Department of Economics Master Thesis February Rule-based Monetary Policy for Developing Countries, Evidence from Developed Countries

Transcription

1 Deparmen of Economics Maser Thesis February 9 Rule-based Moneary Policy for Developing Counries, Evidence from Developed Counries Auhor: Saiful Islam, Mohammed ID 7193T15 Supervisor: Klas Freger

2

3 C O N T E N T S Page Absrac 5 1. Inroducion. Moneary Policy for Developing Counries.1. Targeing Rules.. Insrumen Rules 9 Taylor Rule 1 3. Daa and Mehodology 1. Esimaion Resuls of Taylor Rule Model Specificaion and Simulaion 1. Simulaion Resuls.1. Developing Counries 3.. Developed Counries 7. Modificaion of Taylor Rule 9. Conclusion 3 Appendix A 3 Appendix B 3 Appendix C Appendix D References 5 Lis of Tables Table 1. Componens of Exising Social Loss in Six Counries for he period 19-7 Table. Esimaed Taylor Rule, Goodness-of-fi and Tes Resuls for Uni Roo 15 Table 3. Acual and Taylor Rule-induced Shor erm Ineres Rae in Pakisan 1 Table. Acual and Taylor Rule-induced Shor erm Ineres Rae in India 17 Table 5. Acual and Taylor Rule-induced Shor erm Ineres Rae in Sweden Table. Acual and Simulaed Social Loss in Pakisan Table 7. Simulaion of Pakisan Economy wih Differen Ses of Parameer Values 5 Table. Simulaion of Indian Economy wih Differen Ses of Parameer Values 7 Table 9. Simulaion of Swedish Economy wih Differen Ses of Parameer Values 9 Table 1. Decomposiion of Social Loss in Pakisan wih Ineres Rae Smoohing 31 3

4 Lis of Figures Page Figure 1. Ineres Rae Behaviour of Three Developed Counries Figure. Ineres Rae Behaviour of Three Developing Counries 7 Figure 3. Call Money Rae and Treasury Bill Rae in UK 1 Figure. Acual vs. Rule Induced Ineres Rae in Pakisan 17 Figure 5. Acual vs. Taylor Rule-induced Ineres Rae in India 1 Figure. Acual vs. Rule Based Ineres Rae in Bangladesh 1 Figure 7. Acual vs. Rule-based Ineres in Sweden (19-) 19 Figure. Acual vs. Rule-based Ineres in Sweden (1993-) 19 Figure 9. Acual vs. Taylor Rule-based Ineres Rae in UK 1 Figure 1. Acual vs. Taylor-Rule Based Ineres Rae in USA 1

5 Absrac: Price and oupu sabiliies deermine he success of moneary policy in eiher economy. This paper briefly examines he moneary policy sraegies of hree developed counries (USA, UK, Sweden) and hree developing counries (Bangladesh, India, Pakisan). I is found ha he developed counries follow some rule-based moneary policy whereas he developing counries wih ill-organised moneary sysem do no follow he rulebased policy, raher hey ofen formulae and launch policies under some discreionary framework. The fundamenal objecive of his sudy is o examine he performance of rulebased moneary policy in developing counries by exracing experience from developed ones. Since is incepion in 1993, Taylor rule has become synonymous o moneary policy. Bu i is a maer of fac ha his rule was grounded on he developed economies and numerous researches have been carried ou wih he same respec disregarding he applicabiliy of his rule o he developing economies. In his paper, I use one simple macroeconomic model o simulae he economies wih he Taylor rule as moneary policy. Counerfacual simulaion confirms ha macroeconomic performance of developing economies can be improved, in erms of sabiliy in inflaion and oupu, when simple Taylor ype rule is followed and i furher improves he performance wih some degree of smoohing in he insrumen. Using daa for he period 19-, his sudy proposes a se of opimal parameer values for Taylor rule and coefficiens of lagged ineres rae for differen counries. Keywords: Developing Counries, Moneary Policy, Taylor Rule, Counerfacual Simulaion Acknowledgemen: Umos paience and guidance of Klas Freger enabled me o carry ou his research. I am ever-graeful o him for his inensive and generous supervision. 5

6 1. Inroducion John B. Taylor defined moneary policy rule as a descripion- expressed algebraically, numerically, graphically- of how he insrumens of policy such as he moneary base or he ineres rae, change in response o economic variables. In he conex of developing counry, Rangarajan (1997) views moneary policy as jus a ool o achieve he broad economic policy objecives of faser rae of economic growh, a reasonable degree of price sabiliy and promoion of disribuive jusice. Mos of he developing counries formulae moneary policy employing heir discreion. Because of he complex srucure of he economy i is ofen difficul o follow some simple rules. Developing counries have weak insiuions, small informaion se, low capaciy of professionals and moneary policy having muliple objecives wihou clear prioriisaion (Malik and Aher, 7). Calvo and Mishkin (3) indenify five fundamenal insiuional problems in developing counries: weak financial insiuions, low credibiliy of moneary insiuions, currency subsiuion, liabiliy dollarisaion and sudden sops in capial inflows. These pracicaliies induce he policy makers of developing counries employ heir discreion bu Kydland, Pressco (1977) argue ha policy makers should follow rules, raher han have discreion. The reason ha hey should no have discreion is no ha hey are supid or evil bu, raher, ha discreion implies selecing he decision which is bes, given he curren siuaion. Such behaviour eiher resuls in consisen bu subopimal planning or in economic insabiliy. The ineres rae behaviour of six differen counries of his sudy shows ha developed economies follow some rules ha generae frequen changes in ineres rae whereas he developing counries keep ineres rae unchanged for quie longer period of ime alhough he economic condiions change. Figure 1 Ineres Rae Behaviour of Three Developed Counries 1 1 rae of ineres (in percen) 1 1 Ineres rae in UK Ineres rae in Sweden Ineres rae in USA Q Q15 Q3 Q31 Q99 Q19 Q9 Q39 Q9 Q191 Q9 Q37 Q5 Q1 period

7 Figure Ineres Rae Behaviour of Three Developing Counries 5 rae of ineres (in percen) Ineres rae in Bangladesh Ineres rae in India Ineres rae in Pakisan Q Q15 Q3 Q31 Q99 Q19 Q9 Q39 Q9 Q191 Q9 Q37 Q5 Q1 peirod Figure 1 plos ineres rae of hree developed counries and Figure of developing counries. I is clear ha here are repeaed changes in ineres rae of he developed counries relaive o he developing ones. Of course, developing counries have o pay for keeping heir discoun raes less reacive o he changes in economic condiions. Following able demonsraes he social loss (SL) defined over he variabiliy in inflaion and oupu gap. Table 1 Componens of Exising Social Loss in Six Counries for he period 19- Bangladesh India Pakisan Sweden UK USA σ y σ π Social Loss The underlying social loss funcion is SL = σ y + σ π, where σ y and σ π represen variance of oupu gap and variance of inflaion respecively. Social loss of Bangladesh seems o be comparaively small bu his is surely due o he small size of daa on inflaion. Due o he unavailabiliy of daa, inflaion raes of Bangladesh have been calculaed during he periods beween 1997:Q1 and :Q1. All oher variables have been evaluaed by using he daa for 7

8 he period 19:Q1-:Q1. I is clear ha social loss of hree developing counries is larger han hree developed counries. USA has he minimum social loss which is followed by UK and ou of six counries hese wo counries follow Taylor rule as moneary policy. This finding inclines me o examine wheher Taylor rule can be adoped as he moneary policy rule by he developing economies. The organisaion of his paper is as follows. In he nex secion, I describe differen moneary policy rules where in a subsecion he insighs of Taylor rule are discussed. Secion 3 presens he source of daa and mehodology used. Esimaed Taylor rules for all he sample counries are given in secion. Secion 5 discusses he model ha is used for simulaion purpose and secion is abou simulaion resuls. Secion 7 focuses he performance of augmened Taylor rule wih ineres rae smoohing and secion concludes he paper.. Moneary policy for developing counries There is lile evidence ha developing counries follow any specific rule in implemening moneary policy even here is a dispue over if he developed counries follow any rule of sric naure. Bu whaever he case, i is srongly argued ha he policy rules have greaer advanages over discreion in improving economic performance. In he academic lieraure on moneary policy in general, and inflaion-argeing sraegies in paricular, wo differen models of rule-based moneary policy have been applied (Berg, Jansson, Verdin ). Those are argeing rules and insrumen rules..1. Targeing rules This approach describes moneary policy in erms of objecives and consrains he policy makers face. The argeing-rules approach has been advocaed by Svensson () on he grounds ha i beer capures he essence of moneary policy making in inflaion-argeing counries. Inflaion argeing was firs inroduced in New Zealand in 199 wihou any prior specific academic research. I spread very quickly o an increasing number of counries: Canada 1991, he UK 199, Sweden, Finland and Ausralia Brazil was he firs developing counry o inroduce full-fledged inflaion argeing. Israel and Chile have gradually developed ino inflaion argeers. To be inflaion argeer, here should have a numerical inflaion arge, in he form of eiher a poin arge or a arge range. Achieving inflaion arge is he primary objecive of moneary policy. There is no oher nominal anchor, like an exchange rae arge or a money-growh arge. The cenral bank is accounable for

9 achieving he inflaion arge and provides ransparen and explici moneary-policy repors presening is forecass and explaining and moivaing is forecass. According o Svensson (1997) and Bernake e al. (1999), among ohers, he essence of inflaion argeing is o formulae explici objecives and o creae insiuional mechanism in order o achieve hose objecives. Such a sraegy may be difficul o capure in erms of simple insrumen rules, like Taylor s. There are wo ypes of argeing rules, general argeing rule and specific argeing rule. A general argeing rule specifies an operaional loss funcion, which he moneary policy is commied o minimise. In specific argeing rule, a condiion for seing he insrumen is specified (Malik and Ahmed, 7). I gives an implici reacion funcion of he moneary auhoriy ha needs no o be announced. According o his ype of framework, cenral banks collec large amoun of daa and hen formulae he policy in a complex way. Such a framework can bes describe he sraegy adoped by mos of he inflaion argeing cenral banks. The Bank of England and Sweden s Riksbank have formulaed a simple specific argeing rule o guide policy which can be expressed as se ineres-raes so he inflaion forecas abou wo years ahead is on arge (Goodhar, 1 and Heikensen, 1999). This ype of rule has good heoreical base, as here is no simple represenaion of reacion funcion. Specific argeing rule is boh simple and operaional, i is no necessarily opimal 1... Insrumen Rules Insrumen rules define policy insrumen as an explici funcion of he informaion available o he cenral bank. McCallum (19) rule is one of his variey where he money-base growh rae changes in response o deviaion of he nominal GDP growh rae (or he level) from a desired arge value ha grows a a specified rae. In his rule policy insrumen is he money base. Many researchers sugges ha he McCallum money-base argeing rule has undesirable sabilisaion properies. Blinder (199) provides argumens agains he money-base rule. Apar from McCallum rule, here are some oher insrumen rules namely Melzer (197), Taylor (1993), Henderson and McKibbin (1993). Taylor rule araced he researchers ineres enormously and my sudy is also o examine wheher such a rule can be proposed for he developing economies. 1 See Lars E.O. Svensson (), NBER working paper 95, p. 9

10 Taylor rule Taylor rules recommend a seing for he level of nominal ineres rae based on he sae of he economy. The rules assume ha policy makers seek o sabilise oupu and prices abou pahs ha are hough o be opimal. For insance, hey may recommend raising he federal fund rae when inflaion is above arge or lowering he federal fund rae when recession appears o be more of a hrea. The famous Taylor rule suggesed by John Taylor in 1993 represens a moneary policy sraegy for achieving price sabiliy and maximum employmen by considering ineres rae as he policy insrumen. In paricular, Taylor rule is a linear algebraic rule described by equaion (1) below ha specifies how he Federal Reserve mus adjus is funds rae according o he inflaion rae and he oupu gap: * i = r + π + h( π π ) + by. (1) where, i is he nominal rae of ineres, r is he long run equilibrium real rae of ineres, π is * he year on year inflaion rae, π is arge inflaion rae and y is percenage deviaion of real oupu from poenial oupu. Equaion 1 recommends a arge level of nominal ineres rae ha is equal o he sum of equilibrium real rae of ineres and curren rae of inflaion provided ha curren rae of inflaion is equal o arge rae and here is no oupu gap i.e., real GDP is same as poenial GDP. More formally, seing π = π and y = ino equaion 1, i follows i = r + π. This rae of ineres can be ermed as he benchmark recommendaion for he nominal rae of ineres. If his benchmark recommendaion is denoed as If π π, π π ib, hen i b = r + π. Equaion 1 becomes i = ib + ( posiiveerm). In words, his equaion illusraes ha he curren rae of inflaion being above argeed inflaion, Taylor s recommended ineres rae should be above he benchmark ineres and vice versa. The same holds rue for a nonzero oupu gap. If he economy is operaing a is poenial i.e., i π = π and y = hen i = ib = r + π = 1. This explains he imporan feaure of Taylor rule. A one percen increase in curren π inflaion rae calls for a one percen increase in nominal ineres rae such ha he real ineres rae is consan. This is real rae of ineres ha should be cared when moneary policy is formulaed because he real rae of ineres affecs real economic aciviy. Real rae of ineres, however, is no he policy insrumen, raher policy makers adjus nominal rae of ineres as 1

11 policy insrumen. If here is inflaionary pressure, ineres rae should be increased such ha i dampens demand and hence inflaion. Similarly, a posiive oupu gap can be vanished hrough he increase in rae of ineres and a negaive oupu gap hrough he decrease in rae of ineres. Taylor rule incorporaes boh shor-run and long-run goals of moneary policy. The shor-run goal is refleced in oupu gap adjusmen facor and long-run goal in inflaion gap adjusmen facor. Taylor (1993) ses boh he long run equilibrium real ineres rae and he arge inflaion rae equal o, and h and b are se equal o.5. Using hese values, equaion (1) can be rewrien as i i = π +.5y......(), i follows ha = 1.5 > 1. This indicaes a one π percen increase in inflaion rae resuls in more han one percen increase in nominal rae of ineres and vice versa. This is ermed as Taylor s principle insrucing ha cenral bank should reac more han 1-1 o inflaion in order o lower he curren inflaionary pressure. The mechanism is sraighforward. If nominal rae of ineres is increased more han one percen following a one percen increase in inflaion, real rae of ineres will rise and demand will fall ha evenually will dampen inflaionary pressure. Alhough Taylor rule incorporaes many imporan aspecs of policy, i also is based on several oversimplified assumpions. Assumpions are embedded in all componens of he rule. Following paragraphs discuss he conemporaneousness of Taylor s variables insead of eiher lagged or forecased variables and ignorance of lagged ineres rae in curren seup. Conemporaneous versus lagged daa: Taylor rule recommendaions in a given quarer are based on he conemporaneous oupu gap and on inflaion over he four quarers ending in he same quarer. In his specificaion, i is assumed ha he cenral bank knows he curren quarer values of real GDP and a price index when i ses he nominal rae of ineres for ha quarer. In mos of he counries, cenral banks do no have daa on real GDP for a cerain quarer unil a monh or a couple of monhs afer he end of ha quarer. The unavailabiliy of curren informaion o policy makers a he ime decisions are made has led o a debae abou wheher o use curren or lagged daa in esimaing ineres-rae rules. To address his iming problem, here are some sudies viewing ineres rae in a given quarer as a funcion of oupu and inflaion gap in he previous quarer. In general, he empirical evidence does no show a subsanial loss in performance when lagged daa are used insead of curren daa (Levin e al. (1999), McCallum and Nelson (1999), Rudebusch and Svensson 11

12 (1999)). The coss are small because inflaion and oupu are persisen enough such ha lags of he inflaion rae and he oupu gap are good proxies for curren values (Hamalainen, ). Also, i is reasonable o claim ha he cenral bank has more informaion abou he sae of he economy a he ime ineres rae changes are made han is capured by inflaion and oupu alone (Baini and Haldane (1999)). I has been argued on his premise ha using conemporaneous daa insead of lagged daa can be hough o implicily include informaion ha is no refleced in inflaion and oupu measures (Kozicki (1999), Rudebusch and Svensson (1999)). Conemporaneous versus forecased daa: Some researchers prefer forward looking, or forecas-based ineres rae rules, o conemporaneous rules like Taylor s. They argue ha moneary auhoriies generally make policy decisions based on economic condiions expeced in he fuure. Bu he same argumen of persisence of inflaion and oupu holds rue in supporing conemporaneous rules over he forward-looking rules. I is no clear ha forward-looking esimaes have any advanage over conemporaneous or backward-looking versions of he rule. Alhough here are concepual benefis o using forecas-based rules, he choice of an opimal rule sill depends on he srucure of he model under consideraion, paricularly he specific wage-price conracing process. In cases where wage bargaining is backward looking, Baini and Haldane show ha forward-looking rules serve as sabilising mechanisms o couner-balance he backward-looking behaviour of he privae secor. On he oher hand, when wages are fully flexible, here is no need for forwardlooking elemens in he policy rule. However, Baini and Haldane cauion ha in exreme cases where he moneary auhoriy and he privae secor have an excessive degree of forward-looking behaviour, forecas-based rules could be desabilising. Using a backwardlooking model, Rudebushch and Svensson (1999) find forecas-based rules ouperforms he conemporaneous Taylor-ype rule. The advanage, however, is only marginal. In fac, heir resuls sugges ha conemporaneous rule is a very close second o he mos favourable forecas rules. Smes (199) uses a model based on ha of Rudebusch and Svensson (1999) wih he excepion ha poenial oupu is endogenous. Wih his modificaion, he finds ha conemporaneous rules perform similar, and marginally superior, o forecas rules. Taylor (1999a) also concludes ha forecas-based rules have lile advanage over conemporaneous rules afer he fails o find much difference beween he performance of inflaion forecass and acual inflaion in his policy rule. Taylor () noes ha as long as forecass are no oo far ou ino he fuure, hey will be very close o heir conemporaneous counerpars. 1

13 Ineres Rae Smoohing: Taylor rule does no ake ino accoun he smoohing behaviour of cenral bank. Bu here is evidence ha cenral banks have he endency of smoohing ineres rae such ha he possibiliy of insabiliy in he financial markes is minimised. Absence of smoohing implies a policy acion on he basis of curren inflaion and oupu gap. Smoohing, on he oher hand, relaes curren policy acion o inflaion and oupu gaps of several quarers raher han jus a single quarer. Someimes i may happen ha he economic impac of change in nominal ineres is uncerain. In such case i is reasonable o recommend smoohing. Taylor-ype rules are commonly modified o incorporae ineres rae smoohing by including a lagged ineres rae erm. Sack and Wieland (1999) quesion wheher ineres rae smoohing is deliberae or simply he resul of moneary policy reacing o persisen macroeconomic condiions. If ineres rae smoohing reflecs he reacion of moneary auhoriies o persisen macroeconomic variables, one would expec he coefficien on he lagged ineres rae o be small or insignifican. However, esimaed Taylor-ype rules ha include a lagged ineres rae ypically yield large and significan coefficiens, indicaing ha ineres rae smoohing is deliberae. The Taylor rule, however, incorporaes many of he feaures of good moneary policy like ransparency, accounabiliy and credibiliy. Especially, a cenral bank ha adheres o a Taylor rule reveals o he public ha i is commied o price sabiliy and sysemaically akes sep o achieve i. The public herefore keeps is expecaion of inflaion low and sable, and financial markes, in addiion anicipae he Federal Reserve s nex move and increase marke ineres raes immediaely when inflaion goes up. The origin of Taylor rule is really federal specific and i is rue ha many of he developed economies follow Taylor-ype moneary policy rule implicily or explicily. Surprisingly, here is a vas amoun of lieraure dealing wih suiabiliy and implemenabiliy of Taylor rule in developed counries bu lile is known abou developing counries. For example, here is no sudy specific o Bangladesh, only one sudy by Malik and Ahmed (7) abou Pakisan and one sudy by Vinee Virmani (199-1) abou India have been underaken so far. Besides, some researches have been performed on individual underdeveloped counries wihou any generalisaion. However, examining he feaures of all developing economies is beyond he capaciy of one shor hesis, raher effor has been pu o draw conclusion on he basis of hree counries. 3. Daa and Mehodology Quarerly daa for six counries have been rerieved from he online version of Inernaional Financial Saisics (IFS) from IMF for he period beween 19:Q1 and 13

14 :Q1. Ineres rae used for he purposes of his analysis is he bank raes of USA (Federal fund rae), Sweden, Bangladesh and India, call money raes of Pakisan and UK. Figure 3 Call Money Rae and Treasury Bill Rae in UK 1 1 rae of ineres (in percen) 1 1 call money rae reasury bill rae Q1 Q3 Q15 Q33 Q1 Q3 Q199 Q397 Q19 Q39 Q193 Q391 Q19 Q3 Q17 Q35 Q1 period I could use Treasury bill rae as he insrumen for UK bu daa demonsrae ha reasury bill rae and call money rae have approximaely same plo ha is revealed in Figure 3. CPI series of enire sample period are available for all he counries excep for Bangladesh. For his counry CPIs are available from 1993:Q3. Inflaion rae for each quarer has been compued by using CPI on yearly basis. Four quarer inflaions have been averaged and called i π in order o esimae he Taylor rule. Seasonally adjused indusrial producion index has been used o consruc oupu gap. Following he previous lieraures on he esimaion of Taylor rules, I consruced he oupu gap in percen using he Hodrick-Presco cyclical componen of he logarihm of indusrial producion. Consrucions of differen series and esimaions have been carried ou in EViews.. Esimaion Resuls of Taylor Rule In his secion I esimae Taylor rule for six counries as a posiive analysis in order o examine which counries are following his rule and if i is followed hen wha proporion of he variaion in insrumen is explained or unexplained. Esimable form of he Taylor rule is β i β + β1π + β = y which is derived from equaion (1) by seing * r hπ ; β (1 + h) β = b. Table conains esimaed Taylor rules of he six sample 1 ; counries. Esimaion resuls primarily show ha UK and USA follow Taylor rule explicily. 1

15 Low R may quesion he goodness-of-fi in case of USA, in paricular, because Taylor rule is Federal specific hence one may argue ha R should be as large as possible. Bu Judd and Rudebusch (199) underake esimaes for he US, using various sub-samples and alernaive measures of inflaion and real aciviy over he period All in all, hey esimae differen rules for he changes in he Federal funds rae. Their op-performing rule has an R a 7 percen. Mos of heir rules, however, have R around 5 percen. Thus by his measure, he explanaory power of my esimaed rule, has o be judged o be plausible. Table Esimaed Taylor Rule, Goodness-of-fi and Tes Resuls for Uni Roo Counry Pakisan Esimaed Taylor Rule Adjused R ADF es saisic for residuals i = π.5 y..(3) (5.9) (.5) (.7) India Bangladesh i =.7+.1π. y.() (1.) (1.1) (.35) i =.9.1π. y (5) (1.7) ( 1.1) (.7) Sweden i =.1+ 1.π.11 y () (.7) (1.) ( 1.9) i = π.1 y (1993-) (.71) (5.) ( 1.95) UK USA i = π +.19 y (7) (.5) (17.7) (1.3) i = π +.5 y () (.) (.7) (.1) Equaion (3) in Table is he esimaed Taylor rule wih adjused daa beween 195:Q and :Q1 for Pakisan ha is consisen wih he finding of Malik and Ahmed (7) where hey used he daa beween 1991 and 5. Before esimaing he equaion, call money rae of Pakisan has been seasonally adjused in order o avoid he influence of any seasonal componen since i is mos likely ha in some underdeveloped counries his rae shows unusual flucuaion on he eve of public holidays. Residuals series from his esimaion is saionary as he null of he uni roo in ADF es can easily be rejeced a any level of significance since he P-value of ADF es saisic is.5. As Enders () argues if he esimaed residuals are saionary hen OLS esimaes are super consisen and inegraion of he variables in he equaion does no creae any problem. Esimaed equaion confirms ha 15

16 Sae Bank of Pakisan (SBP) which is he cenral bank does no follow Taylor rule. Because, o be consisen wih Taylor rule coefficien of π should be larger han uniy and coefficien of y should be posiive. Low value of R indicaes ha oupu gap and inflaion can poorly explain he overall variaion in shor erm ineres rae. Furhermore, he value of Durbin- Wason saisic is.55 which clearly indicaes ha here is high degree of auocorrelaion in he esimaed ineres rae reacion funcion. This is a clear signal ha eiher SBP has he objecive of ineres rae smoohing or here are missing variables in he above regression. Table 3 along wih Figure shows ha he rule induced and he acual shor erm ineres rae have visibly differen behaviour. In he sample period, acual rae of ineres has smaller mean and variaion han he rule induced ineres. I implies ha Sae Bank of Pakisan responded less aggressively o oupu and inflaion han wha rule would have suggesed. This is no unrealisic for a developing counry because developing counries have differen moneary policy objecives like, ineres rae smoohing, exchange rae sabiliy, financial secor sabiliy ec. Table 3 Acual and Taylor Rule-induced Shor erm Ineres Rae in Pakisan Acual Rule Induced Mean Median Maximum Minimum Range Sd. Dev..9.3 numbers in parenheses indicae -saisics 1

17 Figure Acual vs Rule Induced Ineres Rae in Pakisan rae of ineres (in percen) 1 1 acual ineres rule based ineres year I is clear from equaion () ha India is no following Taylor rule in seing rae of ineres. There is no consensus if hey follow any specific rule for conducing moneary policy. In an aemp, Vinee Virmani ried o operaionalise Taylor-ype rules for he Indian economy by using he daa beween 199Q3 and 1Q bu he conclusion was ha McCallum rule augurs well for he conduc of moneary policy in he Indian conex bu here is ambiguiy if hey really follow any such rule. Had hey followed Taylor rule, ime pah of ineres rae would have been like he dashed line in Figure-5. In consrucing he following * Table and diagram, Taylor s original formulaion is used, i.e., r =, π =, b =.5, h =. 5. Table Acual and Taylor Rule-induced Shor erm Ineres Rae in India Acual Ineres Rule Induced Ineres Mean Median Maximum Minimum Range Sd. Dev

18 Figure 5 Acual vs Taylor rule-induced Ineres Rae in India rae of ineres (in percen) 1 1 acual ineres rule induced ineres year Equaion (5) is he esimaed Taylor rule of Bangladesh for he period 199Q- Q1. Negaive coefficiens of boh inflaion and oupu gap imply ha cenral bank does no follow Taylor rule in seing ineres rae. Adjused R is. implying ha inflaion and oupu sabilisaion ogeher explain merely % of aggregae variabiliy in moneary policy insrumen. There may have also he possibiliy of spurious regression because ADF es for residuals canno rejec he null of uni roo wih subsanial confidence. Advanced regression echniques can be employed wih necessary correcion in he ime series properies of ineres rae, inflaion and oupu gap in order o esimae he Taylor rule for Bangladesh bu he moneary policy saemen of he cenral bank does no reveal any informaion supporing Taylor rule in heir seup. Figure Acual vs Rule Based Ineres Rae in Bangladesh 1 1 rae of ineres (in percen) bank rae rule-based ineres year 1

19 Since my objecive is o examine he suiabiliy of Taylor rule, I have creaed one series of ineres rae by mechanically following he rule and graphed i joinly wih acual rae of ineres in Figure. I is quie obvious ha rule-based ineres is far from acual rae. Esimaed Taylor rules wih eiher full sample or sub-sample do no show ha Sweden follows a generalised Taylor 3. This finding is no surprising because previous researches do no conclude ha Sweden is he follower of Taylor rule being even successful inflaion-argeing since This is in line wih he argumen of Svensson (1) ha he simple Taylor rule does no give righ insighs abou wha inflaion-argeing cenral banks are doing. In a working paper How Useful are Simple Rules for Moneary Policy? The Swedish Experience Berg, Jansson and Verdin (199) concluded ha on cerain occasions, policy seems o have been more expansionary or conracionary han wha is implied by mos relevan simple rules. They found ha Riksbank from ime o anoher deviaes from he simple rules for reasons ha are usually negleced in models of moneary policy. To hem, when uncerainy abou he macroeconomic condiions has been perceived o be unusually large, a cauious policy has been followed and he repo rae has been lef unchanged despie changes in he expeced rae of inflaion. Figure 7 Figure Acual vs Rule-based Ineres in Sweden (19-) 1 Acual vs Rule-based Ineres in Sweden (1993-) 1 1 ineres rae (in percen) 1 rae of ineres (in percen) acual ineres Taylor rule-based ineres acual ineres rule-based ineres Diagrams above demonsrae ha in recen years acual raes of ineres are close o he simple Taylor rule-based ineres. Maybe Sweden is now following Taylor-ype moneary policy rule. Table 5 also gives some signs of convergence of Sweden oward 3 Coefficien of inflaion is maching wih Taylor rule bu negaive coefficien of oupu gap is no. I can sill be argued ha Sweden follows Taylor rule wih zero weigh on oupu sabilisaion because coefficien of oupu gap is no saisically differen from zero. 19

20 Table 5 Acual and Taylor Rule-induced Shor erm Ineres Rae in Sweden Mean Median Maximum Minimum Range S. dev. Acual ineres Rule-based ineres Taylor rule-based policy because average acual ineres raes and average rule-based ineres raes are almos equal wih alike sandard deviaions in boh cases. Table 5 uses daa beween 1993 and. If he whole sample period is aken ino accoun, i is found ha sandard deviaion of acual ineres rae is 3. and of rule-based ineres rae is. ha may auhenicae he phenomenon saed earlier ha recen moneary policy of Sweden is in he ne of Taylor s alhough no from he beginning. Equaion (7) is he esimaed Taylor rule for UK using he whole sample consising of 97 quarerly observaions. This equaion is simplified below o make i more explici as he simple Taylor rule: i y = + π, comparing his equaion wih i = β + β1π + β y where, β1 1+ h = 1. 5 h =. 5 ; β = b =. 19 and * * β r hπ 1.9 ; hπ = r 1.9; π * = (using sample r = 3. 5 and h =.5 ). = Finally, equaion (7) akes he form of Taylor rule i = π +.5( π ) +. 19y ha is equivalen o * i = r + π + h( π π ) + by. Underlying arge rae of inflaion is calculaed percen which is no oo unrealisic because he las observed inflaion rae is 3.9 percen. I is clear ha UK follows Taylor-ype moneary policy rule. This resul is no conroversial because in an invesigaion by Mark P. Taylor and Emmanuel Davradakis () i is concluded ha ineres rae seing behaviour of Bank of England appears o be well capured by Taylor rule wih some degree of nonlineariy in policy seing. I is unimporan o esimae he Taylor rule for USA because his rule is federal specific ha recommends a seing for he level of he federal funds rae based on he sae of he economy. Neverheless, I esimaed he equaion depiced as () in Table using whole sample and here is no deviaion from Taylor s suggesion a leas wih respec o sign of he coefficiens. In addiion o economeric evidence, figures 9 and 1 illusrae he fac ha boh USA and UK are consisenly following Taylor rule since he rule induced ineres and acual ineres have close plo. Empirical finding displays ha rule-based moneary policy in UK and in USA keeps inflaion and oupu variabiliy quie low. I can be convincingly

21 Figure 9 Figure 1 Acual vs Taylor Rule based ineres rae in UK Acual vs Taylor-rule Based Ineres Rae in USA rae of ineres (in percen) 1 1 rae of ineres (in percen) acual ineres rule-induced ineres rule-based ineres acual ineres argued ha his is he virue of Taylor rule ha could keep combined variabiliy of wo fundamenal macroeconomic indicaors a a minimum. Wih his moivaion, his paper examines if such a simple moneary policy rule can be proposed for he developing counries. 5. Model Specificaion and Simulaion This secion deals wih he normaive analysis of his sudy. In order o check wheher macroeconomic performance of developing counries can be improved hrough he implemenaion of Taylor rule as moneary policy, economies have been simulaed on he basis of Rudebusch-Svensson model along wih he Taylor rule. y = β y π = γ π β y + γ π β3( i + γ π π 1) + ε... + γ y + η (9).. (1) Rudebusch and Svensson (199) used his small Neo-keynesian ype empirical model of he US economy consising of equaions (9) and (1) o examine he performance of policy rules, where y is oupu gap in percen, i is four-quarer average ineres in percen a an annual rae, π is quarerly inflaion in percen a an annual rae and π is four-quarer average inflaion. ε and η are demand and supply shocks respecively. I have found ha for some counries he model works reasonably well wih one lag in boh oupu gap and inflaion hence i has been furher simplified as This model is shorer version of Svensson (1997) where he used he following hree equaions: = β β ( π ) + β + ε ; = γ + θ y π = π y α y i α x η ; α, β > ; β < 1; γ < 1 x 1 x 1 x 1 1

22 y = β y β ( i + η... ( ε andη are i.i.d. shocks) 1 1 π = γ π γ y 1 1 π 1 ) + ε......(9.1); ( < β < 1, β > ) (1.1); 1 ( < γ 1, γ > ) The firs equaion relaes oupu gap o is own lag and o he difference beween average ineres rae and average inflaion over he previous four quarers-an approximae ex pos real rae. This equaion can be reaed as he IS equaion or aggregae demand equaion showing an inverse relaion beween real rae of ineres and oupu. The second equaion, in conras, can be reaed as he aggregae supply equaion relaing inflaion o a lagged oupu gap and o lag(s) of inflaion. In his model π is quarerly inflaion in percen a an annual 1 rae, i.e., (lncpi lncpi 1) ; π is four-quarer inflaion, i.e., 1 3 j = π ; is quarerly j i average ineres rae in percen a an annual rae and i is four-quarer average ineres rae, i.e., 1 3 i j.the above specificaion would make sense only if j= β is posiive because an increase in real average rae of ineres rouinely lowers oupu ha is a simple represenaion of he moneary ransmission mechanism. The lags of inflaion are auoregressive or adapive represenaion of inflaion expecaion. In heir empirical analysis, Rudebusch and Svensson imposed he resricion ha coefficiens of inflaion lags sum o one. Since here is only one lag in he simpler model, I have performed he Wald es H γ 1 for each counry. : 1 = Nonrejecion of his null corresponds o Rudebusch-Svensson model s resricion. This resricion, however, only applies if here is no consan erm in he righ hand side of equaion (1.1). Equaion (1.1) shows ha oupu affecs inflaion wih one period lag and (9.1) shows ha ineres rae affecs oupu wih one period lag, i.e., ineres rae affecs inflaion wih woperiod lag. The crucial propery of he model is ha he insrumen (ineres rae) affecs inflaion wih a longer lag han i affecs oupu. According o Svensson (1997), alhough simple, he model has good heoreical properies and capures essenial feaures of he more elaborae models, which some of he cenral banks are using for policy analysis (Malik, 7). I have simulaed all he economies in sample excep UK and USA in order o invesigae macroeconomic performance wih Taylor rule. Simulaion was accomplished hrough Excel ha required he following seps: 1) Equaions (9.1) and (1.1) have been esimaed in Eviews; x is an exogenous variable. ε, η and θ are i.i.d. shocks a period ha are no known a period 1

23 ) Residuals (demand and supply shocks) have been sored in wo columns of excel shee; 3) Ineres rae ( i) has been compued by employing he Taylor rule wih differen plausible and y ; * r, π, h and b. Simulaed π and y have been used insead of acual π ) Curren π and y have been compued by using he coefficiens obained in sep 1, residuals in sep and compued values of lagged i, π and y. 5) Acual social loss and simulaed social loss have been compued by using he following social loss (SL) funcion: SL = σ y + κσ π......(11) This social loss funcion is robus in ordering for differen values of κ ha measures he degree o which sociey values sable inflaion relaive o oupu sabiliy. In his sudy κ is assumed equal o. Lower social loss is an indicaor of improved macroeconomic performance hence parameer values of Taylor rule are se in a way ha minimises social loss.. Simulaion Resuls.1. Developing Counries Pakisan Esimaed Rudebusch-Svensson (1999) model for Pakisan is described by equaions (1) and (13). Before esimaion, I checked he ime series properies of he variables by examining if hey are saionary or no. Mos of he variables are saionary alhough no all bu residuals from OLS regression are saionary ha may perhaps ensure nonspuriousness. y.y 1.( i 1 π 1) (1) = (.3) (-1.3) S.E.=3.7 DW=. π.π 1 +.1y 1... = (13) (15.7) (1.7) S.E.=.3 DW=.13 All he coefficiens have righ signs and magniudes o be compaible wih he economic heory. Wih his esimaed model and assuming Taylor rule as moneary policy sraegy, I have simulaed he economy, incorporaing in each period he esimaed shocks (o oupu and inflaion) from equaions (1) and (13). Counerfacual simulaion shows ha Taylor rule performs beer in he conex of macroeconomic performance of Pakisan because 3

24 social loss subsanially falls once he rule is implemened. Following able depics he acual and simulaed values of inflaion, oupu variabiliy and corresponding social loss: Table Acual and Simulaed Social Loss in Pakisan (b*=.71, h*=1.5) 5 Acual Simulaed Variance of inflaion Variance of oupu gap Social Loss In he above compuaion arge rae of inflaion is assumed %. Opimal weighs on inflaion and oupu gap sabilisaion have been evaluaed by using he Solver funcion in Excel. Those values are 1.5 and.71 respecively. Real rae of ineres is assumed.5 percen. I is clear ha boh oupu and inflaion variabiliy are lower under rule based policy. Depending on he sae of he economy, i may be desirable o keep he inflaion arge as small as possible. Bu none of he cenral banks, wih any moneary policy sraegy, arges zero inflaion as cenral banks are no inflaion nuers in King (1997) erminology. Too low inflaion may badly affec economic growh of a developing counry. Therefore, inflaion arge should be se a a posiive level. However, Khan and Senhadji (1) esimaed hreshold level of inflaion for developing counries ha ranges beween 7 and 11 percen. Mubarik (5) esimaed 9 percen hreshold level of inflaion for Pakisan. Acual daa reveals real ineres rae equal o. percen and hus simulaion is done wih and around his value. Apar from he opimal weighs on inflaion and oupu, simulaion is done wih differen pairs. Simulaion resuls are summarised in Table 7. There are several salien feaures of his able: (1) Rule induced average rae of ineres is smaller han he acual average whereas rulebased ineres rae has higher variabiliy. This implies ha moneary auhoriy in Pakisan kep ineres rae higher wih a low response o he changes in economic condiions. This was perhaps o avoid any sudden insabiliy arising from frequen changes in ineres rae. Rule, on he oher hand, proposes aggressive changes in insrumen wih respec o he economic condiions. 5 * indicaes opimum values. See Malik, Ahmed 7

25 Table 7 Simulaion of Pakisan Economy wih Differen Ses of Parameer Values wih opimal weighs (b=.71, h=1.5) wih equal weighs (b=.5, h=.5) Full weigh on inflaion (b=, h=1.5) Full weigh on oupu (b=.71, h=) acual Ineres rae Average Sandard deviaion Ouupu gap Average 3.E Sandard deviaion Inflaion Average Sandard deviaion Social Loss () Rule induced ineres rae has maximum sandard deviaion when all he weighs are given o inflaion. This indicaes ha if he cenral bank s objecive is o sabilise inflaion around he arge by disregarding oupu, hey should change ineres rae repeaedly ha evenually yields bigger sandard deviaion. (3) Table 7 also signals ha if he aim of moneary policy is o achieve lowes oupu variabiliy hen whole weighs should be given o oupu gap wih zero weigh o inflaion and vice versa. Figures A5 and A7 in Appendix A correspond hese cases. () In any case, loss o he sociey is smaller if rule based policy is in pracice. This is in accordance wih he findings of Malik and Ahmed (7) who used Pakisani daa for he period and came o he conclusion ha macroeconomic performance, in erms of less variabiliy of oupu and inflaion and small value of he loss o he sociey, could be improved significanly. In heir calculaion loss drops from 1.3 o 7.. Bu hey compued social loss as he simple arihmeic mean of wo variances. If he same formula is followed hen in my calculaion social loss drops from 11.7 o 9.79 for he whole sample period 19-. I find i reasonable for Pakisan o follow he Taylor rule as Moneary policy because previous sudy wih 15 years of daa and my sudy wih years of daa exhibi he common feaure of improvemen in performance of he economy under Taylor rule. This is worh menioning ha counerfacual hisorical simulaion gives he expeced resul bu sochasic simulaion does no. I le he Random Number Generaor of Microsof Excel generae repeaed series of shocks for Svensson model s inflaion and oupu wih sandard deviaions 5

26 being equal o sandard errors of regression bu he resuling loss o he sociey wih hese generaed shocks appears larger han he acual ones in mos of he cases. This is conflicing wih he finding of Malik and Ahmed (7) where hey used Pakisani daa beween 1991 and 5. Their claim was ha only ou of 1 imes, sandard deviaion of simulaed oupu gap greaer han or equal o ha of he acual daa and for 1 imes for inflaion series. This conradicion may arise from mehodological difference as hey performed i by boosrapping he sandard deviaion of oupu and inflaion which is an advanced economeric echnique raher han wha is done by Random Number Generaor. India In his sudy I esimae he Rudebusch-Svensson model for India by using he full sample 19-. Esimaed y and π are described as equaion (1) and (15). y.1y 1 +.( i 1 π 1) (1) = π.9π 1 +.y 1... = (15) Equaion (15) is consisen wih economic heory bu equaion (1) is no because i indicaes an increase in oupu following an increase in real rae of ineres. I reminds me one feaure of underdeveloped economy where a low rae of ineres increases he availabiliy of credi o he unproducive secor hence reduces credi expansion o acual invesors ha evenually lowers overall oupu. If cenral bank, on he oher hand, imposes some resricions on credi expansion oward unproducive secors, i can work successfully in favour of incremen in oupu. When he economy is simulaed wih he Taylor rule joinly wih (1) and (15), i shows an improvemen in macroeconomic performance wih respec o loss o he sociey. In his case Excel Solver gives some negaive opimal parameer values for India which is clear conradicion o Taylor s suggesion. Daa from -, however, give meaningful resuls hroughou. Esimaed Svensson model is described by he following wo equaions: y.7 y 1.3( i 1 π 1) = (.1) (.7) (1) π 7..1π 1.55π +. y 1 = (.) ( 1.35) ( 3.5) (.59) (17) Taylor rule wih opimal weighs on inflaion and oupu sabilisaion resuls in a loss o he sociey equal o whereas acual loss is Since he weighs are bigger in magniudes one can argue o use he relaive weighs 7. In line wih his reasoning, when I pu.5 weigh on inflaion and weigh on oupu sabilisaion social loss is Table 7 Opimal weighs are found as h = 9.3 and b = 5. 13

27 conains differen values of explained and explanaory variables, and accompanying social loss compued for a variey of parameer values in Taylor rule. In his consrucion, I se real equilibrium ineres and arge inflaion rae equal o and 9 percen respecively. For India Table Simulaion of Indian Economy wih Differen Ses of Parameer Values wih opimal weighs (b=, h=.5) wih equal weighs (b=.5, h=.5) Full weigh on inflaion (b=, h=.5) Full weigh on oupu (b=, h=) acual Ineres rae Average Sandard deviaion Ouupu gap Average Sandard deviaion Inflaion Average Sandard deviaion Social Loss observed daa reveal real rae of ineres equal o 1.3 and I find infiniesimal change in social loss around his value. Whaever he cases regarding parameer values, real equilibrium ineres and arge inflaion rae, i is quie obvious ha Taylor rule can perform beer han he exising policy in India since i reduces boh inflaion and oupu variabiliy wih high degree of robusness. Bangladesh Bangladesh is a less developed counry (LDC) wih an ill-organised and poorly developed moneary sysem. The moneary secor of Bangladesh is small relaive o he size of he oal economy. Abjec povery of majoriy of he people means ha hey rarely make ransacion of moneary naure. Moreover, Bangladesh inheried a number of primiive moneary and fiscal insiuions since is independence in 197. There has been no fundamenal improvemen in hese insiuions during he ensuing wo decades (Wahid, 1993). Approximaely 5 percen of he gross naional produc sill originaes in he non-moneised subsisence secor. In his secor he bulk of he oupu is reained for self-consumpion and is 7

28 no markeed. Under such condiion i is widely recognised ha he pracice of moneary policy may produce disappoining resuls. y. y 1.3( i π ) (1) = (1.5) (.5) π = (.39) (.) (.3) π 1 +. y (19) Equaion (1) is he esimaed IS equaion and (19) is he Phillips curve equaion given by Rudebusch and Svensson. Real rae of ineres and oupu gap wih one lag do no give righ economic sense neiher hey are saisically significan hence no included in he equaions above. I is quie differen from eiher he Pakisan case or Indian case ha IS equaion becomes sensible if real rae of ineres is aken ino accoun wih wo-period lag and supply equaion ges saisical and economic significance if oupu gap is incorporaed wih wo-period lag as well. This newer phenomenon of he model indicaes ha policy insrumen affecs inflaion wih four-period lag which is he sympom of rigidiy in financial secor of Bangladesh or i can be viewed as Friedman s argumen ha moneary policy ends o affec he real economy wih long and variable lags. Simulaion wih Taylor rule, however, shows an improvemen in macroeconomic performance in erms of aggregae variabiliy in oupu and inflaion. Opimal weighs for inflaion and oupu are 9.75 and.9 respecively ha reduce loss o he sociey from 1. o.79 wih an inflaion arge of 1 percen and zero real rae of ineres. This is noiceable ha he above parameer values produce some exreme values in he series of rule-based ineres rae. In such case, i can be recommended o use he relaive weighs insead. In special case of Bangladesh, i is a maer of fac ha counerfacual simulaion wih Taylor rule does no show a large change in overall performance alhough here is some. There may have several reasons behind, like small size of daa, leas-developed moneary sysem and oo resricive and prudenial 9 policy formulaion... Developed Counries Sweden Simulaion of he economy wih Rudebusch-Svensson model described by equaions () and (1) shows smaller variabiliy in boh 1 inflaion and oupu if Taylor rule Milon Friedman and Anna Schwarz, A Moneary Hisory of he Unied Saes, 17-19, Princeon, NJ, Prudenial in he sense ha cenral bank of Bangladesh seldom changes policy decision wih he fear of any unforeseen insabiliy. 1 Taylor suggesed equal weighs on boh inflaion and oupu sabilisaion. If his suggesion is followed boh oupu and inflaion variabiliy fall bu wih opimal weighs here is a rade-off.

29 is he ineres rae reacion funcion. Opimal weighs on inflaion and oupu are. and. respecively. y.5 y.( i π ) () = (.13) (.9) π.. (1) =.π 1 +.1y 1 (.) (.9) Social loss drops from o 1. if inflaion arge is se 3% and real equilibrium rae of ineres is.35%. Possibiliy of spurious regression is ruled ou because residuals are I() in each of he above hree equaions. Table 9 compares acual social loss wih rule-based loss. I appears ha here is a rade-off beween inflaion and oupu variabiliy had he opimal se of parameers been chosen. The same holds rue if moneary policy is framed only for inflaion sabilisaion. In boh cases variabiliy in oupu increases and of inflaion decreases under rule-based policy. Table 9 Simulaion of Swedish Economy wih Differen Ses of Parameer Values wih opimal weighs (b=., h=.) wih equal weighs (b=.5, h=.5) Full weigh on inflaion (b=, h=.) Full weigh on oupu (b=., h=) acual Ineres rae Average Sandard deviaion Ouupu gap Average 1.5E Sandard deviaion Inflaion Average Sandard deviaion Social Loss Simulaion resuls show ha Taylor rule wih opimal parameer values reduces social loss of Pakisan, India and Sweden by 19%, 1% and 1.5% respecively. Significan fall in loss o he sociey ulimaely indicaes he improvemen in macroeconomic performance under Taylor rule. 7. Modificaion of Taylor rule The findings so far observed reflec ha Taylor rule as moneary policy can improve macroeconomic performance by lowering loss o he sociey. Such improvemen can 9