Back to Current Issues 
Exchange Rate Regimes and the Linkage between Money and Output in GreeceAthanasios P. Papadopoulos and Gregory T. PapanikosUniversity of Crete, Department of Economics Rethymno, GR
741 00 GREECE

i)  Friedman and Schwartz (1963) have shown that a positive correlation exists between innovations in nominal money stock and innovations in real output. 
ii)  Sims (1972, 1980) pointed out that output innovations lag innovations in the nominal money stock. 
iii)  The inclusion of the interest rate in these studies indicates that their innovations lead the innovations in nominal money stock and real output, (Sims, 1980, Litterman and Weiss, 1985). 
The abovestylised facts demonstrate a definite link between money and output whatever the exchange rate regime, but they also suggest that the transmission mechanism has to be examined. In this paper, the maximum likelihood method proposed by Johansen is used to estimate cointegration vectors, i.e. identify longrun relationships. Johansen's method is not the only procedure to test for the existence of cointegrated variables. However, as Dickey, Jansen and Thornton (1991) have argued, Johansen's cointegration tests outperform other tests. Although this procedure does not explicitly require a structural model, recent studies have developed such models that relate the structure of the economy to the cointegration method.
The Vector Error Correction Model proposed by Johansen can be represented by the following equation:
X_{t} = + _{1} X_{t1} + _{2}X_{t2} ... + _{k1}X_{tk+1} + X_{t1} + D_{t} + e_{t}
where
X:  a vector of difference stationary variables (output, money, price, trade balance, and exchange rate) 
:  a vector of constant terms 
D:  seasonal dummies 
e:  a mean zero i.i.d. term 
The shortrun dynamics of the variable are captured by the terms _{1}X_{t1} + /sub>X_{t2} ... + _{k1}X_{tk+1} while the longrun relations between the variables are captured by the coefficient matrix . This matrix is factored into '. Matrix a consists of the response coefficients of the dynamic adjustment of the X variables. The columns of matrix represent the cointegrating vectors.
The empirical process involves the following five steps^{[6]}. The first step involves the test for unit roots because only variables with comparable stationary properties can be co integrated. The second step consists of the determination of the optimum tag for each period in the VAR. This involves the estimation of the first differenced variables of the VAR with alternative tag lengths. The Akaike Information Criterion (AIC) is utilised to determine the significant number of tags to include in the regression. The third step contains the estimation of the cointegrating vectors. The fourth step uses Johansen's likelihood ratio trace criterion to test for the significance of the cointegrating relations. The fifth step comprises the estimation of variance decompositions for all variables in the system (output, money, price, trade balance, and the exchange rate). This step estimates the proportion of forecast error variance of one variable that can be explained by its own innovations and by the innovations of other variables. For example, a central issue of this study is to infer whether there is a difference between the proportion of money innovation that explains the output forecast error variance under flexible and fixed exchange rate regimes. This issue can be addressed by variance decomposition analysis.
Greece participated in the Bretton Woods System from 1953. In this system, the drachmae performed well, being fixed at the arranged parity vis a vis the dollar, with no deviations above the world rate of inflation and no deficit pressures in the balance of payments. This period was characterised by tight monetary and fiscal policies in an attempt to maintain the exchange rate target (see Papadopoulos, 1991). The disintegration of the system allowed the policy makers to disengage the drachmae from the dollar and reverse the policy stance towards expansionary policies. This has resulted in high inflation rates and persistent current account deficits.
The emergence of the European Monetary System provided the opportunity to return to a form of quasifixed exchange rate. However, Greece did not the Exchange Rate Mechanism (ERM) in its early stage, although it has been a full member of the European Economic Community since 1981, with the drachmae participating in the basket of ECU currencies since 1984^{[7]}. This decision is attributed to politicians who argued forcefully that ERM participation was both economically and politically undesirable because of the high inflation differentials between Greece and the rest of the EEC countries.
The Central Bank of Greece, by disengaging the drachmae from the dollar in November 1973, implemented defensive policy measures by adopting a form of a crawling peg against a basket of currencies in an effort to maintain the competitive position of the traded goods sector of the economy. This has been verified by Katseli's (1983) investigation of on the exchange rate policy in Greece during the 1970s. In addition, it was thought that the speculative pressures in the home currency could be prevented, if the instituted restrictions on capital flows were retained. However, this policy has been unsuccessful in securing either the drachmae's competitiveness e.g. Karfakis and Moschos, 1989; Karfakis, 1991) and therefore the external position of the country, or in inducing convergence in the rate of inflation to the rest of the EEC countries. Finally, Brissimis and Leventakis (1984) argue that in Greece monetary disturbances have a significant effect on exchange rate changes.
In this study monthly data have been used for the Greek economy for the period 1960.1  1994.03. Details for data definitions and sources are found in Appendix I. The sample is separated into two periods that distinguish the fixed exchange regime from the crawling peg one. The collapse of Bretton Woods on August 1971 is regarded as the critical point for splitting the sample. From this point until March 1975, Greece did not disengage from the dollar, but the Greek currency fluctuated against other major currencies.
Real output y is proxied by the Industrial Production Index and the price level measure p corresponds to the Consumer Price Index. The money supply aggregate used is either m1 or m1q. The exchange rate er corresponds to the average market rate of drachmae per dollar and the trade balance b is defined as the foreign trade balance in drachmae. Nominal interest rates have demonstrated little variation for both periods, due to administrative controls, and therefore they were excluded.
The first step of the empirical process involves a test for unit roots. This is necessary because the cointegration tests can be applied only to variables that are nonstationary in levels (contain a unit root). Dickey and Fuller (1979, 1981) and Said and Dickey (1984) have developed a method to determine whet er a variable contains a unit root. This method involves the estimation of the following equation:
X_{t} = + t + X_{t1} + e_{t}
The DickeyFuller (DF) test uses the tstatistic from OLS to test the null hypothesis that the y coefficient is equal to zero. If additional lags are included in the above equation, it takes the form
X_{t} = + t + X_{t1} + _{i} X_{t1} + e_{t}
and then the test becomes the Augmented DickeyFuller test (ADF). The variable X is stationary if y is negative and significantly different from zero.
Tables 1 and 2 provide the DF and ADF statistics for unit roots for the fixed and flexible exchange rate periods respectively. In each table, test results are presented for the variables in levels and in first differences, with and without a trend. The number of lags required to remove autocorrelation from the residuals is also reported. Six variables are tested for unit root: two measures of money supply (Ml and Mlq), output, price, trade balance and the exchange rate. The exchange rate is excluded from the first period because it was fixed due to Bretton Woods.
The tests do not reject the hypothesis of a unit root for the relevant variables in each period. However, the results reject the hypothesis for a second unit root. This suggests that all variables are integrated of order one, I(1), i.e. their first differences are stationary.
Cointegration implies that stationary linear combinations of nonstationary variables exist. The central concept of cointegration is the specification of models that include the long run movements of one variable relative to others. The cointegration tests require an appropriate VAR specification among the variables of interest.
In this work, the selection of alternative ordering in the variables of the VAR relies on different theoretical assumptions and empirical applications. The first model is based upon the assumptions of the monetary approach to the balance of payments and the theory of the exchange rate. According to this theory, in a fixed exchange rate regime, the price level is determined by the rest of the world and the money supply becomes fully endogenous. Therefore, deviations from the world price level create balance of payments disequilibrium. Accordingly, the ordering is as follows: price (p), money supply (m1) or (m1q), output (y), trade balance (b). In the flexible exchange rate period, the ordering takes the form: exchange rate er, p, ml (mlq), y, b. This is because during the second period, policy makers have used the exchange rate as an instrument to stabilise external imbalances. Therefore, the exchange rate is not determined by the market forces, but it is determined by factors considered important by the policy makers. Three more orderings are used as alternatives to the monetary model. First, a traditional classical model is used with ordering ml (m1q), er, b, y, p. Second, a model similar to Sims (1980) with ordering m1 (m1q), p, y, b, er. Finally comes the real business cycles model with the following ordering: y, b, er, p, ml (m1q).
Table A presents a summary of the models and their predictions. The monetary model predicts different effects under alternative exchange rate regimes, while the other models anticipate different moneyincome causation independent of the exchange rate regime.
The determination of the optimum lag length for each period in the different VARs is rationalised with the utilisation of the Akaike Information Criterion. Tables 3 to 6 report results of the cointegration analysis for the two periods and for optimum lags in the monetary VARS. Using the tables provided by Hansen and Juselius (1995) the hypothesis of no cointegration is rejected at the 5 percent confidence level for both periods.
The maximal eigenvalue test and the trace statistic provide evidence that at least one cointegrating vector exists among the variables in each period and for the given tagged structure. The existence of cointegrating vectors allows for further tests regarding the elements of the P and a matrixes.
Tests for variable inclusion in the cointegrating vector ( 0) or not ( = 0) for both periods using ml and mlq respectively, were carried out^{[8]}. The hypothesis that an individual variable does not belong to the cointegrating vector is rejected. Thus, a longrun equilibrium relationship exists between the set of variables included in the present analysis. Tests for weak exogeneity ( = 0) for the two definitions of money supply and for both periods were conducted^{[9]}. With two exceptions, the hypothesis of weak exogeneity can be rejected in all circumstances. First, the hypothesis of weak exogeneity cannot be rejected for the trade balance in the flexible exchange rate period when m1 is used. Second, if m1q is used, then the hypothesis of weak exogeneity for the mlq cannot be rejected for money during the fixed exchange rate period. This implies that these variables are weakly exogenous to the system and can enter on the righthand side of the vector error correction model (VECM). Therefore the equations containing b and m1q provide no information about the long run relationships for the flexible and fixed exchange rate period respectively. For the remaining variables, the longrun relationships among them are further reinforced by the variance decomposition analysis presented below.
The cointegrated variables are related in a unique way in the sense that the forecast error variance of one variable is related to its own innovation and the innovations of other variables. How strong this relation is depends on the proportion of the forecast error variance that can be explained by the innovation of a specific variable. Since the I(1) variables are cointegrated, the appropriate model for estimation is a VECM, i.e. a VAR in first differences with the vector of cointegrating residuals as an additional set of righthandside variables. The VECMs are estimated with the appropriate lags suggested by the AIC.
The results of the variance decompositions are very sensitive to the ordering of the variables. In this study, four orderings are used that correspond to what are considered a monetary model, a classical model (CM), Sim's (1980) model (S) and a real business cycle (RBC) model. The variance decomposition results will also explain the possible linkages between real and nominal variables of the model. Before the presentation of the results, taking into account the outcome of weak exogeneity of m1q during the fixed exchange rate period and b during the flexible exchange rate period, it is suggested that the interpretation of the models' behaviour must be treated with caution, for at least the monetary model during the fixed exchange rates period. A thorough look at the results^{[10]} suggests the following.
The variance decompositions of output show that money has a greater effect under a flexible exchange rate regime, though the magnitude of the effect differs and depends on the definition of money used and the ordering of the variables. The wider definition of money has a greater impact. In the CM model, 22.2% of output forecast error variance can be explained by money innovations under a flexible exchange rate regime and only 7.7% under a fixed exchange rate regime. This result is consistent with the monetary approach. If a monetary model is used, then, when ml is utilised, there is no change in the effect of money innovations on output. This is probably due to the money demand shifts (ml) as reported in Papadopoulos and Zis (1997). If mlq is used, then, under a flexible exchange rate regime, money innovations explain 16% of the output variance and 7% under a fixed exchange rate regime. Taking into consideration the nonavailability of an interest rate variable, money innovations would have explained a smaller proportion (less than 16%) of the output variance. Furthermore, output is affected by unexpected price level variations especially in the period under flexible exchange rates in all models, independent of the money definition. The quantitative effect is significantly higher in the real business cycle model. On the other hand, the effect of exchange rate variations on output is small, amounting to 5% 10%, when mlq is used in the CM and in the RBC modes. Finally, output is nearly equally and significantly affected by the balance of trade variations in all of the models using any money definition and exchange rate regime. The only exception is the CM model when mlq is used.
Forecast error variances of external variables (trade balance and the exchange rate under a flexible exchange rate regime) are not affected much by money innovations. The greatest impact occurs when mlq is used and under a flexible exchange rate regime (16.2% for the trade balance and 6.2% for the exchange rate). A surprising result is obtained when ml is used. The effect on trade balance is greater under a fixed exchange rate regime. One possible explanation for the small or opposite effect of money innovations on real and external variables might be the role money played during the two periods due to foreign capital mobility restrictions imposed by monetary authorities which limited the effect of money adjustments on any trade balance disequilibrium. The opposite effect can be explained by the component of mlq that is not included in the ml definition (e.g. savings account).
Variance decompositions of the money supply show that a large proportion of money variance is explained by its own innovation under fixed and flexible exchange rate regimes with the exception of the RBC Model under fixed exchange rates.
Finally, variance decompositions of the price level indicate that money innovations significantly affect the prices in the flexible exchange rate period in all models with the exception of the S model, when m1q is used. In the same period, unanticipated output innovations have considerable effects on price variations. On the contrary, prices were affected by unexpected balance of trade innovations during the fixed exchange rate period in all of the models.
This paper used cointegration analysis and variance decompositions to examine the impact of money on real domestic and external variables under a fixed and a flexible exchange rate regime. The application of unit root and cointegration tests indicates that despite the exchange rate regime, a longrun equilibrium relationship exists among output, money, the trade balance and price. Moreover, variance decompositions predict a larger effect of money on output under a flexible exchange rate regime relative to a period of fixed exchange rates. However, the extent of the impact depends on the definition of money and the ordering of the variables. A wider definition of money has a larger impact. In addition, monetary variables do not explain much of the variance of the forecast error of external variables, trade balance and the exchange rate under the flexible exchange rate regime. The impact is larger when a wider definition of money is used. The overall empirical results do not entirely support any theoretical model but this may be due to the role of policymaking shifts throughout the period.
It is worth noting that the findings of longrun relationships are robust during this period because of the occurrence of policymaking shifts and the international disturbances^{[11]}, thus making them important for policy prescriptions. The empirical evidence presented in this paper supports the view that price innovations and output innovations are outcomes of monetary disturbances under the flexible exchange rate period. This suggests that "Money matters" and Greek policy makers should adopt a pragmatic approach in attaining the economic targets.
The option of the Greek policy makers for drachma participation in the ERM II is on the right track. However, they have to redefine the scope of the monetary policy by shifting it from the role of an instrument which attains the simultaneous inflation and exchange rate targets. In this context monetary policy should focus on maintaining its pursuit to of exchange rate target. The existence of a technology commitment (ERM 11) provides credibility to the policy makers and renews the expectations for a rapid reduction in inflation that is comparable to the Euro countries' rates. As Apergis (1997) has empirically demonstrated, changes in the exchange rates has an impact on the diversification of monetary holdings by the economic agents. The intemationalisation of financial markets makes this diversification much stronger, making it more difficult for national authorities to implement an efficient monetary and/or fiscal policy.
Although the role of fiscal variables have not been examined in this work, it can be argued that recent research demonstrates that the cumulative deficits in the light of the late fiscal developments in Greece, do not undermine the stabilisation process followed since 1992 to attain the criteria set by the Maastricht Treaty^{[12]}. With monetary policy being confined to serving exchange rate targets, the policy makers are asked to achieve low and stable inflation rates, by continuing the present restrictive fiscal stance and taking policy measures that assist the liberalisation of tabour markets in an effort to reduce the unit tabour cost.
Alexakis, P. and Apergis, N. (1994) The FeldsteinHorioka Puzzle and Exchange Rate Regimes: Evidence from Cointegration Tests. Journal of Policy Modeling 16: 459 472.
Alogoskoufis, George S. (1982) "Unanticipated Money, Output and Prices in Greece", Eu ropean Economic Review, 19(2/3) 289303.
Apergis, N. (1997) Domestic and Eurocurrency Yields: Any Exchange Rate Link? Evidence from a VAR Model. Journal ofpolicy Modeling 19:4149.
BahmaniOskooee, M. and Alse, J. (1995) Is there any LongRun Relation Between the Terms of Trade and Trade Balance. Journal of Policy Modeling 17: 199205.
Brissimis, S. N. and Leventakis, J. A. (1984) "An Empirical Inquiry into the ShortRun Dynamics of Output, Prices and Exchange Market Pressure", Journal oflnternational Money and Finance, 3:7589.
Cifarelli, Giulio (1995) "Fundamentals, Regime Shifts, and Dollar Behavior in the 1980s", Open Economies Review 6:2948.
Cuthbertson, Keith, Hall, Stephen G. and Taylor, Mark P., (1992) Applied Econometric Techniques, New York: Philip Allan.
Dickey, David and Fuller, Wayne (1979) "Distribution of the Estimators for Autoregressive Time Series with a Unit Root" Journal of the American Statistical Association, 74: 427431.
Dickey, David and Fuller, Wayne (1981) "Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root", Economettica, 49: 10571072.
Dickey, David, Jansen, Dennis and Thorton, Daniel (1991) "A Primer on Cointegration with an Application to Money and Income", Federal Reserve Bank of St. Louis Review 73: 5878.
Edwards, Sebastin (1994) "Real and Monetary Determinants of Real Exchange Rate Behavior: Theory and Evidence from Developing Countries" in John Williamson's edition Estimating Equilibrium Exchange Rates Washington, DC: Institute for International Economics, pp. 6192.
Faruqee, Hamid, (1995) "LongRun Determinants of the Real Exchange Rate: A StockFlow Perspective", IMF Staff Papers, 42:80107.
Flood, Robert P., (1979) "Capital Mobility and The Choice of Exchange Rate System", International Economic Review, 20(2): 405416.
Frenkel, Jacob A., Goldstein Morris, and Masson Paul R., (1991) Characteristics of a Successful Exchange Rate System, International Monetary Fund, Occasional Paper No 82.
Friedman, Milton, (1953) "The Case for Flexible Exchange Rates", in Essays in Positive Economics, Chicago: University of Chicago Press, 1953.
Friedman, Milton, and Schwartz Anna (1963) A Monetary History of the United States. Princeton, NJ: Princeton University Press.
Hansen, Henrik and Katarina Juselius, (1995) CATS in RATS. Cointegration Analysis of Time Series, Institute of Economics, University of Copenhagen, 1995
Isard, Peter (1995) Exchange rate Economics, Cambridge: Cambridge University Press, 1995.
Johansen, Soren, (1988) "Statistical Analysis of Cointegration Vectors", Journal o@f Dynamics and Control, 12:231254.
Johansen, Soren, (1991) "Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models", Econometfica, 59:15511580.
Johansen, Soren, (1992) "Determination of Cointegration Rank in the Presence of a Linear Trend", Oxford Bulletin ofeconomics and Statistics, 54:383397.
Johansen, Soren and Katarina Juselius, (1990) "Maximum Likelihood Estimation and Inference in Cointegration  With Applications to the Demand for Money", Oxford Bulletin ofeconomics and Statistics, 52:169209.
Joyce, Joseph P. and Kamas Linda, (1994) "Money and Output Under Alternative Exchange Rate Regimes in the USA", Journal of International Money and Finance, 13:679 697.
Karfakis, Costas J., (1991) "A model of exchange rate policy: evidence for US dollarGreek drachma rate 197587 ", Applied Economics, 23, pp. 815820
Karfakis, Costas J. and Moschos Demetrios, (1989) "Testing for Long Run Purchasing Power Parity: A Time Series Analysis for the Greek Drachma", EconomicLetters, 30(3), pp. 24548.
Katseli, Louka, (1983) "Adjustment and exchange rate policy in middle income countries: Greece and Spain in 1970s", in International Economic Adjustment, (ed) M. de Cecco, Blackwell, Oxford.
Litterrnan, Robert, and Laurence Weiss, (1985) "Money, Real Interest Rates and Output: A Reinterpretation of U.S. Postwar Data." Econometfica, 53:12956
McKinnon, Ronald 1, (1981) "The Exchange Rate and Macroeconomic Policy: Changing Postwar Perceptions", Journal of Economic Literature, June 1981, 19:53557.
Mussa, Michael L., "Exchange Rates in Theory and in Reality", (1990) Essays in International Finance (Department of Economics, Princeton University), No. 179.
Papadopoulos, A. P. and Sidiropoulos, M. (1999) "The Sustainability of Fiscal Policies in the European Union: Evidence from Selected Countries", International Advances in Economic Research, forthcoming.
Papadopoulos, A. P. and Zis G. (1997) "The demand for money in Greece. Further empirical evidence and policy implications", The Manchester School of Economic and Social Studies, 65(l) pp. 7189.
Papadopoulos, A. P., (1991) "An Interpretation of PostWar Stabilization Policies in the Greek Economy: 19531981 ", The Journal of European Economic History, 20:187 205.
Papadopoulos, A. P., (1993) "The Effect of Monetary, Fiscal and Exchange rate policies on Output, Prices and the Balance of Payments in the Open Economy of Greece: 1955 90 ", Applied Economics, 25:879890.
Said, Said and Dickey, David (1984) "Testing for Unit Roots in Autoregressive Moving Average Models with Unknown Order", Biomettica, 71: 599607.
Sims, Christopher (1972) "Money, Income and Causality." American Economic Review, 62:540542.
Sims, Christopher (1980) "Comparison of Interwar and Postwar Cycles: Monetarism Reconsidered." American Economic Review, 70:250275.
Taylor, Mark, (1 995) "The Economics of Exchange Rates", Journal of Economic Literature, 33:1347.
er = exchange rate, market rate of drachmas per dollar, IFS line aa. Log form.
m1 = Notes and Coins plus demand deposits in circulation, OECD, Main economic indicators, Greece. Log form.
m1q = Ml plus quasimoney, OECD, Main economic indicators, Greece. Log form.
y = Industrial Production at constant 1980 prices, N.A.G., OECD, Main economic indicators, Greece. Log form.
p = Consumer Price Index at constant 1985 prices, OECD, Main economic indicators, Greece. Log form.
b = Foreign Trade Balance in Greek currency, OECD, Main economic indicators, Greece.
BoG = Bank of Greece, Monthly Statistical Bulletin, various issues, and Time Series of the Greek Economy, 1989.
N.A.G. =National Accounts of Greece, various issues. I.F.S. = International Financial Statistics.
O.E.C.D. = Organisation of Economic Cooperation and Development
^{[1]} See for example Alogoskoufis (1982), Papadopoulos (1991,
1993) and references cited.
^{[2]} A recent exception is the study by Edwards (1994) that
used quarterly Greek data from 1960 to 1985 but they were pooled into a
group of 12 countries.
^{[3]} McKinnon (1981) examines the developments in the 1970s.
See also the survey by Mussa (1990).
^{[4]} An extended version of the monetary approach with capital
mobility has been developed by Flood (1979).
^{[5]} Frenkel, Goldstein and Masson (1991) concluded that the
choice of a successful exchange rate system should: (a) link price
stability to a nominal anchor; (b) facilitate international adjustment due
to trade imbalances, and (c) allow for different exchange rate regimes.
^{[6]} The applied econometric techniques of the cointegration
analysis are presented by Cuthbertson et al (1992).
^{[7]} Greece joined ERM on March 1998, after a 14% devaluation
of the Greek drachma against the ECU.
^{[8]} Details are available on request, or in the electronic
edition of the Journal.
^{[9]} See footnote 8
^{[10]} See footnote 8
^{[11]} Specifically in 1987 the "hard Drachrna" policy
replaced the "dirty floating" regime which in addition to the
gradual depreciation of the Drachma involved two large devaluations in
1982 and 1985, and the 199293 ERM turmoil.
^{[12]} See for example Papadopoulos and Sidiropoulos (1999)