화폐수요의 안정성에 관한 연구- 공적분과 오차수정모형을 중심으로

Abstract

본고는 화폐수요와 국민소득, 이자율, 환율과의 관계를 분석함으로서 과거의 통화금융정책을 재조명하여 향후의 정책에 이론적인 근거를 제공하는 것을 목적으로 한다. 이러한 목적을 달성하기 위해서는 화폐수요가 안정적이어야 한다. 그것은 화폐수요함수가 불안정할 경우 균형화폐수요로부터의 괴리는 더욱 증폭되고, 이러한 상황에서 실시된 통화정책은 경제적 교란을 더욱 확대시킬 가능성이 크기 때문이다. 이에 따라 본고는 M1, M2, M3 3개 통화량의 실질잔고에 대해 실질GDP, 이자율(콜금리와 상업어음할인율), 실질환율을 이용한 4개의 모형을 구성하여 이들 모형들이 안정적인가를 조사하였다. 그러나 모형의 안정성을 조사하기 위해서는 투입변수들의 안정성을 체크하여야 하므로 이를 위해 Dickey-Fuller의 단위근 검정을 실시하였다. 단위근 검정결과 투입변수들의 수준변수가 각기 다른 시차에서 단위근을 갖는다는 가설을 기각하는 데 실패하였다. 그러나 1차차분 시계열은 모두 귀무가설을 기각하여 안정성을 갖는 것으로 나타났다. 이와 같이 안정성을 갖기 위하여 1차차분을 필요로 하는 시계열로 확인됨에 따라, 다음 단계로 시계열간의 선형결합에 대한 분석을 위해 공적분 검정을 실시하였다. EG 공적분 검정 결과 모든 모형에서 잔차가 단위근을 갖는다는 가설을 기각하는 것으로 나타났다. 그러나 EG검정은 낮은 검정력을 비롯하여 몇 가지 문제점을 안고 있기 때문에 Johansen 다변량공적분 검정을 실시한 결과 모형과 통화량에 따라 혼합된 결과가 도출되었다. 이에 따라 GPH 검정기법을 이용하여 본고가 도입한 모형이 안정적임을 알 수 있었다. 이와 같이 모형이 안정적인 것으로 나타남에 따라 화폐수요함수를 추정하여 모든 통화량과 모형에서 실질소득과 환율은 양(+)의 부호로 유의하며, 상업어음할인율은 음(-)의 부호로 유의함을 보였다. 오차수정방정식을 도출하여 오차수정항이 음의 부호로 유의함으로써 공적분 관계가 존재한다는 확인할 수 있었으며, 시장불균형이 상당히 빠르게 조정되고, 단기적인 조정 역할을 이자율이 아닌 실질국민소득이 수행하고 있음을 알 수 있었다. 전반적으로 불균형 조정속도는 m2가 가장 빠르며 그 뒤를 m3와 m1이 뒤따르고 있는 것으로 나타났다. 또한 전향적 이동회귀를 통해 전반적으로 실질소득과 이자율이 화폐수요에 미치는 영향력이 감소하는 경향이 있음을 알 수 있었다. 모형의 예측가능성을 보기 위해 역사적 분해를 실시하여 경제변수들로 구성된 본고의 모형에 의한 예측에 문제가 없다는 것을 알 수 있었다. 그리고 소득증가 충격은 화폐수요를 증가시키고 이자율 상승충격은 화폐수요를 감소시키는 반응을 유발하였으며, m1과 m3의 경우 소득충격의 효과가 일찍 사라지나 m2에 있어서는 소득효과가 비교적 장기간 지속되었다. 또한 이자율의 충격이 소득의 충격보다 크고 오래 지속되고 있으며 전반적으로 충격이 단기에 그친다는 것도 알 수 있었다. 마지막으로 예측력을 비교하여 예측오류는 m1의 경우 모형 2에서, m2의 경우 모형 1에서, m3의 경우 모형 3에서 가장 낮았다. 이와 동시에 소득과 이자율의 영향력이 모형에 따라 그리고 통화량에 따라 약간씩 상이하게 나타나고 있으며, 경제변수 충격에 대한 반응의 지속도와 크기 역시 모형과 통화량에 따라 다르게 나타나고 있다. 따라서 정책을 실시할 경우 한국 경제에 가장 적합한 화폐수요모형을 선택하는 것이 필요하며, 특정한 모형을 통해 통화정책의 효과를 기대하는 것은 무리이라는 것을 밝힐 수 있었다.|A stable money demand function linking real balances, real income and interest rates is essential to many macroeconomic models and monetary policy. For a number of years, stable long-run aggregate demand for real balances has been the central proposition of monetarist models. It is also an important element in New Classical models, in New Keynesian analyses, in some empirical real business cycle models, and in the monetary approach to exchange-rate determination. Money can be a useful tool in the determination of monetary policy only when there is a predictable correspondence between fluctuations in money and fluctuations in income, interest rates, exchange rates, prices or whatever economic variables the central banks seek to influence. If money demand is not stable, then an observed departure of money from ex-ante paths does not bear a systematic implication for future income, interest rates, exchange rates and prices. As such, a proper monetary policy conduct will be misguided ex post and a period of learning might be necessary before the policy reaches its proper course. Moreover, some studies indicate that the issues of stationarity and stability should not necessarily be treated independently. A non-stationary money demand implies that the central bank may not be able to use such a relationship to target money growth accurately. In this case, the central bank would have difficulty in preventing money market disequilibria from affecting the economy. The body of empirical evidence on the issue of stable money demand is mixed. Earlier tests of money demand stability typically centered on whether the coefficient estimates were subject to structural changes. The single-equation partial adjustment model has now largely been abandoned. Linear cointegration analysis has recently been the mainstream approach in examining the long-run behavior of money demand. The primary purpose of this paper is to estimate an appropriate money demand function for Korea and to examine its stability. The data used in this study consist of seasonally adjusted, quarterly observations of real M1, M2 and M3, real output, the real exchange rate, the commercial paper rate, and the call money rate. To achieve the purpose of this study, a number of analytical and technical issues warrant comment. First, the unit root tests suggested by Dickey and Fuller are employed to examine the time-series properties of income, money demand, interest rate and exchange rate. Perhaps the most widely used of these tests in the money demand literature are DF tests. Second issue relates to testing for cointegration among a number of variables. The cointegration test requires that the different economic series are integrated of order one. In other words, the data should be stationary in their first differences but not in levels. To determine the order of integration, this paper uses the procedures suggested by Dickey and Fuller. Each of the different series used in this study was tested for unit roots. The hypothesis of a unit root in the level data is not rejected at the 5 percent critical level across every series. The results using first-differenced data unanimously reject the hypothesis of a unit root. Then this study tests for cointegration on the specification that uses real money balances, real income, interest rate, and exchange rate. The next step in the test for cointegration is to estimate the cointegrating regressions and to conduct unit root tests upon the residuals from these regressions. The augmented Dickey-Fuller test procedure is used to test for the presence of a unit root for the residuals from the cointegrating regression. The lag lengths entering money demand equations are chosen to be shortest for which, using the Box-Ljung Q-statistic, the hypothesis of white noise of the residuals cannot be rejected at the 5 percent significance level. The results show that the null hypothesis of a unit root for the residuals from the cointegration cannot be rejected at the 5 percent level. However, the test, which relies on the Engle-Granger two step methodology, suffers from a number of deficiencies. The Johansen procedure poses several advantages over the popular residual-based Engle-Granger two-step approach in testing for cointegration. However, previous tests, which rely on the Engle and Granger two-step methodology, suffer form a number of deficiencies. The multivariate cointegration technique is superior to the simpler regression-based technique because it fully captures the underlying time series properties of the data, provides estimates of all of the cointegrating vectors that may exist among a vector of variables, and offers a test statistic for the number of cointegrating vectors which has an exact limiting distribution. This test may therefore be viewed as more discerning in its ability to reject a false null hypothesis. The trace test allows us to evaluate the null hypothesis that there are or fewer cointegrating vectors against a general alternative. Looking at the trace test outcomes, it is clear that the null hypothesis ofandare not rejected for all money demand models, whereas the hypothesisis rejected at 5 percent level. These test results suggest that a reliable money demand function exists for all monetary aggregates over time periods examined. Given the presence of a unique cointegrating vector, this then provides us with one error-correction term for constructing model. The empirical results are satisfactory as indicated by the values of determination coefficient andvalue for testing the null hypothesis that all variables are as a group have a zero coefficient. The coefficient on the error-correction term indicates what proportion of the discrepancy between the actual and long-run or equilibrium value of money demand is eliminated or corrected each month. The significance of the lagged error-correction term displaying the appropriate negative sign supports the cointegration findings and implies a valid equilibrium relationship between the variables in the cointegrating equations. This means that excluding the cointegration relationship would lead to misspecification in the dynamic structure of the model. The results show that income has positive effects on money demand but the increase of interest rate reduces money demand in the short term. This study also employs the impulse response functions to get additional information regarding the responses of the variables to the shocks in the other variable. These impulse response functions show the effect of a one standard deviation shock applied to one of the equations, on both the short and long-run responses of all variables in the system. Impulse response functions essentially map out the dynamic response path of a variable due to a one-period standard deviation shock to another variable. The impulse response functions indicate that while the money demand increases sharply before peaking one quarter after the shocks to income and declines quickly to its pre-shock level, it responds negatively to the shocks in interest rate. However, the shocks are small and shortlived.