr much more hn he depreciion brough bou by corresponding chnges in he exchnge re fundmenls o mke up for he sicky commodiy prices hereby ensuring momenry equilibrium in he goods mrke during he shor run. As goods prices increse grdully in response o he posiive nominl shock brough bou by he moneized cpil inflows, he exchnge re will be ppreciing grdully o in he iniilly oversho long run equilibrium vlue. As o which beween he wo compeing empiricl oucomes bes chrcerizes given counry, one cnno ell priori. While ps reserch effors hve led o imporn findings such s he Duch Disese nd he rilemm of monery policy mngemen (O Connell e l., 1), he need for counry-specific evidence regrding he cusl relionship beween cpil flows nd exchnge res is even greer now hn ever before. This is becuse, over he yers, mny counries hve opened heir cpil ccouns nd herefore esed resricions on cross-border cpil flows such h inernionlizion of inernionl finnce hs ssumed unprecedened volumes s discussed in Ole Rummel (1). Ye gin, he chllenge of mnging he dverse implicions of cpil Pge 3 of 53

4 flows on exchnge res is priculrly enormous for smll open economies wih shllow finncil mrke nd herefore relively more suscepible o finncil crises cused by dverse mrke senimen. The speed nd mgniude of impc of ne cpil flows on exchnge res is lso expeced o vry cross counries ccording o he previling differences in he srucure of cpil flows which differen counries rc nd he vried levels of finncil secor developmen cross counries. Generlly, long erm cpil flows re populr for filling in he domesic svings-invesmen gp nd for being clyic in promoing echnologicl progress in he cpil flow desinion counries, shor erm cpil flows (which re predominnly porfolio flows) re noorious for cusing finncil nd economic crises consequen o sudden flow reversls nd herefore deserve specil enion in empiricl nlysis nd mngemen of cpil flows. Avilble evidence on he impc of cpil flows on exchnge res is mixed nd seems o vry wih he nlyicl frmework pplied o rrive he empiricl resuls. For insnce, ne equiy cross-border flows re shown o consisenly led o currency ppreciion unlike cross-border bond flows which re relively more menble o hedging gins poenil deful or foreign exchnge re risk. I is repored, for insnce, in Hu nd Rey () h well over 9% of cross-border bond holdings re hedged compred o up o 1% hedging of equiy holdings. According o Siourounis (3), lso, ne purchses of USA equiies consisenly impc he Briish Pound Serling-USA dollr, Germny Deuschmrk-USA dollr nd he Swiss Frnc-USA dollr exchnge res whils purchses of USA or oher foreign bonds do no; presumbly becuse of hedging gins he foreign exchnge re risk. More specificlly, Siourounis (3) finds h 1 sndrd deviion posiive shock o ne USA equiy purchses leds o n ppreciion of he USA dollr by bou 1% nd he ppreciion is significn for up o 13 monhs. Similr resuls re repored by Hu nd Rey () whereby purchses of USA sses by foreigners pprecie he USA dollr. Moreover, in heir empiricl invesigion of he cusl implicions for mjor currencies bilerl exchnge res vis--vis he US dollr of he curren ccoun defici nd cpil flows, on he one hnd, nd he rdiionl underlying fcors simuling foreign cpil inflows, on he oher, Brooks e l. Pge of 53

5 (1) find h ne equiy reled cpil flows exer significn effec on he exchnge res. Also found o be imporn re long erm bilerl ineres re differenils nd relive sock reurns. Due o he limied vilbiliy of forwrd conrcs for hedging gins finncil risks in developing counries, one would expec h even medium erm nd long erm ne cpil inflows re imporn in influencing exchnge re movemens in he developing counries including Keny. In spie of his, however, i is useful o empiriclly nlyze he role of ne cpil inflows by is componens o obin resuls which re specific o he componens of ne cpil inflows h hve unique perns h enil unique policy responses. For insnce, shor erm cross-border cpil flows such s porfolio cpil flows re menble o sudden reversls nd herefore pose significn domesic finncil secor vulnerbiliy. O Connell e l. (1) conends h he srucure of ne cpil inflows mers. Keny presens priculrly exciing conex wihin which o derive comprive evidence regrding he impc of ne cpil inflows on he developing counries currency exchnge res. Keny hs liberlized finncil secor h hs been developing sedily over he ls one decde. Keny s cpil ccoun in he blnce of pymens is open nd she is signory o he Inernionl Monery Fund Aricle VIII which discourges member counries from imposing resricions on curren ccoun flows. Moreover, he Nirobi Sock Exchnge is one of he mos vibrn in sub-shrn Afric nd hs n cive foreign rding bord nd herefore rcs foreign porfolio invesmen. Mos impornly, since her dopion of flexible exchnge regime in June 1995, Keny hs experienced episodes of incresed shilling exchnge re voliliy which require empiricl invesigion o find ou he role plyed by ne shor erm cpil inflows s he usul suspec. 3 While vilble evidence, in for insnce Ndung u () nd O Connell e l. (1) sugges h ne foreign cpil inflows significnly influence shor run shilling exchnge re movemens hereby, brodly speking, suggesing h speculive foreign cpil flows end o 3 The Foreign Exchnge Conrol Ac (197) ws repeled in June 1995 (O Connell e l., 1). Pge 5 of 53

6 pile up pressure on he shilling exchnge re, here is need for upding he evidence o esime he specific role plyed by ne shor erm flows. The evidence is poenilly useful in ppreciing circumsnces during which he Cenrl Bnk of Keny my ke policy cion o rid he domesic inerbnk foreign exchnge mrke of excessive exchnge re insbiliy cused by speculive civiy. In he recen ps he Bnk hs hd o ke cion o discipline mrke pricipns o desis from such speculive behvior mid proess from some plyers in he foreign exchnge mrke. Adducing such evidence is poenilly useful o he Cenrl Bnk in supporing is policy cions in deling wih speculive civiies in he domesic foreign exchnge mrke. This sudy herefore empiriclly nlyzes he role of ne shor erm cpil inflows in he deerminion of Keny shilling-usa dollr exchnge re. More specificlly, we ddress wo reserch quesions nd which re: do chnges in ne shor erm cpil inflows exer sisiclly significn impc on he Keny Shilling-USA Dollr nominl exchnge re? Wh is he relive impornce of ne shor erm cpil inflows in explining he Keny Shilling-USA Dollr nominl exchnge re flucuions? We hve chosen he shilling-usa dollr nominl exchnge re s he bsis of he empiricl nlysis minly becuse he USA Dollr commnds he lrges weigh in Keny s officil reserves. I is lso he currency used mos by Keny in selemen of her inernionl finncil obligions. The USA dollr is lso noble in erms of is weigh in he rde weighed nominl or rel effecive shilling exchnge res. Is weigh is les 5%. I is lso he domesic foreign exchnge mrke inervenion currency. Among he key empiricl resuls of his sudy is, rher surprisingly, h 1 sndrd deviion shock o ne shor erm cpil flows hs significn depreciing impc on he Shilling-US Dollr exchnge re. The depreciing effec is followed by grdul correcion in erms of n ppreciion relive o he iniil depreciing effec which suggess h he Shilling-USA Dollr exchnge re movemens mnifes he exchnge re overshooing phenomenon. We lso find, bsed on he vrince decomposiion resuls, h ne shor erm cpil flows plys limied direc role in explining he exchnge re movemens. Pge of 53

7 In conrs, posiive 1 sndrd deviion shock o he risk djused ineres re differenils hs sisiclly significn nd immedie incresing effec on ne shor erm cpil inflows. The risk djused ineres re differenils ccouns for 71.% of he one period hed forecs error in he exchnge re. We hve orgnized his pper ino six secions. Upon review of he lierure in secion, we ouline he Byesin Vecor Auo-regressive (BVAR) mehodology which we use in his sudy. We hen, in secion, briefly explin how we ke he BVAR model o he qurerly d drwn from Q1-1Q. In priculr, we explin he choice of vribles h we hve used in he empiricl nlysis. We presen nd discuss he empiricl resuls in secion 5 nd hen summrize resuls nd conclude he pper in secion.. Lierure Review During he firs decde of he generlized flo of he mjor currencies, mny iniiives ken mde o model exchnge re deerminion for number of resons. Chief mong he resons ws he need for enhncing undersnding of he fcors which drove exchnge res movemens so policy mkers could choose pproprie policy mesures when he exchnge re were eiher misligned or excessively volile. See for insnce Edwrds (199) nd he exensive coverge of he nlyicl nd empiricl models of exchnge re mislignmen in Hickle nd Moniel (1999). The exchnge re models of he 197s, 19s nd he 199s, which include he monery model in is wo vrins of flexible nd sicky prices, he porfolio blnce model, he rel ineres re differenils model nd he rel produciviy differenils model, nd which re generlly referred o s rdiionl models of exchnge re deerminion urned ou no o be sufficienly useful in explining exchnge re movemens. In priculr, he models ou-ofsmple forecsing power ws severely limied (Meese nd Rogoff, 193) o he exen h hey were ou-performed by he rndom wlk model; nive model in which he relized vlue of he exchnge re is he bes prediced vlue of he exchnge re in he subsequen period. Pge 7 of 53

8 Chriss forecss could lso ouperform exchnge re forecss bsed on he economic fundmenls. This resul hs been difficul o overurn excep in isoled cse such s in Koedijk nd Sochmn (199). Among he resons dvnced o explin filure of he economic models of exchnge re deerminion is model miss-specificion whereby liner relionships beween he exchnge res seleced economic fundmenls re ssumed when he rue relionship is nonliner. Lck of sufficienly long ime spns of d on exchnge res nd economic fundmenls lso severely limied efficien esimion of he models using sympoic esimors such s Ordinry Les Squres (OLS). The success of Koedijk nd Sochmn (199), for insnce, follows he uhors dopion of wo key innovions in model esimion. The uhors expnd he informion spce used o esime he exchnge re models by exploiing he concep of cross-res nd lso pooling exchnge re d nd using pnel regression nlysis insed of sndrd regression model. The experience of empiricl modeling exchnge res during he hree decdes following he generlized flo cn be sid o hve lso idenified news o be criicl fcor in he deerminion of, especilly shor run, exchnge re movemens. I ws however duning sk finding correc mesure of he unobservble vrible h news ws nd hs been. I is lso resonble o rgue h, perhps one of he resons, he mcroeconomic fundmenls bsed exchnge re models underperformed ws due o indeque incorporion of expecions nd s such he models mus hve succumbed o he Lucs criique (Lucs, 197; nd Srgen, 1979). This specific limiion of he economic models is inferred from he beer performnce of he vinge dynmic sochsic generl equilibrium (DSGE) models, whose key feure is deque incorporion of rionl expecions. Techniclly, we could sy h news is bsiclly he process nd se of becoming wre of relevn informion h ws hihero unknown, bou fcors which drive he exchnge re. In priculr, i is he expecions error in he relized vlues of he exchnge re economic fundmenls mde by economic gens priciping in he foreign exchnge mrke. Wih he rrivl of he news, economic gens mke decisions which, mong oher consequences, upde heir expecions bou fuure exchnge res nd by cing consisen wih he revised exchnge re expecions, corresponding movemens in he exchnge res ensue. Pge of 53

9 For insnce, Thoenissen (3) pplies clibred DSGE model in which incomplee exchnge re pss-hrough is ssumed, o find h he curren ccoun, which bsiclly reflecs cpil nd finncil ccoun flows, significnly impc he rel exchnge re. Wheher chnge in he curren ccoun defici leds o n ppreciion or depreciion of he rel exchnge re will depend on he home counry s iniil ne foreign sses posiion. We should emphsize h he mos crucil conribuion in bridging he gp beween he rdiionl Keynesin pproch o mcroeconomics s encpsuled in he Mundell-Fleming model which is inroduced in Mundell (193,19) nd Fleming (19) nd underpinned by he ssumpion of perfecly compeiive mrkes, nd he inroducion of he new open economy mcroeconomics in which mrkes feure fricion s, for insnce, shown by he Dornbusch s exchnge re overshooing pper (Dornbusch, 197), is he Exchnge Re Dynmics Redux pper of Obsfeld nd Rogoff (199). An ccessible discussion of cenrl poins in he exchnge re dynmics redux pper is provided by Rogoff (). An imporn nlyicl resul deriving from open economy DSGE models is uncovered ineres priy (UIP) condiion which is criicl elemen, if no fundmenl exchnge re model iself. The microsrucure pproch, in which invesor order flows re criicl fcor in he deerminion of exchnge re movemens, coninues being pplied successfully in modeling exchnge re movemens. Once informion bou invesor order flows is reveled, pricipns in he foreign exchnge mrke c ccordingly by eiher buying or selling foreign exchnge hereby exering pressure on he exchnge re which hs o ssume new vlue in ligh of he new informion. According o Evns nd Lyons (), for insnce, % of dily exchnge re movemens re due o dily iner-deler order flow. Rime (1) lso dduces evidence bsed on weekly order flows h is supporive of he microsrucure pproch nd herefore collbore he evidence in Evns nd Lyons (). Froo nd Rmdori () provides furher insighs regrding he role of invesor order flows in driving exchnge res. I is shown h invesor order flows re priculrly imporn for shor run exchnge re movemens. Pge 9 of 53

10 The microsrucure pproch is lso used in Brooks e l., (1) whereby n empiricl nlysis of he role of differen ypes of ne foreign cpil inflows (s well s he role of fcors h underlie cpil flows) in driving seleced mjor currency exchnge res. Bsed on he bivrie regression model resuls, h sudy shows h equiy flows nd long-run ineres re differenils significnly influenced movemens in he Euro-US dollr nd he Yen-US dollr bilerl exchnge res. A key limiion of he sudy is h simple bivrie regression models re pplied in he nlyses. Wih his kind of model, poenilly useful informion in he form of oher fcors is no conrolled for in he esimed models. This rises issues of he omied vrible kind of model miss-specificion. Siourounis (3) shows how o overcome he limiion by, no only using he comprehensive VAR mehodology, bu lso using liner regression models in which, in ddiion o he equiy flows, imporn fcors known o drive exchnge res re incorpored. Siourounis (3) nlyzes he cusl relionship beween cpil flows nd he bilerl nominl exchnges res of 5 OECD counries h cross-border equiy flows predic exchnge re movemens unlike cross-border bond flows which do no. I is lso found h shock o domesic equiy reurn differenils induces shor run ppreciion of he domesic currency. Mos recenly, Were, Kmu nd Kisinguh (13) crried ou n empiricl invesigion of he role of he curren ccoun blnce in he deerminion of he Keny shilling exchnge re nd find h n improvemen in he curren ccoun [which reflecs improvemen in he ne cpil inflows, when ccommoding flows surpss uonomous flows in he cpil nd finncil ccoun] is ssocied wih n ppreciion of he exchnge re. More specificlly, i follows from he long run exchnge re model resuls h 1 percenge poin improvemen in he curren ccoun blnce, on verge, leds o n ppreciion of he shilling exchnge re by 5.9 percenge poins. I cn lso be inferred from he sudy findings h n increse in rel ineres re differenils [which induce improved ne cpil inflows] leds o n ppreciion of he domesic currency. These evidence from Were, Kmu nd Kisinguh (13) is consisen wih recen developmens in he shilling exchnge re. Wih he persisen nd incresingly widening curren ccoun Pge 1 of 53

11 defici, he shilling exchnge re hs ended o deprecie in nominl erms. A sudy by he World Bnk however predics n ppreciion of he rel effecive exchnge re in he ls couple of yers (World Bnk, 13). The sudy finding in Were, Kmu nd Kisinguh (13) h rel ineres re differenils ply consisen role in he deerminion of he shilling exchnge re is consisen wih erlier evidence in Ndung u (). In priculr, Ndung u () finds h rel ineres re differenils explin shor run exchnge re movemens. The empiricl resuls presened nd discussed in O Connell e l. (1) show h rel ne foreign direc invesmen nd porfolio flows were generlly subdued during he mos pr of And so were officil developmen ssisnce flows nd reminces. During he sme period, shor erm exernl deb s proporion of expors, nd lso s proporion of ol exernl deb were generlly beween 5-5%. Mos impornly, deviions from uncovered ineres priy re esimed o be lrges in Keny mong comprive counries nd he implicion of his is h risk premium djused rel ineres re differenils should be he beer vrible o pply when empiriclly nlyzing he role of ineres re differenils (s he underlying fcor for ne cpil inflows) in exchnge re movemens. Esimes of he foreign exchnge mrke pressure in O Connell e l. (1) for he period 199- show h while he voliliy in exchnge re pressure hrdly exceeded sndrd deviions, he pressure ws relively higher during nionl elecion periods. This suggess h poliicl risks hve dverse implicions for exchnge re. I is explined h such poliicl risks iniilly dversely impc he curren ccoun blnce which in urn dversely ffec he domesic foreign exchnge mrke in erms of incresed exchnge re voliliy. Applying vecor uo-regressive (VAR) model, O Connell e l. (1), obin resuls which show h he consequen o 1 sndrd deviion innovion in domesic nominl ineres res, he response of he exchnge mrke pressure (EMP) is mixed. Moreover, shock o he exchnge mrke pressure (EMP) significnly nd persisenly increses domesic ineres res Pge 11 of 53

12 including he repo re. 5 This would sugges h whenever exchnge mrke pressures mouned, he Cenrl Bnk ook policy cion o ese he pressure. 3. Mehodology 3.1. Vecor Auoregressive (VAR) Model We ssume h he economy is dequely represened by he srucurl model provided by (1) nd which is reduced o obin (). Ay Ly Z D ; ~ iid B, (1) l Z Q u ; u ~ iid y C L y l, () Whereby, 1 C L A B L ; 1 A ; A 1 ; nd u A 1 (3) The moving verge represenions of he srucurl model re useful in compuing impulse responses of seleced economic vribles o shocks nd hese represenions re provided by () nd (5) o correspond o (1) nd (). While () forms he bsis for genering he impulse responses of y wih respec o shocks o he innovions which re componens of, (5) does he sme wih respec o he componens of he regression residuls, responses re, respecively, nd y ŷ s s,1,,... seps nd seps. u. The impulse () 5 This is he re which he Cenrl Bnk of Keny mops up excess liquidiy using is open mrke operions. OMOs involve compeiive ucion nd he repo re is he rde weighed ineres re for he successful OMOs bids. Pge 1 of 53

13 y ŷ u (5) s s Idelly, he impulse responses of ineres o us re. However, becuse (1) is no direcly esimble due o indeque informion, we cnno lso direcly esime (). By convenion, he reduced form provided by () is esimed o obin he covrince mrix,, nd he vecor of regression residuls, u. For ny inverible mrix A, () holds. Using (), herefore, we cn re-wrie (5) o obin (7). A 1 A I () y 1 ŷ sa Aus (7) s A direc comprison of () nd (7) suggess h if we were o rnsle (7) ino (), he following relionships mus hold. () s 1 sa Au s (9) When evlued s, (9) is bsiclly he re-prmeerizion of he innovion process o he regressions residuls vecor provided by (3) nd s such we cn obin (3) from (9) by invering (9) nd evluing he resuln expression s. The implicion of his nlyicl resul is h he opiml choice of A in (7) is coningen upon upholding he srucurl model provided by (1). In oher words, A should be h mrix which cn be used o decompose he poenilly conemporneously correled vecor of regression residuls, u, ino he conemporneously uncorreled innovion process vecor,,in line wih (3) or (9). In order herefore o recover from u, we need o know A nd he problem hnd is how we cn use he informion vilble o us upon esimion of he reduced form provided by () o esime A. If we squre boh sides of (3) nd hen ke expecions of boh sides of he resuln expression, we will obin (1) which is re-orgnized o obin (11) nd hen solved o obin (1). Pge 13 of 53

14 E E T 1 T 1 T u u E A A T 1 T 1 u u A E A T (1) (11) 1 1 A A T (1) T T Whereby, E u u nd E By invering (1), we obin (13). T AA (13) For simpliciy, i is ssumed h in (13), he digonl covrince mrix is n ideniy mrix. This is nmoun o ssuming h he componens of he srucurl shocks vecor re ideniclly nd independenly disribued wih men zero nd uniy vrince. Using In (13), we obin (1) which is he fundmenl sysem from which A is o be solved. I follows from (1) h, he much needed esime of A is essenilly fcorizion of he covrince mrix,, which is one of he oupus from esiming he reduced form. T AA (1) Due o he symmericl nure of he covrince mrix,, he sysem represened in (1) is no solvble s i is becuse i involves fewer number of independen equions hn he number of unknown elemens of A. In prcice, minimum number of vlues of he elemens of A mus be se priori o pve wy o esimion of he res of he unknown elemens of A from he resriced vrin of (1). The minimum number of resricions is jus equl o he number of symmericl elemens of A nd which is resricions hn n n in n. One cn impose more n nd he over-idenifying resricions cn be esed for sisicl significnce using he likelihood rio es sisic whose disribuion is kin o disribuion wih he number of degrees of freedom equl o he number of overidenifying resricions. This wy, one cn es economic heory relevn o he empiricl nlysis under considerion. Pge 1 of 53

15 In imposing he idenifying resricions, one would use eiher he Choleski or he srucurl idenificion pproch. We hve used he Choleski idenificion in his sudy in line wih Muru nd Ndirng u (13). Under he Choleski idenificion pproch which ws inroduced by Sims (19), one imposes recursive ordering of he endogenous vribles so h he resuln mrix A is lower-ringulr which resuls in jus idenified sysem. Unlike he Choleski idenificion pproch, resricions under he srucurl idenificion pproch re imposed ccording o heoreicl predicions or ccording o vilble relevn empiricl sudy findings jus like in clibring economic model. The srucurl idenificion pproch is ribued o Bernnke (19) nd Sims (19) lhough oher pplicions cn be found in for insnce Blnchrd nd Quh (199). T Under he simplifying ssumpions h E nd h E s, mening h he componens of he regression residuls vecor obined upon esimion of he unresriced reduced form VAR model re conemporneously uncorreled cross he VAR equions nd ech one of he componens is devoid of uocorrelion, he reduced form model provided by (), is esimed one equion ime using ordinry les squres (OLS) o obin consisen esimes of he VAR prmeers nd he VAR covrince mrix, pplied in (1). Use of OLS in his cse is however pproprie when he degrees of freedom re sufficienly lrge. When he VAR model involves mny endogenous vribles nd long opiml lg s i is he cse in his sudy, esimion of he VAR using OLS resuls in model prmeer over fiing whereby he esimed model s in-smple predicion power is usully good whils is ou-of-smple predicion power is miserbly poor. One should use Byesin esimion echniques, insed. s 3.. Byesin Vecor Auoregressive (BVAR) Model A BVAR model is simply VAR model in se-spce represenion wih priors ssigned o he prmeers nd he prmeers vrince-covrince mrix. The ide is h he chllenge of Pge 15 of 53

16 indeque degrees of freedom nd which leds o model over-fiing for lrge dimensioned VAR models is overcome by inroducing ddiionl prior informion o ugmen observed d. Inroduced by Don (19) nd dvnced by Liermn (19) nd Don, Liermn nd Sims (19), his echnique of ugmening observed d involves specifying prior join densiy funcion for he prmeers. This mens h bsed on simplifying ssumpions nd oher prior informion, he modeler ssigns vlues of he prmeers s prior mens nd he confidence he modeler hs h he ssigned prior mens re s good s he rue vlues of he prmeers. The ssigned confidence in he chosen se of prior mens colescences ino prior vrince-covrince mrix for he prmeers. The se-spce represenion of liner sysem comprises mesuremen nd rnsiion equion. For he unresriced VAR model provided by (), he mesuremen nd rnsiion equions re (15) nd (1). I is ssumed h he mesuremen nd rnsiion equions vrinces re independen of ech oher nd his permis evluion of ech equion ime nd in recursive mnner; beginning wih he rnsiion equion which feeds ino he mesuremen equion. Thus, shock o he rnsiion equion ffecs he mesuremen equion indirecly hrough he shock s iniil effec on he se vecor,, nd no direcly hroughu. y u x u ; Vr (15) 1 M ; Vr (1) I my furher be ssumed, for simpliciy purposes, h nd M re, known. More specificlly, 1nd h M (i.e. M is he ppropriely dimensioned null vecor). In which cse hen we will be deling wih BVAR in which he prmeers or he se vecor is ssumed o be consn during he esimion period. In periods of mjor srucurl breks, i is obviously no pproprie o ssume consn prmeers nd we herefore es for model sbiliy during he smple period. Sims (199) provides n empiricl sudy in which hese ssumpions hve been relxed o work wih ime-vrying prmeers nd obins improved resuls. Pge 1 of 53

17 3.3. Esimion of he BVAR Model Being liner sysems, BVAR models cn be effecively esimed using liner Klmn Filer which involves mking one-period hed predicion bou he se vecor nd is vrincecovrince mrix which is hen used in compuing he simuled vlues of ll he endogenous vribles for h period. The filer hen updes he iniilly prediced se vecor in ligh of he predicion errors. The upded nd finl simuled vlues of he endogenous vribles for h period re hen compued using he upded se vecor. In he subsequen ierion of he Klimn Filer, he preceding period s upded se vecor is used s if i were he iniil se vecor for he subsequen period nd s such i is used o mke he iniil predicion of he endogenous vribles vlues for he subsequen period. Once gin he iniil se vecor for he second period is upded o incorpore informion bou he predicion errors nd he upded finl vlues of he endogenous vribles compued. This ierive process herefore is crried ou for ll subsequen period for he enire smple period. Once he Klmn Filer lgorihm is fully execued, he poserior join densiy funcion nd simuled ime series for he endogenous vribles re obined mong oher resuls. The poserior join densiy funcion whose elemens re he poserior se vecor mens nd he poserior covrince mrix is priculrly imporn for furher empiricl nlysis. We pply he poserior mens in he unresriced VAR model o esime he regression residuls, u. We hen lso fcorize he poserior covrince mrix in line wih (1) o obin he fcorizion mrix A. We re hen in he posiion o, in urn, fcorize esimes of he regression residuls, u, in line wih (3) or (9) o obin esimes of he much needed conemporneously uncorreled componens,. The ide is o genere he orhogonl vecor of he esimes of he underlying srucurl shocks,, which we use in () for s o The simuled ime series will include hose for unobserved vribles incorpored ino he nlysis nd his is when he Klmn Filer is useful in esiming unobserved componens of given model. Pge 17 of 53

18 compue poin esimes of impulse responses s well s genering he vrince decomposiion resuls. I is useful noing h he liner Klmn Filer upding-equions re generlly provided by (17), (1) nd (19). 7 M 1 1 (17) 1 1 T T 1 x 1x x 1 x (1) 1 1 T T 1 x x y x x (19),1,,...,T 1 1 In order o se he Klmn Filer in moion, iniil vlues of he se vecor,, covrince mrix,, mesuremen equion residuls vrince,, nd he rnsiion equion residuls vrince, M, should be chosen nd pplied ccordingly so h he Klmn Filer is iniilized nd execued by choosing nd pplying: 1, 1, nd M M in (17), (1) nd (19),1,,..., T covrince will hen be nd, respecively. T T. We denoe he poserior se vecor men nd 3.. Assigning Priors We hve ssigned he prior mens nd prior vrinces o he se vecor using he originl Minneso Prior which ws inroduced by Don, Liermn nd Sims (19) nd Liermn (19). The originl Minneso Prior join densiy funcion is summrized in () nd (1) whereby () is he bsis for ssigning prior mens nd (1) he bsis for ssigning prior vrinces. Accordingly, under he belief h endogenous vribles were dequely represened s rndom wlk processes, 1, i j 1 nd k 1. i 1 7 Deiled explnion of hese equions is provided in Muru nd Ndirng u (13). Pge 1 of 53

19 In (1), is he overll ighness of he prior men (i.e. he srengh of our belief h he ssigned prior men is he rue men vlue of he prmeer under considerion. In he exreme cse of, i will men h we belief h he prior men nd he rue vlue of he prmeer re idenicl nd s such we need no use he observed d o esime he prmeer whose vlue we cn simply fix. Conversely, when, he ssumpion is h we re compleely ignorn bou wh he rue vlue of he prmeer of ineres should be nd s such he prmeer should be esimed using he observed d. For he inermedie cses when nd do no hold represen siuions whereby prior informion nd he observed d re combined in line wih Byes heorem o esime he relevn prmeers. i i j nd k 1 i,j E k () i j nd k 1 i,j 1 i ; 1,,..., p Vr k j (1) p y p y p y () p Whereby, p = Prior densiy funcion; p y = Log likelihood funcion; p T y = Poserior densiy funcion; nd p y= mrginl likelihood of he observed d. In siuions whereby we hve, we will hve very igh priors oherwise for non-informive priors. we hve fl or Pge 19 of 53

20 The lernives o he originl Minneso Prior include he sum of coefficiens Prior inroduced by Sims (199) nd Sims nd Zh (199). In his cse, i is ssumed h B B.. B 1 (i.e. VAR coefficiens for own lgs in ny VAR equion dd up o uniy). 1 p Oherwise, one my lso use he symmeric Minneso Prior (lso clled he nurl conjuge prior) dvnced by Roberson nd Tllmn (1999), Bnbur e l. (1) nd Beuchemin nd Zmn (1) Opimizing he Hyper-prmeers The originl Minneso prior ckles he problem of limied degrees of freedom in wopronged mnner. I ugmens he observed d by ssigning he prior join probbiliy funcion nd by so doing, in he mnner h i does i, reduces he number of unknowns from n pn k k d indexed. In npn k k d o much smller nd mngeble number o which he prior vrinces re, n is he number of endogenous vribles, p is he vlue of he opiml lg lengh used in he VAR, k is he number of exogenous vribles in he VAR nd d is he number of deerminisic erms in he VAR nd which include consn nd se of sesonl dummies (if he d is no sesonlly djused) in ech equion. We hve ssumed, in npn k k d h hough, by heir very nure, he exogenous vribles re no ssigned equions s pr of he unresriced VAR, hey cn (nd re, in his cse) lgged o he exen h he endogenous vribles. Choosing opiml vlues of he hyper-prmeers (i.e. he prmeers in () nd (1) o which he prior mens nd he prior vrinces re indexed) is n imporn pr of he esimion of he BVAR model. One pproch, which we hve used in our nlysis nd lso in Muru nd Ndirngu (13), is o clibre he hyper-prmeers on he bsis of resuls in ps sudies. The lernive is o esime he hyper-prmeers using he mximum likelihood esimor condiionl o he prior mens. Opimizion under he mximum likelihood esimor is usully done on he bsis of minimizion of he ou-of-smple forecs errors. Pge of 53

21 For insnce, Beuchemin nd Zmn (1) in which he hyper-prmeer opimizion problem o be solved kes he form provided by (3). Noice h in (3), pyis he mrginl likelihood of he observed d which is inegred ou of he poserior join probbiliy densiy funcion provided by () wih respec o he se vecor. Noice h by opimizing over *, *, he ssumpion is h he res of he unknown hyper-prmeers in (1) re clibred. p y * *, rg Mx n (3), y py p d p (). Empiricl Anlysis An imporn pr of model specificion is o selec he se of relevn vribles o be included in he model. We lso need o choose he opiml lg for he model. For ese of reference, we hve explined he selecion of vribles under sub-secion.1. In choosing he vribles, we ssume h Keny is semi-smll open economy which is lrgely influenced by foreign economic developmen bu in number of simplifying ssumpions, in priculr he locl currency pricing ssumpion does lso influence foreign economic developmens. See for insnce, Cuche-curi, Dells nd Nl (9) on he working of he semi-smll open-economy nd locl currency pricing ssumpions in he conex of dynmic sochsic generl equilibrium modeling. Like Sims (199), we do lso include vribles o ensure we cn ell consisen economic sory bsed on our empiricl resuls. In priculr, we drw from ps relevn sudies o idenify he vribles. Pge 1 of 53

22 .1. Choosing Model Vribles.1.1. Endogenous Vribles Since he objecive of his sudy is o empiriclly nlyze he role plyed by shor erm cpil flows,caps, on he Keny Shilling US Dollr exchnge re, LSHS, ne shor erm cpil inflows nd he Keny shilling US Dollr exchnge re re mndory inclusions in he relevn se of d h we use in he nlysis. We follow Siouriounis (3), which is sudy h is prey similr o ours, o include equiy reurn differenils, differenils, IRD, in he endogenous vribles vecor, Y, s shown in (5). ' NSE, nd ineres re Inclusion of ne equiy flows nd ineres re differenil is jusified furher by he supporive empiricl evidence in Brooks e l. (). Alhough equiy reurn differenils re no suppored by he evidence in Brooks e l. (), i is fcor worhwhile considering in sudy focusing on developing counry such s Keny in which here is cive rding of equiies nd bonds he Nirobi Securiies Mrke (NSM). We noice h he risk premium vrible is no incorpored in some of he ps sudies including Brooks e l. () nd Siouriounis (3) nd we herefore exend he endogenous vribles vecor o include he risk premium vrible, RISK. In so doing, we effecively use he risk djused ineres re differenil, IRD _ ADJ, hereby relisiclly conrolling for he counry-specific risk which mers for foreign invesors. We lso ssume h invesors no suffer from money illusion nd s such hey re conscious of he role inflion plys in eroding he rel vlue of heir cross-border gross invesmen reurns. This herefore mens h inflion differenils re imporn o cross-border invesors. This explins furher exension of he endogenous vribles vecor o include he inflion differenils. We hve however oped o conrol for inflion differenils by including domesic consumer price index, LP, s pr of he endogenous vribles while he foreign consumer price index, LPF,is included s pr of he exogenous vribles. Pge of 53

23 A shock o, sy, he domesic ineres re such s could rise from monery policy shock, mid sicky goods prices, is prediced o induce corresponding more hn proporione insnneous chnge in he curren vlue of he exchnge re o equilibre finncil mrkes. Wih ime, however, when he goods prices djus grdully in response o he monery shock, he exchnge re which hd over-djused o mke up for he sicky prices will pprecie in ndem wih he grdul djusmen in he goods prices. See for insnce he seminl pper on he exchnge re overshooing phenomenon by Dornbusch (197). This resul is generlized o siuion of presumed iner-emporl budge consrin for n open economy so h while goods prices remin sicky mid monery policy shock, hen we would hve chnge in no only he nominl bu lso he rel ineres res which will consequenly impc rel oupu wih implicions for he counry s curren ccoun. As discussed in Obsfeld nd Rogoff (199), he welfre effec will hen hve lsing effec on he exchnge re. The implicion of his nlyicl conclusion is h he curren ccoun hs poenil o impc he domesic currency exchnge re nd his, for insnce, jusifies inclusion of he curren ccoun s n explnory vrible in he exchnge re deerminion model formuled by Were, Kmu nd Kisinguh (13). In order herefore o conrol for he welfre effecs rnsmied hrough he curren ccoun, nd his would be embodied in reminces nd oher curren rnsfers, o he Shilling exchnge re, we include he curren ccoun blnce, vribles in our VAR model. ACAB, s one of he endogenous We complee wriing he lis of endogenous vribles wih he inclusion of monery condiions vrible o proxy he poenil role of monery nd fiscl policy in driving he Shilling exchnge re movemens. The inclusion of monery condiions vrible, LRM, is jusified by he nlyicl finding in monery shocks, mid, goods price rigidiies no only cuse exchnge res o overshoo nd s such drive exchnge res in he shor run bu lso cn induce welfre effecs leding o permnen impc on he exchnge re. I my s well be h he policy shocks emne from he Cenrl Bnk Re, CBR, which is he policy re. Pge 3 of 53

24 The empiricl nlysis in O Connell e l. (1) shows h exchnge mrke pressure is fully bsorbed in he exchnge re if here is no esing of he pressure hrough chnges in he ne foreign sses on ccoun of finncing. Which suggess lso h i is useful conrolling our nlysis for chnges in ne foreign sses nd for his reson, we hve included ' DNFA. Y CACB LY LP LRM CBR DNFA IRD_ ADJ NSE CAPS LSHS (5) Whereby, CACB = Curren ccoun blnce s proporion of GDP; LY = Log of rel oupu mesured s GDP; LP = Log of he domesic consumer price index; LRM = Log nominl reserve money; CBR = Cenrl Bnk Re; DNFA = Chnge in ne foreign sses for finncing purposes; IRD _ ADJ = Risk djused nominl ineres re differenils mesured in erms of he excess of he 91-dy Tresury bills nominl ineres re over he USA 9-dy Tresury bills nominl ineres re; NSE = Equiy reurn differenils mesured s he excess of relized cpil gin he Nirobi Securiies Exchnge (NSE) Mrke over relized cpil gin he New York Sock Exchnge Mrke; CAPS = Ne shor erm ne cpil inflows; nd LSHS = Log of he nominl bilerl exchnge re mesured s he number of unis of he Keny Shilling h buy US dollr. Pge of 53

25 .1.. Exogenous Vribles We lso conrol for he broder world economy developmens wih implicions for Keny s economy nd in priculr, Keny Shilling exchnge re. We herefore include in he lis of exogenous vribles, mesure of he world economy rel economic performnce esimed s he USA rel gross domesic produc, LYF. Similrly, we hve lso ken ino ccoun developmens in world money mrkes by including money mrke ineres re differenils mesured in erms of previling nd ps monery policy snce differenils. While hve included he Cenrl Bnk Re s n endogenous vrible, we include he foreign policy re, which we consider o be he USA Federl Reserve Re, FEDRATE, s pr of he exogenous vribles. We do lso conrol for he world demnd nd supply environmens by including he world oil prices, LPRICE, nd he foreign consumer prices, LPF. Keny s impor bill is no only significn pr of he overll impor bill bu world oil price shocks pose significn poenil impc on Keny s economy nd more specificlly, he shilling exchnge re which hs o ber he pressure piled up by delers, from ime o ime, seeking foreign exchnge o impor oil. In nushell, he exogenous vribles vecor is provided by (). ' x LYF LPF LPRICE FEDRATE ().1.3. Deerminisic Vribles The wo usul suspec vribles which we include s pr of he deerminisic erms re he consn nd sesonl fcors. We herefore hve he vecor of deerminisic erms provided by (7). ' 1 3 d CONS SEAS SEAS SEAS (7) Pge 5 of 53

26 .. Solving he Idenificion Problem For simpliciy purposes, Sims (19) pproch which is essenilly he Choleski fcorizion of he covrince mrix nd hence he recursive pproch o decomposiion or orhogonlizion of he esimed VAR regression residuls. Rher hn ordering he endogenous vribles rbirrily, we do he ordering bsed on preliminry compuion of he conemporneous correlion coefficien mrix of he endogenous vribles vecor. The endogenous vribles re hen lised heir order of incresing conemporneous correlion wih ech oher wih he mos correled vrible being lised ls. The esimble decomposiion model is he jus idenified sysem provided by (). y y y y y y y y y y y u y u y u y u y u y u y u y u y u y u () Whereby, A Pge of 53

27 Noice h he empiricl problem involves esiming A s he fcorizion of he covrince mrix in line wih (1) nd hen using he esimed A in () o decompose esimes of u o recover is orhogonl componens which re essenilly esimes of he innovion process componens,. Hving esimed A nd, from (1) nd (), respecively, we re in posiion o compue he impulse response funcions from () (or, equivlenly, from (7)). The impulse responses re hen ploed s impulse response funcions wih superimposed confidence inervls o guge he sisicl significnce of he poin esimes of he impulse responses. The confidence inervls re derived from simuled impulse responses using he Mone Crlo Inegrion s in Muru nd Ndirngu (13). We lso obin vrince decomposiion resuls which re useful in evluing he relive impornce of shor erm cpil inflows in explining shilling exchnge re movemens..3. Preliminry D Anlysis Figure 1 shows h during he period Q1-5Q1, ne cpil inflows were predominnly comprised of ne shor erm flows, generl limied nd for subsnil pr of he sub-period, declining. As one would expec, oher fcors remining equl, he shilling USA dollr exchnge re ws generlly sble. As he exchnge re generlly pprecied during he second sub-period of 5Q-Q, ne cpil inflows generlly picked up whereby shor erm cpil inflows were sill predominn. This pprenly provides consisen poenil relionship beween he ne foreign flows nd he exchnge re. However during he remining sub-period of Q3-1Q he relionship beween ne cpil inflows, ne shor erm inflows in priculr, nd he exchnge re is rher prdoxicl becuse s he ne cpil inflows picked up rher srongly relive o is hisoricl ps, he exchnge re ws lso depreciing rher remrkbly. This could sugges h here mus be oher fcors explining he rher eleved depreciion of he exchnge re in spie of he incresing rend in ne cpil inflows. We suspec h one of hese oher fcors is he inflion differenils. Pge 7 of 53

28 Mr- Sep- Mr-1 Sep-1 Mr- Sep- Mr-3 Sep-3 Mr- Sep- Mr-5 Sep-5 Mr- Sep- Mr-7 Sep-7 Mr- Sep- Mr-9 Sep-9 Mr-1 Sep-1 Mr-11 Sep-11 Mr-1 Sep-1 Mr- Sep- Mr-1 Sep-1 Mr- Sep- Mr-3 Sep-3 Mr- Sep- Mr-5 Sep-5 Mr- Sep- Mr-7 Sep-7 Mr- Sep- Mr-9 Sep-9 Mr-1 Sep-1 Mr-11 Sep-11 Mr-1 Sep-1 Figure 1: Srucure of Ne Cpil Inflows nd he Exchnge Re Ne Medium Term nd Long Term Cpil Inflows (LHDS, US$ Million) Ne Commercil Bnks Flows (LHDS, US$ Million) Ne Shor-erm cpil (LHDS,US$ Million) Ne Cpil Flows (LHDS, US$ Million) Shilling USA Dollr Re (RHDS, KShs Per USA Dollr) Figure : Ineres Re Differenils nd he Exchnge Re IRD RISK IRD_ADJ KESRATE Pge of 53

29 Mr- Sep- Mr-1 Sep-1 Mr- Sep- Mr-3 Sep-3 Mr- Sep- Mr-5 Sep-5 Mr- Sep- Mr-7 Sep-7 Mr- Sep- Mr-9 Sep-9 Mr-1 Sep-1 Mr-11 Sep-11 Mr-1 Sep-1 Mr- Sep- Mr-1 Sep-1 Mr- Sep- Mr-3 Sep-3 Mr- Sep- Mr-5 Sep-5 Mr- Sep- Mr-7 Sep-7 Mr- Sep- Mr-9 Sep-9 Mr-1 Sep-1 Mr-11 Sep-11 Mr-1 Sep-1 Figure 3: Curren Accoun Blnce nd he Exchnge Re Curren Accoun Blnce (LHDS, US$, Million) Shilling USA Dollr Re (RHDS, KShs Per USA Dollr) Figure : Ne Commercil Bnk Flows nd he Exchnge Re Ne Commercil Bnks Flows (LHDS, US$ Million) Shilling USA Dollr Re (RHDS, KShs Per USA Dollr) Pge 9 of 53

30 5. Empiricl Resuls Using qurerly d for he period Q1-1: from he Cenrl Bnk Dbse, we crried ou preliminry nlysis o deermine he opiml lg for he unresriced VAR model. Using he Theil U es sisics while llowing for sufficien dynmics in he sysem, we doped lgs o be bes suied for he empiricl nlysis. Vriion of he lg o 3 or 5 does no led o rdiclly differen empiricl resuls. These oher resuls re presened he Appendix. We presen he preferred empiricl resuls in secions 5.1 nd 5.. A brief discussion of he resuls is provided in secion Impulse Responses The Impulse response resuls presened in Figure show, rher surprisingly, h 1 sndrd deviion shock o ne shor erm cpil flows hs significn depreciing impc on he shilling-dollr exchnge re. The resuls which re sisiclly significn on impc remin significn hrough he h qurer when he impc dies off. The resuls re surprising becuse one would expec, generlly, ne cpil inflows o dd o he supply of foreign exchnge in he domesic foreign exchnge mrke so h, oher fcors remining equl, n ppreciion of he shilling ensues. The resuls in he figure lso show h shock o he risk djused rel ineres re differenils will immediely nd unmbiguously led o n ppreciion of he exchnge re. The ppreciion is hen correced for fully wihin bou 5 qurers. The exchnge re djusmen shown in hese resuls re consisen wih he exchnge re overshooing phenomenon. Consisenly lso, posiive shock o he risk djused ineres re differenils hs significn immedie incresing effec on ne shor erm cpil inflows. The effec however swiches o being reducing one wihin yer beyond which i urns ou o be reducing nd significn. Pge 3 of 53

31 Responses of This would sugges h indeed shor erm cpil flows re ho flows h seek quick rel reurns nd re herefore subjec o poenil sudden sops or reversls. Also found o be imporn for exchnge re movemens is he curren ccoun blnce whereby posiive shock o he curren ccoun blnce, which represens improvemen in he curren ccoun blnce, is show o, on verge, significnly pprecie he shilling exchnge re. The significn effec is however shor-lived s i lss for bou 3 qurers. Figure 5.1.1: Preferred Impulse Response Funcion Resuls CACB LY LP LRM CBR DNFA IRD_ADJ NSE CAPS LSHS CACB LY LP LRM CBR DNFA IRD_ADJ NSE CAPS LSHS CACB LY LP LRM CBR DNFA IRD_ADJ NSE CAPS LSHS Pge 31 of 53