Money, Income, Prices, and Causality in Pakistan: A Trivariate Analysis. Fazal Husain & Kalbe Abbas

Transcription

1 Money, Income, Prices, and Causaliy in Pakisan: A Trivariae Analysis Fazal Husain & Kalbe Abbas I. INTRODUCTION There has been a long debae in economics regarding he role of money in an economy paricularly in he deerminaion of income and prices. The Moneariss claim ha money plays an acive role and leads o changes in income and prices. In oher words, changes in income and prices in an economy are mainly caused by he changes in money socks. Hence, he direcion of causaion runs from money o income and prices wihou any feedback, i.e., unidirecional causaion. The Keynesians, on he oher hand, argue ha money does no play an acive role in changing income and prices. In fac, changes in income cause changes in money socks via demand for money implying ha he direcion of causaion runs from income o money wihou any feedback. Similarly, changes in prices are mainly caused by srucural facors. The causal relaionships beween money and income and beween money and prices have been an acive area of invesigaion in economics paricularly afer he provocaive paper by Sims (1972). Based on Granger causaliy, he developed a es of causaliy and applied i o he U.S. daa o examine he causal relaionship beween money and income. He found he evidence of unidirecional causaliy from money o income as claimed by he Moneariss. However his resuls were no suppored by subsequen sudies. Barh and Benne (1974) replicaing Sims es in Canadian economy found a bidirecional causaliy beween income and money, whereas, Williams, Goodhar, and Gowland (1976) applying Sims procedure in he U.K. found he evidence of unidirecional causaliy from income o money, opposie o Sims findings. They also found he evidence of unidirecional causaliy from money o prices. On he oher hand, Brillembourg and Khan (1979) using a longer daa se suppored Sims findings and found a unidirecional causaliy from money o income and prices in he U.S. However, Dyreyes, Sarleaf, and Wang (1980), examining he paern of causaliy beween money and income for six indusrialized counries, found differen resuls. For example, hey found bidirecional causaliy in he U.S., conrary o Sims (1972) and Brillembourg and Khan (1979). Similarly, hey found unidirecional causaliy from money o income in Canada, conrary o Barh and Banne (1974). However, heir finding of unidirecional causaliy from income o money in he U.K. was in line wih Williams e al. (1976). In he case of developing counries, Lee and Li (1983) examined causaliy among money, income, and prices in Singapore and found bidirecional causaliy beween income and money and unidirecional from money o prices. Joshi and

2 2 Joshi (1985) found a bidirecional causaliy beween money and income in India. Khan and Siddiqui (1990) found unidirecional causaliy from income o money and bidirecional beween money and prices in Pakisan. Abbas (1991) performed causaliy es beween money and income for Asian counries and found bidirecional causaliy in Pakisan, Malaysia and Thailand. In an unpublished paper, Bengali, Khan, and Sadaqa found a bidirecional causaliy beween money and income and unidirecional from money o prices in Pakisan. The above discussion indicaes ha he empirical evidence regarding causal relaions beween money and he oher wo variables, income and prices, remain inconclusive. Moreover, he papers reviewed above invesigaed causaliy beween wo variables, i.e., beween money and income and/or beween money and prices. However, an economic variable is generally influenced by more han one variables, herefore, models involving more variables may be more useful. Ho (1982) invesigaed causaliy among money, domesic prices and impor prices in Hong Kong using a rivariae causaliy approach. He found a unidirecional causaliy from domesic prices o money as well as significan effecs of impor prices on domesic prices. However, his resuls were no differen in bivariae and rivariae case. The purpose of his paper is o re-examine he causal relaionship beween money and income and beween money and prices in Pakisan. We use a longer daa se from o covering almos he enire hisory of he counry. Furher, we ake care of he sochasic properies of he variables used, no done earlier wih he excepion of Bengali e al. (n.d.). In addiion, we also invesigae he causal relaionship hrough rivariae approach no aemped before. The res of he paper is organised as follows. The nex secion describes he daa sources as well as he limiaions of he analysis. Secion III oulines he mehodology o es he sochasic properies of he variables and heir inerrelaionship. Secion IV presens and discusses he empirical resuls. The final secion conains he summary and conclusions. II. DATA SOURCES AND LIMITATIONS We use annual daa from o o invesigae he causal relaions beween money and income and beween money and prices in Pakisan. Gross Naional Produc (GNP) a curren prices, broad measure of money (M2), and Consumer Price Index (CPI) wih base , are used as Income, Money and Prices, respecively. The principal daa source is 50 Years of Pakisan in Saisics, prepared by he Federal Bureau of Saisics. 1 The oher daa sources include he regular issues of Economic Survey by Finance Division and Monhly Bullein by Sae Bank. 1 The source adjuss and presens he daa which perain o Wes Pakisan for he periods prior o he separaion of Eas Pakisan.

3 The daa for GNP a curren prices are no available for earlier periods ( o ). These are generaed hrough GNP a consan prices using CPI as proxy for GDP deflaor. Before proceeding furher, we would like o poin ou ha he analysis is based on fify years of Pakisan during which he counry has undergone a series of economic and poliical changes. In paricular, here have been significan improvemens in he moneary secor as well as is impac on economy in he 1990 s. Examining he causal relaionships in his period may provide differen conclusions regarding he role of money in Pakisan s economy. Since GNP is no available oher han on annual basis and 10 or 11 observaions are oo shor for he kind of analysis carried ou in his sudy we choose o use he presen daa se. Hence, he conclusions of his sudy mus be aken wih care. 3 III. METHODOLOGY We sar by examining he sochasic properies of he variables used in he analysis before applying formal ess of causaion. Hence, he Uni Roo Tes is performed on he variables o es for he saionariy of variables. In his conex, Phillips-Perron (1988) es (PP) is used, which is robus o a wide variey of serial correlaion and heeroskedasiciy. The es deecs he presence of a uni roo in a series, say Y, by esimaing. * 1 Y = α + ρ Y +e (1) * 1 Y = α+ β + ρ Y +e (2) where he second equaion includes a rend variable. The PP es is he -value associaed wih he esimaed coefficien of ρ *. The series is saionary if ρ * is negaive and significan. The es is performed for all he variables where boh he original series and he differences of he series are esed for saionariy. The co-inegraion beween he wo series, X and Y, is esed by conducing he PP es on residuals obained from running he OLS regression, called he coinegraing regression: Y = α + β X + e... (3) The causal relaionship beween hese variables is examined hrough Granger causaliy and Error Correcion Models (ECM) as, p q + j= 1 Y = α1 + ρ e 1+ β Y i δ j X j (4) 1 i= 1 p i q + j= 1 X = α2 + ρ e 1+ β Y i δ j X j (5) 2 i= 1 i

4 4 where e 1 is an error correcion erm represening he long run relaionship. A negaive and significan coefficien indicaes he presence of long run causal relaionship. If boh coefficiens are significan, his will sugges he bidirecional causaliy. If, e.g, only ρ 1 is significan, his will sugges a unidirecional causaliy from X o Y, implying ha X drives Y oward long run equilibrium bu no he oher way around. On he oher hand, he lagged erms of ΔY and ΔX, appeared as explanaory variables, indicae shor run cause and effec relaionship beween he wo series. Thus, if he lagged coefficiens of ΔX appear o be significan in he regression of ΔY, his means ha X causes Y. If we omi he error correcion erms from he equaions we will ge he convenional Granger causaliy model, widely used o invesigae causal relaions. As poined ou in he lieraure, Granger model is very sensiive o lag lenghs. In his conex, he convenional pracice is o choose lags on he basis of minimum Final Predicion Error (FPE). We follow he pracice and deermine he lag, e.g for ΔY, as follows, 1. Regress ΔY on a consan erm and is own pas values for p=1 o 5, and choose p which gives he min FPE(p). 2. Given he value of p run he regression again by including pas values of ΔX, for q=1 o 5, and choose he equaion wih min FPE(p,q). Trivariae Causaliy One purpose of his paper is o examine he causal relaions using rivariae causaliy approach. In his conex, firs we examine he long run relaions among variables by conducing he PP es on residuals obained from following coinegraing regression, Y = α + β X + γ Z + e (6) The rivariae causaliy approach involves inclusion of anoher variable, say ΔZ, in Equaions (4) and (5), i.e., ΔY p q = α1 + ρ e 1 + β Y i δ j X j γ Z r (7) 1 i= 1 p i Δ + j= 1 X = α2 + ρ e 1+ β Y i δ j X j γ Z r (8) 2 i= 1 i Δ + r k= 1 q r + + k j= 1 k= 1 This model examines he causal relaionship beween X and Y condiional on he presence of Z. In his case he procedure for he deerminaion of lags is, 1. Regress ΔY on a consan erm and is own pas values for p=1 o 5, and choose he wo bes equaions on he basis of min FPE(p). Similarly, k Δ

5 Regress ΔY on a consan erm and pas values of ΔZ for r=1 o 5, and choose he wo bes equaion on he basis of min FPE(r). 2. Regress ΔY on a consan erm, is own pas values and pas values of ΔZ for four differen combinaions of bes lags chosen in sep 1. Now choose he equaion wih min FPE(p,r). 3. Given he values of p and r run he regression again by including pas values of ΔX, for q=1 o 5, and choose he equaion wih min FPE(p,r,q). The wo seps procedure is used o examine bivariae causaliy beween money and income and beween money and prices. Whereas, he hree seps procedure is used o examine rivariae causaliy beween money and income condiional on he presence of prices and beween money and prices condiional on he presence of income. IV. EMPIRICAL RESULTS A he firs sep, he variables used in he analysis are esed for he uni roos suggesed by Phillips-Perron. The es is applied o boh he original series (in log) and o he firs differences. Furher, boh he models wih and wihou rend are ried. The runcaion lag parameers are deermined following Schwer s (1987). The resuls are repored in Table 1 which indicae he presence of uni roos in he original series. The resuls furher sugges ha aking firs differences remove hese roos implying ha hese variables are firs differenced saionary. 5 Table 1 Uni Roo Tess (Phillips Perron) for he Period o Series in Levels Firs Differences Truncaion Lag Parameers Truncaion Lag Parameers L 4=3 L 12=10 L 4=3 L 12=10 Wihou Trend Income(Y) ** 4.615** Money(M) ** 4.879** Prices(P) ** 4.150** Wih Trend Income(Y) ** 4.790** Money(M) ** 5.046** Prices(P) ** 4.231**

6 6 A he second sep, co-inegraing regressions, o examine long run relaions beween wo variables, are esimaed. Then, he series of residuals are obained from each regression and he PP es is applied o es he presence of uni roos in hese residuals. Table 2 shows he resuls of he PP es on residuals. The able shows he rejecion of hypohesis of no co-inegraion in boh cases of wo variables indicaing he exisence of long run relaionship beween money and he oher wo variables, ha is, income and prices. Nex, he Granger causaliy and Error Correcion Models are employed o explore he direcion of bivariae causaliy. The resuls are repored in Table 3. The able shows he lags deermined as explained above, he F-values for he lags of independen variable, he -values for he error correcion erm, and he paern of causaion. I can be seen ha boh he Granger causaliy and ECM provide similar resuls regarding he direcion of causaion. These resuls show a unidirecional causaliy running from income o money as argued by he Keynesians. The ECM indicaes he same direcion of causaion in he long run. Regarding he money-price relaionship, he resuls sugges a bidirecional causaliy beween hem. However, he ECM shows a one way causaion from money o prices in he long run. This implies ha he growh in money sock is affeced by inflaion in he shor run bu no in he long run. Trivariae Causaliy Finally, we examined he causal relaionship using rivariae causaliy approach. Firs, he long run relaions among money, income, and prices are examined hrough co-inegraion analysis, repored in Table 2. The able shows he exisence of long run relaions among hese variables. The rivariae causaliy analysis is shown in Table 4. The able shows he causal relaionship beween money and income condiional on he presence of prices. Similarly, i shows he causal relaionship beween money and prices condiional on he presence of income. Table 2 Resuls from Co-inegraion Tess Truncaion Lag Parameers L 4=3 L 12=10 Two Variables M on Y 2.613** 2.568** M on P 2.498** 2.494** Three Variables M on Y, P 2.39** 2.36**

7 7 Table 3 Granger Causaliy and Error Correcion Model: Bivariae Case Granger Error Correcion Y on X Lags(y,x) F-values Causaion (err) F-values Causaion Y on M (1,1) M Y M Y M on Y (2,3) 9.384*** Y M 2.283** 6.25*** Y M M on P (2,2) 6.394*** P M *** P M P on M (2,1) *** M P 2.61** *** M P Table 4 Granger Causaliy and Error Correcion Model: Trivariae Case Granger Error Correcion Y on X/Z (y,z,x) F-values Causaion (err) F-values Causaion Y on M/P (1,5,1) M Y M Y M on Y/P (1,2,3) 5.249*** Y M 2.866*** 6.628*** Y M M on P/Y (1,3,1) 3.685* P M 2.911*** 7.453*** P M P on M/Y (2,2,1) 7.236** M P 2.243** 8.386*** M P I can be seen ha he resuls are similar o hose found in he bivariae case, i.e., a unidirecional causaliy from income o money and bidirecional causaliy beween money and prices. However, now he ECM indicaes he bidirecional causaliy beween money and prices in he long run. The findings of his sudy, a unidirecional causaliy from income o money and bidirecional causaliy beween money and prices, are in line wih hose of Khan and Siddiqui (1990) who found similar resuls using quarerly daa from 1972:I o 1981:IV. V. SUMMARY AND CONCLUSIONS The objecive of he paper is o re-examine he causal relaionship beween money and he wo variables, i.e., income and prices. For his purpose, annual daa on money (M2), Income (GNP) and Prices (CPI) from o are used. The Granger causaliy and Error Correcion Models are employed aking care of sochasic properies of he variables. The analyses indicae he long run relaionship beween money and oher wo variables, income and prices. The analyses furher sugges a one way causaion

8 8 from income o money indicaing ha probably real facors raher han money supply has played a major role in he growh of naional income of Pakisan. A closer look a he resuls reveals ha income affecs money a hird lag, i.e., he significan impac of income on money appears afer hree years. This resul may be useful in esimaing demand for money where income acs as an exogenous variable. Regarding he causal relaionship beween money and prices, he analyses sugges a boh way causaion beween hem. This implies ha moneary expansion increases, and is also increased by, inflaion in Pakisan. In oher words, he increase in money supply raises he general price level which in urn increases he demand for money which resuls furher increase in money supply. A closer look reveals ha money affecs price afer one year bu price affecs money afer wo years. Hence i is he money ha akes lead in increasing inflaion which in urn cause increase in money supply. The increase in money supply resuled from inflaion is mainly due o, as discussed in lieraure, increase in governmen borrowing o finance is expendiure which in urn resuls furher increase in inflaion. This suggess ha fiscal policy should also be conduced wih care. REFERENCES Abbas, Kalbe (1991) Causaliy Tes Beween Money and Income: A Case Sudy of Seleced Developing Asian Counries ( ). The Pakisan Developmen Review 30: Barh, J., and J. Banne (1974) The Role of Money in he Canadian Economy: An Empirical Tes. Canadian Journal of Economics, May, Bengali, K., A. Khan, and M. Saddaqa (n.d.) Money, Income, Prices and Causaliy: The Pakisani Experience. (Unpublished.) Brillembourg, A., and M. Khan (1979) The Relaionship Beween Money, Income, and Prices: Has Money Maered Hisorically? Journal of Money, Credi, and Banking, Augus, Dereyes, F., D. Sarleaf, and G. Wang (1980) Tes of Direcion of Causaion Beween Money and Income in Six Counries. Souhern Economic Journal, Ocober, Ho, Yan-ki (1982) A Trivariae Sochasic Model for Examining he Cause of Inflaion in a Small Open Economy: Hong Kong. The Developing Economies, Sepember, Joshi, K., and S. Joshi (1985) Money, Income, and Causaliy: A Case Sudy for India. Arhavikas. Khan, A., and A. Siddiqui (1990) Money, Prices and Economic Aciviy in Pakisan: A Tes of Causal Relaion. Pakisan Economic and Social Review, Winer,

9 Lee, S., and W. Li (1983) Money, Income, and Prices and heir Lead-lag Relaionship in Singapore. Singapore Economic Review, April, Pakisan, Governmen of (1998) 50 Years of Pakisan in Saisics. Islamabad: Federal Bureau of Saisics. Pakisan, Governmen of (Various Issues) Economic Survey. Minisry of Finance. Pakisan, Governmen of (Various Issues) Monhly Saisical Bullein. Sae Bank of Pakisan. Phillips, R., and P. Perron (1988) Tesing for a Uni Roo in Time Series Regression. Biomerika, June, Schwer, W. (1987) Effecs of Model Specificaion on Tess for Uni Roos in Macroeconomic Daa. Journal of Moneary Economics, July, Sims, C. (1972) Money, Income, and Causaliy. American Economic Review, Sepember, Williams, W., C. Goodhar, and D. Gowland (1976) Money, Income, and Causaliy: The U.K. Experience. American Economic Review, June,

10 10 ABSTRACT This paper re-examines he causal relaionship beween money and income and beween money and prices in Pakisan using a longer annual daa se from o and employing Granger causaliy and Error Correcion Models. We also invesigae he causal relaionships hrough rivariae approach. The analyses indicae he long run relaionship among money, income, and prices. The analyses furher sugges a one way causaion from income o money implying ha probably real facors raher han money supply has played a major role in increasing Pakisan s naional income. Regarding he causal relaionship beween money and prices, he analyses sugges a boh way causaion beween hem. This implies ha moneary expansion increases, and is also increased by, inflaion in Pakisan.