International and Development Economics

Transcription

1 Asa Pacfc School of Economcs and Governmen WORKING PAPERS AUSTRALIAN NATIONAL UNIVERSITY Inernaonal and Develomen Economcs M oneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach 5 9 Mohammad-Yusuf Tashrfov Asa Pacfc School of Economcs and Governmen THE AUSTRALIAN NATIONAL UNIVERSITY h://aseg.anu.edu.au

2 Mohammad-Yusuf Tashrfov 25 Absrac Usng he Srucural Vecor Auoregresson (SVAR) mehod hs aer analyses he effecs of moneary olcy on Tajksan s economy for he erod 996 o 23. A number of resrcons are mosed and he conemoraneous and long-run resrcons model are used o denfy he dynamc resonse of nflaon and ouu o he moneary and exchange rae nnovaons. As a resul hese shocks are used o generae he srucural mulse resonse and forecas error varance decomoson funcons for assessng he dynamc macs of moneary and exchange rae olces on counry s real secor varables.

3 MONETARY POLICY MODEL OF TAJIKISTAN: A STRUCTURAL VECTOR AUTOREGRESSION APPROACH M. Yusuf Tashrfov* Asa Pacfc School of Economcs and Governmen Ausralan Naonal Unversy Canberra ACT 2 Absrac Usng he Srucural Vecor Auoregresson (SVAR) mehod hs aer analyses he effecs of moneary olcy on Tajksan s economy for he erod 996 o 23. A number of resrcons are mosed and he conemoraneous and long-run resrcons model are used o denfy he dynamc resonse of nflaon and ouu o he moneary and exchange rae nnovaons. As a resul hese shocks are used o generae he srucural mulse resonse and forecas error varance decomoson funcons for assessng he dynamc macs of moneary and exchange rae olces on counry s real secor varables. JEL Classfcaon: E3, E42, E52, E58 and E6. Key Words: Moneary and Exchange Rae Polces, Inflaon, Ouu, Srucural Vecor Auoregresson (SVAR) and Tajksan. All corresondence o: M.Yusuf Tashrfov The Naonal Bank of Tajksan 23/2 Rudak Ave., Dushanbe Tajksan e-mal: yusuf_anu@yahoo.com * Ths aer s drawn from my PhD dsseraon a he Ausralan Naonal Unversy. I would lke o han k Dr. Tom Komas, Prof. Raghbendra Jha and Prof. Heaher Anderson for her useful commens and suggesons.

4 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Inroducon In he frs few years afer s ndeendence, Tajksan s real GDP droed o 3 er cen and nflaon rae rose above er cen annually (Goscomsa of Tajksan, 2). A mehodcal undersandng of he dynamcs of GDP, nflaon and moneary nsrumens s moran, esecally n erms of he correc drecons for economc olcy durng ranson. Ths aer analyses he mac of moneary and exchange rae olces o rce and ouu and her relave morance n deermnng rce levels and he growh rae. The sandard framework o nvesgae he role of nnovaons on he economy and her ossble deermnaon s a Srucural Vecor Auoregresson (SVAR) model. A srucural VAR can be used o descrbe he dynamc effecs of nnovaons n moneary olcy on dfferen macroeconomc varables and o esmae he effec of moneary olcy nnovaons n accounng for arcular macroeconomc evens n revous mes. Ths echnque has been used by Sms (982, 986), Blanchard (989), 3

5 Bernanke and Blnder (992) and Crsano e al. (994) for he US economy, and Gannn e al. (995) for he Ialan economy. Noe ha mos of he emrcal research n hs area relaes o he US, whle some emrcal sudes refer o oher economes as well, for examle, hose of Sms (992), Cushman and Zha (997) and Chrsano e al. (998). Accordng o her sudes, an unforeseen ghenng of moneary olcy n he frs hase reduces moneary aggregaes and varous economc acves and, n he nex sage, he rae of nflaon falls. An unforeseen ghenng of moneary olcy n a small oen economy always leads o local-currency arecaon as does n large economes. However he reacon of rces s faser n small economes because of he qucker resonse of he exchange rae o changes n moneary olcy (Cushman and Zha, 997). The urose of hs research s: ) o evaluae nflaon argeng n he secfc conex of a small ranson economy, such as Tajksan; and 2) o examne he shor run and long run effecs of he Naonal Bank of Tajksan s (NBT) moneary and exchange rae olces on he counry s economy, n arcular on he level of rces and real ouu durng he ranson erod Usng monhly macroeconomc daa, a sysem of fve equaons for he SVAR model s consruced. These varables are, real ouu growh rae, nflaon rae, growh rae of money suly (M), nomnal neres rae and growh rae of nomnal exchange (local-currency derecaon). I should be noed ha, o he bes of my knowledge, hs s he frs ever model for Tajksan. Gven s shor hsory whou a cenral lannng srucure, he consrucon of such a model s dffcul. However, n vew of he fac ha he Tajk economy needs urgen sablsaon and hgher growh, he need for such a model s ressng. The rocedure for esmang a sraghforward SVAR nvolves a number of searae ses. The frs se nvolves esmang he reduced form VAR usng OLS, ensurng ha 4

6 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach enough lags are ncluded o ensure no seral correlaon from he resduals. The nex se s mosng suffcen resrcons o denfy he srucural arameers of he model. Then, n he case where he shocks are assumed o have emorary effecs he shor-run resrcon SVAR model s used and, n conras, where he shocks are assumed o have ermanen effecs, he long-run resrcon SVAR model s used. In he fnal se he orhogonalsed and srucural mulse-resonse funcon and forecas-error varance decomoson are analysed. In he model secfcaon ar of hs sudy, a leas en resrcons are mosed o comleely denfy he srucural model. Followng revous sudes, such as Sharo and Wason (988), Blanchard and Quah (989), and Malszewsk (2, 23), and akng no accoun he ransonal suaon of Tajksan s economy, several assumons are made and a number of resrcons are mosed o dfferenae he srucural models of hs sudy. Assumng ha moneary and exchange rae olces have conemoraneous effecs on he nflaon rae, necessary resrcons for he frs model can be denfed. In he second model, he long-run resrcons have been assumed for real GDP and rce resonses o moneary nnovaons. Hence moneary shocks have been used o generae he mulse resonse and forecas error varance decomoson funcons o assess he dynamc macs on he dfferen economc varables. The esmaed resuls show ha moneary olcy nnovaons do conan an moran cause of nflaon varably n he shor-run resrcon model raher han he long-run. However he nomnal exchange rae and rce shocks accoun for he major redcve ower of rce varably n he shor-run and long-run resrcon models. On he oher hand he morance of moneary and exchange rae nnovaons as causes of real ouu varably ncreases n he long-run resrcon model. Neverheless, moneary and 5

7 exchange rae nnovaons conrbue subsanally o varably of rce level and real ouu durng he ranson erods. Overall he analyses suggess ha a beer erformance of mlemenng moneary and exchange rae olces, n arcular money suly, neres rae and nomnal exchange rae, are he key nsrumens for any ransonal develong economy o aan lower nflaon. Ths aer s organsed as follows. Secon 2 descrbes he role of moneary olcy and nflaon argeng n Tajksan. Secon 3 resens mehodology and model secfcaons for srucural VAR. Secon 4 rovdes he daa and varables, and dscusses he esmaon resuls. Secon 5 concludes he aer. 2 Moneary Polcy and Inflaon Targeng n Tajksan Durng he years under he cenral lannng sysem (under he USSR) Tajksan had a closed economy. Afer he fall of he Sove Unon n 99, Tajksans s economy wen hrough several hgh nflaon hases, caused by he counry s economc and olcal nsably; cvl war, enormous budge defcs, rce shocks and weak moneary olces. As a resul of hese dsurbances consumer rces n some erods rose 2-5 er cen er monh and annual real GDP fell abou 2-3 er cen. The balance of aymens crss was one of he man facors ha caused hgh nflaon and oher roblems n he moneary sysem of he counry. Throughou he erod 99-95, he nflaon rae acceleraed due a shar devaluaon of he Russan ruble. Tajksan was he las former USSR reublc ha was oerang wh he Russan ruble as he naonal currency, unl May 995. Ths brough abou hgh levels of flucuaon n key macroeconomc varables, n arcular rces and real ouu. In addon, here was a large exanson n The frs currency reform was n May 995, swchng from he Russan ruble o he Tajk ruble. As a resul of hs reform he oulaon had consderable fnancal losses. 6

8 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach he oal cred value n he economy, mosly n he ndusral and agrculural secors ha led sgnfcanly o erssence of hgh levels of nflaon. Fgure (a) Inflaon rae varaon n Tajksan beween 996 and 23 (%) nflaon year Source: Auhor s own calculaon (usng Goscomsa of Tajksan daa). Fgure (b) Real ouu varaon n Tajksan beween 996 and 23 (%) real ouu year Source: Auhor s own calculaon (usng Goscomsa of Tajksan daa). The counry s cvl war, a decrease n ouu and emloymen, fas derecaon of he domesc currency, balance of aymens errors and monearsaon of budge defcs were he major causes of hgh nflaon and economc crss. From md 996 a lberalsaon and sablsaon rogram wh Inernaonal Moneary Fund (IMF) and 7

9 World Bank (WB) asssance was nroduced. The man arges of he rogram were he reducon of he nflaon rae, ncreased growh of GDP and rce lberalsaon. However as a resul of he nroducon of hese olces, nflaon became he key concern of he economy. Thus mananng rce sably became he rmary arge of he NBT ha had earler amed a mananng economc develomen, (Fgure (a) and (b)). Afer facng a erod of hyernflaon, he frs years of he rogram ( ) were no very successful n erms of achevng economc growh or lower nflaon. Temorary fxng of he Tajk ruble agans he US dollar was one of he more sraghforward ways o reduce he hgh level of nflaon whch, s one of he uroses of sablsaon olces for ranson economes. The Naonal Bank of Tajksan (NBT) also reduced he amoun of cred exended o he ublc secor. A he early sage of ranson, he NBT conrolled money suly hrough cred celngs, however n he lae 99s he NBT urned o mlemenng ndrec moneary olcy nsrumens. Gradually, wh he hel of he IMF and WB, moneary and budgeary regulaon has been resored and domesc rces have sablsed comared o revous erods. Oen marke oeraons have become one of he olcy nsrumens snce her nroducon by he jon effors of he NBT and he Mnsry of Fnance n 998. A he end of 998 he effec of he Russan fnancal crss and he devaluaon of he local currency brough some new nflaonary endences, however he NBT aled measures o avod a quck fall n he domesc currency. The exchange rae was consdered as he man nsrumen for he romoon of exors and macroeconomc sably. Thus he NBT swched from a fxed o a managed floang exchange rae sysem n order o mrove comeveness of he naonal currency. A he same me he NBT mananed gh 8

10 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach lqudy managemen, holdng u money marke raes n order o reven serous derecaon of he naonal currency. The macroeconomc sablsaon rogram nroduced nflaon argeng n he moneary olcy of Tajksan. As a resul of he sablsaon rogram nflaon declned more sharly han argeed. For nsance, he rae of change of consumer rces declned from 59 er cen durng 997 o 2.7 er cen n 998 (NBT, 998, 999 and IMF, 2). The man reasons for he sgnfcan declne n nflaon were he hgh neres raes and lower real wages ha brough domesc demand under conrol. Dese he manenance of a gh moneary olcy n early 999 o deress he effecs of he exchange rae derecaon of lae 998, nflaon connued o be sensve and sared o rse beween 999 o 2 due o exended creds o he rvae secor and ublc borrowng o fnance he defcs. The hgh and connung nflaon has affeced he economy negavely, deerorang ncome dsrbuon and holdng consan he low level of nvesmen from abroad. As regards moneary olcy, dese he fac ha reserve money remaned close o he arge, ne domesc asses were sgnfcanly hgher han execed as ne creds o he governmen exceeded he redced arges. Moneary auhores mlcly sared o use exchange rae as ar of a dsnflaon rogram by holdng hgher neres raes on domesc asses and a lower derecaon rae for rsng nvesmen. However, noe ha snce 999 he NBT has gradually reduced neres raes by srcly monorng develomens on he foregn exchange marke. The NBT has also raccally sared exendng all cred hrough he cred aucon mechansm. 9

11 In June 998, he governmen of Tajksan adoed a new medum-erm economc adjusmen and reform rogram for he erod , suored by he World Bank s Srucural Adjusmen Creds, n order o combne he gans n fnancal sablsaon and nensfy srucural reforms. The key medum-erm macroeconomc objecves of he rogram were o: ) decrease he nflaon rae o 7 er cen; 2) ncrease real GDP growh o 6-7 er cen a year; and 3) ncrease foregn exchange reserves of he NBT o abou 3.5 monhs of mors by he end of 22. To acheve he nflaon and foregn reserve ams, he rogram was based on arorae gh moneary and fscal olces. The new Tajk currency somon whch relaced he Tajk ruble was nroduced on Ocober 3, 2, wh he equvalen of one somon o Tajk rubles. Ths currency reform was assessed as osve as: ) was he frs currency reform ha dd no cause losses o he counry oulaon; 2) he nroducon of he new currency was based on mrovemen of he economc and bankng sysems; and 3) mananed he lnk o hsorcal radons (Governmen of Tajksan, 2, and NBT 2). In commenng on Tajksan s currency reform, Eduardo Anna, IMF Deuy Managng Drecor, sad: currency reform s an moran elemen of Tajksan s economc rogram, whch calls for ruden moneary olcy by he cenral bank, connued fscal consoldaon, and acceleraed srucural reforms, o mrove he nvesmen envronmen n he counry (IMF News Bref, 2). As a resul of he gh moneary olcy framework, Tajksan has acheved srong macroeconomc erformance over he as hree years (2-3) as growh has been susaned a a relavely hgh rae, whle nflaon has declned as well. Real GDP exanded by 3 er cen n 2, over 9 er cen n 22 and abou er cen n 23.

12 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Meanwhle, nflaon was less han 3 er cen n 2, abou 4.5 er cen n 22 and 3.7 er cen n 23 (NBT, 2-23; Goscomsa of Tajksan, 22). The exchange rae remaned sable ll November 2, bu weakened afer ha, reflecng an exanson of lqudy and olcal nsables over he suaon n neghbourng Afghansan. Dese some derecaon n he exchange rae, he nomnal exchange rae has remaned mosly sable over he las wo years. Whle he mlemenaon of varous sablsaon rograms snce 997 lowered he level of nflaon, was only emorary, and, as was shown above, nflaon remans a key dlemma for he develomen of he Tajk economy. The ossble reasons for oday s hgh nflaon raes n Tajksan are: ) nflaonary nfluences of rasng exchange raes hrough rces of mored goods; 2) ncreases n world rces of major mored nus (such as ol, gas, whea); 3) regonal olcal nsably (Afghansan, Russa and Cenral Asa) and, o some exen, he exsng hgh ublc secor budge defcs and her monearsaon; and 4) he rse n he rces of ublc secor roducs ha are used by he domesc rvae secor. There s also he queson relaed o ndeendence of he NBT. I s no easy for he NBT o carry ou s oblgaons ndeendenly from he governmen or arlamen, whch conradcs n resolvng some argeng asks. To ensure he successful mlemenaon of moneary olcy, he NBT has mroved s shor-erm lqudy conrol and develoed s ools for ndrec moneary managemen. Thus as a resul, a sable moneary seng could have subsanally ncreased foregn nvesmen and domesc savngs n 23. Furher derecaon of he local currency, dese hgh neres raes and nervenons n he foregn exchange marke, has ke he moneary seng gh and mroved erformance of he moneary and exchange rae argeng framework, or, n oher words, overall he moneary and exchange rae olces

13 have amed a followng nflaon argeng. However one of he requremens for nflaon argeng s rce sably n he economy. Hence one of he rmary goals of he NBT s o acheve and manan rce sably hrough careful mlemenaon of moneary olcy nsrumens ha can asss unemloymen and growh olces of he governmen n aanng economc develomen. 3 Mehodology and Model Secfcaon of Srucural VAR 3. Mehodology The srucural VAR s an arorae mehodology o brng mulle me-seres analyss and economc heory ogeher o deermne he dynamc resonse of esmaed varables o varous shocks ha ake lace n he economy. Ths sudy s based on he Sms model (98, 982, 986) and he general model of SVAR u forward by Gannn (992), Amsano and Gannn (997). 2 Sms (98) n hs semnal sudy se he bass for vecor auoregressons n economercs. 3 He made ossble o drec boh he relave meanng and he dynamc effec of varous dsurbances on macroeconomc varables, by descrbng how a se of me seres daa was generaed by random nnovaons n varables of neres. However, Cooley and Leroy (985) crcsed he VAR mehodology because of s aheorecal denfcaon sysem. They argued ha Sms dd no oenly jusfy he denfcaon resrcons and clamed ha a model recognsed by hs arbrary rocedure canno be nerreed as a srucural model, for he reason ha a dfferen varable arrangemen 2 Amsano and Gannn s (997) mehodologcal framework ncludes all he dfferen models used n he aled SVAR sudy. They show hree dfferen ways n whch he SVAR roceeds. Srucures n he VAR model are referred o as he K model, he C model and he AB model. Acually he AB model can be ransformed no a K model (shor-run resrcons) or a C model (long-run resrcons). 3 Inroducon o vecor auoregressons (VAR) s gven n Aendx A. 2

14 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach generaes dfferen srucural arameers. Thus, mrovng he srucural arameers from an esmaon rocedure requres some resrcons. Snce Cooley and Leroy s crque, macroeconomss have begun o concenrae more on he maer of denfyng resrcons. Sms (98, 986), Bernanke (986) and Sharo and Wason (988) u forward a new caegory of economerc model, non-recursve resrcons on he conemoraneous neracons among varables for denfcaon, ha s now known as he srucural VAR aroach. As economc heory ofen does no rovde enough sgnfcan conemoraneous resrcons, Sharo and Wason (988) and Blanchard and Quah (989) nroduced resrcons for long-run SVAR models. Followng revous sudes we consder a sysem of smulaneous equaons mled n vecor form as: Βy y ( L) + Μ = γ + Α ε, () Where y s a vecor of endogenous varables, γ s he fxed consan, y - s a vecor of her lagged values, ε s a vecor of random error of he dsurbance erms for every varable whch caures any exogenous facors n he model, B s he square marx of dmenson n n, where n s a number of varables, and conans he srucural arameers of he conemoraneous endogenous varables, A(L) s a marx olynomal n he lag oeraor L of lengh, and M s he square n n marx, whch conans he conemoraneous resonse of he varables o he nnovaons (dsurbances). The frs sage of srucural VAR analyss s he esmaon of he reduced form. As he coeffcens n he marces of () are unknown and he varables have emorary effecs on each oher he model n hs form canno be comleely denfed. Bu, s lkely o 3

15 ransform () no a reduced-form model by mullyng boh sdes of he equaon by he nverse marces of B, whch brngs abou he sandard VAR reresenaon as: Υ ( L) y e = α + D +, (2) ( L) Β Α( L) where α, D e = Β = Β γ and Με. = The error erms e are lnear combnaons of he orhogonalsed shocks (ε ), such ha each ndvdual error erm s serally uncorrelaed wh a zero mean and a consan varance. Whle dfferen from he dsurbance erms ε, he error erms n e are correlaed wh each oher. Hence, hs rases a dlemma n recoverng he underlyng srucural dsurbances from he esmaed VAR. In VAR analyss he only source of varaon of y varables s random dsurbances ha n he reduced form are ndcaed by a vecor of whe nose e, usually called a vecor of nnovaons (Amsano and Gannn, 997). Ths research emloys he models used n he aled SVAR sudes, such as: reduced form of VAR; he shor-erm SVAR models; he long-run effec; he mulse-resonse funcon (IRF); and he forecas error varance of decomoson (FEVD). 4 The SVAR framework s generally focused on how he nnovaons o one endogenous varable affec oher endogenous varables. Also srucural VAR analyss s focused on he drecon of nsan correlaon beween nnovaon varables. Havng shor-run or long-run resrcons n he model only deends on wheher shocks are emorary or ermanen. Hence, hs sudy examnes boh he shor-run and long-run SVAR models. 4 In a shor-erm SVAR, A and B marces model, all he nformaon s abou conemoraneous correlaons. P sr denfes he srucural mulse-resonse funcons, and Psr self s denfed by he resrcons laced on he arameers n A and B. More nformaon abou shor-run and long-run resrcon models s gven n Aendx. 4

16 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach 3.2 Model Secfcaon and Idenfcaon of Resrcons The emrcal work aled n hs sudy s n he form of a small, oen-economy srucural vecor auoregresson (SVAR). Ths SVAR model s comosed of a sysem of fve equaons, decng he relaonsh beween he man macroeconomc ndcaors of Tajksan, he growh rae of real GDP, he nflaon rae, he growh rae of nomnal money suly (M), he nomnal neres rae, and local currency derecaon. The above equaon () s called a srucural VAR as s assumed o be deermned by some underlyng economc heory. Thus he srucural model of hs sudy s descrbed by he followng dynamc sysem of smulaneous equaons (.-.5): er= a r = a m = a g = a = a b2r b3m b4g b5 + a er + a 2r + a 3m + a 4g + a 5 + = = = = = 2 b2er b23m b24g b25 + a 2er + a 22r + a 23m + a 24g + a 25 + = = = = = 3 b3er b32r b34g b35 + a 3er + a 32r + a 33m + a 34g + a 35 + = = = = = 4 b4er b42r b43m b45 + a 4er + a 42r + a 43m + a 44g + a 45 + = = = = = 5 b5er b52r b53m b54g + a 5er + a 52r + a 53m + a 54g + a 55 + = = = = = er ε, (.) r ε, (.2) m ε, (.3) g ε, (.4) ε,(.5) er ε 2 σ er r ε 2 σ r m where ε ~..d. 2, σ, (.b) g m ε 2 σ g ε 2 σ 5

17 where er, r, m, g, and are endogenous varables [er s he log of he growh rae of nomnal exchange rae (NER), r s he log of he nomnal neres rae(nir) of NBT, m, he log of he growh rae of nomnal money suly (M), g s he log of real ouu growh (GDP), denoes he log of he nflaon rae or growh rae of consumer rce ndces (Prce),]. Here he exogenous error erms ε er, ε r, ε m, ε g ε are ndeenden and are nerreed as srucural nnovaons. For smlcy he seasonal dummy varables are omed n he equaons here bu hey are ncluded n he esmaon rocess. The realsaon of each srucural nnovaon s known as caurng unexeced shocks o s deenden varable (resecvely), whch are uncorrelaed wh he oher unexeced shocks (ε ). In (.-.5), he endogeney of er, r, m, g, and s deermned by he values of coeffcens of b. The model (..5) can be wren n marx form as: b b b b b 2 b b b b b 3 b 23 b b b b 4 b b b b b er r m g er a er a ε a2a3a4a5 r r a2 a2a22a23a24a25 ε m m = a 3 + a 3a32a33a34a35 + ε, (.c) g g a4 a4a42a43a44a45 ε a 5 a5a52a53a54a55 ε where =,2,.,n. The reduced form of he VAR model ha s o be esmaed does no have he nsananeous endogenous varables and, followng (2) above, s shown by he equaons (2.-2.5): 6

18 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach er = α der + d2r + d3m + d4g + d5 + = = = = = + e er, (2.) r = α m = α g = α = α d 2er + d 22r + d 23m + d 24g + d 25 + = = = = = 2 + d 3er + d 32r + d 33m + d 34 g + d 35 + = = = = = 3 + d4er + d42r + d43m + d44g + d45 + = = = = = 4 + d 5er + d 52r + d 53m + d 54 g + d 55 + = = = = = 5 + e e r e e m g, (2.2), (2.3), (2.4), (2.5) Noce ha he reduced form errors e, are lnear combnaons of he srucural errors ε and have a covarance marx E(e e )=Σ. Also he coeffcens, a, of he lagged varables n he srucural VAR model vary from he coeffcens, d, of he lagged varables n he reduced form VAR model. Whou mosng a number of resrcons, he arameers n he SVAR model (..5) canno be denfed. To denfy a moneary and exchange rae olcy shock n Tajksan he shor and long-run arameer resrcons are aled. Idenfcaon of shocks n he sysem descrbed by equaons (.-.5) and (.b) requres mosng a leas n(n-)/2 suffcen resrcons. Table (n Aendx) resens he mosed shor-run arameer resrcons (B marx). In he shor-run verson of he SVAR model, resrcons are requred for secfc denfcaon. Resrcons can be mosed on he srucure of he vecor of error erms, (e ), on he bass of economc heory. Each equaon mus hold an ndeenden srucural dsurbance erm. 7

19 The denfcaon of resrcons n he real secor (rces and real GDP) s obaned by assumng ha moneary secor varables affec he real secor only wh a lag (A marx). The real secor varables have no effec n denfcaon of shocks n he moneary secor. Assume ha nomnal exchange rae does no conemoraneously affec real GDP. Two more resrcon are ha he nomnal varables have no conemoraneous effec on neres rae and he las wo addonal resrcons are ha real ouu does no have a conemoraneous effec on rces, and change n he neres rae may no conemoraneously change M. Dese of NBT s hgh neres rae, he demand for borrowng money dd no declne n he shor-run as he execaon of hgh nflaon was obvous. Thus a hgh nomnal neres rae was less effecve on conemoraneously varaon of NER and M. However change n he nomnal exchange rae has a sgnfcan effec on varaon of money suly (Auhor s ersonal communcaon wh Mr. Samkhon Qurbonov he head of he NBT s Moneary Polcy Dearmen, January 2). Overall, n denfyng he shor-run resrcon model I have only hree overdenfyng resrcons. Noce ha he resrcons denfyng he moneary secor do no rely on a arcular olcy regme. Also he advanage of he shor-run SVAR model s ha he mulse resonse funcons can be aled o check wheher he shocks have an effec on each endogenous varable as economc heory execs. In oher words, he dfference of hs mehod from he Cholesky decomoson s ha he IRF (Imulse Resonse Funcon) and FEVD (Forecas Error Varance Decomoson) effecs from hese shor-run resrcons can resen drec economc meanng from he analyss (Sms, 98; Hamlon, 994; Enders, 995; and Amsano and Gannn, 997). Aar from denfcaon of srucural shocks by shor-run arameer resrcons on A and B marces here s an alernave aroach of mosng resrcons on he long-run arameers for he srucural dsurbances (C marx gven n Aendx). The mehod of 8

20 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach long run srucural VAR analyss nroduced by Sharo and Wason (988), and Blanchard and Quah (989) based on he hyohess ha he long-run effec of arcular shocks on arcular varables s resrced. Ths echnque can be more aracve for macroeconomss, as he long-erm roeres of economc heory caure more undersandng han he shor-erm. Table 2 n Aendx resens he denfyng long run arameer resrcons of C marx for hs sudy. We assume ha n he case of ransonal economes (lke Tajksan) M and NIR have a long run effec on rces and real GDP. The exchange rae shocks wll change oher economc and moneary varables n he long run. In he long-run a rad growh rae was execed as rces became more sable consequenly he nomnal exchange rae nnovaons has been effecng o he varaon of nomnal neres rae and money suly (Auhor s ersonal communcaon wh Mr. Samkhon Qurbonov, he head of he NBT s Moneary Polcy Dearmen, January 2). Thus n denfyng he long-run model we have only one overdenfyng resrcon, whch s shown by a lkelhood rao es value n whch he long-run model s exacly denfed as n he shor run. To summarse hs secon noe ha a srucural VAR s a sandard VAR where he resrcons requred for denfcaon of he srucural model are gven by economc heory. However he resrcons can be shor-run or long-run, mosly relan on economc heory, deendng on wheher he shocks are emorary or ermanen. 4 Esmaon and Resuls Monhly daa beween January 996 and December 23 are used o esmae he srucural VAR model of hs sudy. All he daa, (he monhly growh rae of consumer 9

21 rce ndexes, growh rae of real GDP, growh rae of nomnal money suly (M), growh rae of nomnal exchange rae (a weghed average of USD o domesc currency) and nomnal neres rae (NBT neres rae)), are from he sascal dearmen of he NBT. Followng heory and n order o avod any economerc roblems n he esmaon model, he naural log s aken for all varables, ncludng nomnal neres raes. The me dummy varables are ncluded for he ossbly ha shocks relaed o an unexeced change of varables n ha erod are no deced as oher shocks allocaon n he model. Un roo ess resul n Aendx (Table 3) show ha all he varables are I (), ha s saonary n levels. Snce he NBT s nomnal neres rae does no vary for some erod of me (several monhs) and as well n order no o om a large number of observaons, he lag lengh s se o wo, whch s he omal value accordng o he Akake creron. 5 Resuls are gven n Table 4 (Aendx). Inally I use reduced form VAR, he shor-run and he long-run resrcon of SVAR models, o evaluae nflaon argeng and he resonse of real secor (rces and GDP) o moneary and exchange rae nnovaons n he secfc conex of Tajksan s economy. Hence hese esmaed shocks are used o generae he srucural mulse resonse and forecas error varance of decomoson funcons for assessng he dynamc macs of moneary and exchange rae olcy on real secor varables. 5 Snce he number of observaons s no large, we are neresng only on wo key crera, such as AIC and SBIC, o deermne he omal lag lengh of he VAR sysem n hs sudy. See Enders (995) for more deal on lag lengh and crera s. 2

22 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach 4. Reduced Form In he fve varable reduced form VAR models, NER, NIR, M, real GDP and Prce are esmaed o see he effec of moneary and exchange rae olcy, as well as he conrbuon of he real ouu o rce level (nflaon). All of he varables are n logarhmc form. Also seasonal dummes are ncluded n he esmaons n order o escae he effecs of seasonal aerns observed n he seres. The esmaon of VAR resuls s demonsraed n Table 5 (n Aendx). Real GDP rovdes a lle nformaon abou he rces because coeffcens are very small and sascally nsgnfcan a each esmaed lag. Ineres rae behaves smlarly o GDP, bu dese non-sgnfcance has large coeffcens whch mgh be affeced f we use SVAR model o mose arameer resrcons. Whle, M and NER have a redcve nfluence on he varaon of rces (nflaon). Coeffcens of M are hgh and sgnfcan a he frs lag, alhough he coeffcen of NER s greaer and sgnfcan a he second lag. Ths ndcaes boh ha money suly and nomnal exchange rae do have an moran effec on nflaon varaon. The hghes redcve nformaon abou rces comes from he rces hemselves a he frs lag a whch he coeffcen s hgh and sgnfcan. We also es for sably of he VAR model. The resul shows ha all he egenvalues of he model le nsde he un crcle, whch esablshes ha he VAR sasfes he sably condon. 6 These resuls show ha he nflaon varaon s from nflaonary execaons as well as from moneary and exchange rae olcy nsrumens. However s lkely ha beer redcons could be acheved only by alyng he shor-run and long-run resrcon models of srucural VAR. 6 Secon 4.4, rovdes more deals on VAR / SVAR sably es. 2

23 4.2 The Coeffcens of he Conemoraneous Varables Afer sx monhs of nflaon argeng, he nflaon rae n 998 droed o 2.7 er cen (from 59.6 er cen n 997) and growh rae rose o 5.3 er cen (.7 er cen n 997). Followng good economc achevemens durng 998, from January 999 he NBT sared o gradually decrease he nomnal neres rae by focusng more on oher nsrumens (olces) such as money suly and exchange rae. Followng Sms and Zha (22) and Malszewsk (23) for he esmaon of shor-run arameers, a lmed me-varaon n coeffcens of he model s used n order o observe changes n moneary olcy desgn and nflaon argeng. The subsequen samle for he shor-run model s also analysed o see he level of nconssency of coeffcens. Table 6a and 6b (n Aendx) resens esmaed coeffcens of he conemoraneous (shor-run) endogenous varables (marx B). Noe ha he lkelhood rao (LR) es sascs for he null hyohess of over-denfyng resrcons are 2.7 and 5.6. Under he null hyohess hese sascs have a ch square dsrbuon wh hree degrees of freedom. Thus he denfed resrcons canno be rejeced a any reasonable sgnfcance level. The esmaon resuls ndcae ha he sgn and sgnfcance levels of he coeffcens of he conemoraneous varables vary durng he wo analysed erods exce for he nomnal exchange rae and real ouu. Addonal esmaons show ha only a change n he exchange rae olcy could conemoraneously change he nflaon rae for he whole samle, whle moneary nnovaons do affec he level of rces conemoraneously, s no as much as hey affec exchange rae. However he effecs of money suly (M) and nomnal neres rae on rces are only sgnfcan n he frs esmaon. 22

24 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Overall, from he above esmaed resuls, can be concluded ha n he shor-run resrcon model he exchange rae and moneary olcy have more effecs on he varaon of rces han real ouu. 4.3 The Long-run Resrcons Model For measurng he ermanen effecs of moneary and exchange rae shocks on real secor varables n a ransonal economy, I aly he long-run resrcon SVAR model. Table 7 (n Aendx) resens he resul of he long-run srucural VAR model (marx c). The LR es sasc for he null hyohess of he overdenfyng resrcons s 2.26 and under he null hyohess hs sasc has a ch square dsrbuon wh one degree of freedom. Therefore he denfed resrcons canno be rejeced a any level of sgnfcance. The long-run SVAR model esmaon shows he ermanen effec of moneary and exchange rae olces on nflaon, as well as he conrbuon of real ouu o rce level. The obaned coeffcens are sascally sgnfcan a esmaed lags. When comared o he shor run he real GDP n he long run rovdes a clear effec on rces and s relevan o he heory ha ncreases n ouu growh wll decrease rces n he long run. Dese havng low coeffcens he moneary olcy nsrumens have conrbued sgnfcanly o a changng nflaon rae n he long-run model. However, he hghes redcve nfluence for rces n he long run s he coeffcen from he nomnal exchange rae. Wh resec o he growh rae argeng he esmaed long-run model reveals ha moneary nnovaons have greaer effecs han exchange rae. The resul shows ha NBT neres rae and money suly (M) have conrbued 23

25 subsanally o ouu growh n he long run. Local currency derecaon has negavely nfluenced and dmnshed he growh of real GDP n he long-run resrced model. Ths mles ha he NBT has been mlemenng beer moneary olcy nsrumens for nflaon and growh rae argeng, and ermng a managed floa for he exchange rae. Overall, he moneary and exchange rae shocks have nfluenced he real secor of he economy (rce level and real ouu), bu her conrbuon o varaon of real ouu s hgh, whch s dfferen from he shor-run resrcon model. 4.4 Sably and LM ess For he condon of sably of he VAR / SVAR models he sably of models was measured o fnd ou wheher egenvalues n hs model le whn he un crcle and how VAR / SVAR model sasfes sably condons. The resuls n Table 8 (Aendx) shows ha all he egenvalues of he shor-run and long-run resrcons model le nsde he un crcle, whch ells us ha srucural VAR sasfes sably condons. The Lagrange muller (LM) es s conduced o see ha dsurbances are no auocorrelaed n os analyss of VAR and SVAR models (Johansen, 995). One of he assumons uon whch nference and os analyss afer VAR and SVAR are redced s ha he errors s no auocorrelaed. The obaned LM sascs for resdual auocorrelaon afer he srucural VAR model show ha here s no auocorrelaon a esed lag order or 2, snce we canno rejec he null hyohess, hs es does no rovde any hn of model mssecfcaon. The above es resuls are summarsed n Table 9 (n Aendx). 24

26 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach 4.5 Imulse Resonse Analyss 7 The mulse resonse analyss descrbes how nnovaons (shocks) o one varable affec anoher varable afer a gven erod of me. Sms (98) Cholesky decomoson s one mehod o denfy he mulse-resonse funcons n a VAR model. Hence, he Cholesky decomoson denfcaon mehod corresonds o srucural VAR. However he am of he srucural VAR s o aly economc heory (raher han Cholesky decomoson) o beer oban he srucural nnovaons from he resduals e (Enders 995). As we are esmang he shor-run and long-run resrcons of he srucural VAR model, aar from usng orhogonalsed IRF s necessary o observe n (srucural) mulse resonse funcons for each ndeenden shock for beer analysng esmaed models. The orhogonalsed (Cholesky) and srucural mulse resonse analyses only nclude resuls for model b and model c for 2 erods ahead of he real secor varables (rces and real GDP) and some moneary varables ha normally allow for a sensble economc nerreaon. The esmaed orhogonalsed and srucural IRF for boh shor-run and long-run resrcon SVAR models s resened n Fgure 2 (n Aendx). The grahs reveal ha he shaes of he funcons are very smlar over he wo models. Furher, grahs llumnae only one man dfference over he wo models. In model b, he esmaed orhogonalsed and srucural IRF move alongsde each oher almos n all grahs. In model c here s a ga beween orhogonalsed and srucural IRF, whch means ha shor-run and long-run resrcon models of esmaed SVAR could show 7 Accordng o Amsano and Gannn (997) here s one key dfference beween long-run and shor-run resrcon SVAR models. In he shor-run model, he consrans are aled drecly o he arameers n he A and B marces. However, n he long-run model, he consrans are mosed on funcons of he esmaed VAR () arameers. Esmaon and nference of he arameers n C s sraghforward, obanng he asymoc sandard errors of he SIRF needs unenable assumons. For hs reason, varrf esmaes only boosraed sandard errors for he long-run SVAR model (Saa (8.2), 24). 25

27 dfferen effecs (mulse-resonse) of esmaed varables. Noce ha he shae of he orhogonalsed IRF s abou he same n boh models. Accordng o Fgure 2(a), model b ndcaes ha a osve shock o nomnal exchange rae causes an ncrease n rces and he effec decreases afer 8 erods. In conras o model b, n he model c grah he nal effec value s lower and mles ha rce resonse o NER nnovaons s hgh n he shor-run, whch means ha he NBT s floang-managed exchange rae olcy has conrbued sgnfcanly o fallng nflaon raes. In 2(b) he srucural resonse of rces o NIR caures a osve and negave shock effec n model b and model c resecvely. I mles ha changes n NBT nomnal neres rae conemoraneously has a osve effec on rces bu n he longrun model rce resonds negavely. Fgure 2(c) shows ha M shocks n boh models have almos he same effec and s clear ha rces are affeced sgnfcanly only beween he second and hrd erods. However, he effec slowly des ou. In Fgure 2(d) rces resond conemoraneously o real ouu nnovaons, and a negave shock o real ouu causes rces o vary for abou 8 erods wh some sgnfcan ons. Grah 2(e) s of he orhogonalsed and srucural mulse-resonse funcon from he shocks o rces and resonse of rces. We see ha he denfcaon resrcons aled n model b and model c mly ha a osve shock o rces causes rces o ncrease for a very shor me and he effecs slow for abou erods, afer whch he effec dssaes. Real ouu growh argeng was denfed as he key ssue of Tajksan s macroeconomc sably rogram, whch was mlemened by he IMF, WB and Governmen. Therefore n erms of real ouu resonses o moneary and exchange rae nnovaons, as well as o rce and s own dsurbances, we examne Fgure 2(f)-(j). 26

28 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Sarng from Grah 2(f), nomnal exchange rae shock does no have a conemoraneous effec on real ouu. Consequenly nal value s equal o zero. However he effecs change gradually. Even a negave shock o NER n model b aears only afer he frs erods whle mmedaely aears n model c, and nnovaons have sgnfcan mac on real ouu growh, whch mles ha he NBT s exchange rae olcy has conrbued o real GDP growh durng he erod 997 o 23. Neverheless he NBT s nomnal neres rae, Fgure 2(g), does have a conemoraneous effec on real ouu, and a osve shock o NIR n model b sgnfcanly ersss only for he frs wo erods, hough n model c a hgh osve shock affecs real ouu sgnfcanly for abou erods. In 2(h), real ouu conemoraneously resonds o money suly bu only a osve shock has a sgnfcan effec beween erods 2 and 3. Bu n he same grah (model c), a osve shock o M affecs real ouu sgnfcanly for abou 8 erods. A hgh osve shock effec of rces o real ouu s shown n model c, Grah 2(), and afer 2 erods he effecs become negave. Whle n he shor-run resrcons model ouu does no resond mmedaely o rce shock. Smlarly Grah 2(e) o 2(j) shows a srucural mulseresonse funcon from he nnovaons o ouu o self. The man dfference s ha n model b he nal effec of srucural shock s greaer han n model c. I also mles ha a osve shock o real ouu causes he counry s real ouu o ncrease radly n he shor run and he effecs are only mnmsed n he long-run erod. In erms of moneary and exchange rae analyss, Fgure 2(k) ndcaes ha he denfcaon resrcons used n model b and model c, mly ha a osve shock o NER frs causes ncreases n money suly, hen decreases, and hen ncreases, and so on, wh effecs geng smaller n boh models. The money suly (M) resonds quckly o he nomnal neres rae shocks n model c bu a conemoraneous effec, 27

29 whch we see n model b, does no exs. Fgure 2(m) shows how he nnovaons o local currency derecaon affec he NBT s neres rae. Immedae osve shocks n he models became smooher smulaneously, however n model c, nomnal neres rae resonds sgnfcanly durng he erods esmaed. Smlar o wha s shown n Grahs 2(e) and 2(j) he srucural mulse-resonse funcon from he nnovaons o moneary and exchange rae olces caures osve shocks n boh models. The man dfference beween models s ha n he shor run he nal effec of he shocks s hgh. Analysng he srucural mulse resonse funcon for he above models of Tajksan s economy we can conclude ha he nflaon argeng resonses of rces o moneary and exchange rae nnovaons are dfferen n he wo models. Frs, he resonse of rces o he nomnal neres rae shocks s osve n model b and negave n model c, whch means ha rasng he NBT s nomnal neres rae conemoraneously does no affec rsng rces bu has a sgnfcan mac o decrease rces n he long run. Second, he effec of money suly nnovaons o rce level s smaller n he long run, whle he effecs n he shor-run model are hgh. Ths mles ha ncreasng money suly (esecally he amoun of cred gven o he rvae and ublc secors) has an effec on rsng rces n he shor run, bu had a very slow effec n he long run. As analyses show, and also based on he moneary and ransonal suaon of he Tajk economy over he as years, s mos lkely ha long-run M has more nfluence on he growh rae han rces. Fnally, rce resonds srongly o he nomnal exchange rae nnovaons n boh he shor-run and he long-run resrcon models, whch agan roves ha flucuaon of rces durng hese years s mosly relaed o varaon n he nomnal exchange rae and sably of he naonal currency. 28

30 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Overall, he srucural mulse-resonse analyses show ha he effec of moneary and exchange rae nnovaons s hgh n he conemoraneous model. These analyses sugges ha beer erformng moneary and exchange rae olces, n arcular money suly, neres rae and nomnal exchange rae, are he bes nsrumens for any ransonal develong economy as seeks o lower nflaon. 4.6 Forecas Error Varance Decomoson Analyss The Forecas Error Varance Decomoson (FEVD/SFEVD) mehod gves nformaon abou dynamc relaonshs among jonly analysed VAR and SVAR sysem varables. One more measure of he effec of he nnovaons n varable n on varable j s he forecas error varance decomoson (FEVD). Ths mehod, whch s also known as nnovaon accounng, measures he fracon of he error n forecasng varable j, afer some erod, ha s due o he uncorrelaed nnovaons n varable n. The FEVD and srucural FEVD for he shor-run and long-run resrced arameer SVAR models are shown n Fgure 3 (n Aendx) of he real secor (rces and real GDP) and moneary varables. They suor he resuls mled by mulse resonse analyss. There s a sgnfcan dfference beween model b and model c of he srucural forecas error varances (SFEVD), bu a mnor dfference can be seen beween he wo models forecas error varances (Cholesky s and he shor-run resrcon models). I can be seen n Fgure 3(a)-3(e) ha n he resen model he forecas error varance (FEVD/SFEVD) of he rce derecaon seres n he shor-run resrcon model (model b) s manly deermned by nomnal exchange rae shock, s own shocks, and o some exen by neres rae, money suly and real ouu shocks. In oher words, for model b a he 2 erods horzon, around 48 er cen of he varance n rces s 29

31 accouned for by rce shocks whle around 42 er cen resuls from currency derecaon shocks. The neres rae, money suly and real ouu conan he remanng er cen of redcve nformaon abou rces. However n he long-run resrcon model (model c), he redcve ower of exchange rae, money suly, real ouu and s own nnovaons on rces are low, exce for he srucural fracon of mean squared error (MSE) due o nomnal neres rae. The redcve ower of NER on rces declnes o 36 er cen, whle only 3 er cen of he varance n rces s caured by rce shocks. In he long-run resrcon model he conrbuon of nomnal neres rae nnovaons o rce varaon ncreases o 28 er cen over 2 erods. The rooron of varance n rces arbuable o money suly and real ouu nnovaons decreases furher because mos of he varaon n rces s due o currency derecaon shocks. Fgures 3(f)-3(j) show how he moneary and exchange rae nnovaons have redcve ower on real ouu varances n hese models. Frs, he Cholesky forecas error varance decomoson (FEVD) for boh models has he same shae and almos smlar level of varances n real ouu. Alhough he srucural FEVD are dfferen beween he wo models hey gve furher llumnaon o he relaonshs among he real secor (ouu, rce) and moneary (money suly, neres rae and nomnal exchange rae) varables. In model b, a all me horzons, exacly 6 er cen of he varance n real ouu s exlaned by s own nnovaons whle nomnal exchange rae exlans abou 28 er cen, nomnal neres rae exlans abou 7 er cen and only 5 er cen of ouu varances s accouned for by money suly and rce shocks. In he long-run resrcon model (model c), only 2 er cen of he varance n real ouu s accouned for by s own shocks, abou 32 er cen s accouned for by NER, 3 er cen by NIR, 2 er cen by money suly and only 6 er cen by rce nnovaons. These resuls are conssen 3

32 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach wh he evdence from he srucural mulse resonse funcons, showng ha he lnk beween rces and real ouu s no robus. The SFEVD analyss suggess ha exchange rae nnovaons accoun for rce flucuaons n he shor run (model b) more han moneary nnovaons, bu n he longrun resrcon model (model c) he moneary nnovaons (n arcular nomnal neres rae) accoun more for rce varances. Ths mles ha beer managemen of moneary and exchange rae olces, n arcular nomnal exchange rae, neres rae and, o some exen, he money suly, are he bes nsrumens for nflaon and growh argeng for ransonal develong counres. 5 Concluson Tajksan has exerenced an enormous ncrease n s nflaon rae and declne n real ouu growh over he las decades. However, as a resul of he successful mlemenaon of moneary and exchange rae olces amed a achevng low nflaon and srenghenng he balance of aymens, n he lae 99s and he las four years (2 o 23) he nflaon rae has droed and real GDP growh has rsen sgnfcanly. The NBT s sable moneary envronmen has also suored ncreased domesc nvesmen, whch n urn s he key facor for rad growh. Ths research analyses he effec of moneary and exchanges rae olcy nnovaons as a source of consderable change n nflaon and real ouu n Tajksan. Gven hese shar changes n he macroeconomc erformance of Tajksan we need a framework o evaluae he conduc of macroeconomc olcy. Ths research rovdes such a framework n he shae of an SVAR model and s he frs o do so for Tajksan. Our 3

33 man neres has been o nvesgae how ceran shocks affec nflaon argeng and o examne he effec of moneary and exchange rae olcy nnovaons on he counry s real secor economy, n arcular on he change of rces and real ouu. Usng monhly macroeconomc daa, an SVAR model s consruced, whch ncludes daa for Tajksan s ranson erod, beween 996 and 23. The reduced form VAR, followed by a shor-run and long-run resrcon SVAR model, allows denfcaon of moneary and exchange raes olcy nnovaons wh a conssen dynamc resonse of key macroeconomc varables. The dynamcs of he esmaon models are resened by analyss of srucural mulse resonse funcons and he srucural forecas error varance decomoson. The coeffcens of he shor-run resrcon model reveal ha exchange rae olcy nnovaons are more effecve han moneary olcy nnovaons on he nflaon argeng rocess n Tajksan. The moneary and exchange rae nnovaons have nfluenced he real secor varables, n arcular rce level and real ouu n he longrun resrcon model. Imulse resonse analyses for he above models show ha he nflaon argeng resonses of rces o moneary and exchange rae shocks are dfferen n he wo models. The resonse of rces o he nomnal neres rae shocks s osve n he shor-run and negave n he long-run resrcon models. The effec of money suly nnovaons on rce level s hgh n he shor run. The rce resonse o nomnal exchange rae nnovaons s hgh n boh models. The SFEVD analyss suggess ha exchange rae nnovaons accoun for rce flucuaons n he shor run (model b) more han moneary nnovaons. However n he long-run resrcon model (model c) he moneary nnovaons (n arcular nomnal neres rae) accoun more for rce varances. 32

34 M.Yusuf Tashrfov - Moneary Polcy Model of Tajksan: A Srucural Vecor Auoregresson Aroach Overall our esmaons show ha moneary olcy nnovaons are a more moran cause of nflaon varably n he shor-run resrcon model. However he nomnal exchange rae and rce shocks accoun for he major redcve ower of nflaon varably n he shor-run and he long-run resrcon models. On he oher hand he morance of moneary and exchange rae nnovaons as causes of real ouu varably ncreases n he long-run resrcon model. The man dfference beween he shor-run and long-run resrcons of he models s ha n he shor run moneary and exchange rae nnovaons have a greaer mac on varaon n nflaon, whle n he long run hese olcy nnovaons are more effecve n enhancng growh. Thus, he NBT s moneary and exchange rae olces have conrbued sgnfcanly n aanng low nflaon and hgh real ouu of Tajksan s ransonal economy beween 998 and 23. Illumnang he resonse of rces and real ouu o moneary and exchange rae nnovaons s ceranly he key objecve of hs research. Hence he ncluson of moneary secor nnovaons no he srucural VAR model and analysng shor-run and long-run resrcons, as well as he dynamc analyss of SIRF and SFEVD, hel us o undersand he ransmsson rocess of Tajksan s moneary and exchange rae olces. 33

35 References Amsano, G., and Gannn, C., 997. Tocs n Srucural VAR Economercs, Second, revsed and Enlarged Edon, Hedelberg: Srnger, Berln. Bernanke, B Alernave Exlanaons of he Money-Income Correlaon, Carnege-Rocheser Conference seres on Publc olcy, 25: 49-. Bernanke, B. and Blnder, A., 992. The Federal Funds Rae and he Channels of Moneary Transmsson, Amercan Economc Revew, Blanchard, O.J., 989. A radonal Inerreaon of Macroeconomc flucuaons, Amercan Economc Revew, 79: Blanchard, O.J., and Quah, D., 989. The Dynamc Effec of Aggregae Demand and Suly Dsurbances, Amercan Economc Revew, 79: Chrsano L., Echenbaum M., and Evans C., 994. Idenfcaon and he Effecs of Moneary Polcy Shocks, FRB Chcago Workng Paer 94/7., 998. Moneary Polcy Shocks: Wha have we Learned and o wha End? NBER Workng Paer seres, no.64. Cooley, T. and Le Roy S., 985. Aheorecal Macroeconomcs: A crque, Journal of Moneary Economcs, Cushman D., and Zha T., 997. Idenfyng Moneary Polcy n a Small Oen Economy under Flexble exchange raes, Journal of Moneary economcs, 39: Enders, W Aled Economerc Tme Seres. Wley seres n robably and mahemacal sascs. USA Gannn, C., 992. Tocs n Srucural VAR Economercs, Srnger-Verlag Berln Hedelberg. Germany Gannn, C., Lanzaro, A. and Segheln, M., 995. A radonal nerreaon of macroeconomc flucuaons: The case of Ialy, Euroean Journal of Polcal Economy : Goscomsa of Tajksan, 2. Annual reor, Dushanbe Goscomsa of Tajksan, 22. Annual reor, Dushanbe Governmen of Tajksan, 2. Reor, Dushanbe Hamlon, J. D., 994. Tme Seres Analyss. Prnceon Unversy Press. IMF, 2. Counry reor, Washngon. Inernaonal Moneary Fund Inernaonal Moneary Fund, News Bref No. /97, Ocober 26, 2 34