South Africa. Mehmet Balcilar. Tel: +27. University of Pretoria. Working June 2016

Transcription

1 Universiy of Preoria Deparmen of Economics Working Paper Series The Effeciveness of Moneary Policy in Souh Africa under Inflaion Targeing: Evidence from a Time-Varying Facor-Augmened Vecor Auoregressivee Model Goodness C Aye Universiy of Preoria Mehme Balcilar Easern Medierranean Universiy and Universiy of Preoria Rangann Gupa Universiy of Preoria Working Paper: 63 June 6 Deparmen of Economics Universiy of Preoria, Preoria Souh Africa Tel: +7 3

2 The Effeciveness of Moneary Policy in Souh Africa under Inflaion Targeing: Evidence from a Time-Varying Facor-Augmened Vecor Auoregressive Model Goodness C Aye *, Mehme Balcilar ** and Rangan Gupa *** Absrac This paper examines he ransmission mechanism of shocks o moneary policy in Souh Africa using quarerly daa from 98: o : We also in addiion idenify demand and supply shocks Our analyses are based on a facor-augmened vecor auoregression wih ime-varying coefficiens and sochasic volailiy (TVP-FAVAR), which allows us o simulaneously analyse he changing impulse responses of a se of 77 macroeconomic variables Our resuls based on he impulse response funcions, are consisen wih economic heory as we observe no price puzzle ha is ofen associaed wih he sandard VAR models We find evidence of modes ime variaion in he ransmission of shocks Overall, he macroeconomic variables seemed o have responded slighly more o he moneary policy shocks in he pos (inflaion argeing) sub-period han he pre period, albei he differences in he effecs are saisically insignifican Demand shocks are found o have conribued more o changes in macroeconomic variables in Souh Africa han moneary policy and supply shocks Our resuls sugges he need for a more efficien role of he moneary auhoriy as his will boh improve is credibiliy and greaer economic sabiliy Inroducion In he 97s and 98s, he economies of several developed and developing counries winessed much volailiy in oupu growh, inflaion and oher macroeconomic variables Thereafer, hese became considerably low and more sable This laer developmen has been named he grea moderaion wih firs observaion in he US (Sock and Wason, 3, 5) and comparable evidence for oher developed economies Du Plessis, e al (7) and Burger (8) also idenified a similar moderaion for he Souh African economy since he early o mid-nineies Ellis e al () noed ha over he pas five decades, major indusrialised economies have undergone deep srucural changes including dramaic shifs in macroeconomic policy and globalisaion induced changes in echnological advances, compeiion and financial innovaion As for Souh Africa, Du Plessis and Koze () observed ha here is a considerable srucural change in mos macroeconomic variables ha would impac on he business cycle during he mid-98s The srucural changes in many economies and he recen disrupion of he era of grea moderaion by he financial crisis underlines he imporance of undersanding he naure and causes of he grea moderaion This may assis policy makers in aiding is resumpion Some have ascribed he causes o good policy, an evoluion of he economic srucure, good luck, or a combinaion of hese facors (Gali and Gambei, 9; Du Plessis and Koze, ) Three possible causes of he grea moderaion in Souh Africa were hypohesized by Burger (8) These include beer moneary policy, a more efficien financial secor, and improved invenory managemen These hree could be classified under he srucure and good policy headings * Corresponding auhor Deparmen of Economics, Universiy of Preoria, Preoria,, Souh Africa goodnessaye@gmailcom ** Deparmen of Economics, Easern Medierranean Universiy, Famagusa, via Mersin, Norhern Cyprus, Turkey and Deparmen of Economics, Universiy of Preoria, Preoria,, Souh Africa; IPAG Business School, Paris, France mehme@mbalcilarne *** Corresponding auhor Deparmen of Economics, Universiy of Preoria, Preoria,, Souh Africa; IPAG Business School, Paris, France rangangupa@upacza

3 His evidence suppored wo of he hypoheses, ha is, beer moneary policy and a more efficien financial secor conribued o greaer economic sabiliy, while beer invenory managemen did no This is consisen wih earlier empirical evidence ha moneary policy explains a large par of he grea moderaion in Souh Africa, wih fiscal policy also conribuing o his even (Du Plessis, 6; Du Plessis e al, 7, 8) Ellis e al () noed ha he developmens being winessed could have affeced he ransmission hroughou he economy of shocks o moneary policy, demand and supply In his sudy, we focus on he effecs of moneary policy shocks on he Souh African economy alhough our benchmark idenificaion scheme allows us o analyse changes in he ransmission of demand and supply shocks We will be considering he effeciveness of moneary policy pre- and pos- he inflaion argeing era Inflaion argeing is a moneary policy sraegy ha has been adoped by a number of indusrialized and now emerging marke counries as a preferred alernaive o pegged exchange rae regimes This became paricularly appealing since he unhappy experience of Lain American and Eas Asian counries found hemselves in deep financial crises in he 99s under he pegged exchange rae regime (Mishkin, ) Inflaion arge provides a nominal anchor for moneary policy whereby, an inflaion objecive is announced and a clear commimen o achieve his objecive is spel ou This assiss in shaping he public s expecaions, aiding planning and hence providing an anchor for fuure inflaion expecaions and wage seing decisions The Souh African Reserve Bank (SARB) has been in pursui of low inflaion (implici inflaion argeing) including seing some pre-announced M3 and oher inermediae arges (such as he exchange raes) from 986 unil (Jonsson ) The formal and explici inflaion argeing framework was only adoped in February following he announcemen by he Miniser of Finance during he budge speech Since hen, he sole objecive of he SARB has been o achieve and mainain price sabiliy by aiming o keep he CPIX inflaion rae wihin he arge band of 3 6%, using discreionary changes in he Repurchase (Repo) rae as is main policy insrumen Since he grea moderaion was no limied o oupu only bu also affeced he dynamics of he yield curve (Bianchi e al, 9) as well as money, credi and asse prices (Mumaz,, Ellis e al, ), i is imporan ha empirical analysis should incorporae hese variables hrough which moneary policy would be expeced o operae (Mishkin, 7) Consisen wih his, his paper assesses he impac of moneary policy shocks on he real, moneary and financial secors of he economy by exploiing a daa-rich environmen ha includes 77 quarerly series, during he pre-inflaion and pos-inflaion period Given he large number of variables involved, one of he economerically feasible mehods would be a facor-augmened vecor auoregression (FAVAR) Moreover, Bernanke e al (5) noed ha he moneary auhoriies monior and possibly respond o housands of variables in heir decision-making process, hus making i abnormal o ignore his fac if one inends o mimic he acions of a cenral bank Anoher reason including more daa is of vial imporance for correcly idenifying macroeconomic dynamics is ha he mismach beween informaion ses makes i likely ha he srucural shocks canno be recovered using a small-scale VAR as indicaed by Hansen and Sargen (99) and Lippi and Reichlin (99) Also he FAVAR overcomes he possibiliy of price puzzle ofen observed in small scale VARs Tha is, a moneary policy ighening (conracionary moneary policy) is followed by an increase in he price level and/or inflaion rae Using he FAVAR approach is inuiive as i amouns o exracing a few laen common facors from a large marix of many economic variables, wih he former preserving he same informaion as in he original daa se wihou having problem of degrees of freedom (Gupa e al, ) A number of sudies (see for example Cecchei e al, 6; Boivin and Giannoni, 6; Boivin e al, ; Korobilis, 3; Eickmeier e al, ; Liu e al, ; Barakchian and Crowe, 3;

4 Baumeiser e al, 3; Cloyne and Hürgen, ; Ellis e al, ; Evgenidis and Siriopoulos, 5) have invesigaed he effec of moneary policy on cerain macroeconomic variables in some advanced counries bu mosly in he US The resuls are a bes mixed as is eviden in he lieraure review secion As far as Souh Africa, an emerging economy, is concerned some sudies have also been conduced including Gupa e al (a) who used he consan FAVAR o examine he effec of moneary policy on macroeconomic variables and find ha he abiliy of moneary policy in affecing key macroeconomic variables, including inflaion, has increased in he pos-argeing period albei he effecs are insignifican in general Also Gupa e al (b) used a consan FAVAR and find ha in general, house price inflaion responds negaively o moneary policy shock, wih he responses being heerogeneous across middle-, luxury- and affordable-segmens of he housing marke To he bes of our knowledge, his is he firs sudy in Souh Africa o compare he effeciveness of moneary policy on he dynamics of macroeconomic economic variables pre- and posinflaion argeing periods using a ime-varying facor augmened VAR (TVP-FAVAR) approach In conras o he consan FAVAR developed by Bernanke e al (5), he TVP-FAVAR allows for ime-varying coefficiens in he VAR ransiion equaion and for heeroscedasic shocks and is also capable of accommodaing variaions in he volailiy of he underlying series (Ellis e al ) This is moivaed based on he claim by Boivin e al () ha par of he conflicing US evidence in heir survey migh be due o he fac ha he evoluion of he moneary ransmission is more complex han wha can be capured by he spli sample esimaion sraegy and ha random walk coefficiens [] allow for a much richer evoluion of he ransmission of moneary policy Evidence by Naraidoo and Rapusoane (5) and Balcilar e al (6), show ha he Souh African macroeconomic daa incorporaes a number of srucural breaks due in par o poliical ransiions, changes in policy frameworks and economic crises This furher jusifies our choice of a TVP-FAVAR model The remainder of he paper is organized as follows: Secion presens he relaed lieraure The daa and economeric model is discussed in secion 3 Empirical resuls are presened and discussed in secion while secion 5 concludes Lieraure Review As our sudy is no he firs o examine he moneary ransmission channels, we provide in his secion a review of relaed sudies For insance Cecchei e al (6) develop a mehod for allocaing performance changes among moneary policy making, a reducion in he variabiliy of supply shocks, and changes in he srucure of he economy o examine which has been more responsible for fallen volailiy of inflaion and oupu in boh indusrialised and developing counries over he pas years For of he counries sudied, more efficien moneary policy is a major conribuing facor behind improved performance Boivin and Giannoni (6) using a VAR over he pre- and pos-98 periods, find a reduced effec of moneary policy shocks in he laer period for he US economy They find ha by responding more srongly o inflaion expecaions, moneary policy has sabilized he economy more effecively in he pos-98 period Also for he US economy, Boivin e al () use boh a relaively unresriced facor-augmened vecor auoregression (FAVAR) and a DSGE model and find ha moneary policy innovaions have a more mued effec on real aciviy and inflaion pos-98 compared o pre-98 Using a FAVAR model wih ime-varying coefficiens and sochasic volailiy, and 3 pos- World War II quarerly variables, Korobilis (3), show ha boh endogenous and exogenous shocks resuled in he high inflaion volailiy during he 97s and early 98s in he US He also 3

5 find ha for oher indicaors, GDP, invesmen, exchange raes and money showed he larges response o unanicipaed moneary policy shock Eickmeier e al () examine he inernaional ransmission of US financial shocks (defined as unexpeced changes of he US financial condiions index FCI) on nine major advanced counries using a ime-varying FAVAR wih a large se of macroeconomic, financial and rade variables over he period 979 They show ha posiive (negaive) US financial shocks have a considerable posiive (negaive) impac on growh in he nine counries; he ransmission o GDP growh in European counries has increased gradually since he 98s and a more marked increase is deeced in he early 98s in he US iself They also find ha he size of US financial shocks varies srongly over ime, wih he `global financial crisis shock explaining 3% of he variaion in GDP growh on average over all counries in 89, compared o a lile less han % over he 977 period Using a new measure of moneary policy shocks derived from Fed Funds fuures conracs and a 3 monhly variable VAR esimaed for he period 988:8:6, Barakchian and Crowe (3) find ha over he pos-988, here is a small bu saisically significan negaive effec of conracionary moneary policy on oupu wih maximum impac a he wo year horizon and also almos 5% of oupu variabiliy (a a 3 year horizon) can be explained by moneary policy shocks Baumeiser e al (3) use a ime-varying FAVAR, and find evidence of a subsanial decline in he dispersion of disaggregae price responses over ime They also find ha many disaggregae prices rise emporarily in response o a moneary ighening in he early par of he sample, whereas no evidence of a price puzzle was found a he aggregae level However, he share of disaggregae prices ha exhibi he price puzzle diminishes from he early 98s onwards There is also evidence of a subsanial decline in he dispersion of disaggregae price responses over ime Liu e al () using quarerly daa from 97 o 5 and a ime-varying open economy FAVAR analysed he response of a large se of domesic variables o foreign money supply, demand and supply shocks The key resuls show ha a foreign moneary policy ighening resembles he classic beggar-hy-neighbour scenario for he UK in he period whereas he response is negaive bu largely insignifican in more recen periods Ellis e al () use a ime-varying facor-augmened VAR and find ha moneary policy shocks had a bigger impac on inflaion, equiy prices and he exchange rae during he inflaion argeing period; and he median of disaggregaed prices became more negaive and cross-secional paerns poining o a decrease in he role of cos channels Evgenidis and Siriopoulos (5) examine he inernaional ransmission of US moneary shocks across euro area and Eas Asian counries using a ime varying FAVAR model In general, he ransmission of he shock hides considerable heerogeneiy across he counries wih an increase in he size of he shock o GDP, inflaion, rade and credi variables during he pos crisis period in he majoriy of he counries Cloyne and Hürgen () use daa from 975 o 7 o examine wheher he effecs of moneary policy afer-99 change markedly from he effecs using he full sample Evidence based on he ARDL and VAR models sugges ha he response of GDP pos-99 is a bi smaller, a around per cen raher han 5 per cen whereas, he indusrial producion response is larger The overall effec of a moneary policy conracion on prices and GDP is similar across he wo samples hough he response of he price level is faser in he pos-99 period While he responses from ARDL model are highly significan hose from he VAR are no quie significan

6 Aside Gupa e al (a, b) on Souh Africa, oher sudies have basically focused on he effec of moneary policy on inflaion and/or oupu (Woglom, 3; Burger and Marinkov, 8; Gupa and Uwilingiye,, ); Gupa and Naraidoo, ) Woglom (3) use simple Taylor rule funcions and conclude ha inflaion argeing has led o modes changes in he conduc of moneary policy and also ha he credibiliy of he inflaion arge has probably fallen as uni labour coss acceleraed from he 55 per cen range in he firs half of o over per cen in he hird quarer While Gupa and Uwilingiye (, ), use cosine-squared cepsrum and modified Barro-Gordon model of ime inconsisency respecively o provide evidence ha CPI inflaion in Souh Africa has become more volaile in he inflaion argeing regime, Burger and Marinkov (8) find ha he explici regime has been marginally more successful in keeping inflaion a lower level based on resuls from he Taylor rule hree-equaion model Gupa and Naraidoo () used simple nonlinear srucural Taylor rule ype framework and find ha he adopion of he inflaion arge has led o significan changes in moneary policy and moneary policy is more responsive o inflaion when i is furher from he arge, herefore implying nonlinear adjusmen of he ineres rae Alhough some of he above sudies have also employed a ime varying FAVAR in he analysis of moneary policy ransmission mechanisms, mos of he sudies were conduced for he US Moreover, wih mixed findings, he effeciveness of moneary policy on an economy is inconclusive This herefore requires ha more sudies be conduced in differen economies for comparison 3 Daa and Empirical Model We use a quarerly daa se for he Souh African economy which spans he period from 98: o : The ime varying FAVAR esimaion is based on 77 macroeconomic variables which cover he inflaion, real aciviy, asse prices and moneary series We also have inangible variables, such as confidence indices, and survey variables The deailed lis of variables is included in he appendix as Table A These daa capure he broad rends in he Souh African economy The relevan ransformaions o achieve saionariy are also indicaed in Table A The daa used were obained from differen daa providers including he SARB, Saisics Souh Africa, Bureau for Economic Research (BER), Souh Africa, ABSA Group Limied, Souh Africa, Naional Associaion of Auomobile Manufacurers of Souh Africa, OECD saisics, Oxford Economics, and IMF Inernaional Financial Saisics The FAVAR model we use is wrien in sae space form following Ellis e al () Consider he observaion equaion: X X R F F F R e e e, K N, N KN N K N () 5

7 where x ( X,, X N, ) is a se of variables ha conains informaion abou real aciviy, K inflaion, money and asse prices in he Souh Africa, while F o F denoe he K laen facors Incorporaing large amoun of informaion helps o avoid he problem of nonfundamenalness e ~ N(, U) where e ( e,, e N ) and U is a diagonal marix The dynamics of he Souh African economy is assumed o be capured by hese K facors whereas s denoe heir loadings Following Bernanke e al (5) and Ellis e al (), he only observed facor in he model is he Bank Rae, R Since some of he variables are allowed o have a conemporaneous relaionship wih he shor-erm ineres rae,, for daa series ha are expeced o reac promply o moneary policy acions The ransiion equaion of he sysem is a ime-varying VAR model given as: L z c l, z l v, v ~ N(, Ω ) () l K where z ( F, F, F, R ) and L is fixed a consisen wih Ellis e al () The law of moion for he coefficiens φ vec ({ c ; β,,, β L, }) is posulaed as: φ φ η, η ~ N(, Q) wih he innovaion ( v ) covariance ime-varying marix Ω facored as Ω E[( v Ev )( v Ev ) ] A H ( A ) (3) where he ime-varying marices H and A are assumed o evolve as random walks To inroduce ime variaion ino he model, we allow for drif in he coefficiens and he error covariance marix of he ransiion equaion This flexible specificaion ensures ha our model can accoun for he possible srucural breaks Alhough we are basically ineresed in he moneary policy shock, our idenificaion scheme allows us o addiionally idenify demand and supply shocks Following Canova and Nicolo () and Uhlig (5), he idenificaion scheme places sign resricions on he conemporaneous response of some of he variables in x o he srucural shocks The choice is moivaed based on he wo-counry model described in Lubik and Schorfheide (6) and adoped by Ellis e al (), which is an esimaed, open-economy exension of he sandard, hree-equaion New Keynesian se-up The resricions imposed on he conemporaneous responses from he FAVAR are summarised in Table Imposing hese sign resricions requires ha: we le Ω P P be a Cholesky decomposiion of he VAR covariance marix Ω, and we draw an N N marix, J, from he N(, ) disribuion Nex, we ake he QR decomposiion of J, which provides us wih a candidae srucural impac marix as A P Q, The advanages of his conemporaneous sign resricion over he recursive scheme used by Bernanke e al (5) are clearly enumeraed in Ellis e al () (i) expansionary moneary policy shocks are assumed o decrease he nominal ineres rae, increase GDP growh, increase CPI inflaion and lead o he depreciaion of a nominal effecive exchange rae (NEER) (a decrease based on our definiion of NEER bu an increase on he one in Lubik and Schorfheide (6)) on impac; (ii) posiive demand shocks have a posiive conemporaneous impac on GDP growh, inflaion and he nominal ineres rae; and (iii) posiive supply shocks decrease CPI inflaion and increase GDP growh on impac 6

8 X,, following which we compue he conemporaneous impulse response of,,, X N o he shocks as irf irf irf X X R, N, N K KN Ψ ψ N A, () where irf X i, represens he response of he ih variable a horizon If hese responses saisfy our sign resricions, we sore A, and repea he procedure a each ime period unil we have obained A, marices ha saisfy he sign resricions Ou of he se of admissible models, he A, marix wih elemens closes o he median across hese esimaes is seleced (Fry and Pagan, 7) Table : Sign Resricions GDP growh CPI inflaion Nominal exchange rae Shor-erm ineres rae Moneary policy Demand Supply Noes A ( ) symbol denoes a non-negaive (non-posiive) on-impac response of he corresponding variable; he symbol denoes ha no resricion is imposed on he response sign To esimae our model, we employ he sandard Markov chain Mone Carlo (MCMC) Bayesian esimaion algorihm The enire algorihm is execued 55, imes wih he firs 5, draws discarded as burn-in 3 K Once we calculae he impulse responses of F, F, F and R o he moneary policy shock, he supply shock and he demand shock for each quarer, he imevarying impulse responses of each underlying variable can be obained using he observaion equaion of he model () Specifically, he impulse responses of X,, X, are compued as N K KN Ψ ψ N irf J irf K F J irf R J F X N, 3 Deails of he MCMC procedure we follow as well as he prior and poserior disribuions can be found in Ellis e al () and heir working paper version (Munaz e al, ) 7

9 where J is he impulse response horizon Following Koop e al (996), we define he IRF of he endogenous variables in he ransiion equaion for z a each dae as E( z j Ξ, ) E( z j ) j Ξ (5) j where Ξ denoes all he parameers and hyperparameers of he VAR Equaion (5) saes ha he IRFs are calculaed as he difference beween wo condiional expecaions The firs erm in (5) denoes a forecas of he endogenous variables condiioned on a srucural shock The second erm is he baseline forecas, ha is, he one condiioned on he scenario where he srucural shock equals zero Therefore, in effec, (5) inegraes ou fuure uncerainy from he parameers of he ransiion equaion The condiional expecaions in (5) are compued via Mone Carlo inegraion for, replicaions of he Gibbs sampler 5 Empirical Resuls Esimaed Facors and Sochasic Volailiy The esimaed facors (facor and facor ) indicaed in blue line and he associaed 68% highes poserior densiy inervals (HPDIs) (in red lines) are shown on he op lef panels of Figure while he observed facor (Bank Rae) is presened in he op righ panel of Figure Facors and appear o be quie accuraely esimaed Facor looks far less volaile han facor wih he former being much lower from he s compared o periods before The volailiy of he bank rae has also been low and even close o zero in recen imes compared o he earlier par of he sample On one hand, facor is posiively and highly correlaed wih financial variables 6 Aside oher ineres rae variables, i has he highes posiive correlaion (99) wih he 3-monh Treasury bill rae followed by o 3 year governmen bond (95) Some real aciviy also have high posiive correlaion (beween 7 and 8) wih facor namely producion in he mining secor and employmen On he oher hand, facor has negaive bu high correlaion wih basically all he variables The highes correlaion (-87) is observed for he composie business cycle leading indicaor wih GDP, CPI, consumpion, invesmen and house price having a correlaion of abou 8 The correlaion of facor wih he variables included in he analysis is no quie high However, he highes posiive correlaion is observed for producion in he mining secor (6) followed by long-daed governmen bond yield (9) while he highes inverse correlaion is observed for credi exended o he privae secor for invesmen (6) Given ha he hree facors summarise he dynamics of all he series in our daa se and, as all hree sochasic volailiies associaed wih hem appear somehow lower in he s compared o pre, one can o some exen claim o have firs impression evidence ha he Souh African daa is characerised by he grea moderaion The volailiy figures also provide a naural spli of he sample ino wo sub-periods: pre and pos which corresponds o he inroducion of he inflaion argeing framework in Souh Africa Therefore subsequen presenaions and discussions of he resuls will be mosly done using he pre- and pos- sub-periods This does no however imply ha our resuls are condiional on he assumpion of a single srucural break as none of our parameers are esimaed using only sub-sample daa This This specificaion is imporan since he compuaion of IRFs has o ake ino accoun he possibiliy of parameer drif over he IRF horizon as we have allowed he presence of ime-varying parameers in he ransiion equaion 5 See Koop e al (996) for deails on he Mone Carlo inegraion procedure 6 See Figure A in he appendix 8

10 is in line wih Ellis e al () In he discussion ha follows, we will concenrae on he responses of he macroeconomic variables o a moneary policy shock hough we will examine he conribuion of hree shocks via forecas error variance decomposions (FEVDs) 5 Facor 5 Facor 5 Bank Rae Sochasic Volailiy of Shock o he s Equaion 3 Sochasic Volailiy of Shock o he nd Equaion Sochasic x Volailiy of Shock o he 3rd Equaion Figure : The Esimaed Facors, he Bank Rae and he Sochasic Volailiy of Shocks Transmission of Moneary Policy Shocks o Macroeconomic Variables Inflaion Measures The poserior disribuion of he cumulaed responses of inflaion measures o a conracionary moneary policy shock, normalised o increase he Bank Rae by basis poins on impac is presened in Figure The median response in each quarer is presened in he lef panel of he Figure Our firs obervaion is ha he immediae response of all he inflaion measures o a conracionary moneary policy shock is negaive showing ha our ime varying FAVAR model is capable of avoiding he commonly winessed price puzzle in small scale VARs In general, he responses of inflaion measures o moneary policy shocks appear no o have changed much over ime hough he IRFs of CPI and wage deflaor seem larger pos To examine his furher, we presen he median responses along wih he 68% HPDI over he pre and pos periods in he nex wo panels While here is no marked change in he response of he CPI and consumpion deflaor over he wo sub-periods, he response of wage inflaion appear o have become more negaive in he pos period We es wheher he responses of he inflaion measures o moneary policy shock is saisically differen from zero following Ellis e al () The fourh panel of Figure displays he plo of he median difference in impulse responses along wih he associaed HPDI In all hree cases, he he median difference is essenially zero suggesing ha he inroducion of inflaion argeing has no had a saisically significan effec on inflaion in Souh Africa over he years An alernaive perspecive on he responses of changes in impulse responses was given by Cogley e al () In his approach, he plo of he esimaed join disribuion of he cumulaed impulse response could provide evidence of asymmeric disribuion of poins around he 5 degree line if here are sysemanic differences beween he wo sub-periods We presen he plos for he -quarer horizon (ha is one year) in he fifh panel of Figure and observe ha he disribuion is more concenraed o he righ of he line suggesing ha inflaion measures have responded more o moneary policy shock in pos despie he fac ha our saisical es in he HPDI sense did no provide evdence of significan difference 9

11 Difference in impulse responses Accumulaed Response CPI Time Consumpion Deflaor Time Wages Time Figure : Impulse Responses of Inflaion o a Moneary Policy Shock Ellis e al () was of he view ha impulse responses may no necessarily be indicaive of changes in he rasnsmission of policy shocks They argued ha alhough he policy rae shock is normalised o an iniial basis poin rise, is persisence may have increased, and ha increase could mechanically accoun for he greaer responses of inflaion indicaors in heir resuls Based on his we also presen he impluse response of he Bank rae o a moneary policy shock in figure As can be clearly seen, here appear o be no change in he responses o unanicipaed moneary conracion prior and afer he inflaion argeing period

12 Figure 3: Time-varying Impulse Response of he Bank Rae o a Moneary Policy Shock To deermine which of moneary, demand and supply shock have driven inflaion measures mos, we use he forecas error variance decomposiions (FEVDs) The plos are presened in Figure wih he shaded area in black colour showing he conribuion of moneary policy shock, grey relaes o demand shock while blue relaes o supply shock Similar o Ellis e al (), we observe ha he impac of moneary policy o inflaion slighly and slowly increased over ime However, in conras o heir findings, we find ha he grea inflaion and subsequen grea moderaion is mainly caused by demand shock This finding should no be inerpreed o mean ha he changes in moneary policy did no conribue o falls in inflaion, bu may have perhaps have more dominan effec on he elemens of he VAR covariance marix, wih a comparaively milder effec on he VAR coefficiens as argued by Benai and Surico (9)

13 CPI Consumpion Deflaor Wages Figure : Forecas Variance Decomposiion of Inflaion Noe: Shaded areas in black relae o moneary policy shock, grey relae o demand shock while blue relaes o supply shock Asse Prices To undersanding wheher changes in asse price responses can help idenify moneary policy ransmission channels in Souh Africa, we presen he impulses responses in Figure 5 We firs sar wih he expecaions channel Following Boivin e al () and Ellis e al (), we use he response of -year governmen bond yield o conracionary moneary shocks as a measure of he responsiveness of inflaion expecaions o policy The responses are presened in he second row of Figure 5 The response pos period appears slighly larger hough he difference beween he wo sub-periods is no saisically differen from zero Similarly, he responses of nominal effecive exchange rae (NEER) and share prices as presened in he firs, hird and fourh rows appear o be larger pos Sandard heory suggess ha increases in ineres raes raise he reurn on domesic asses and lead o a currency appreciaion The currency appreciaion is hen argued o dampen domesic demand via expendiure swiching (Ellis e al, ) Our resuls provide evidence of posiive response of exchange rae o a conracionary moneary policy Consisen wih he classic q heory of Tobin (969) and he balance shee ransmission channel (Bernanke and Gerler, 989; Bernanke e al, 999), share prices respond negaively o conracionary moneary policy shocks boh pre and pos period Expecaions can have an imporan effec on he user cos of capial and housing spending (Case and Shiller, 3), implying ha house prices should be expeced o conrac by more in response o a negaive policy shock afer However, growing compeiion in he housing marke, combined wih inroducion of new financial producs, may proec consumers from he immediae impac of changes in ineres raes (Benne e al, ), implying ha house prices should become less sensiive o moneary conracions The responses of house prices in our sudy seem o be consan over ime This implies ha house prices are no very sensiive o moneary policy and his is consisen wih he finding in Ellis e al () and Gahergood () Evidence based on he esimaed join disribuion of he accumulaed responses shows ha while house prices and NEER have responded more o moneary policy shock during he pre period, share prices and year governmen bond has responded more in he pos- era As wih he inflaion measures, he FEVDs displayed in Figure 6 shows ha he

14 demand shocks have conribued more o changes in asse prices and he conribuion of moneary policy do no paricularly seem o have changed much Y Gov Bond Yield House Prices Time Time Difference in impulse responses Accumulaed Response NEER 99Time Share Prices 99Time Figure 5: Impulse Responses of Asse Prices o a Moneary Policy Shock House Prices Y Gov Bond Yield NEER Share Prices Figure 6: Forecas Variance Decomposiion of Asse Prices Noe: Shaded areas in black relae o moneary policy shock, grey relae o demand shock while blue relaes o supply shock 3 Real Aciviy 3

15 We presen he impulse responses of changes in he ransmission of policy shocks o measures of real aciviy, in Figure 7 The immediae observaion in Figure 7 is ha all real variables responded negaively o a conracionary moneary policy shock An imporan moneary policy ransmission channel o invesmen is hrough changes in he user cos of capial (Jorgenson, 963) One could also hink of channel of ransmission via Tobin s q Our resuls show ha he response of invesmen o unanicipaed moneary policy shock is small boh pre- and pos sub-periods consisen wih Boivin e al () However, he IRFs in he second and hird panel of row 3 are slighly larger in he pos period Alhough he difference beween he wo periods are no saisically differen from zero (as he HPDI includes zero), he accumulaed response a he -quarer horizon shows ha he disribuion is concenraed on he righ hand side, confirming ha he response is more in he pos period Some cauions on inerpreing hese resuls is vial given ha we focus on unanicipaed policy shock (surprises), which end o be associaed wih shor-lived shifs in ineres raes Aside he effec of ineres raes on consumpion via iner-emporal subsiuion (Boivin e al, ), anoher imporan ransmission channel of policy shocks o consumpion is via he lifecycle hypohesis of Ando and Modigliani (963) ha relaes changes in consumpion o flucuaions in lifeime resources including sock and real esae wealh Similar o he equiy and house prices, he changes in he response of consumpion o a moneary policy shock is very modes suggesing ha he people have low propensiies o consume ou of financial wealh (Cae e al, ) The shor-run response of consumpion o policy conracions is equal o zero bu negaive in boh periods Again his finding is in line wih hose of Ellis e al () for he UK As wih invesmen and consumpion, GDP, indusrial producion and employmen responses appear relaively similar in boh periods and very close o zero wih all HPDIs of differences including zero (ie consisen wih no-change) We noe however ha alhough he disribuion of he cumulaed response for GDP appear o concenrae around he 5 degree line, hose of indusrial producion and employmen are slighly skewed o he righ showing ha he laer may have responded slighly more in he pos period despie he difference no being significan The plos of he FEVDs are presened in Figure 8 We noe ha while demand shock remain he predominan driver of changes in he real aciviy measures followed by supply shock, moneary policy shock has conribued less wih lile or no change in is conribuion beween he wo sub-periods hough he conribuion seems o have changed over ime in each sub-period

16 Real GDP Time Difference in impulse responses Accumulaed Response Consumpion Time Invesmen Time Employmen Time Manuf Prod Time Figure 7: Impulse Responses of Real Aciviy o a Moneary Policy Shock Real GDP Consumpion Invesmen Employmen Manuf Prod Figure 8: Forecas Variance Decomposiion of Real Aciviy Noe: Shaded areas in black relae o moneary policy shock, grey relae o demand shock while blue relaes o supply shock 3 Moneary Variables Moneary policy may affec bank loan supply via he credi channel The credi channel operaes eiher via he bank lending channel or he balance shee channel (Black and Rosen, ) If conracionary moneary policy increases banks exernal finance premium, banks may respond by reducing he oal amoun of credi hey are willing o supply (Sein, 998) This is he bank 5

17 lending channel Alernaively, conracionary moneary policy can reduce he ne worh of he banks borrowers, hereby increasing agency coss in lending This effec is likely o be inversely relaed o firm ne worh, and may lead banks o reallocae heir loan supply oward higher-neworh firms This implies ha moneary policy impacs he disribuion of loan supply across firms of differen sizes (Bernanke, e al, 996) This is he balance shee channel Changes in he willingness and abiliy of banks o exend credi may have implicaions for aggregae economic aciviy Hence, he need o also examine he response of money measures o an unanicipaed money shock The impulse responses of he moneary variables o a shock in he moneary policy shock are presened in Figure 9 Overall, he moneary variables appear o have responded negaively o a conracionary moneary policy shock in boh pre and pos periods Alhough he HPDIs includes zero in all cases indicaing no significan change in he responses of moneary variables beween he wo periods, he join disribuion shown in he las panel of column 5 sugges ha all he variables wih excepion of credi responded relaively more o moneary policy shock in he pos period Resuls presened in Figure poins o a greaer conribuion of demand shock o changes in he moneary variables whereas he conribuion of moneary policy hough very small, appear o have in general changed slighly over ime M Time Difference in impulse responses Accumulaed Response M Time Credis Time M T B ill R a e Time Inerbank Rae Time Figure 9: Impulse Responses of Moneary Variables o a Moneary Policy Shock 6

18 M M Credis M T Bill Rae Inerbank Rae Figure : Forecas Variance Decomposiion of Moneary Variables Noe: Shaded areas in black relae o moneary policy shock, grey relae o demand shock while blue relaes o supply shock 5 Impulse Response of Seleced Variables o a Moneary Policy Shock To be sure ha he correc responses of he various macroeconomic variables are no masked by he averages for pre- and pos responses and since he averages may also smooh he ime variaion, we decided o presen he individual impulse responses of seleced variables for four differen horizons:,, 6 and 6 The plos for horizons,, 6 and 6 are shown in Figures o respecively Alhough we provide modes evidence of variaion in impulse responses of hese variables over ime for each horizon, he responses are in general no saisically significan given ha he HPDIs include zero in almos all cases excep for consumpion and invesmen in horizon, and employmen in horizon 6 These resuls are basically consisen wih he earlier presenaions 7

19 -6-8 CPI Consumpion House Prices Share Prices Wages Invesmen Y Gov Bond Yield M Real GDP Employmen 8 NEER 3 Credis Figure : Impulse Response of Seleced Variables o a Moneary Policy Shock a Quarer - CPI Consumpion House Prices Share Prices Wages Invesmen - Y Gov Bond Yield - M Real GDP - -3 Employmen 8 6 NEER 3 Credis Figure : Impulse Response of Seleced Variables o a Moneary Policy Shock a Quarer 8

20 CPI 5 Consumpion House Prices 5 Share Prices Wages Invesmen Y Gov Bond Yield M Real GDP - -3 Employmen 8 6 NEER Credis Figure 3: Impulse Response of Seleced Variables o a Moneary Policy Shock a 8 Quarer CPI Consumpion 6 House Prices 5 Share Prices Wages Invesmen Y Gov Bond Yield M Real GDP Employmen NEER Credis Figure : Impulse Response of Seleced Variables o a Moneary Policy Shock a 6 Quarer Conclusion This paper aemps o idenify he ransmission mechanism of shocks o moneary policy in Souh Africa over he period 98: o : Our idenificaion scheme also allows us o idenify demand and supply shocks We use a facor-augmened VAR wih ime-varying coefficiens and sochasic volailiy (TVP-FAVAR), which allows us o simulaneously analyse he changing impulse responses of a se of 77 macroeconomic variables We examined he impulse responses of seleced inflaion measures, real aciviy, asse prices and moneary variables Our resuls based on impulse response funcions, are consisen wih economic heory as we observe no price puzzle which is ofen associaed wih small-scale moneary VAR models All hree inflaion measures (CPI, wage and consumpion deflaors) respond negaively o an unanicipaed conracionary moneary policy We find ha he response of real aciviy measures 9

21 (GDP, consumpion, invesmen, employmen and manufacuring producion) is negaive consisen wih heory For he class of asse prices, while he response of share prices and year governmen bond is negaive, ha of nominal effecive exchange rae is posiive while house prices appear o be less sensiive o changes in moneary policy shocks The response of moneary variables (M, M3, credis, 3 monh Treasury bill and inerbank rae) o a conracionary moneary policy is in general negaive Our resuls also show modes evidence of ime variaion in he ransmission of shocks Overall, he macroeconomic variables seemed o have responded slighly more o he moneary policy shocks in he pos sub-period (official inflaion argeing period in Souh Africa) han he period before However, he difference in he responses beween he wo periods is no saisically significan in he sense ha he 68% highes poserior densiy inervals (HPDIs) include zero for mos cases Evidence from he forecas error variance decomposiion suggess ha demand shocks play greaer role in he changes in he macroeconomic variables hough he conribuion of moneary policy increased slighly over ime Our finding of insignifican effec of moneary policy shocks are no srange given ha hey are consisen wih hose of Ellis e al () for mos of he UK macroeconomic variables and Gupa e al (a) for Souh Africa bu using consan FAVAR This however does no in any way connoe ha moneary policy in Souh Africa is no effecive since he responses of he variables appear o be slighly more during his inflaion argeing period Perhaps our bes suggesion would be in line wih Gupa and Uwilingiye (, ), of a narrower, and possibly a lower, arge band o improve he SARB s credibiliy and effeciveness in sabilizing inflaion and hence he Souh African economy References Ando, A and Modigliani, F (963) The life cycle hypohesis of saving: aggregae implicaions and ess American Economic Review, 53(), 55 8 Balcilar M, Gupa, R and Kozé, K (6) Forecasing Souh African macroeconomic variables wih a Markov-swiching small open-economy dynamic sochasic general equilibrium model Deparmen of Economics, Universiy of Preoria, Working paper series, 6- Barakchian, MS and Crowe, C (3) Moneary policy maers: Evidence from new shocks daa Journal of Moneary Economics, 6, Baumeiser, C, Liu, P and Mumaz, H (3) Changes in he effecs of moneary policy on disaggregae price dynamics Journal of Economic Dynamics and Conrol, 37(3), 536 Benai, L and Surico, P (9) VAR analysis and he grea moderaion American Economic Review, 99(), Bernanke, B and Gerler, M (989) Agency coss, ne worh, and business flucuaions American Economic Review, 79(), 3 Bernanke, B, Gerler, M and Gilchris, S (996) The financial acceleraor and he fligh o qualiy The Review of Economics and Saisics, 78, Bernanke, BS, Gerler, M and Gilchris, S (999) The financial acceleraor in a quaniaive business cycle framework in (JB Taylor and M Woodford, eds), Handbook of Macroeconomics, vol, pp 3 93, Amserdam: Elsevier

22 Bernanke B, Boivin J and Eliasz P (5) Measuring he effecs of moneary policy: a facoraugmened vecor auoregressive (FAVAR) approach The Quarerly Journal of Economics, (), 387 Benne, P, Peach, R and Perisiani, S () Srucural change in he morgage marke and he propensiy o refinance Journal of Money, Credi and Banking, 33(), Bianchi, F, Mumaz, H and Surico, P (9) The grea moderaion of he erm srucure of UK ineres raes Journal of Moneary Economics, 56(6), Black, L and Rosen, RJ () The effec of moneary policy on he availabiliy of credi: How he credi channel works FRB of Chicago Working Paper No 7-3 (revised) Boivin, J and Giannoni, MP (6), Has moneary policy become more effecive? The Review of Economics and Saisics, 88(3), 5 6 Boivin, M, Kiley, T and Mishkin, FS () How has he moneary ransmission mechanism evolved over ime? NBER Working Paper No 5879 Burger, P (8) The changing volailiy of he Souh African economy Souh African Journal of Economics, 76(3), Burger P and Marinkov, M (8) Inflaion argeing and inflaion performance in Souh Africa TIPS working paper Canova, F and Gambei, L (9) Srucural changes in he US economy: is here a role for moneary policy? Journal of Economic Dynamics and Conrol, 33(), 77 9 Case, KE and Shiller, RJ (3) Is here a bubble in he housing marke? Brookings Papers on Economic Aciviy, 3(), Cae, P, Girouard, N, Price, RW and Andre, C () Housing markes, wealh and he business cycle, Working Paper, OECD Economics Deparmen Cecchei, S G, Flores-Lagunes, A and Krause, S (6) Has moneary policy become more efficien? A cross-counry analysis The Economic Journal, 6, 8 33 Cogley, T, Primiceri, GE and Sargen, TJ () Inflaion-gap persisence in he US American Economic Journal: Macroeconomics, (), 3 69 Cloyne, J and Hürgen, P (forhcoming) The macroeconomic effecs of moneary policy: A new measure for he Unied Kingdom American Economic Journal Du Plessis, S (6) Reconsidering he business cycle and sabilisaion policies in Souh Africa Economic Modelling 3(5), Du Plessis, S and Koze, K () Trends and srucural changes in Souh African macroeconomic volailiy ERSA Working Paper No 97 Du Plessis, S, Smi, B and Surzenegger, F (7) The cyclicaliy of moneary and fiscal policy in Souh Africa since 99 Souh African Journal of Economics 75(3), 39

23 Du Plessis, S, Smi, B W and Surzenegger, F (8) Idenifying aggregae supply and demand shocks in Souh Africa Journal of African Economies, 7(5), Eickmeier, S, Lemke, W and Marcellino, M () The changing inernaional ransmission of financial shocks: Evidence from a classical ime-varying FAVAR Deusche Bundesbank, Germany, Discussion Paper Series : Economic Sudies No 5 Ellis, C, Mumaz, H and Zabczyk, P () Wha lies beneah? A ime-varying FAVAR model for he UK ransmission mechanism The Economic Journal,, Evgenidis, A and Siriopoulos, C (5) Wha are he inernaional channels hrough which a US policy shock is ransmied o he world economies? Evidence from a Time Varying FAVAR Bank of Greece Working Papers, No 9 Gali, J and Gambei, L (9) On he sources of he grea moderaion American Economic Journal: Macroeconomics (), 6 57 Gahergood, J () How do consumers respond o house price declines? Economics Leers, 5(), 79 8 Gupa, R and Uwilingiye, J () Dynamic ime inconsisency and he Souh African Reserve Bank Souh African Journal of Economics, 78(), Gupa, R and Uwilingiye, J () Comparing Souh African inflaion volailiy across moneary policy regimes: an applicaion of Saphe cracking Journal of Developing Areas, 6(), 5 Gupa, R, Kabundi, A, Modise, MP (a) Has he SARB become more effecive pos inflaion argeing? Econ Change Resrucuring, 3:87 Gupa, R, Jurgilas, M, and Kabundi, A (b) The effec of moneary policy on real house price growh in Souh Africa: A facor-augmened vecor auoregression (FAVAR) approach Economic Modelling, 7, Hansen, LP and Sargen, TJ (99) Two difficulies in inerpreing vecor auoregressions, in (L P Hansen and TJ Sargen, eds) Raional Expecaions Economerics, 77 9, Boulder: Wesview Jonsson G () Inflaion, Money Demand, and Purchasing Power Pariy in Souh Africa IMF Saff Papers, 8 (), 365 Jorgenson, DW (963) Capial heory and invesmen behaviour American Economic Review, 53(), 7 59 Koop, G, Pesaran, MH and Poer, SM (996) Impulse response analysis in nonlinear mulivariae models Journal of Economerics, 7(), 9 7 Korobilis, D (3) Assessing he ransmission of moneary policy shocks using dynamic facor models Oxford Bullein of Economics and Saisics, 75 (), 57-79

24 Lippi, M and Reichlin, L (99) VAR analysis, nonfundamenal represenaions, Blaschke marices Journal of Economerics, 63(), 37 5 Liu, P, Mumaz, H and Theophilopoulou, A () Inernaional ransmission of shocks: a ime-varying facor-augmened VAR approach o he open economy Bank of England, Working Paper No 5 Lubik, T and Schorfheide, F (6) A Bayesian look a he new open economy macroeconomics NBER Macroeconomics Annual 5,, 33 8, Cambridge MA: Naional Bureau of Economic Research Mishkin, FS () Inflaion argeing in emerging-marke counries The American Economic Review, 9(), 5-9 Mishkin, F S (7) The ransmission mechanism and he role of asse prices in moneary policy, in(fs Mishkin, ed), Moneary Policy Sraegy, 59 7, Cambridge, MA: MIT Press Mumaz, H () Evolving UK macroeconomic dynamics: a ime-varying Facor Augmened VAR, Working Paper, Bank of England Mumaz, H, Zabczyk, P and Ellis, C () Wha lies beneah? A ime-varying FAVAR model for he UK ransmission mechanism Working Paper, European Cenral Bank Naraidoo, R and Gupa, R () Modelling moneary policy in Souh Africa: Focus on inflaion argeing era using a simple learning rule Inernaional Business and Economics Research Journal, 9(), Naraidoo, R and Rapusoane, L (5) Financial markes and he response of moneary policy o uncerainy in Souh Africa Empirical Economics, 9(), Sein, J (998) An adverse selecion model of bank asse and liabiliy managemen wih implicaions for he ransmission of moneary policy Rand Journal of Economics, 9, 6686 Sock, J H and Wason, M W (3) Has he business cycle changed and why?, in M Gerler and K Rogoff, eds, NBER Macroeconomics Annual, Volume 7, MIT Press, Cambridge, Ma, 59 3 Sock, J H and Wason, M W (5) Undersanding changes in inernaional business cycle dynamics Journal of he European Economic Associaion, 3(5), Tobin, J (969) A general equilibrium approach o moneary heory Journal of Money, Credi and Banking, (), 5 9 Uhlig, H (5) Wha are he effecs of moneary policy on oupu? Resuls from an agnosic idenificaion procedure Journal of Moneary Economics, 5(), 38 9 Woglom, G (3) How has inflaion argeing affeced moneary policy in Souh Africa? SAJE 7(), 98 Appendices 3