THE ACADEMY OF ECONOMIC STUDIES BUCHAREST

Transcription

1 THE ACADEMY OF ECONOMIC STUDIES BUCHAREST DOCTORAL SCHOOL OF FINANCE AND BANKING DISSERTATION PAPER Money Demand in Romania Suden: DUMITRU IONUŢ Supervisor: PhD. Professor MOISĂ ALTĂR BUCHAREST, JUNE 00

2 ABSTRACT Finding a sable money demand relaionship is considered essenial for he formulaion and conducion of an efficien moneary policy. Consequenly, numerous heoreical and empirical sudies have been conduced in boh developed and developing counries o evaluae he deerminans and he sabiliy of he demand for money funcion for various moneary aggregaes. This paper briefly reviews he heoreical work, racing he conribuions of several researchers beginning from he classical economiss, and explains relevan empirical issues in modelling and esimaing money demand funcion for Romania. The paper models he empirical relaionship beween broader definiion of money, oupu, ineres raes, inflaion and exchange rae in Romania and examines he consancy of his relaionship, especially in he ligh of financial reform and deregulaion of financial markes. The demand for broad money in Romania has been sable beween 1996 and 00 despie of a pronounced financial liberalizaion. The analysis suggess ha, in Romania in he long run he inflaion is weakly exogenous for he money demand, which means ha inflaion is no a moneary phenomenon.

3 CONTENTS 1. INTRODUCTION THE ROLE OF THE DEMAND FOR MONEY IN THE TRANSMISSION MECHANISM OF MONETARY POLICY BASIC THEORETICAL APPROACHES TO EXAMINATION OF THE DEMAND FOR MONEY Quaniy heory Keynesian heory Neo-Keynesian heory of money demand Pos-Keynesian heories of money demand Modern monearis approach MONEY DEMAND ESTIMATES IN EASTERN EUROPE MONEY DEMAND FOR ROMANIA Background Money demand modeling in Romania Esimaion resuls Discussion on esimaed coefficiens Weak-exogeneiy ess The sabiliy of parameers in he long run equilibrium Shor run error correcion model (ECM) Forecasing abiliy A sae space model for he money demand in Romania CONCLUSIONS REFERENCES... 4 APPENDIX I APPENDIX II TABLES Table 1 Time series used... 1 Table Tes for uni roo... Table 3 Long run coinegraion relaionship (seasonally adjused daa)... 3 Table 4 Weak exogeneiy ess... 3 Table 5 Weak exogeneiy joinly ess... 6 Table 6 Long run coinegraing relaion (sasonally unadjused daa)... 7 Table 7 Parsimonious error correcion model - model I

4 FIGURES Figure 1 The evoluion of broad money and inflaion Figure Money velociy and inflaion Figure 3 The evoluion of real deposi ineres rae, T-bills ineres rae and inflaion Figure 4 Weigh of FCD in oal M Figure 5 Weigh of marke capializaion in M (%); Figure 6 Weigh of governmen credi in M Figure 7 Real exchange rae, depreciaion and ineres raes... 5 Figure 8 The unresriced coinegraing relaion... 8 Figure 9 Recursive diagnosic graphs of he shor run unresriced ECM - model I... 9 Figure 10 Recursive diagnosic graphs of he shor run unresriced ECM - model II... 9 Figure 11 Recursive diagnosic graphs of he shor run unresriced ECM - model III... 9 Figure 1 Graphs of he recursive coefficiens of he shor run unresriced ECM - model I. 30 Figure 13 Normaliy es of residual for parsimonious ECM model I... 3 Figure 14 Normaliy es of residual for unresriced ECM model I Figure 15 Recursive diagnosic graphs of he shor run parsimonious ECM - model I Figure 16 Graphs of he recursive coefficiens of he shor run parsimonious ECM - model I34 Figure 17 M acual vs. fied model I Figure 18 M acual vs. fied model II Figure 19 M acual vs. fied model III Figure 0 M acual vs. fied model ECM parsimonious Figure 1 Residual from Kalman filer Figure The deposi rae elasiciy of he money demand (Kalman)

5 1. INTRODUCTION The modelling of he demand for money has been a major focus of ineres in macro economerics since he early 1970s. This is no surprising considering is imporance for moneary policy and is role in modern economies. The demand for money is one of he mos imporan componens of he ransmission mechanism of moneary policy in a marke economy. A sable money demand funcion is a condiion in he conduc of moneary policy as i enables a policy-driven change in moneary aggregaes o have predicable influences on oupu, ineres rae, and ulimaely price. The analysis of he demand for money plays an imporan role in he decision-making process of cenral banks including he European Cenral Bank which has been working on a demand-formoney analysis inensively. The analysis of money demand is complicaed by he developmen of new financial producs like derivaives, changes in paymens agreemens, he developmen of non-banking financial insiuions, financial crises, as well as oher facors. The purpose of his paper is o poin ou he developmens in he demand for money in Romania beween 1996 and 00 1 and is deerminans. The paper is organized as follows. In he firs par, a brief heoreical background on he examinaion of he demand for money is oulined. The second par includes daa, a mehodological deerminaion of he analysis, and an economeric analysis of he problem. The examinaion is based on he Johansen procedure for coinegraion. Thereafer, a imevarying parameers model is esimaed using Kalman filer. Finally, he hird par conains some concluding remarks. 1 The sudy does no include an analysis before 1996 due o he irregular moneary and economic developmens in his period. 5

6 . THE ROLE OF THE DEMAND FOR MONEY IN THE TRANSMISSION MECHANISM OF MONETARY POLICY The demand for money reflecs he desirabiliy o hold money for firms, households, individuals and oher economic eniies. In nominal erms, i indicaes he araciveness of a cerain amoun of money; in real erms i shows how aracive is o hold money corresponding o he number of unis of asses and services ha may be acquired wih he money. Alhough abou he necessiy of dealing wih he demand for money is no fundamenal doubs, opinions concerning is specific impac on he economy differ depending on he heoreical bases aken ino accoun. Thus, abou long erm view here are few basic approaches: he Keynesian approach, emphasizing he imporance of he demand for money in he economy and moives for holding real money balances, and he monearis approach, sressing he effecs of he exchange area on demand for money developmens as represened by developmens in nominal GDP. A presen hese approaches represen alernaive heoreical conceps based on differen mehodological viewpoins. 3. BASIC THEORETICAL APPROACHES TO EXAMINATION OF THE DEMAND FOR MONEY 3.1 Quaniy heory According o classical economics, all markes are in equilibrium and is always a full employmen. The role of money is simple: i serves as he numeraire, ha is, a commodiy whose uni is used in order o express prices and values, bu whose own value remains unaffeced by his role (Sriram, 1999). I also faciliaes he exchange of goods. Money is neural wih no consequences for real economic aciviy. 6

7 The quaniy heory emphasized a direc and proporional relaionship beween money and price level. This relaionship was developed in classical equilibrium framework by wo alernaive bu equivalen expressions: 1. Equaion of exchange associaed wih Irving Fischer s equaion: MV = PT (1) where M is he quaniy of money in circulaion, V is ransacions velociy of circulaion of money, T is he volume of ransacions and P is he price level. Money is held only o faciliae ransacions and has no inrinsic uiliy.. Cambridge approach or cash balance approach - associaed wih he Cambridge Universiy economiss, especially A.C. Pigou. This alernaive paradigm relaes he quaniy of money o nominal income and sresses he role and imporance of money demand in deermining he effec of money supply on he price level. Money is held no only as a medium of exchange as in Fischer s case, bu also as a sore of value ha provides saisfacion o is holder by adding convenience and securiy. Cambridge economiss poined ou he role of wealh and he ineres rae in deermining he demand for money. 3. Keynesian heory In a Keynesian economy, he mos imporan relaionship is he relaionship beween economic growh and he level of invesmens. This relaionship is relaed o demand for money, where demand for money induces he money supply. In he long run, money demand and money supply are balanced. In comparison wih moneary approach, Keynesian heory assigns o he moneary policy a lower efficiency in he effecs on economic developmen. Keynes posulaed ha he individuals hold money wih hree moives: The ransacions-moive, i.e. he need of cash for he curren ransacion of personal and business exchanges. The ransacions demand for money arises because of he no synchronizaion of paymens and receips. An comprehensive explanaion of he principles of Keynesian, neo-keynesian, and pos-keynesian heory exceeds he scope of his sudy. 7

8 The precauionary-moive provides a coningency plan for unscheduled expendiures during unforeseen circumsances. The speculaive-moive i.e. he objec of securing profi from knowing beer han he marke wha he fuure will bring forh 3. The speculaive demand for money is wha Keynes called as liquidiy preference. The heory of liquidiy preference provides an answer o why economic eniies demand and hold money ha does no yield any ineres, insead of securiies or similar asses. Keynes adoped he ransacional moive from he monearis approach of he Cambridge school (A. Marshall, A. Pigou, e al) and considered he fac ha a par of he demand for money is associaed wih ransacions relaed o income developmens. The speculaive moive of money possession is inroduced by Keynes. Formal, Keynes s approach can be wrien as follows: M = L Y ) + L ( ) () 1( i where L 1 expresses he ransacional and precauionary moive, L expresses he speculaive moive of liquidiy preference, Y is nominal GDP and i is he ineres rae (Keynes, 1953). These moives exer influence simulaneously and are muually independen and consequenly M is a oal money demand. Keynes considered only nominal level of money demand. Afer Keynes, according o Dornbusch and Fischer 4 people possess money because of is purchasing power, i.e. he quaniy of goods and services ha hey can purchase wih money, wha means ha we mus consider he real level of money demand. 3. Neo-Keynesian heory of money demand The neo-keynesian inerpreaion of he money demand is based on Keynes s principles. The ransacional moive and precauionary moive are expressed as direcly 3 Keynes, J. M Keynes J. M.: The General Theory of Employmen, Ineres, and Money, Dornbusch R., Fischer S. (1994)- Macroeconomy, McGraw Hill Inc 8

9 proporional o income. The demand for money for he speculaive moive is dependen o ineres raes. Formally, such dependence can be formulaed as: M da = ky and α βi M ds = (3) where M is demand for acive balances, k is he share of acive balances in GDP, Y is da nominal GDP, M ds is speculaive demand for money, α and β are parameers and i is he ineres rae. The relaionship beween GDP and precauionary demand for money should be formulaed as ani-cyclical insead of pro-cyclical, similar o he ransacional moive. Thus, he demand for money can be expressed as follows: where M d = L( Y, i) (4) M d is demand for money, L is he liquidiy preference funcion, Y is nominal GDP, and i is an ineres rae. This approach was developed by Baumol (195) and Tobin (1956) o an approach based on he possession of money as invenory, where he ransacional moive of liquidiy preference is paricularly emphasized. Resuls of such consideraions lead o he well-known formula: where M d / P = cy / i M d is demand for real balances, c is ransacional coss, Y is real GDP and i is he ineres rae. The Bauman-Tobin model assumpion of cos sabiliy in a ransacion (c parameer) is no realisic in he long run. (5) 3.3 Pos-Keynesian heories of money demand Two characerisics of money demand provide he saring poin for many of hese heories. In ransacions models invenory models assume he level of ransacions o be known and cerain and in he precauionary demand models ne inflows are cerain. The special characerisics of money lead o formulaion of heories ha are based on explici moives o holding i. 9

10 Pos-Keynesian economics emphasizes he role of uncerainy associaed wih he hisorical developmens of he economy and pus he demand-for-money concep ino a broader conex. The volume of money in he economy is he resul of a demand and supply process ineracion. Through is insrumens, he cenral bank is able o influence he condiions for issuing loans due o he impac of such insrumens on ineres rae developmens. Addiionally, he behaviour of he banking secor owards economic eniies applying for loans is significanly influenced by insiuional characerisics of he banking secor. In his conex, an imporan role is mainained by banking regulaion and banking supervision funcions (see Dow, Rodríguez-Fuenes in Aresis, Sawyer, 1998). Pos-Keynesian economics differs from neo-keynesian especially in he inclusion of he financial moive in he demand for money. The financial moive reflecs he fac ha enrepreneurs mus mainain cerain money balances in he course of ime, so ha hey are able o mee heir liabiliies when enering fuure conracs associaed wih he purchase of inpus necessary for he producion. If he planned invesmens do no change, he money balances will remain permanen; if hey increase addiional financial demand for money is creaed. In his approach, he demand for money is usually expressed in nominal erms. For ransformaion o he real demand for money form, i is necessary o consider inflaion. Mos economiss, however, ignore he fourh moive of holding money balances (i.e. financial moive). Philip Aresis is one imporan pos-keynesian scholars working on he demand for money heory. In his aricle 5 discusses he demand for money in a small, open economy. His approach o he demand for money can be expressed using he following equaion: a e b c e d M = K( Y ) ( P ) ( CR) ( ER ) u (6) where d r M d are real money balances, K is he Cambridge coefficien, which is a funcion of GDP growh, prices and he volume of money in circulaion and is expressed by a reversed value of money velociy, Y r is real GDP, e P is he expeced rae of inflaion, CR is an 5 Aresis Ph.: The Demand for Money in Small Developing Economies: An Applicaion of he Error Correcion Mechanism (1988). 10

11 e esimaed variable for credi limiaions, ER is he expeced appreciaion or depreciaion rae of he currency, u is a non-sysemaic componen and a, b, c, and d are elasiciy values. Aresis s model in he previous expression, however, is no ideally suied for condiions prevailing in Romania. A presen, quanifiable credi limiaions do no exis in he Romanian economy. 3.4 Modern monearis approach The monearis approach analysis is based on he assumed direc influence of he volume of money in he economy and nominal income, usually expressed by nominal GDP. In moneary approach of he economy, money plays a primary role wih he money supply being a decisive facor. Modern moneariss wihdrew from he noion of an exclusive ie beween he demand for money and nominal income. They emphasize he influence of boh ineres raes and yields of oher angible and financial asses. Among of modern moneariss, Milon Friedman refreshed he radiional quaniaive money heory in he Cambridge version. According o Friedman, developmen of he demand for money depends on he overall wealh of sociey in various forms (money, bonds, securiies, maerial and human resources) as well as on he ase and preferences of holders of he wealh. Sabiliy of demand-for-money developmen is an imporan assumpion on which Friedman and oher moneariss base heir expansions of he heory. Formally, he demand for money in Friedman s concep may be expressed as follows: 1 dp M d = F( Y, W, rm, re rm, rb rm,, u) (7) P d where M d is demand for real money balances, Y is he overall wealh, W is a share of accumulaed human resources in he overall wealh, r m is he expeced money yield, r b is he expeced yield of bonds, r e is he expeced yield of securiies, in commodiy prices and u is he influence of oher facors. 1 dp P d is he expeced change 11

12 issues. The equaion (7) indicaes he wide range of Friedman s view of demand-for-money 4. MONEY DEMAND ESTIMATES IN EASTERN EUROPE In his secion we will poin ou several empirical works which deal wih he money demand in some of he ransiion couries from Easern Europe. Klacek and Smidkova (1995) esimaed he long-run demand for broad and narrow money in he Czech Republic since ransiion. The auhors iniially include GDP as a scale, bu he esimaed funcion did no characerize a money demand funcion due o incorrecly signed parameers. Privae consumpion was hen used, since i may give a beer approximae of he volume of ransacions. This esimaed model had he correc signs, wih privae consumpion having a significan effec. The inflaion erm was significan for narrow money, while he ineres rae on foreign (German) bonds was significan for broad money. Van Aarle and Budina (1996) esimaed money demand and specifically he effec of currency subsiuion using he porfolio balance approach for Poland, Hungary, Romania, and Bulgaria during ransiion. As a resul of he reform aking place in former cenrally planned economies, his has led o he liberalizaion of foreign exchange resricions and so legally allows he possibiliy of foreign currency o replace domesic as a means of paymen and a sore of value. The auhors in mos cases found a long-run relaionship beween money and income and ineres raes. An imporan conribuion o his paper is ha he auhors invesigae he impac of currency subsiuion on money demand. Arl, Guba, Radkovsky, Sojka and Siller (001) esimaed money demand for Czech Republic in period I is clear from hem resuls of he analysis ha, in is wide concep, he real demand for money in he Czech Republic had developed mosly under he influence of real GDP and nominal ineres rae developmen. The influence of an exernal economic environmen in he developmen of he demand for money has no been economerically proved. 1

13 Erwin Nijsse and Elmer Serken (1996) esimaed a household money demand funcion for Poland from 1969 o Conrary o heoreical belief and earlier empirical evidence porfolio argumens are found o be significan. Despie regime shifs during he 1980s and full liberalizaion of he Polish economy in he beginning of 1990, coinegraion relaionships are found beween broad real money holdings, real household income, ineres rae on an alernaive asse, he inflaion rae and shorage of goods. The auhor s ess confirm sabiliy of long-run income and ineres elasiciies. Also, Anoni Chawluk (000) analysed he same household money demand and consumpion for Poland. Variables ha measure shorage and expecaions abou is fuure course are inroduced o capure he effecs of he ransiion from cenrally planned o marke economy. The Johansen procedure issued o idenify a sysem of he wo coinegraing vecors. The repored resuls show ha disequilibrium in household secor money holdings has a srong influence on consumpion. 5. MONEY DEMAND IN ROMANIA 5.1 Background The demand for money in Romania beween 1996 and 00 has o be analyzed in he broader conex of he ransiion o he marke economy, which implies he ransformaion of he insiuional srucures and changes in moneary policy (box 1): in he beginning of 1997 exchange rae and prices were liberalized and NBR adoped a conrolled floaing regime for he exchange rae; in 1997, NBR simulaneously argeed he exchange rae and reserve money (M0); a he end of 1998 and in 1999 ried o resore exernal compeiiveness by using a real depreciaion of exchange rae; from 1999 o presen, he moneary policy framework seeks o srike a balance beween wo poenially conflicing objecives of: reducing inflaion hrough a degree of exchange rae sabiliy, and safeguarding he exernal posiion. 13

14 Box 1 Trends in he moneary and exchange rae policy in Romania during Gradual renouncing o direc moneary policy insrumens and orienaion owards indirec moneary insrumens. Afer 1997, he open marke operaions have become he main insrumen used by NBR, alhough i is more cosly. Improving insiuional framework of moneary policy. In 1998 crucial laws for he cenral and commercial bank aciviy were adoped: The Naional bank of Romania Ac, The Banking Ac and The Bank Insolvency Ac. These laws had an imporan impac on he conducing of moneary policy: hey esablished he cenral bank auonomy and independence vs. he oher sae insiuions, he price sabiliy became he moneary policy primary objecive and he ransparency of he moneary ransmission mechanism increased. The conducing of moneary policy had o overcome hree major consrains: he weak corporae managemen from he public secor, he fragiliy of he banking sysem and he exernal deb siuaion. Weak corporae governance of he large sae-owned enerprises leaded o inflaionary pressures caused by iner-enerprises arrears. Over he las years, he conducing of moneary policy has been complicaed by he fragile condiion of he banking sysem. I is he case of he wo former large insolven sae-owned banks Bancorex and Banca Agricola. The exernal deb crisis occurred in 1999 when Romania had larger repaymens on he previously conraced loans. The problem of insolvency was successfully surpassed wihou foreign suppor given he rising hosiliy of he inernaional financial markes. The moneary policy in 1999 was focused primordially on he repaymens of he exernal deb. As a main anchor for he moneary policy (inermediary arge), he NBR used he broad money M. The goal was an increase in he quaniy of money lower han he increase in naional income in order o miigae inflaionary pressures. Figure 1 shows he evoluion of broad money less foreign currency deposis in real erms. The broad money in real erms (in logarihm LMR), as well as is growh, are seasonally adjused by he Tramo-Seas procedure 6. The evoluion of he moneary aggregaes afer 1990 recorded numerous climbs and descends as he NBR was forced o accommodae large fiscal and quasi-fiscal deficis, or loss making sae-owned enerprises in general and of he agriculural secor in paricular. 6 In he esimaions we used he economeric program Eviews

15 Figure 1 suggess ha relaionship beween money and prices (monhly inflaion-pm) was raher loose beween 1996 and 00, hough a posiive correlaion has become somewha clear over he pas hree-four years. Figure 1 The evoluion of broad money and inflaion during LMR_SA MRgrowh_SA PM The higher inflaion rae in 1997 associaed wih price liberalizaion from January and wih exchange rae liberalizaion from March leaded o a sharp decline of he broad money in real erms. Using a igh moneary policy, he NBR achieved he miigaion of inflaionary pressure, bu broad money resumed is upward rend saring from lae 000. The use of moneary anchor M was hampered by he insable money velociy (figure ). Saring wih 1996, money velociy has had large flucuaions, recording is highes levels in 1997 and 001. Figure Money velociy and inflaion velociy_sa(index) p 15

16 These high levels of money velociy express a loss of confidence in domesic currency and a decrease in money demand. They were accompanied by high levels of inflaion in and a faser expand of oupu relaive o money growh in he conex of coninuously decrease in inflaion in Alhough he inflaion was brough under conrol afer he shocks from January-March 1997, an increase in he rae of inflaion in early 1999, bu no as higher as i was in 1997, produced an upward rend for money velociy. Financial markes in Romania remained relaively undeveloped, so he financial inermediaion has been low, he capial marke plays a less imporan role han i should, and for many periods he real ineres raes were negaive. Figure 3 The evoluion of real deposi ineres rae (DP), T-bills ineres rae (DTS) and inflaion (P) DP DTS P 5. Money demand modelling in Romania The empirical modelling of he money demand ypically 78 has as a saring poin a general specificaion for he long run 9 money demand as follows: M d = f ( y; r; x) d where, M is he money demand in real erms, y is a scale variable measuring he level of economic aciviy, r is a vecor of variables poining ou he opporuniy cos of holding 7 Sriram, S.S. (1999a) 8 Ericsson, N.R. ( In his paper he long run does no mean a very long period. We are ineresed in a period covering five years and hree monhs and we use monhly daa. 16

17 money, and x is a vecor of oher variables (including dummy variables) which will be included in he model. The relaion assumes an insan adjusmen of he acual money holdings o heir desired long erm level, which implies equilibrium beween he real demand and supply of money. This is no very plausible given he coss of he ransacions and he uncerainy. More, he desired level of money balances is unobservable. Due o he d s marke clearing mechanism, we can assume ha M = M = M (money supply is equal o money demand). So, we can use in money demand analysis series of daa for money supply. The money balances are measured as Lei M defined as lei currency ouside banks plus demand deposis plus household savings plus ime and resriced deposis. The foreign currency deposis (FCD) are excluded from he definiion of broad money in par due o he lack of daa on foreign currency cash holdings of populaion which are suspeced o be significan. Alhough he weigh of foreign currency deposis in M is significan, accouning for abou 30% in he las years, here is no empirical prove ha he foreign currency was used significanly as a mean of paymen or uni of accoun. FCD are asses used by he populaion in a high inflaion and volaile exchange rae environmen o subsiue he naional currency deposis. As a scale variable we use he real indusrial producion index (deflaed by he consumer price index CPI) as a proxy for GDP which is no calculaed monhly in Romania. The appropriae measure of he opporuniy cos of holding money in Romania is difficul o deermine a priori due o he limied and evolving availabiliy of alernaive domesic and foreign financial asses during he sample period: (box ). In our analysis we used he following coss of opporuniy: Deposi ineres rae for non-bank cliens measures he rae of reurn on lei ime deposis ( R ). own T-bills ineres rae measures he rae of reurn on asses ouside of M ( R ). Expeced inflaion rae proxied by curren inflaion rae p capure he reurn on real asses. The inclusion of he expeced rae of inflaion was necessary because in ou 17

18 developing economies wih weak financial sysems, he real asses offer proecion agains inflaion and hey are an alernaive asse for he nonbank agens 10. Box Alernaive financial asses in Romania The relaive imporance of he relaive asses o money has varied grealy over he las years. Foreign currency denominaed deposis has consiued an imporan alernaive for he domesic denominaed money balances (figure 4) noably afer he liberalizaion of he foreign currency marke in March Figure 4 Weigh of FCD in oal M Jan-96 Jul-96 Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Jul-99 Jan-00 Jul-00 Jan-01 Jul-01 Jan-0 The developmen of he capial marke in Romania provided alernaives o bank deposis: securiies, invesmen funds and T-bills. However, he marke capializaion for hese asses coninues o be very low (figure 5). The marke capializaion of he equiies accouned for only % of GDP in he las hree years. The invesmen funds asses for less han 1% of GDP. The governmen securiies had an ineresing evoluion (figure 6). For many periods, he governmen securiies had beer yields han bank deposis. Over he las years a secondary marke for his ype of securiies has developed in Romania. A sharp decline in he yield of he T-bills since 000 was accompanied by a sharp decline in he invesmen in hese asses. Figure 5 Weigh of marke capializaion in M (%); Jan-96 Jul-96 Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Jul-99 Jan-00 Jul-00 Jan-01 Jul-01 Jan-0 10 The basic idea is ha in developing economies wih limied alernaives of invesmens on he capial marke, asses subsiuion usually means replacemen of he money balances wih real asses raher han financial asses. This idea is no very plausible for he analyzed period in Romania, where he governmen securiies play an imporan role in deermining he demand for money in he long run, while inflaion has a greaer influence in he shor run. 18

19 Figure 6 Weigh of governmen credi in M Jan-96 Jul-96 Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Jul-99 Jan-00 Jul-00 Jan-01 Jul-01 Jan-0 Expeced depreciaion of he lei-dollar exchange rae. Capures he reurn on holding US dollars, imporan asses ouside of M. The acual depreciaion was used as proxy for he expeced rae. We will analyze hree specificaions (models): 1. The firs model essenially a closed economy model in which he opporuniy cos variables are limied o hose on lei asses. In our esimaions we will use a (semi) log linear form: d own ou m γ 0 + γ 1y + γ R + γ 3R + γ 4 = p (8) where variables wih small caps are in logarihm, and d m is real money demand, own R and ou R are nominal rae of reurn on asses included, respecively excluded from he moneary aggregae, and p is he annualized rae of inflaion. Relaion (8) assumes price homogeneiy of money demand in he long run. In equaion (8), γ 1 measures he log run income elasiciy of he money demand, whileγ, γ 3 and γ 4 are semi-elasiciies. We expec, according o economic heory ha γ 1 > 0, γ > 0, γ 3 < 0, γ 4 < 0 and possibly, γ γ 3 =. In he las case, he long run money ou own R R, which can be inerpreed as demand can be expressed as funcion of he spread an opporuniy cos of money holdings. In general he coefficien of inflaion has o be negaive. However, i is possible ha inflaion has a posiive coefficien in he long run relaion of money demand because when agens expec a rise in inflaion, hen increase heir money balances expecing a rise in planed expendiures (Jusoh (1987)). 19

20 Values for γ 1 greaer han one can be found in many empirical sudies of he demand for broad money (M) because of he presence of wealh effecs.. The second model a model for an open economy in which opporuniy coss variables include he rae of reurn on foreign asses measured by he exchange rae depreciaion. In our esimaions we will also use a (semi) log linear form: d own ou m = γ + γ y + γ R + γ R + γ p + ED (9) γ 5 where ED is he exchange rae depreciaion calculaed as E E E 1 1, E being he exchange rae a he momen expressed in lei per US dollar. According o he economic heory we can expec haγ 5 < 0, which means ha a rise in he expeced depreciaion of he exchange rae will lead o rise in he rae of reurn on foreign asses and consequenly he agens will subsiue asses in domesic currency wih foreign asses (Simmons 11 (199)). 3. The hird model includes he level of exchange rae as a proxy for converibiliy risk. The form used in our esimaes will be: d own ou m = γ + γ y + γ R + γ R + γ p + E (10) γ 6 The variables used are presened in able 1. As many of he series exhibi regular seasonal paerns, i is necessary o ake accoun of he seasonal facors in he esimaion. This is done in wo ways: firs we will seasonal adjus he daa using Tramo-Seas procedure; and second, we will use he raw series wih monhly seasonal dummies 1. I is noiceable ha if sandard 0-1 dummy variables are included, hey will affec boh he mean and he rend of he series. To handle his we used cenred seasonal dummy variables as Johansen suggesed. These variables affec he mean bu have no influence on rend. 11 Simmons emphasizes he possibiliy of obaining boh a posiive and negaive relaion beween he exchange rae depreciaion and he domesic money balances. The impac can be negaive if he depreciaion of he domesic currency will lead o anicipaions for fuure depreciaion. On he oher hand, he impac can be posiive if he depreciaion leads o expecaions for a fuure appreciaion. 1 A priori, i is difficul o choose beween hese wo echniques. The use of seasonally adjused daa may impac dynamic specificaion (Ericsson, Hendry and Tran (1994)). The alernaive approach of including seasonal dummy in he esimaions is no wihou coss as i imposes consan seasonal facors (unlike Tramo-Seas which permis he facors o evolve hrough ime) and uses up degrees of freedom, hereby reducing he power of ess saisics. Tramo-Seas has he advanage, unlike oher mehods o give beer resuls in presence of ouliers. 0

21 Table 1 Time series used VARIABLE DESCRIPTION LMR Real broad money (logarihm) LMR_SA Real broad money (logarihm) seasonally adjused LYRIBF Indusrial producion index (volume index, logarihm, December 1995=1) LYRIBF_SA p Indusrial producion index seasonally adjused Monhly inflaion (annualized) p_sa LE ED DP DTS Monhly inflaion seasonally adjused Nominal exchange rae USD/ROL (logarihm) Exchange rae depreciaion (annualized) Deposi ineres rae for non-bank cusomers (percen per annum) Average reurn for T-bills (ineres-bearing T-bills and discoun T-bills) 5.3 Esimaion resuls The esimaions are conduced in a number of seps. Firs, uni roo ess are applied o all variables of ineres o deermine he saionariy of he individual variables. As in mos oher sudies of money demand, real money is found o have a single uni roo, implying ha i is saionary in firs differences. The esimaions were performing using monhly daa for January 1996 unil March 00. Daa hrough end-sepember 001 were used for esimaions, wih he remaining observaions reserved for ou-of-sample forecasing. The saionariy properies of he series were examined using boh Augmened Dickey Fuller and Philips Perron ess (appendix I). The resuls are repored in able 13. The number of lags used in hese ess was chosen using Akaike informaion crierion and Schwarz crierion. 13 The resuls of saionariy ess mus be regarded wih circumspecion, given he low power of such ess in presence of srucural breaks. 1

22 Wih he excepion of he exchange rae depreciaion and inflaion, he variables were found o be inegraed of order one in levels, which is consisen wih a saionary represenaion in firs differences. Table Tes for uni roo Variable ADF es PP es Real broad money* I(1) C I(1) C Real oupu* I(1) C I(1) C Exchange rae* I(1) C T I(1) C T Exchange rae depreciaion I(1) C or I(0) C I(0) C Inflaion I(1) C or I(0) C I(1) C or I(0) C Deposi ineres rae I(1) C T I(1) C T T-bills ineres rae I(1) C T I(1) C T Seasonally adjused variables Real broad money* I(1) C I(1) C Inflaion I(1) C or I(0) C I(1) C or I(0) C Real oupu* I(1) C I(1) C *variables are in logarihm As uni roo ess shows ha he variables are I(1), he coinegraion echnique is appropriae o esimae he long run money demand. Table sugges ha none of he variables is inegraed of order (I()) or more. Inflaion and depreciaion of exchange rae is probably I(0). Even in his case i does no mean ha hey mus be excluded from coinegraing vecor (Dickey and Rossana, 1994). We chosen he number of lags included in coinegraion ess (appendix II) esimaing a VAR wih variables of ineres and using crieria like LR, FPE, AIC, SC and HQ. If he opimum lag in VAR is p, hen we esimae he VEC wih p-1 lags. Firs, he ess were conduced wih seasonally adjused daa and including dummy variables for he 1997 shocks (dummy9701-ake value 1 in January 1997 and 0 in res and dummy9703-ake value 1 in March 1997 and zero in res 14 ). The resuls including dummy variables were unsaisfacory, he coefficiens of dummy variables being saisically insignifican, and consequenly we reesimaed he relaions wihou dummy (able 3). 14 The resuls of a VEC wih sandard 0-1 dummy mus be inerpreed wih cauion.

23 Table 3 Long run coinegraion relaionship 1/ Oupu Deposi ineres rae T-bills ineres rae Inflaion Coef. SE / 3/ Coef. SE Coef. SE Coef. SE I 6/ 1.39* * * * II 7/ 1.33* * * * III 8/ 1.46* * * * Depreciaion Exchange rae Speed of adjusmen RMSE 4/ Coef. SE Coef. SE Coef. SE saic dynamic I 6/ -0.04* II 7/ -0.46* * III 8/ -0.34* * * significan a 5%; ** significan a 1% Table 4 Weak exogeneiy ess 1/ ( α i = 0, α i represen speed of adjusmen) LMR_SA LYRIBF_SA DP DTS P_SA ED LE I χ (1) = 5.5 [0.018] 5/ * χ (1) = 1.99 [0.16] χ (1) = 0. [0.64] χ (1) = 8.9 [0.00]** χ (1) = 0.01 [0.9] II χ (1) = 1.4 [0.00]** χ (1) = 0.77 [0.38] χ (1) = 1.80[0.18] χ (1) = 4.4[0.04]* χ (1) = 0.13 [0.7] χ (1) = 1.8 [0.5] III χ (1) = 0.03 [0.00]** χ (1) = 0.60 [0.43] χ (1) = 0.45 [0.50] χ (1) = [0.00]** χ (1) = 0.00 [0.98] χ (1) = 0.00 [0.98] 1/ Seasonally adjused daa; / Sandard errors; 3/ T-saisic; 4/ Roo mean square error for forecas; 5/ null hypohesis is ha here is weak exogeneiy (in squared brackes - probabiliy); 6/ VEC has be esimaed wih 4 lags; 7/ VEC has be esimaed wih 3 lags; 8/ VEC has be esimaed wih 4 lags; ** and * indicaes rejecion of null hypohesis a 1% respecively 5%. 3

24 5.3.1 Discussion on esimaed coefficiens Esimaes for each of he models on he basis of seasonally adjused daa have he anicipaed signs (according wih economic heory) and here is relaively lile variaion in he esimaes of key parameers 15. The oupu elasiciy in he long run relaionship is greaer han one like we expeced for broad money aggregae. Oupu decrease in Romania, especially in he firs half of he las decade was accompanied by a demoneizaion. A he same ime when he oupu decrease sopped a remoneizaion ook place, bu he iniial level of moneizaion was no reached. Elasiciy for oupu greaer han one can be he resul of he omission of some imporan facors for he money demand. Two imporan facors for money demand dynamics in Romania migh be he financial arrears or barer, bu due o he lack of daa we couldn include hem in our analysis. Sof budgeary consrains on he enerprises in ransiion counries lead o a proliferaion of arrears, which were used as a subsiue for money. The level of iner-enerprise arrears in Romania, according o an IMF repor (January 001) were equivalen o 4 percen of GDP, and apar from a fall in 1997, has risen seadily each year from around 0 percen of GDP as a end Like magniude, income elasiciy does no differ significanly from 1 (for model I, if we assume B(1,1)=1 and B(1,)=1 we obain χ (1) =.8[0.13]), which is consisen wih he quaniy heory of money. The fac ha we canno rejec uni income elasiciy of money demand suggess ha in he analyzed period, real oupu change produced a proporional change in he real money demand. This does no necessary imply ha money velociy is saionary. As a maer of fac, as figure shows, money velociy was no consan due o shocks during he analyzed period. The differences in size beween ineres raes coefficiens in models I and II may due o he fac ha muli-collineariy problem become more acue in he case of model II due o he close correlaion (appendix II) beween exchange rae depreciaion and inflaion. 15 Like magniude, he esimaed coefficiens are similar wih hose obained in oher empirical works performed in oher counries from Cenral and Easern Europe - Ane Babic for Croaia, Guba, Sojka, Siller and Don Bredin for Cehia, or Ericsson and Sharma (1998) for Grecia and Nachega (001) for Camerun. 4

25 Deposi ineres rae semi-elasiciy of money demand is posiive and has an opposie sign o he T-bills ineres rae semi-elasiciy. I is also greaer han he T-bills ineres rae semi-elasiciy, which suggess a smaller elasiciy of money demand o a change in T-bills ineres rae. A es imposing equals semi-elasiciies (in absolue value) is saisically rejeced for models II ( (1) = χ 11.01[0.00] and III ( χ (1) =.8[0.00], while for model I i canno be rejeced ( χ (1) = 1.05[0.30]). The exchange rae depreciaion coefficien has a negaive sign (consisen wih economic heory) and is saisically significan, which indicaes he exisence of a currency subsiuion in Romania. This is oulined also by he increase in weigh of foreign currency deposis in oal broad money (figure 4). The coefficien of depreciaion is relaively small (he exchange rae depreciaion elasiciy of money demand being 0.46*0.43=0.19, where 0.43 is an annualized average depreciaion of he exchange rae). This is explained by he fac ha on average, he naional currency appreciaed in real erms (figure 7) exceping he 1997 shocks (price and exchange rae liberalizaion) and 1999 (he pressure induced on he real exchange rae depreciaion by a peak in he paymen of exernal deb), and he deposi ineres rae was on average greaer han he yield for foreign exchange deposis, which was very obvious since 001. The small coefficien of exchange rae migh be he cause of dollarizaion hyseresis and/or a high risk premium in Romania. Figure 7 Real exchange rae, depreciaion and ineres raes real exchange rae DP DTS ED The inflaion semi-elasiciy of money demand is 0.47, which involves a long erm elasiciy of 0.33 (figure obained by muliplying he annualized monhly average inflaion by 0.47), considered relaively high. This is no surprising for a counry like Romania in which 5

26 financial asses ouside broad money are limied and he agens hold significan quaniies of real asses Weak-exogeneiy ess Weak exogeneiy hypohesis is acceped boh separaely (able 4 - for oupu, deposi ineres rae, inflaion, depreciaion and exchange rae) and joinly (able 5). I is rejeced for broad money and T-bills. The deposi ineres rae weak exogeneiy suggess ha i is deermined ouside he sysem (i is no caused by money demand, bu i deermines money demand). The T-bills ineres rae is no weak exogenous and i adjus o he money demand disequilibria from he long erm level. The relaionship beween inflaion and broad money is from inflaion o broad money and no vice versa, he inflaion being weak exogenous for he money demand. I is no he rise in broad money which generaes inflaion, bu i is he broad money which is accommodaed o inflaion as his moneary aggregae rises o reach he equilibrium. Consequenly, we can sae ha inflaion is no a moneary phenomenon. Table 5 Weak exogeneiy joinly ess ~ 1/ (1,,3,4=0) (1,,4=0) (1,,3,4,5=0) (1,,4,5=0) (1,,3,4,5,6=0) I χ (4) = 9.6 [0.054] χ (3) =.49 [0.48] II χ (4) = 8.30 [0.08] χ (3) = 3.4 [0.35] χ (5) = [0.057] χ (4) = 3.93 [0.41] III χ (4) = 3.1 χ (3) = =1.01 χ (6) = 3.33 [0.00]* [0.79] [0.00]* ~Seasonally adjused daa; 1/ 1,,3,4,5,6 represen speed of adjusmen (α ) for oupu, deposi ineres rae, T- bills ineres rae, inflaion, depreciaion and exchange rae; ** and * indicae rejecion of null hypohesis (here is weak exogeneiy, in square brackes p-value) a 1% respecively 5%. The coefficiens represening speed of adjusmen (able 3) indicae relaively rapid adjusmen of real money demand o disequilibria. The negaive sign of speed of adjusmen suggess ha if in he previous monh he money demand exceeded he long erm level in he curren monh money demand would decrease. We can say ha he cenral bank accommodaes relaively quickly hese disequilibria (for model II and III 10% from he previous monh disequilibrium is adjused in he curren monh, leading o an accommodaion 6

27 of his disequilibria in abou 10 monhs; for model I, 4% from he previous monh disequilibrium is adjused in he curren monh). As an alernaive for he above esimaes, we esimaed he models using unadjused daa and inroducing seasonally cenred (orhogonalised) dummy variables in order o poin ou seasonaliy. The resuls were no consisen in economical and economerical erms unless in he case of closed economy model (able 6), he oher models leading o saisically insignifican coefficiens. Table 6 Long run coinegraing relaion 1/ Oupu Deposi T-bills ineres Inflaion Dummy S1 17 ineres rae rae Coef. 1.07* 7.84* -4.8* -0.8* * SE / / S3 S10 S11 S1 Speed of adjusmen Coef * * SE / / RMSE 4/ saic dynamic * Significan a 5%; 1/ seasonally unadjused daa; / Sandard error; 3/ T-saisic; 4/ Roo mean square error for forecas; The key parameers esimaed in able 6 are consisen wih oher research sudies relaed o money demand and hey are similar as size wih hose esimaed for oher ransiion economies. The dummy9701 variable has a negaive coefficien implying ha money demand was lower in January 1997, and S1 has a posiive coefficien showing he rise in money demand in December each year. Figure 8 shows he coinegraing vecor obained following Johansen procedure. The coinegraing vecors show he deviaion of he money demand from is long erm level. In he conex of money demand models, he posiive (negaive) difference beween acual holdings of money and he long erm level can be inerpreed as a measure of moneary overhang (shorfall). 16 Dummy for price liberalizaion from January 1997 is 1 for January 1997 and 0 in res. 17 S-seasonally cenered dummy for January (S1 - he increase in money supply following he paymen of he 13- h wage), March (S3 - he increase in prices in March), Ocober and November (S10 and S11-he increase in oupu), and December (S1-he increase in money demand). 7

28 Figure 8 The unresriced coinegraing relaion (used as error correcion erm in dynamic model) ECM1 ECM ECM3 ECM1dummy Figure 8 shows ha during he deviaion of money demand from is long erm level for he hree models and model I wih unadjused daa is saionary, so we can use i in an error correcion mechanism. More, his deviaion is relaively small, exceping 1997 when in he firs hree monhs here was an obvious moneary overhang, and in he nex wo monhs a shorfall occurred due o pressures made by NBR o rise he ineres raes, afer which he siuaion came back o normal The sabiliy of parameers in he long run equilibrium The sabiliy of he parameers is essenial for a good specificaion of money demand. Insabiliy can occur during and immediaely afer a financial crisis, and he main facors of money demands can change. In order o asses he sabiliy of he parameers we will recursively re-esimae he parameers from he unresriced ECM. In figures 9-1 are presened he CUSUM ess, recursive residuals, N-sep forecas es and he recursive coefficiens. 8

29 Figure 9 Recursive diagnosic graphs of he shor run unresriced ECM - model I Recursive Residuals ± S.E. CUSUM 5% Significance N-Sep Probabiliy Recursive Residuals Figure 10 Recursive diagnosic graphs of he shor run unresriced ECM - model II Recursive Residuals ± S.E. CUSUM 5% Significance N-Sep Probabiliy Recursive Residuals Figure 11 Recursive diagnosic graphs of he shor run unresriced ECM - model III :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01:07 Recursive Residuals ± S.E. CUSUM 5% Significance N-Sep Probabiliy Recursive Residuals According o he ess performed, he esimaed coefficiens are consan over ime, alhough in1997 here are some signs of insabiliy. The sabiliy of parameers suggess ha he deerminans of he money demand were consan over he analyzed period. This sabiliy is quie remarkable if we ake accoun of he period of ransiion and he changes in he economic policy which ook place in Romania. 9

30 Figure 1 Graphs of he recursive coefficiens of he shor run unresriced ECM - model I :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01:07 ECM1_1 ± S.E. DLMR_SA_1 ± S.E. DLMR_SA_ ± S.E. DLMR_SA_3 ± S.E. DLMR_SA_4 ± S.E :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01:07 LYRIBF_SA_1 ± S.E. LYRIBF_SA_ ± S.E. LYRIBF_SA_3 ± S.E. LYRIBF_SA_4 ± S.E. DDP_1 ± S.E :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01:07 DDP_3 ± S.E. DDP_4 ± S.E. DDTS_1 ± S.E. DDTS_ ± S.E. DDTS_3 ± S.E :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01: :01 99:07 00:01 00:07 01:01 01:07 DDTS_4 ± S.E. DP_SA_1 ± S.E. DP_SA_ ± S.E. DP_SA_3 ± S.E. DP_SA_4 ± S.E. 30

31 An explanaion for he sabiliy of he parameers in he case of seasonally adjused daa may be he use of he Tramo-Seas procedure which gives good resuls in he presence of ouliers and incorporaes very well srucural breaks. In spie of his, he sabiliy of parameers could be explained. A series of facors conribued o he sabilizaion of he economic environmen including a resricive moneary policy conduced by he NBR. For example, when in 1997 a sharp decline in money demand occurred, he NBR simulaed is recovery hrough a rise in he level of ineres raes. Because of he insufficien number of observaions unil 1997:01 (1 observaions), he models canno be esimaed separaely for he wo sub-periods; his is he reason why a Chow es for a srucural break in 1997 canno be performed. In order o es if he 1997 shock produced only a one ime jump in he deerminans of money demand and lef hem unmodified, we re-esimaed he coinegraing relaion for he 1997:07-001:09 period. For he log run we obained parameers similar o hose obained for he enire sample. For example, in model II, if we resric all he long run coefficiens esimaed for he 1997:06-001:09 sample o be he same wih hose esimaed for he enire sample we canno rejec he null hypohesis: χ (5) = 5.73[0.34] Shor run error correcion model (ECM) The esimaed coinegraing equaions include he facors affecing he real money demand in he long run. In he shor run deviaions from hese relaions can occur reflecing shocks in he relevan variables. More, he shor run elasiciies differ from he long run elasiciies. Engle and Granger (1987) proved ha if here is a coinegraing relaion beween non-saionary variables, hen here is a correcion represenaion owards equilibrium. In his secion we will esimae a shor run parsimonious error correcion model as follows: 5 6 ji V j, i + γ 1ECM 1 j= 1 i= 1 LM R = C + γ + ε (11) where V is he vecor of variables (broad money, oupu, deposi ineres rae, T-bills ineres rae and inflaion). Using a general-o-specific mehodology (D. Hendry) by eliminaing insignifican lags we obain he parsimonious model from able 7. 31