In BEAC, the following variables have no unit root in level: M1D, M2D, CPI, TDBEAC; the same result is achieved by M1, M2, M1D, M2D, CPI and TDBCEAO in the BCEAO. As concern the other variables, test statistics shows that the hypothesis of a unit root cannot be rejected at the one or five percent level after one differencing. According to ADF results, we specify a VAR model in level; in the VAR specification, the differentiated form is taken for the variables which are not stationary in level.

The evaluation is made by the variance decomposition analysis drawing from VAR models. In our specification, we considered Sims’s classic 4-variables VAR, consisting in the interest rate, the logged money supply, the logged price level, and logged real GDP in that order. Table 2 and table 3 report the result of variance decomposition analysis respectively in BEAC and BCEAO. The results in table 2 show that the contribution of monetary aggregate to price level variation is less than 15%. At all horizons, inflation rate is better explained by Divisia M2. However, variance decompositions show that innovations in money explain small percentages of the variance of real GDP; but traditional M2 perform better. Table 3 point out that the contribution of money to inflation and GDP variability is worst in the BCEAO. One exception is the percentage of simple sum monetary aggregate which explain price level variability. In fact, M2 explains 15.65%, 19.40%, 22.18% and 24.55% respectively at 4, 6, 8 and 10 quarters. Figures 13 to 20 present the actual and the forecasted values of price level and real GDP from the alternatives aggregates in each Central Bank. For example, figure 13 compares the forecast of price level from the M1 VAR while figure 14 compares the forecast of real GDP from the M1 VAR in BEAC. Financial services markets

Considering figure 14 specifically, it appears that the two M1 aggregates are similar in their ability to predict real GDP. Also, there is no difference between traditional M1 and Divisia M1 in their capacity to forecast price in BEAC (figure 13).

**Table 2: Variance decomposition results, BEAC**

Horizon | Price Level | Real GDP | ||||||

M1 | M1D | M2 | M2D | M1 | M1D | M2 | M2D | |

4 | 10.3017 | 12.6373 | 8.3768 | 13.5339 | 14.9778 | 11.7257 | 16.7692 | 11.8419 |

6 | 13.6883 | 14.1389 | 13.8017 | 14.5561 | 12.3833 | 11.5384 | 13.6203 | 11.4513 |

8 | 14.3484 | 14.2021 | 14.4780 | 14.6007 | 11.4779 | 11.9837 | 12.8557 | 10.5718 |

10 | 14.3629 | 14.3724 | 14.6090 | 14.7612 | 10.3424 | 12.9290 | 11.6844 | 11.5715 |

**Table 3: Variance decomposition results, BCEAO**

Horizon | Price Level | Real GDP | ||||||

M1 | M1D | M2 | M2D | M1 | M1D | M2 | M2D | |

4 | 8.2436 | 0.4422 | 15.6546 | 3.3649 | 2.0627 | 1.6602 | 2.1967 | 1.5199 |

6 | 7.2827 | 0.4190 | 19.4067 | 3.5028 | 2.0833 | 1.7086 | 2.1937 | 1.5174 |

8 | 6.2503 | 0.7283 | 22.1840 | 3.4111 | 2.1032 | 1.7649 | 2.1945 | 1.5242 |

10 | 5.3790 | 1.2286 | 24.5590 | 3.2751 | 2.1136 | 1.8021 | 2.1955 | 1.5276 |

*Figure 13: Price Level Forecasts, M1 BEAC*

*Figure 14: Real GDP Forecasts, M1 BEAC*

*Figure 15: Price Level Forecasts, M1, BCEAO*

*Figure 16: Real GDP Forecasts, M1 BCEAO*

*Figure 17: Price Level Forecasts, M2 BEAC*

*Figure 18: Real GDP Forecasts, M2 BEAC*

*Figure 19: Price Level Forecasts, M2, BCEAO*

*Figure 20: Real GDP Forecasts, M2 BCEAO*

In BEAC, the following variables have no unit root in level: M1D, M2D, CPI, TDBEAC; the same result is achieved by M1, M2, M1D, M2D, CPI and TDBCEAO in the BCEAO. As concern the other variables, test statistics shows that the hypothesis of a unit root cannot be rejected at the one or five percent level after one differencing. According to ADF results, we specify a VAR model in level; in the VAR specification, the differentiated form is taken for the variables which are not stationary in level. The evaluation is made by the variance decomposition analysis drawing from VAR models. In our specification, we considered Sims’s classic 4-variables VAR, consisting in the interest rate, the logged money supply, the