A Multivariate Long-Run Money Neutrality Investigation: Empirical Evidence for CAMEU

We examine the long run neutrality of money, in Central Africa Monetary and Economic Union (CAMEU) economies applying the multivariate methodology of King and Watson, using M2 and real output during the period 1970-2008. Tests consistently reject the long run money neutrality hypothesis. It is found that M2 has significant and positive impacts on real output of all CAMEU countries except for Gabon. The results are robust under other monetary aggregate variables and various sub-periods. In addition, the estimated coefficients are stable under two breakpoints corresponding to the dates of central bank reforms and devaluation of the local currency.


Introduction
The hypothesis of long-run neutrality of money, i.e., the absence of long-term effects of money supply on the production level is widely accepted in modern macroeconomics since the work of [1][2][3][4].Indeed, as the velocity of money is constant and that of the activity is constrained by the capacity to supply goods, changes in the money supply lead to price changes.This leads [1] asserted that "inflation is always and everywhere a monetary phenomenon".Moreover, according to the authors of the New Classical Economics, perfectly anticipated monetary policy has no effect, even in the short term on the activity.Thus, permanent changes in the level of money supply do not affect real variables in the long term ( [5][6][7]).However, [8] noted that the test of this proposal is subtle since [9] provides an example in which it is impossible to test the long-run neutrality using a reduced form model: it is a rational expectations model integrating rational expectations with non-neutrality in the short term and exogenous variables which follow stationary processes.Thus, the data generated by this model are not likely to contain sustainable changes necessary to directly test the long-run money neutrality hypothesis.
Despite this example, [9] argues that it is necessary to build fully articulated behavioral models to test the neutrality hypothesis.The specific criticism developed focuses on the issue of stationarity.Indeed, in models in which nominal variables are integrated, the long-run neutrality can be defined and tested without a complete knowledge of the model.But even when the variables are integrated, neutrality long term cannot be tested using a reduced-form model.Instead, it is necessary to test the model of "final form", highlighting the dynamic response variables underlying structural disturbances.
Taking into account this assumption, [8] show that the test of neutrality can be built if the nominal and real variables satisfy certain non-stationarity conditions.The fundamental reason is that monetary neutrality induces permanent changes in the level of money and cannot effectively be tested without strong evidence that the current level of money supply is affected by permanent changes.
Several studies used the long-term neutrality testing procedure developed by [8].[10], for example, show that the long-run neutrality can be tested with limited structural information when nominal variables are integrated.Using quarterly data from Japan, Sweden and Italy, they test the long-term neutrality hypothesis between the inflation and the nominal interest rates.Similarly, Chen [11], in order to test the hypothesis of long-term money neutrality using quarterly data for South Korea and Taiwan, pay particular attention to the stationarity and cointegration of the variables involved.They show that the test procedure of long run neutrality cannot be performed if the variables are not cointegrated.Based on the estimated results, the hypothesis of long-run neutrality of money with respect to the actual output is supported in the case of South Korea, but is rejected in the case of Taiwan.
In Africa, [12] analyzes the impact of the hypothesis of money neutrality in the Franc zone.The analytical framework used by the author is the same as mentioned by [8].However, to account for bias due to omission of certain variables, the author introduced in the model some key interest rates and inflation variables.Using quarterly data covering the period 1994:Q2-2006:Q4, the author observes a trend towards long run non-neutrality of money in Senegal and Cote-d'Ivoire.
In this paper, to investigate the hypothesis of long-run neutrality between monetary aggregates and real output in CAMEU 1 , we consider the methodology developed by [8].The paper endeavors to fill a gap as most empirical studies on long-run neutrality of money have focused on industrialized economies with very little attention, if any, to developing countries.
The remainder of the paper is organized as follows: Section 2 develops the multivariate econometric methodology.Section 3 is related to empirical estimation results.Section 4 concludes the paper.

Multivariate Econometric Methodology
First, we consider the simultaneous equations models developed by [8].The model is of order p and is defined in first differences.To estimate the long run effect of nominal money supply on real output where t  and y t  are respectively structural shocks of money supply and real output which can have permanent effects on the levels of the endogenous variables and .
t m t y In vector form, the above equations can be written as, where   Our objective is to analyze the long run effect of monetary shock m t  on the real output t y .We can express this by usin e long run multip g th lier:     In eir study, [8] show that the endogeneity of t m and t y implies that equation ( 3) is non-identified econometrically.This t can easily be seen by re-writing the reduced form as, where . The matrices  and   are obtained from the following equatio (5) ns, Equation ( 5 order to com letely identify the odel.
To get an appropriate estimation of the model, we adopt the approach of [8].They assume known by a parallel study or eq variate VAR: -The impact elasticity ect to the revenue or; -The impact elasticit onetary indicator or; -The long run elastici the output or finally; -The long run ela e same indicator.
In our framework, the estimated value of the long-run elasticity of output with respect to money depends mainly on what is assumed about the elasticities of the other three elasticities: 1) the impact elasticity of production with respect to money; 2) the impact elasticity of money with respect to production; or 3) the asticity of money with respect to the output.We present the results of tests of neutrality for a wide range of

Data and Results
the selected series have unit roots.To verify this, we used three tests: the ADF test (Augmented Dickey-Fuller), the PP (Phillips-Perron) test and the KPSS test (Kwiatkowski-Phillips-Schmidt-Shin).The tests are made on the basis of the model with constant, and constant and trend.

Data
In our study, we use annual data on the actual product at constant prices and the value of the real money supply

Cointegration Tests
Before testing the long run neutrality assumption, it is of interest to examine the existence of a long-term relationship between real output and nominal money.To this end, we use the [13] test.
The results of this test on the actual product and the nominal money supply are shown in Table 2.We use the trace and the maximum eigenvalue tests for the null hypothesis of no cointegration.The results indicate that the null hypothesis is rejected for all countries except for Gabon.These results are also confirmed by [14] cointegration test.Thus, the long run neutrality hypothesis is examined only for Cameroon, Central African Republic, Chad and Congo.

Final Remarks
We examined the lon actual output in the ca g-term money neutrality on the se of Cameroon, Central African Republic, Chad, Congo and Gabon are using annual data covering the period 1970-2008.To this end, we used the approach of [8], with a specific emphasis on unit root and cointegration issues.The results indicate a cointegrating relationship between money and real output only for Cameroon, the Central African Republic, Chad and Congo.
On the other hand, the empirical evidence shows that the assumption of long run money neutrality is rejected fo that characterizes C of increase in real output for each point of increase in the percentage of the nominal money supply resulting from a permanent monetary shock.In this case, the supp money ly long run neutrality implies the condition: 0 ym   .
to zero in the context of a bi of a monetary indicator with resp y of revenue with respect to a m ty of this indicator with respect to sticity of the output with respect to th long run el values for these elasticities, using graphical methods.

Figures 1 - 4
Figures 1-4 present estimates of ym  for a wide range of values of my  (Panel A), ym  (Panel B) and my  (Panel C) for confidence interval at 95%.Long run

Figure 4 .re reported in Figure 2 .
Figure 4. Confidence intervals for Congo.eutrality is not rejected at a level of 5% if n 0 ym   is within the confidence interval at 95%.As for pa it states that when the assumption of long-run neutrality is not rejected, the ellipse covers the actual values of   ,

3. 2 . 3 . 4 . 1 my
Chad sults are reported in Figure 3.The long run lity assumption is once again rejected for money money neutrality for Congo are prere Based on Panel A results, we reject Results for long run sented in Figu the long run money neutrality assumption since 1.

Unit root test in level; (b) Unit root test in first differences.
Notes: Model 3, model with a trend and constant; Model 2, model with only a constant; and Model 1 has no trend and no constant.