Asymmetric Shocks Patterns in the Central African Economic and Monetary Community

Assessing the economic efficiency of countries’ participation to a currency union has become a relevant topic since the introduction of the Optimum Currency Area (OCA) theory by Mundell (1961). This paper attempts to evaluate the performance of the Central African Economic and Monetary Community (CAEMC) as a currency union in the context of exposure to asymmetric shocks. We first identify structural macroeconomic shocks within the region using the Blanchard and Quah Method. We find that aggregate demand shocks fluctuations display more symmetric patterns than those of aggregate supply shocks. Chad is the apparent outlier, as it is the only economy in the monetary union to experience negative supply shocks. This suggests that the loss of monetary sovereignty might result in significant adjustment costs.

rate uncertainty and transaction costs between countries comes only if intraregional trade is substantial and economic integration high. Finally, the Kenen (1969) criterion implies that the impact of sector specific shocks is reduced if countries produce a wide range of products and therefore have a low degree of specialization. In the same vein, Kalemli-Ozcan, Sorensen and Yosha (2001) compare a cross-section of European countries to a cross-section of US states and find that greater industrial specialization leads to lower synchronization of GDP fluctuations, thus creating more asymmetry.
Unlike the EMU, studies on the CFA franc zone monetary integration are scarce, the reason being the lack of availability of economic data in the least developed CFA franc zone countries. In this context, assessing the degree of openness in trade for the region becomes challenging as much of the cross-border trade between countries goes unreported. In his attempt to apply the OCA theory to the franc zone, Strauss-Kahn (2003) notes the lack of production diversification of member countries, as most of those are specialized in exporting raw commodities. The same goes for economic integration, as intra-regional trade accounts for 6 percent of total trade in CAEMC and 12 percent in WAEMU.
Besides he confirms Gurtner's (1999) observation that heterogeneity of production remains very high, the export base very narrow and intraregional trade remains very low.
Empirical testing using various techniques thus far has shown that the Central and West African monetary unions do not fit criteria designed to reduce exposure to asymmetric shocks, which could threaten the sustainability of the CFA franc zone. The common monetary policy may differ sharply from the optimal policy one country should implement in case of idiosyncratic shocks and the only country-specific response available is through fiscal policy. Even in this case, Chambas (1994) finds francophone Africa fiscal instruments too unwieldy to be considered as reliable stabilization tools.
Thereby, testing that the zones have developed a functioning adjustment mechanism to absorb imbalances and smooth adverse macroeconomic shocks can help one determine whether or not the CFA zone is at the very least sustainable. In this regard, Gurtner (1999) finds it difficult to evaluate the degree of labor mobility of both zones, and conclude that neither WAEMU nor CAEMC fit the classical criteria of OCA designed for adjustment facilitation, as defined in the literature. Devarajan and Boccara (1993) came to the same conclusion, and show that factor mobility is low within the CFA franc zone.
Some searchers share this view that even if the CFA franc zone cannot technically be considered as an Blanchard and Quah (1989) permanent and transitory shocks decomposition of output inspired many existing papers on the identification of structural shocks. Bayoumi and Eichengreen (1992), use this approach of VAR decomposition when analyzing data on output and prices for eleven European countries to identify aggregate supply and demand disturbances. In order to identify all relevant sources of output fluctuations, they impose a set of restrictions on the dynamic responses of variables in the long-run. In this study, a reduced form VAR is estimated for output and prices and the estimated innovations in this VAR will be interpreted as aggregate supply and demand shocks, motivated by the traditional Keynesian AD-AS model view of fluctuation.

Methodology and Estimation Techniques
The reason for choosing changes in output (real GDP growth) and changes in prices (inflation) as variables along the lines of Bayoumi and Eichengreen (1992) instead of output and unemployment as Blanchard and Quah (1989) has to do with the fact that in the theoretical framework we use for interpretation, the AD-AS model, aggregate demand and aggregate supply are functions of output and price. A positive shock to aggregate demand (AD) causes the downward sloping curve to shift to the right, causing both output and prices to rise. As the long run aggregate supply (LRAS) curve reaches the natural level of production, output level is no longer able to meet the demand, as a result prices rise even higher. Therefore, demand shock has a transitory effect on real output and price level, also a permanent effect on prices but no permanent effect on output. In the short run, a positive supply shock causes the upward sloping curve (SRAS) to shift to the right along the demand curve (AD) and causing real output to rise and price level to fall. The same dynamic is kept in the long run (LRAS) as supply shocks unlike demand shocks do have a permanent influence on output. We notice that demand and supply disturbances have opposite effects on prices, whether they are permanent or just transitory.
Using the impulse response functions (IRF) generated from an inflation-output VAR to calculate the dynamic effects of a shock to inflation and a shock to output is the common intuition when it comes to study the effect of shocks on the economy. However, it will be more realistic to consider that the shocks generating inflation and output are an aggregate supply shock and an aggregate demand shock and that both of these shocks have a direct effect on both inflation and output. Then, the procedure for decomposing the shocks in order to separate aggregate supply shocks from aggregate demand shocks, as well as their impulse response begins by supposing that structural shocks and reduced-form VAR shocks are related. Bayoumi and Eichengreen's (1992) representation links the influence of output and prices to each other, assuming that ε t is a vector of serially uncorrelated white noise disturbances that represent unexplained components in output growth and inflation movements, thus interpreted as structural shocks. They can be written as linear combinations of supply and demand shocks: In matrix form, we have e t = Cε t , where ε t YD and ε t PD represent the part of the disturbances that is explained by demand shocks in output growth and inflation respectively; ε t YS and ε t PS denote the part of the disturbances that is explained by supply shocks in output growth and inflation respectively. In order to identify the structural shocks, we have to estimate first a structural vector autoregressive (SVAR) model derived from the chosen underlying economic theory. It is likely that there are contemporaneous interactions between variables, thus we shall consider the following bivariate dynamic simultaneous equations for output and price level: Let X t = (Y t , P t )′ denote the vector of endogenous variables where Y t is the real GDP growth and P t is the inflation rate. The dynamic structural representation of the model is written as follows: Where A = [ 1 -a 12 -a 21 1 ] ; φ is a vector of constants; B k represents a 2x2 matrix of structural coefficients and e t denotes a vector of serially uncorrelated structural shocks with E(e t e t ′ ) = E(e Y e P ) E (e P 2 ) ) = I as the covariance matrix. However, because the disturbances are uncorrelated and independent from each other, it makes it difficult to obtain unbiased estimates using the structural form. Consequently, we need to transform the model by multiplying equation 3-4 by A −1 we obtain the reduced VAR to be estimated: Where γ = A -1 φ; C k = A -1 B k ; μ t = A -1 e t and the covariance matrix for the reduced form VAR shocks is E( μ t μ t ′) = A −1 E(e t , e t ′)A −1 ′ = A −1 A −1 ′ = Σ. Thus the observed covariance structure gives us some information about their relation to the structural shocks. Therefore, in order to calculate long-run effect of a given shock on one of the variables, let say Y t for instance, we calculate the sum of its effects on ∆Y t , ∆Y t+1 , ∆Y t+2 and so on. That is to say, the long-run effect of a shock corresponds to the sum of the impulse responses from p periods. We shall then consider the impulse responses for where ∆Y t is the change in output, and ∆P t is the change in prices. The moving average representation of the model is written as follows: The restriction made by Blanchard and Quah (1989) is to set θ as a lower triangular matrix, as only the supply shock has a long-run effect on the level of output.
Now we focus on calculating the Choleski factor of the symmetric matrix θθ ′ . However, estimating the reduced-form VAR will not give us the information we need about the structure of the economy and μ t cannot be referred to as structural shocks. Thus, in order to identifying actual structural shocks, we should use the relation θ = ( I − B(L)) −1 A −1 to identify A −1 . We previously established that μ t = A −1 e t , then e t = Aμ t . Finally, we calculate the impulse response functions to the structural shocks.

Data Description and Sources
When a macroeconomic shock occurs, there is a need for adjustment. There exist two types of adjustment: real or nominal adjustment. Real adjustment happens through quantities, depicted by variables such as output growth and unemployment rate. On the other hand, for adjustment to be qualified as nominal, it should take place through prices, better measured by variables such as interest rate, exchange rate and inflation rate. Therefore, in order to isolate country-specific shocks that are not the result of innovation in monetary policy, but provoked by exogenous aggregate supply and aggregate demand shocks we chose to identify shocks to output growth and inflation. Hence, those variables represent the most important macroeconomic indicators across African countries as stated by Dhonte et al. (1993). Different preferences about inflation and output growth of countries may make the introduction of a common currency costly. A high inflation country will progressively lose its competitiveness by sharing a common currency with a lower inflation country. The same goes with a fast-growing country having to implement deflationary policies, in order to constrain growth, if it forms a monetary union with a slow-growing country. This motivates the comparative analysis of inflation and output growth across CAEMC member states.  Central African States. However, the World Bank is the main source for real GDP growth and prices data series. As displayed in Figure 1, the inflation rate dispersion and growth rate dispersion standard deviation curves have a peak in 1986 and 1997 respectively. It appears that a movement of convergence might exist in the case of inflation with a standard deviation that follows a descending trend, going from 9.89 at the beginning of the sample to 1.65 by the end of the sample. However it is more difficult to observe a movement of convergence in the case of real GDP, because the standard deviation curve does not seem to follow a trend.
Then, we analyze the correlation of output growth and inflation across countries. Tables 1 and 2 display the correlation matrices of output growth and inflation. By observing cross-country coefficients of correlation, we observe that for inflation, correlations are high and only Cameroon and Equatorial Guinea display less than sixty percentage of correlation, confirming the movement of convergence previously observed in Figure 1. However concerning output growth, coefficients of correlation for any pair of countries are very low and even negative, again in concordance with the graph displayed in

Notes.
a. Probabilities for Fisher tests are computed using an asymptotic Chi-square distribution b. Probabilities < 0.05 indicate that the null hypothesis that the variables are cointegrated is rejected.

Dynamic Effects of Structural Shocks
In this study, we use a two-variable VAR for real GDP growth and inflation rate in levels form. The estimated VAR is stationary in the case of each country and zero long-run restrictions have been imposed on the matrix of long-term effects of aggregate demand shocks. The aggregate demand is restricted not to have a permanent effect on output, as long-run output depends entirely on supply factors. However, no restriction is imposed on the long-term effect of supply disturbances on output and price level. We obtain the identified shocks by applying the transformation of section 3 to the reduced form VAR residuals. inflation and output growth in the same direction. This is the case here when price level reacts in accordance with our theoretical framework, meaning it moves following the same direction than the output level. However, there is a puzzle in the case of Equatorial Guinea, because output and price are not moving in the same direction as first sight like the theory suggests in case of aggregate demand shock. Nonetheless, we chose to interpret this particular shock as a demand disturbance because of the non-lasting and very little positive impact it has on the price level, immediately followed by a drop greater in amplitude from the second to the third year. In general, demand innovations have a more lasting impact on output than in prices, as it takes only between two years (the Central African
The present impulse response analysis is informative in order to identify the nature of the shocks, but incomplete. Concerning the amplitude of the disturbances and their real impact on CAEMC countries, we need further observation. However, from this stage, we can already identify an outlier that is Chad which did experienced both adverse demand and supply shocks when the other member states did experiment negative demand shocks and positive aggregate supply shocks, during the sample period.
Further investigation in the next sections will measure the amplitude of the impact as well as the degree of shock asymmetry across countries in the union. Source: Author's computation from time series data

Contribution of Demand and Supply Disturbances
Note. Bold figures denote that the influence of the shock is at least 30 percent greater than the other.
After identifying the nature of the shocks that impacted output growth and inflation in the CAEMC, we proceed on measuring their size and magnitude. Variance decomposition helps determine whether the sources of shock to variables in the models are the same across the region by identifying the sources of variability for each variable in the models for different member states.

Concluding Remarks
Our study aims to identify macroeconomic shocks within the CAEMC, in order to determine whether or not the currency area is exposed to asymmetric shocks, threatening its optimality. First, by observing output growth and inflation impulse responses to structural shocks, we conclude that all the CAEMC countries are hit by adverse demand shocks, causing output and price to move in the same direction.
Five of them, except for Chad, experience positive supply shocks during our sample period. Then, for nearly all the countries within the CAEMC, variance decomposition shows that in accordance with the proportion of variance explained by different shocks, output level is decreasing, except for the Central African Republic where from the second year following impact to the 20 th year, output level is increasing mostly influenced by positive supply shocks. Besides, the proportion of variance in Chad's output growth is explained by adverse supply shocks rather than adverse demand disturbances for Cameroon, Congo, Equatorial Guinea, Gabon on one hand, and the Central African Republic the first year after the shock.
If the Central African Republic do not seem to suffer a lot from adverse demand shocks, as revealed by variance decompositions, Chad appear to be an outlier in the case of aggregate supply being the only country in the region to experience negative supply shocks. Therefore, policies responses aiming to stimulate aggregate demand in order to increase real output, by decreasing interest rates for instance, would likely threaten the fragile price stability in Chad, as the downward pressure of negative aggregate demand shocks is currently counterbalancing the rising inflation rate, caused by the negative supply shocks faced by this economy. There is no easy answer to this situation. Yet, the BEAC has pursued a restrictive monetary policy over the past three years and raised the interest rates to 3.5 percent aiming at maintaining the parity rate between the CFA franc and its reference currency, thus avoiding devaluation. The reason local authorities were able to pursue such a policy in a global context of declining rates, may lie in the current implementation of capital controls to cope with currency crisis in the region. This explains the negative correlation between domestic and foreign real interest rates movements since quantitative restrictions on international capital flows sever any direct link, as demonstrated by Greenwood and Kimbrough (1985). Besides, the presence of excess liquidity in the banking system weakens the monetary transmission mechanism and thus the ability of monetary authorities to stimulate aggregate demand, as demonstrated by Saxegaard (2006) in the case of CAEMC member states.