Demand for Money in Dollarized, Transitional Economy: The Case of Vietnam

Transcription

1 Draf for Commens Demand for Money in Dollarized, Transiional Economy: The Case of Vienam Waanabe Shinichi and Pham Thai Binh 1 Absrac The negaive correlaion beween broad money and inflaion emerged afer he ouburs of Asian financial crisis. I poses a puzzle if i is inerpreed wihin a widely acceped macroeconomic framework in which he growh of money supply and inflaion posiively correlae. This paper solves he puzzle by decomposing he demand for money ino he demand for domesic currency and he demand for foreign currency in view of he fac ha Vienam is highly dollarized. The empirical analysis employs coinegraion, error correcion model, impulse response and variance decomposiion and uses quarerly daa over he period The resuls show ha he long run demand for real broad money of domesic currency is deermined by real income, domesic ineres rae, inflaion rae and rae of reurn of USD deposis, which saisfies he sandard properies of he demand for money. The long run demand for real foreign currency deposis is deermined by real income and he difference beween raes of reurns of foreign and domesic currency deposis, ha represens asse subsiuion. Demand for real foreign currency deposis is found o be very sensiive o he difference beween raes of reurns of foreign and domesic currency deposis, especially exchange rae depreciaion. A posiive difference beween raes of reurns of foreign and domesic currency deposis riggers a shif from domesic financial asses and real asses o foreign financial asses, resuling in a lower inflaion rae and a srong increase in demand for foreign money. When his effec is very srong, hen even an increase in broad money, including foreign currency deposis, may negaively correlae wih inflaion. 1 Waanabe Shinichi is a Professor of Economics and he Vice Presiden of Inernaional Universiy of Japan. Pham Thai Binh is an MA graduae from Inernaional Universiy of Japan. address: pbinh99@yahoo.com. i

2 I. INTRODUCTION Table 1 shows ha in Vienam, during he period of , growh rae of broad money (M2) posiively correlaed wih inflaion (INF_A). However, afer he ouburs of Asian financial crises, he correlaion was negaive, conradicing a widely acceped macroeconomic framework in which he growh rae of money supply should posiively correlae wih inflaion. These facs are also illusraed in Figure 1 and 2. Why he growh rae of broad money negaively correlaed wih inflaion? Table 1: Simple Correlaion beween Annual Changes in Broad Money (M2), Broad Money of VND (M2VND), Foreign Currency Deposi (FCD), Oupu (GDP) wih Domesic Ineres rae (DEPO), Inflaion (INF): Quarerly daa M2 M2VND FCD INF_A GDP During 1993:1-2004:4 INF_A GDP DEPO During 1997:3-2004:4 INF_A GDP DEPO During 2000:1-2004:4 INF_A GDP DEPO % Figure 1: Annual Changes in Price, Broad Money, Broad Money of Vienam Dong and Foreign Currency Deposis (1993:1-2004:4) :1 1994:1 1995:1 1996:1 1997:1 1998:1 1999:1 2000:1 2001:1 2002:1 2003:1 2004:1 INF M2 M2VND FCD 1

3 % 70 Figure 2: Annual Changes in Price, Broad Money, Broad Money of Vienam Dong and Foreign Currency Deposis (1997:3-2004:4) :3 1998:3 1999:3 2000:3 2001:3 2002:3 2003:3 2004:3 INF M2 M2VND FCD Table 1 and Figure 1 & 2 also indicae ha he puzzle of negaive correlaion beween growh rae of broad money and inflaion was caused by he negaive correlaion beween growh rae of foreign currency deposi (FCD) and inflaion since growh rae of broad money of domesic currency (M2VND) has posiive correlaion wih inflaion. In view of he fac ha Vienam is highly dollarized, 2 he above observaions indicae ha, in order o solve he puzzle, we should look a broad money componens: broad money of domesic currency and foreign currency deposis, insead of looking a broad money as a whole. Vienam is a ransiional economy. Since 1986, he economy sared shifing from a cenrally-planned owards a marke one wih high growh rae. However, he financial sysem is no liberaed quickly enough o conform o he economic growh. The banking sysem has been playing a key role in funding he economy. Ineres rae was fixed by cenral bank and has jus liberaed since June However, he moneary policy sill lack money demand model, hus consraining moneary policy s effeciveness in conrolling inflaion. 3 2 The foreign currency deposi (FCD)/M2 raio in Vienam is as follows (%): A general undersanding is ha undersanding abou money demand is a key ool so ha cenral bank can use money supply o conrol inflaion. 2

4 Only a few empirical researches abou money demand in dollarized economies have been conduced. Very few works has ried o find money demand model in Vienam and none has generaed he money demand model by aking fully ino accoun he dollarizaion siuaion in financial sysem in Vienam. This paper aims o esimae he money demand funcion of Vienam and sudies he characerisics of he economy. In order o solve he puzzle of negaive correlaion beween growh rae of broad money and inflaion, i develops and esimaes he money demand model in Vienam by formulaing demand for real broad money of domesic currency and demand for real foreign currency deposi and conducs some analyses o research empirically he economy. Paricularly, i answers he following quesions: i) Wha are he characerisics of broad money of domesic currency and foreign currency deposis? Since his is an open developing counry having high degree of dollarizaion, asse subsiuions is considered in he models. ii) Wha are he characerisics of he economy when here is a shock in rae of reurn of foreign currency deposi, he characerisics of real broad money of domesic currency when here are shocks of ineres rae, inflaion, oupu and foreign rae of reurn, and he characerisics of real foreign currency deposi when here are shocks of oupu and he difference beween raes of reurns of foreign and domesic deposis? iii) Wha are he characerisics of domesic currency, oupu, inflaion, domesic ineres rae and foreign rae of reurn? This paper invesigaes he demand for broad money of using quarerly daa from 1993:1-2004:4. In accomplishing is ask, he paper employs coinegraion, error correcion model, impulse response and variance decomposiion. The paper is organized as follows: Secion 2 describes he heoreical and empirical researches abou demand for money; Secion 3 presens he demand for money model and he analysis of empirical resuls. The las is a shor conclusion. 3

5 II. THEORETICAL AND EMPIRICAL FRAMEWORK Many economiss have ried o heorize he money demand. The noables are Fisher (1911), Pigou (1917), Keynes (1930 and 1936), Baumol (1952), Samuelson (1958), Tobin (1958), Friedmand (1956 and 1970) and McCallum (1989). Keynes (1930 and 1936) developed money demand heory based on moives leading people o hold money: ransacion, precauion and speculaion moives. Tobin (1958) and Friedmand (1970) made a significan advancemen in money demand heory by considering money as a par among a porfolio of asses including money wih oher financial and real asses. Thus, he sandard money demand equaion has he formulaion as M d /P = f (Y/P, OC), where M d /P is real money demand, Y/P is he real ransacions and OC is a vecor of opporuniy coss. Regarding he money demand model in dollarized economies, since foreign money is used popularly as an asse besides domesic money in agens porfolio, Branson and Henderson (1985) shows ha i is necessary o separae broad money ino domesic and foreign money. Each of hem is deermined by income and heir relaive reurns: m + + = f ( y, i, i *, Δ e ) m * + = f ( y, i, i *, Δ e) + + where m and m* are he real domesic and foreign broad money, i and i* are domesic and foreign ineres rae, respecively; y is real income; and e is depreciaion rae. 2.1 General Empirical Research abou Money Demand There are a large sream of empirical researches have been dedicaed o demand for money, especially in developed counries (Hafer and Dennis 1991, Kole and Ellen 1995, Laurence 1998). The aenion in developing counry s cases has been rising in recen years driven by increasing role of hese counries in he world economy and financial marke and effors o improve economic managemen in hese counries (Bossogo 2000, Dekle and Pradhan 1997, Ibrahim 1999, Rober and Mahmood 1997). 4

6 One imporan facor fuelling he empirical researches on money demand is big advancemens in ime series analysis in he pas decade. Sriram (1999a) has shown ha, in order o ge meaningful resuls, he wo mos imporan poins in modeling demand for money are: firs, selecion and represenaion of variables and second, framework of model. Regarding he variable selecion, he moneary aggregaes can be denoed by M0, M1, M2, M3 or broader moneary aggregaes or componens of heirs; while economic ransacions can be represened by Gross Domesic Produc (GDP), Gross Naional Produc (GNP) or Indusrial Index (Laidler, 1993; Laurence, 1998; and McCallum, 1989). Ericsson (1998) and Nachega (2001) have proved ha he choice of opporuniy cos variables urns ou o be he mos imporan facor in geing meaningful resuls. The opporuniy cos of holding money comprises own rae of money and rae of reurn on subsiued asses o money, including domesic and foreign financial asses and real asses. Regarding he own rae of money, when M0, M1 is dependen variable, he own rae of money can be considered as zero and when broad money (M2, M3, ec.) is dependen variable, he own rae should be shor erm deposi ineres rae. The reurn on domesic financial asses is usually represened by yields on governmen bonds, bills, commercial papers or saving deposis. The reurn on real asses may have a proxy like inflaion or, in narrower erm, changes in prices of some popular asses in he economy. The reurn on foreign financial asses is usually represened by ineres raes of foreign bill, foreign currency deposi, or expeced rae of depreciaion of domesic currency. For he framework of he model, coinegraion is employed popularly o capure he long run equilibrium and error-correcion model is used o reflec he shor run dynamic of money demand; boh coinegraion and error correcion model proved o be useful ools in money demand research (Bossogo, 2000; Bank of England, 1999). 2.2 Empirical Research abou Money Demand in Dollarized Economy Conceps of Dollarizaion, Currency Subsiuion and Asse Subsiuion Dollarizaion is denoed as a siuaion where one or some foreign currencies serves one or more funcions of domesic currency- ha includes medium of exchange, accouning and sore of value (Waanabe, 2002). Currency subsiuion is he siuaion 5

7 where foreign currency is used only as means of paymen (Sa and Zamaroczy, 2003). When foreign currencies served as sores of value, i is regarded as asse subsiuion (Berg and Borenszein, 2000). While currency subsiuion usually occurs from high inflaion simulaing people o shif o foreign currency o safeguard value, asse subsiuion is arisen from risks and consideraions beween raes of reurns of domesic and foreign asses (Oomes and Ohnsorge, 2004 and Pozo e all, 2000). Empirical Research abou Money Demand in Dollarized Economies Previous empirical research shows ha in general here are wo ways in modeling he money demand in dollarized economies. The firs is applied in low dollarizaion economies like China and Czech. In hese counries, Huang (1994), Hafer and Kuan (1994), Tseng e all (1994) and Ko (2004) have formulaed he money demand model wih money balance as a funcion of income, domesic ineres rae and inflaion. There, he money balance can be he broad money as a oal of boh domesic and foreign currency. The second formulaion is used in economies wih high degree of dollarizaion. The model formulaion in hese counries is followed Branson and Henderson (1985) and Porqueras e all (1999), which separaes broad money ino domesic and foreign money and esimaes he money demand for each of hem. The following presens more deails abou he empirical research of money demand in highly dollarized economies. Rodriguez and Turner (2003) invesigaed he money demand in Mexico. By using coinegraion echnique, hey found he following formulaion of money demand: M ln( ) P M α + u. (1) P f e d = o + α1 lny1 + α 2 ln i + α 3i + α 4Δe + α5π + α 6 ln( ) 1 where M/P is he real domesic money, Y is he scale variable (income), i is he ineres rae on domesic bonds, i f is he ineres rae on foreign bonds, e e is he expeced rae of depreciaion, π d is he difference beween he inflaion raes of he domesic and foreign economies. For he money demand of foreign currency deposis, hey firs ried o formulae he model as (1) wih nominal domesic money (M) is replaced by nominal foreign money, in erms of domesic currency (em*). Their es go a resul bu signs of 6

8 some coefficiens conradic wih expecaions. Afer some arrangemens, hey go he final and reasonable resul by he following formulaion: em ln( M em α + u.(2) M * * f f e ) = o + α1 lny1 + α 2( i π ) + α3( i π ) + α 4Δe + α5 ln( ) 1 The resul shows ha currency subsiuion evidences significanly in broad money equaion, and exchange rae has imporan effecs on demand for foreign money. Analyzing he high dollarizaion problem in Lain America, Porqueras e all (1999) found ha he developmen of financial markes and he increasing of he openness have furher diversified and increased he available asses, especially he foreign financial asses, hus affecing he agen s porfolio baske. They se up he following model: d m and While R m f i i d log( ) = log( ) = γ 0 + γ 1log( yi) + λ 2Ii + λ 3Ei + ε..... (3) d 1 Ri mi The noaion R as he raio of foreign money over broad money: R=m f /(m f +m d ), f m represened he real broad money demand in domesic and foreign currency. I is ineres rae of domesic deposis; y is real naional oupu, is he d exchange rae depreciaion and denoes he ime. Using yearly daa from , hey used panel daa o conduc he research by using equaion (3) for 13 counries in Lain America (Argenina, Belize, Cosa Rica, Dominica, Ecuador, El Salvador, Honduras, Mexico, Nicaragua, Paraguay, Peru and Uruguay). The resul shows ha income and exchange rae depreciaion srongly affec he raio of foreign currency deposis over broad money. Bas and Nina (1995) invesigaed he money demand in six ransiional economies in Easern Europe (Bulgaria, Poland, Hungary, Romania, Czecho Slovakia, Czech Republic) during around since in his period, hese economies were reformed very quickly. This was accompanied by high degree of moneary insabiliy, huge shocks in price levels and sharp increase in dollarizaion. Following Cagan (1956), when he economy is in hyperinflaions, he real money demand is mainly affeced by inflaion expecaions. Thus, real oupu and real ineres rae can be assumed o be consan and he domesic money demand and he demand for foreign money ( M * ) can be wrien as: E i 7

9 ln(m )= c +δd - βπ e - γe e +ηd E e +ε... (4) ln( * em P ) = ln m = d + D * * E D E e e e χ απ + δ + φ + where m is demand for domesic real money currency, ε (5) is real foreign money demand measured by domesic currency, π *e is he foreign expeced inflaion and π e is he expeced rae of inflaion and E e is he expeced rae of depreciaion. D is he dummy variable represening he financial liberalizaion in ransiional economies during 1990s. Using coinegraion and error correcion model echniques, hey conduced he ess by using currency in circulaion, base money, narrow money and foreign currency deposis as money aggregaes. The resul indicaed ha in each counry wih differen moneary aggregaes, he long run money demand was found. For example, in Bulgaria, Hungary inflaion had a really srong and significan effec on boh currency in circulaion and narrow money, while he exchange rae affeced significanly foreign currency deposis. In Poland, boh inflaion and exchange rae depreciaion had significan effec on narrow money. In Romania, inflaion and exchange rae depreciaion had significan effecs on currency in circulaion while exchange rae depreciaion had significan effec on narrow money and foreign currency deposis. Considering ha Vienam is a dollarized economy, his paper invesigaes he money demand in Vienam by using he model of Branson and Henderson (1985) and Porqueras e all (1999). Paricularly, i separaes he broad money ino broad money of domesic currency and foreign currency deposis and esimae he money demand for each of hese moneary aggregaes. The explaining variables will be carefully chosen based on he analysis of financial operaions in Vienam. * m 8

10 III. MODELS AND ANALYSES OF RESULTS 3.1 Models and Daa Choice of Variables: Real Money Sock (RM2D and RFCD ): Many recen money demand researches have used broad money insead of narrow money since broad money is considered o be beer explained by economic variables han narrow money. Regarding he broad money (M2) in Vienam, his includes currency in circulaion, Vienam Dong (VND) deposis and foreign currency deposis (primarily in USD), in which USD accouns for around 23.5% of M2 in average during Following Porqueras e all (1999), his paper esimaes money demand by esimaing he demand for M2 of domesic currency (M2VND) and demand for foreign currency deposis (FCD). Hence, his paper employs real M2 of domesic currency and real foreign currency deposis as proxies for money socks. M 2ofVND RM 2D = Ln( CPI FCD RFCD = Ln( CPI *100) *100) where M2ofVND is he nominal broad money balance of domesic currency and measured in billion VND, FCD is he nominal foreign currency deposis balance in erms of domesic currency and also measured in billion VND (using exchange rae a he end of periods). CPI is he consumer price index (1990=100) and denoes ime index. Real Gross Domesic Produc (RGDP ) The scale variable is used as a measuremen of ransacions reflecing he affecs of economic aciviies on money demand. This paper employs gross domesic produc as he scale variable since his is he only available daa. NGDP RGDP = Ln( *100) GDPdeflaor 9

11 where NGDP is he Nominal Gross Domesic Produc (billion VND). GDP deflaor has he year of 1994 as he base year (1994=100). Ineres Rae on 6 Monh Domesic Currency Deposi (DEPO ), Rae of reurn of USD (RF) and Annualized Inflaion Rae ( INF_A ) In he model of demand for real broad money of domesic currency, he seleced opporuniy coss of holding money are: (i) he own-rae of money is represened by 6 monh deposi nominal ineres rae (DEPO ). The 6 monh VND deposi ineres raes is chosen as a proxy since his is he mos popular erm of deposis in he economy. (ii) he rae of reurn on financial asse alernaive o money. Bonds and socks are no popular in Vienam; however USD is a very popular financial asse subsiue o VND, so USD is chosen o be he financial asse alernaive o money in model of domesic currency. Rae of reurn of alernaive financial asse is represened by he expeced rae of reurn of foreign currency deposi and he acual rae of reurn of shor erm USD deposi (RF) in erms of domesic currency is aken as he represenaive and calculaed as follows: RF = DEPOF + E f f where DEPOF is he 6 monh USD deposi ineres rae and is he expeced depreciaion rae, boh of which are annualized. Like VND ineres rae, he 6 monh USD deposi ineres raes is chosen as a proxy since his is he mos popular erm of deposis in he economy. (iii) he rae of reurn on real asses alernaive o money, following Nachega (2001), i is represened by inflaion rae. The annualized inflaion rae ( INF_A ) is compued by: INF_A = (lncpi lncpi -1 )*4. The Difference beween Rae of Reurn of Foreign Currency Deposi and Yield of Domesic Currency Deposi (DIFFER ) The difference beween raes of reurns of deposis by USD and VND (DIFFER ) is calculaed as rae of reurn of 6 monh USD deposis in erms of domesic currency minus nominal ineres rae of 6 monh VND deposis. All daa are annualized. DIFFER = RF DEPO = DEPOF + E f DEPO E 10

12 In he model of demand for foreign money, DIFFER sands for speculaion moive. When he difference beween rae of reurn of USD (RF) and rae of reurn of domesic currency (DEPO) is posiive, people are expeced o shif from domesic currency deposi s holdings o USD deposis holdings and, conversely, when he difference is negaive, USD deposis are expeced o be shifed o domesic currency deposis Model Specificaion and Daa As explained in Secion II, in a highly dollarized economy, he money demand can be formulaed by esimaing demand for domesic currency and demand for foreign currency. The money demand models can be wrien as follows: RM 2 D = β o + β 1 RGDP + β 2DEPO + β 3INF _ A + β 4RF + ε (6) RFCD = α o + α 1 RGDP + α 2DEPO + α 3INF _ A + α 4RF + ε. (7) where RM 2 is real M2 of local currency, RFCD is real foreign currency deposis D in erms of domesic currency, respecively; RGDP is real GDP; DEPO is 6 monh annualized domesic deposi ineres rae; and INF_A is he acual annual domesic inflaion rae. RF = DEPOF + E f is he rae of reurn of USD deposi in erms of domesic currency. We may drop he inflaion variable in he RFCD equaion since foreign money serves as inflaion hedge so inflaion does no cause a shif from real asse o foreign money or conversely. Inflaion may lead o a shif from domesic money o foreign money, however his effec is refleced in he effec of exchange rae (in RF) on foreign money (RFCD) as inflaion is parly absorbed by exchange rae. During , he inflaion is generally considered as in a low-medium level as he mean of annual inflaion was 5.74%, aking ino consideraion ha Vienam is a ransiional developing economy. Therefore, we may drop he inflaion variable in he equaion of RFCD. The research has ried many various arrangemens. A he end, he following formulaion is acceped: RFCD = γ + γ 1RGDP + γ 2DIFFER + ε 0. (8) 11

13 where DIFFER = DEPOF + E is he difference beween raes of reurns of foreign and domesic currency deposis. I represens he asse subsiuion by domesic agens beween domesic money and foreign money depending on raes of reurns of hese wo currencies in erms of domesic currency. f DEPO Expeced Signs of Coefficiens in he Domesic Money Demand Model The scale variable (RGDP ) represens he ransacion effec. I is posiively relaed o money demand. Domesic ineres rae is expeced o have posiive sign since he higher he own rae of reurn on money he more incenive o hold money. The inflaion variable (INF_A ) is expeced o have negaive sign since agens prefer o buy real asses as inflaion hedge in he periods of high inflaion. The coefficien of he rae of reurn of foreign financial asse ( RF ) is expeced o be negaive since he higher he rae of reurn of USD deposi, he more people wans o ransfer from VND o USD and conversely. In summary, followed he model formulaion as indicaed in equaion (6), he expeced signs of variable coefficiens can be wrien as follows: β1> 0, β 2 > 0, β 3 < 0, β 4 < 0. Expeced Signs of Coefficiens in he Foreign Money Demand Model, The scale variable (RGDP) represens he wealh effec and is expeced o be posiively relaed o he real money demand. The coefficien of DIFFER is expeced o be posiive since when rae of reurn of USD deposi is higher han yield of domesic currency deposi, people are expeced o shif from domesic currency deposis o USD deposis and conversely. In summary, following he model formulaion as indicaed in equaion (8), he expeced signs of coefficiens can be wrien as follows: γ 1 > 0, γ 2 > 0. Daa The sudy uses quarerly daa during 1993:1-2004:4. The reason o choose his period is ha: (i) daa are available only from 1991; (ii) inflaion is excepionally high during 1991 and 1992 in compared wih laer years. The daa source for moneary, ineres rae, inflaion, exchange rae and GDP are from Inernaional Financial Saisic (IFS) and various issues of Vienam Economic Review. 12

14 3.2 Uni Roo Tess Dickey and Fuller (1979) shows ha if he characerisics of a sochasic process change over ime, he sochasic process is non-saionary and if we use ordinary leas square (OLS) o represen he relaionship beween non-saionary daa series, we may ge spurious resuls, in oher words he OLS would no generae a consisen parameer esimaor. Therefore, we need o do uni roo ess o check wheher daa series are saionary or no. Table 2: Augmened Dickey-Fuller Uni Roo Tess 4 Lag lengh ADF saisics Firs Differences Lag lengh ADF saisics RM2D RM2D ** RFCD RFCD ** RGDP RGDP INF_A RGDP *** RF INF_A ** DEPO RF *** DIFFER DEPO *** DIFFER *** Noe: 1) (**), (***) shows he variables are significan a 5%, and 1% levels, respecively. 2) In he Firs Differences column, all daa series are a he 1 s difference, excep RGDP is he second difference. 3) Based on visual examinaion of daa, he ADF ess are conduced wih inerceps and rends a levels for RGDP, RM2D, RFCD and wih inerceps only for he remaining variables. A he firs difference, he ADF ess are conduced wih inerceps only. 4) The uni roo is conduced using lags lengh from 1 o 10. The repored resul is chosen wih he number of lags having minimum Akaike informaion crierion or Schwarz crierion. 4 Following Dickey and Fuller (1979), he Augmened Dickey-Fuller es is used o run he regression: Y =β 1 +β 2 + δy -1 + λ Δ +μ p j= 1 and hen Y =β 1 + λ Δ +μ p j= 1 j Y j j Y j where Y = Y - Y -1 ; denoes he ime or rend variable; μ is he disurbances. To es H 0 (β 2 =0, δ=0) meaning Y is non-saionary or has uni roo a order one, a F raio is calculaed and check is significance by he Dickey and Fuller saisics. 13

15 The uni roo es resuls are presened in Table 2. The resuls indicae all variables are I(1), excep RGDP is I(2). However a general knowledge is ha GDP should be I(1) or he growh rae of real GDP should be saionary. Herein, he maer ha RGDP daa is I(2) can be semmed from he limied number of observaions as here are only 48 observaions. Therefore, in his case, we will base on general knowledge o consider RGDP as I(1) for our research analysis. 3.3 Coinegraion Tes Engle and Granger (1987) indicaed ha if a linear combinaion of some non-saionary daa series is saionary, i is called coinegraed and considered as long-run equilibrium relaionship among he variables. As resuls of uni roo es indicae ha all he variables are I(1), his paper conducs coinegraion es o esimae he long run demand for money if a coinegraion exiss. The mehod of Johansen (1988) is applied. Following Ibrahim (1999) and Sriram (1999a), since quarerly daa are used, he analysis begins wih eigh lags and he es is repeaed by reducing consecuively one lag a a ime unil he lag lengh reaches one Coinegraion Tes for Demand for M2 of Domesic Currency Variables RM2D, RGDP, INF_A, DEPO, RF are enered as endogenous variables in ha sequence. A consan is also added. Since he variables are already seasonally adjused, no seasonal dummy is necessary o add o he equaion. The resul is presened in Table 3. They indicae ha race es suppors 1 coinegraion vecor a 98% significan level and maximal eigenvalue es suppors 1 coinegarion a 87% significan level. Thus, here is 1 coinegraion a lag lengh of one. The long run demand for money can be wrien in he following equaion (-saisics in he parenheses): RM2D = * RGDP * DEPO * INF_A *RF (8.512) (1.809) (-3.945) (-3.206) 5 Tha he minimum number of lag lengh is 1 is based on Sriram (1999a). In his paper, he coinegraion es of demand for M2 is conduced by monhly daa wih he lag lengh from 1 o 12 and his paper found he bes coinegraion resul a 3 lags. 14

16 The resul shows ha he long run income-elasiciy of money is This can be explained by he fac ha since Vienam is an economy in he process of moneizaion and he financial sysem is developing quickly, 6 he income elasiciy can be much larger han 1. The income-elasiciy of money shows ha, if RGDP grows by 1%, ceeris paribus, he demand for real M2 of domesic currency would increase by 2.76%. This finding is generally in line wih ha obained by Huang (1994), where he found he long run income-elasiciy of money in China is The own rae of money (DEPO) is posiively affeced RM2D, hus indicaing ha if, ceeris paribus, he domesic ineres rae increases by one-percenage-poin, he demand for real M2 of domesic currency increases by 4.26 percens. The semi-elasiciy of inflaion is , indicaing ha agens away from money holding o real asses when inflaion is high. Like many developing counries, expeced inflaion has a srong impac on money demand in Vienam. Table 3: Johansen Coinegraion Tess for Money Demand Trace Saisics Maximum Eigenvalue Equaion H 0 r = 0 r 1 r 2 r 3 r 4 r = 0 r 1 r 2 r 3 r 4 H A r 1 r 2 r 3 r 4 r 5 r 1 r 2 r 3 r 4 r 5 RM2D 76.51* * RFCD 48.7** ** % Criical Value RM2D RFCD Probabiliy (%) RM2D RFCD An evidence of moneizaion process is he raio M2/GDP (%): Huang (1994) conduced coinegraion es for China during using quarerly daa and found he following money demand: m2 = y p i. Where m2 is real M2 (M2 = sum of cash and consumer s savings deposis (excluding deposis of enerprises)), y is real GNP, p is consumer reail sale price and i is real rae of ineres on one-year saving deposis. Excep i, all variables are in logarihm. 15

17 Noes: 1) r denoes he number of co-inegraing vecors. 2) The aserisk (**), and (*) indicae he rejecion of he null hypohesis a he 1 and 5% significance level. If he expeced rae of reurns of USD increases by one-percenage-poin, he demand for real M2 of domesic currency decreases by 1.94%. This resul shows ha asse subsiuion is happening in Vienam. Ineresingly, he long run coefficien on expeced rae of reurns of USD deposi (1.94) is lower han he long run coefficien of domesic ineres rae (4.26) in absolue erms, suggesing ha he own rae exercises more influences on demand for real broad money of domesic currency han rae of reurn of foreign financial asses. Tha implies domesic ineres rae is sill more imporan han rae of foreign reurn in deermining demand for real broad money of domesic currency. This can happen due o domesic ineres rae parly absorbed he effec of foreign rae of reurns Coinegraion Tes for Demand for Real Foreign Currency Deposis Variables RFCD, RGDP, DIFFER are enered as endogenous variables of he coinegraion es in ha order. A consan is also added. The variables are seasonal adjused so no seasonal dummy is needed o add o he equaion. A dummy variable (dum4) is added o he regression o reflec he srong depreciaion of he domesic currency during he Asian financial crises. Dum4 has he value of one for periods of 1997:1-2 and 1997:4-1998:4 and zero for he ohers. The es resul is presened in Table 3. They indicae ha, a 99.99% significan level, boh race es and maximal eigenvalue es suppor here is only one coinegarion, a lag lengh of one. The long run demand for real foreign currency deposis can be wrien in he following equaion (-saisic in he parenheses). RFCD = * RGDP * DIFFER (4.0745) (7.8909) The resul shows ha he long run income elasiciy is , which is higher han one, bu lower han he long run income elasiciy in he demand for real broad money of domesic currency (2.7633). Tha shows domesic currency is sill he prefer currency in economic ransacions. In oher words, when income increases he demand for M2 of VND expands quicker han ha of USD. The coefficien of RGDP (1.87) 16

18 also can be explained by he fac ha he economy is in he moneizaion process so he income elasiciy of money is larger han 1. The coefficien of speculaion variable (DIFFER) srongly and posiively ( ) affecs real foreign currency deposis, hus indicaing ha if he difference beween expeced raes of reurns of USD and VND increases by one-percenage-poin (say from 1% o 2%/year), he demand for RFCD will increase by around 18%. This robusly affirms ha he asse subsiuion is really a maer in he Vienamese financial sysem and i is very imporan for he moneary auhoriy o keep VND deposi ineres rae higher han USD rae of reurn in order o keep he dollarizaion problem no being worsened. Demand for real foreign currency deposis is very sensiive o he difference beween raes of reurns of foreign and domesic currency deposis, especially exchange rae depreciaion. A posiive difference beween raes of reurns of foreign and domesic currency deposis riggers a shif from domesic financial asses and real asses o foreign financial asses, resuling in a lower inflaion rae and a srong increase in demand for foreign money. To summarize, some major findings are as followed: - I is ineresing o see ha he long erm money demand is exising in Vienam. Previously, i was well believed by policy makers in he Vienamese moneary auhoriy as well as various financial researchers ha i is difficul o find a highly saisical adequae long run demand for money model in Vienam. Acually some effors by inernaional financial organizaions as well as governmen agencies had been done, bu failed o ge any good resuls. To ge his resul, he research has carefully analyzed he Vienamese financial marke o find a proper model for his dollarized, in-ransiion economy and carefully choose he explaining variables in he model. The empirical resuls show ha i is appropriae o separae he broad money ino broad money of domesic currency and foreign currency deposis in order o formulae money demand models. - The coinagraion ess idenified only one coinegraion for real broad money of domesic currency and one unique coinegraion for foreign currency deposis. The long run demand for real M2 of domesic currency is significan a 98% levels by race es and 87% levels by maximum aigen value es, while he long run demand for real foreign currency deposis is significan a 99.99% levels by boh ess. Hence, i is 17

19 reasonable o use moneary aggregaes as one imporan facor in he se of inermediaes argeing variables for moneary policy. - In he model of demand for real M2 of domesic currency, he income elasiciy is found o be 2.76, much larger han one, which can be explained ha he economy is in moneizaion process. The variable employing he foreign effec is found o be he rae of reurn of USD deposi in erms of domesic currency. However, is direc effec o real M2 of domesic currency is found o be smaller han ha of oupu, domesic ineres rae and inflaion rae. Tha can happen due o he governmen have used domesic ineres rae for incorporaing some effecs of he rae or reurn of USD deposi in order o conrol exchange rae. In he equaion of demand for real foreign currency deposis, he variable ha exercises he speculaion effec urned ou o be he difference beween raes of reurns of USD and domesic currency deposis. The semi-elasiciy of his variable is very large, implicaing he moneary auhoriy should be very cauious in keeping domesic currency deposi rae higher han USD rae of reurns in erms of domesic currency. Meanwhile, he income elasiciy in RFCD equaion is higher han 1 bu smaller han he income elasiciy in RM2D equaion. Boh of hem can be explained by he fac ha Vienam is in he moneizaion process. The coefficien of income elasiciy in RM2D equaion (2.76) is judged o be more reasonable han ha in RFCD equaion (1.87) since he fac ha he DEPO coefficien is higher han RF coefficien in RM2D equaion shows ha domesic currency, raher han USD, is used as he main currency in he economy. - In order o have a well specified demand for money model, i is necessary o include ineres rae as he own rae of reurn of money and inflaion as rae of reurn of alernaive real asse. Specifically, when a counry like Vienam, herein: (i) ineres rae is raher rigid; (ii) he economy is in ransiion; (iii) inflaion was decreased from a very high o a low level, i is sill imporan o include boh he own rae of money and raes of reurns of alernaive real asse in he model. Furhermore, when a counry experiences a medium level of inflaion, i is sill imporan o include he rae of reurn of real asse since ineres rae may no fully reflec changes in price level and agens end o hold real asses when inflaion is high. 18

20 3.4 Error Correcion Model Engle and Granger (1987) shows ha he shor run model provides informaion abou how adjusmen is aking place o resore he equilibrium in he long run money demand in response o he shor run deviaion. The error correcion erm is calculaed from he vecor error correcion represening he demand for RM2D and RFCD. Since all variables in he shor run model are saionary, he variables in he shor run model will be he firs differences of he variables in long run model, excep he error correcion erm is a level. No seasonal dummy is needed o shor run models as daa is seasonally adjused Error Correcion Model for Real Broad Money of Domesic Currency The error correcion model for real broad money of domesic currency is derived from he following equaion. ΔRM 2D = 4 i= β iδrm 2D i + αect 1 + βjδrgdp j + βkδdepo k + βmδinf _ A m + βnδrf n + ε j= 0 k= 0 m= 0 where error correcion erm ECT-1=RM2D -1 -RM2D* -1, or he error correcion erm is calculaed as RM2D a ime (-1) minus he esimaed RM2D from he coinegraion vecor (RM2D*) a ha ime. This sands for he excess money in he previous period. Based on he limied number of observaions, he maximum number of lags should no exceed four. And he number of lags is chosen o be 4 based on he minimum Akaike informaion crierion. Since all variable are saionary, he above model can be esimaed by ordinary leas squared (OLS). The unresriced reduced form of he error correcion model is reduced o a resriced reduced form of shor run demand for RM2D by excluding he insignifican variables excep he consan erm. The following able displays he resriced error correcion model. The following discussion can be drawn from he analysis: Firsly, he error correcion erm ECT -1 has he expeced negaive sign and very significan (a 0.01% level), which cerifies he long run relaionship in he coinegraion es. The negaive sign implies ha money demand adjused in he following quarer in response o he disequilibrium beween money supply and money demand. If money supply exceeds money demand in he curren quarer, agens will n= 0 19

21 resric heir money demand in he following quarers. The adjusmen akes abou 15 quarers or he exising disequilibrium will be decreased and disappeared afer 15 quarers. This resul is in line wih he radiional view ha money is largely passive in socialis economies (Bas and Nina, 1995). Table 4: Resriced Error Correcion Model for Demand of Real Domesic Money Variable Coefficien Sandard error -saisic C *** ECT(-1) *** RM2D(-1) *** RM2D(-2) *** RM2D(-3) ** RM2D(-4) ** RGDP(-2) *** RGDP(-3) ** DEPO(-1) *** INF_A *** INF_A(-4) *** RF ** RF(-4) ** R 2 = Adjused R 2 = DW = Noe: The aserisk (***), and (**) indicae he rejecion of he null hypohesis a he 1 and 5% significance level. Secondly, all four elasiciy esimaes of lags of demand for money ( RM2D) are significan a 5% levels, showing ha he curren money demand is depended on money demand of las four quarers. The absolue values of he coefficiens are uniformly less han one. Tha indicaes, in he shor run, he growh rae of demand for money is inelasic o hose of las hree quarers. The values of he coefficiens are 20

22 mixed bu in oal, i shows ha changes in money demand in four previous quarers have posiive recursive effecs (0.4426) o changes in money demand in curren period. Thirdly, he signs of real oupu, inflaion are mixed bu heir oal recursive effecs has signs as expeced and all he coefficiens are significan. The coefficiens of domesic currency ineres rae and USD rae of reurn are significan and have expeced signs. RGDP coefficiens have posiive oal recursive effecs (0.139), oal inflaion coefficiens have negaive effec (-0.092). The coefficien of domesic ineres rae is posiive and significan a 0.01% levels. The value of DEPO coefficien is 1.13 showing ha demand for money is elasic o domesic ineres rae even in he shor run. If DEPO coefficien increases by one-percenage-poin, demand for RM2D increases by around 1.13% in nex quarer. The coefficien of DEPO in shor run model (1.13) is smaller han ha in he long run model (4.261), hus reflecing lags in moneary ransmission mechanism. In oher words, a one-percenage-poin increase of domesic currency ineres rae will increase broad money demand by 1.13 percens, bu he effec in shor erm is smaller han he effec in long erm since i needs some periods o pass he full effec of he increase in ineres rae on o money demand. The RF coefficiens are negaive and significan a 5% level, bu he value of coefficiens are small ( and -0.08). Ineresingly, his resul shows ha, in he shor run, he demand for VND is no elasic o rae of foreign reurn. In erms of saisical adequacy of he ECM, he adjused R squared (0.826) is acceped since regressions wih firs differences of he variables usually don have high R squared. 8 The Durbin Wason (DW) saisic indicaes he absence of a serial correlaion in he ECM Error Correcion Model for Real Foreign Currency Deposis The error correcion model of real foreign currency deposis is derived from he following equaion. ΔRFCD = γjδrfcd j + αect 1 + λδdum 4 + γkδrgdp k + γl j= 1 k= 0 l= 0 ΔDIFFER l + ε 8 Chowdhury (1993) found ha he R-squared in shor run money demand model are from 0.49 o He considered hese R-squared are normal for regression based on firs differences in variables. 21

23 where he error correcion erm is calculaed as ECT -1 =RFCD -1-RFCD* -1, in oher words, RFCD a ime -1 minus he esimaed RFCD a ha ime (from he coinegraion vecor). In line wih he long run equaion, a firs difference of he dummy variable, DUM4, is added o he shor run equaion. Since he daa is quarerly and number of observaions is limied, he maximum number of lags should no go beyond four and he lag lengh of 1 is chosen based on he minimum Akaike informaion crierion. Table 5: Resriced Error Correcion Model for Demand of Real Foreign Currency Deposis Variable Coefficien Sandard error -saisic Prob. C *** ECT(-1) ** DUM *** RFCD(-1) *** DIFFER(-1) *** R 2 = Adjused R 2 = DW = Noe: The aserisk (***) and (**) indicae he rejecion of he null hypohesis a he 1 and 5% significance level. The model can be esimaed by ordinary leas squared. The unresriced reduced form of he error correcion model is reduced o a resriced reduced form by eliminaing insignifican coefficiens. The following discussion can be drawn from he analysis: Firsly, he error correcion erm ECT -1 has he expeced negaive sign and is significan a 1% level. Tha validaes he long run relaionship in he coinegraion es. The negaive sign implies ha money demand adjused in he following quarers in response o he disequilibrium. The exising disequilibrium will be decreased and disappeared afer 70 quarers. Secondly, he elasiciy esimae of he firs lag of demand for money ( RFCD) is significan a 1% level. The value of he coefficien is less han one, 22

24 indicaing ha, in he shor run, he growh rae of demand for money is inelasic o is of las quarer as expeced. Thirdly, ineresingly, in shor run, RGDP doesn play an imporan role in deermining demand for money. In he long run, RGDP is significan. However, in he shor run, changes in oupu have no role o explain he changes in he demand for RFCD. In addicion, unlike he long run equaion, he coefficien of DIFFER in he previous period has negaive and significan effec on money demand, alhough he coefficien in shor run model (-0.36) is small. In erms of saisical adequacy of he ECM, boh he adjused R 2 (0.534) and he Durbin Wason (2.342) saisic are acceped. 3.5 Impulse Responses The impulse response maps he effecs of a variable s shock on all variables in he vecor error correcion model. In his secion, we will invesigae hree impulse responses: (i) responses of real broad money of domesic currency, oupu, domesic ineres rae, and inflaion o one ime shock of USD rae of reurns; (ii) responses of RM2D o an unanicipaed change in oher variables; and (iii) responses of RFCD, RGDP o one ime shock of DIFFER. In oher words, we will race he effecs of a one-ime shock o RF and DIFFER on curren and fuure values of oher variables in heir models, respecively as well as map ou he effecs of unanicipaed changes in oher variables on RM2D. Based on sandard macro economic heory, he order of he variables in he impulse responses and variance decomposiions is seleced as follows: RF, DEPO, RM2D, RGDP, INF_A Impulse Responses of Variables in RM2D Model o One Time Shock of USD Rae of Reurn The reason we are ineresed in RF shock is ha his shock can be considered as an exernal shock o he economy and we will see how such a small dollarized ransiion economy reacs o he exernal shock. The shock of RF will increase RF in firs quarer (Figure 3) bu he overshooing increase will decrease in he second quarer. However, he shock resuls in a long erm increase in RF by around 2.4%. The ineres rae of domesic currency 23

25 responds quickly o he shock of rae of foreign reurns by one sandard deviaion. Domesic ineres rae increases immediaely and afer 2 quarers, domesic ineres rae is around 0.47 poins higher han is pre-shock level and flucuaes slighly around ha in he following quarers. The response parly reflecs he uncovered ineres rae pariy in he financial marke and he persisence a a higher level of domesic currency ineres rae in response o USD rae of reurn shock can help o explain he ineres rae policy which aims o keep a higher ineres rae in response o he increase in rae of foreign reurn in order o preven a currency fligh o USD. In response o shock of rae of reurn of USD, demands for M2 of domesic currency and inflaion decrease immediaely (Figure 3&4). In he following quarers, real M2 of domesic currency sill decreases furher since people are averse o shocks in rae of reurn of USD. From he fourh quarer, inflaion is around 0.48%/year lower han is pre-shock level. Figure 3: Responses of RF, DEPO and RM2D o Cholesky One Sandard Deviaion of RF Innovaion (1993: :4) RF DEPO RM2D Real oupu increases a wo firs quarers bu i decreases in laer periods (Figure 4). This can be explained as follows: when here is an unanicipaed change in RF, demand of USD increases and causes cenral bank o keep he foreign reserves by direcly or indirecly limiing he sell of foreign exchange o imporers, hus leading o he emporary decrease in impor and emporary increase in RGDP. Bu in he afermah, RGDP decreases due o he srong depreciaion make people became uncerainy abou fuure and he higher domesic ineres rae has affeced negaively 24

26 he economic growh. In he conrary o J-curve effec, a srong depreciaion in Vienam will increase very slighly GDP in very shor erm bu decrease GDP very slighly in laer periods. The decline of inflaion coincides wih he decline of RGDP and RM2D, besides he higher domesic ineres rae also leads o lower inflaion rae. Figure 4: Responses of Inflaion and Oupu o Cholesky One Sandard Deviaion of RF Innovaion RGDP INF_A Impulse Responses of RM2D o One Time Shock of Oher Variables Nex, we examine he impulse response funcion of RM2D as a way o measure how he real domesic money demand responds o shocks of oher variables (Figure 5). When here is a shock o RGDP, real domesic money demand will increase as he demand for money in ransacion increases. We can see ha he response of RM2D o RGDP innovaion is sronger and sronger, showing ha RM2D has srong response o RGDP shock in long run. In response o he shock in inflaion, he real domesic money demand will decrease as higher inflaion make people shifed from domesic money o real asse. Figure 5: Responses of RM2D o Cholesky One Sandard Deviaion of RF, DEPO, RGDP and INF_A Innovaions (1993: :4) 25

27 RF DEPO RGDP INF_A As he rae of reurn of FCD increases, he demand for real domesic money will decrease. Inflaion and rae of foreign reurn influence on RM2D as srong as GDP does bu in opposie direcion. The increases of inflaion and rae of foreign reurn decrease he demand for real broad money of domesic currency. Iniially, he increase of ineres rae or a conracionary moneary policy will cause he money aggregae o decrease bu his effec lass around 3 quarers. From he 4 h quarer, he increase in ineres rae caused by conracionary moneary policy raises he demand for real broad money of domesic currency, bu he level of increase is small. The effec of DEPO dies ou quickly, which means ha shock of domesic ineres rae only affecs real domesic money demand in shor run. In long run he domesic rae innovaion has no considerable effec on RM2D Impulse Responses of RFCD and RGDP o One Time Shock of DIFFER In he following par, his paper conducs he impulse responses for variables in he demand for real foreign currency deposis equaion o one ime shock of differences in raes of reurns of VND and USD deposis. The reason we are ineresing in his es is ha DIFFER can be arisen from an exernal facor (changes in USD ineres rae in world marke), changes in exchange rae or changes in moneary policy by adjusing he ineres rae, hus hese changes can be arisen from he policy inenions or has poenial srong effecs on a small, open economy. 26