The Dynamic Effects of Monetary Arrangements on Bilateral Trade in the African Franc Zone

Abstract

In this chapter, I estimate the dynamic effects of the sharing of the CFA franc on bilateral exports of member countries of the African Franc Zone (AFZ), distinguishing the results according to its two monetary unions, namely the Central African Economic and Monetary Community (CAEMC) and West African Economic and Monetary Union (WAEMU). While the overall and average effects are well identified in the recent literature, no study has focused on the dynamic effects of monetary integration in Africa. Using data from the World Bank, UNCTAD, and CEPII, I adopt a gravity specification estimated by ordinary least squares (OLS) and the Poisson pseudo-maximum likelihood (PPML) estimator. Our analysis leads to the following results: (i) in CAEMC, the dynamic effects of the CFA franc on bilateral trade of its member countries are delayed, as they are observed from 2010 onward; (ii) in WAEMU, the CFA franc has permanent dynamic effects throughout the study period; and (iii) these results, robust to the use of the PPML, are partially explained by the detour of trade caused by the fact that most of the partner countries belong to other regional groupings. All of these results call for a deep analysis of the future of the AFZ, which requires relevant reforms to ensure its viability and optimality.

1. Introduction

This chapter contributes to the debate on the future of the Franc Zone by assessing the dynamic effects of monetary integration on the trade intensity of countries in this African space. Created in 1939 by France in order to preserve its political pre-square in the world on the eve of World War II, the Franc Zone can be understood as one of the oldest monetary zones in the world. It is made up of geopolitical zones where currencies that were once linked to the French Franc (former colonies or overseas territories) are used and have been linked to the Euro since 1999 by a fixed parity system guaranteed by the French Treasury. These currencies are the result of the monetary cooperation policy of the Banque de France and the Central Banks of the former colonies, which are bound by agreements. The Franc Zone is made up of France (and its overseas territories), the countries of the Economic and Monetary Community of Central Africa—CAEMC (Cameroon, Congo, Central African Republic, Gabon, Equatorial Guinea, and Chad), the countries of the West African Economic and Monetary Union—WAEMU (Benin, Burkina Faso, Ivory Coast, Guinea Bissau, Niger, Mali, Senegal, and Togo), and the Comoros. These countries and groups of countries each use their own currencies.

A unique feature of both currency unions was the involvement of France as the anchor currency country in the monetary policy of the central banks of the WAEMU and CAEMC. France guaranteed the convertibility into their own currency and participated in the executive boards of the central banks with veto power and thus the ability to block any decisions until the adoption of the Euro. In fact, the CFA Franc Zones went beyond the features of a regular currency union. With the devaluation imposed by France in 1994, very similar rules of macroeconomic surveillance to those established in the EMU were introduced and gradually implemented. Nevertheless, while monetary integration is well established, economic integration is still incomplete in the WAEMU and CAEMC areas.

The need to study the effect of the CFA franc on trade in the countries of the African Franc Zone (AFZ) can be justified by the fact that this monetary zone is one of the oldest in the world. This issue has been mainly studied by referring to the Optimum Currency Area (OCA) criteria. However, those studies have been unable to draw clear-cut conclusions on the optimality of the CFA zone. This can be obviously linked to the limitations of the OCA framework in explaining the actual formation of monetary unions. The authorship of the work on the link between the single currency and market integration goes back to Rose [1]. In his analysis, the author demonstrates the explanatory power of the sharing of a single currency on bilateral trade. This path opened by Rose has generated a very fertile field of research. But in Africa, little attention has been paid to the link between the single currency and bilateral trade. However, some recent works confirm globally the existence of the endogenous effects of a single currency on bilateral trade [2, 3].

The major contribution of this paper is to consider the dynamic analysis of the effects of monetary integration on bilateral trade in the AFZ. Unlike previous approaches that focus on the average effect, I adopt an approach that allows this effect to be broken down over time. This makes it possible to identify the periods for which the common monetary history was an essential factor of market integration in the AFZ in order to constitute a rational memory for the future.

Following this introduction, the rest of the paper is organized as follow. Section 2 presents a brief literature review. Section 3 describes the methodology. Section 4 analyses the main findings. Section 5 addresses their sensitivity and Section 6 their robustness. In Section 7, I draw some concluding remarks.

2. Literature review

2.1 The costs/benefits analysis

According to Bean's [4] seminal analysis, joining or belonging to a monetary union is a decision resulting from an optimization of the related benefits and costs. Theoretically, the costs of a monetary union are assessed in terms of loss of national sovereignty, political independence, and cultural authenticity [5]. Furthermore, Belke and Wang [6] identify relative instability as another potential cost. In this regard, rebalancing and stabilization of production and the labor market become difficult to achieve due to the loss of control of the monetary instrument in the conduct of overall economic policy. The abandonment of exchange rate policy [7] and monetary policy as instruments of macroeconomic adjustment by a country exposes it to shocks whose magnitude is proportional to its degree of integration in the union, with the subsequent costs depending on the country’s specific characteristics.

As for the benefits, they are diverse and varied, without claiming to be exhaustive. The direct benefits of a monetary union are related to the cancelation of exchange rate risk and the reduction of hedging costs against this risk [7]. Indirect benefits relate to the synchronization of cycles, which leads to a better response to macroeconomic shocks [8], but also to the intensification of trade [1]. Monetary union can also create a framework that is favorable to the mitigation of inflationary bias [9] and thus ensure price stability [10]. Adopting a typology specific to currency unions, Grubel [5] distinguishes between static gains in terms of reduced exchange costs, lower interest rates and exchange rate risk, and increased welfare and stability, and dynamic gains in terms of expanded trade, better labor market performance, and improved adjustment of economic structures.

2.2 Dynamic and scale effects

According to Rose [1], there has been little work on the dynamic effects of currency unions on trade. They were interested in the nature of the link between the two variables, without addressing the question of the dynamics of this link over time. On the basis of this observation, Katayama and Melatos [11], using the panel dataset constructed by Glick and Rose [12] that covers 217 countries from 1948 to 1997, demonstrate the nonlinear impact of the single currency on bilateral trade. Thus, they show that, contrary to previous studies, the sharing of a single currency does not influence the level of bilateral trade in the same proportion. After him, De Sousa’s [13] study, based on a theoretical gravity model covering a large period (1948–2009), proves that the effect of sharing a single currency on bilateral trade is eroding over time because of the existence of other channels that are commercial and financial globalization.

This result remains robust and confirmed by Miron et al. [14]. The authors restate the result of Rose [1] on the differentiated effects of sharing of single currency and the reduction in volatility of the exchange rate. Moreover, they confirm the hypothesis of the continuous declining effect of currency union on bilateral trade. According to Larch et al. [15], the monetary union effects on trade are dimensional and could be dynamic. Using a structural gravity model, the authors distinguish in the case of the euro zone, bilateral and multilateral effects. They discover that both effects are positive and statistically significant. Globally, this set of results remains consistent with that previously established by Bergin and Lin (2012).

3. Empirical strategy

3.1 Model specification

The empirical framework used in this chapter is the gravity model. This model is based on the Newtonian physics postulate that the force of attraction between two bodies is proportional to the product of their relative masses and inversely proportional to the square of the distance between them. Although this model was introduced by Tinbergen [16], it is Anderson [17], and in particular Anderson and van Wincoop [18], who are responsible for its theoretical foundations, which have been the subject of considerable debate among economists. The analytical framework is a monopolistic model applied to international trade, i.e. a context that assumes increasing returns to scale and product differentiation. This framework is underpinned by three fundamental assumptions: profit maximization by firms in monopolistic competition, utility maximization by consumers, and specialization of the supply of goods between countries [19].

Empirically, the economic formulation of this equation is as follows:

$$X_{\mathit{ijt}}={\varphi }_0\frac{Y_{\mathit{it}}^{{\varphi }_1}.Y_{\mathit{jt}}^{{\varphi }_2}}{D_{\mathit{ij}}^{{\varphi }_3}}{e}^{{\epsilon }_{\mathit{ijt}}}$$E1

$X_{\mathit{ijt}}$ is the bilateral trade flow between two countries in time t, $Y_{\mathit{it}}$ and $Y_{\mathit{jt}}$ are the GDPs, and $D_{\mathit{ij}}$ is the distance between the two countries. The ${\varphi }^{’}$s are coefficients.

In light of Head and Mayer [19] who systematized the foundations and specifications of the gravity equation (naive form, structural form, and multiplicative form), I adopt the following multiplicative general form:

$$X_{\mathit{ijt}}={\varphi }_0{Y}_{\mathit{it}}^{{\varphi }_1}.Y_{\mathit{jt}}^{{\varphi }_2}{\theta }_{\mathit{ij}}^{\gamma }{M}_{i\left(n\right)\left(j\right)}{e}^{{\epsilon }_{\mathit{ijt}}}$$E2

where ${\theta }_{\mathit{ij}}$ is the common characteristics of $i$ and $j$, and $M$ is a proxy for multilateral resistances, which are generally captured by origin country fixed effects, destination country fixed effects, or by time fixed effects [20].

By explicitly noting the effect of sharing the single currency, I retain the following semi-log-linear specification:

$${\mathit{lnX}}_{\mathit{ijt}}=\left\{\begin{array}{c}{\varphi }_0+{\varphi }_1\mathit{ln}{Y}_{\mathit{it}}+{\varphi }_2\mathit{ln}{Y}_{\mathit{jt}}+{\varphi }_3\mathit{ln}{\mathit{Pop}}_{\mathit{it}}+{\varphi }_4\mathit{ln}{\mathit{Pop}}_{\mathit{jt}}\\ +{\varphi }_5\mathit{ln}{\mathit{\text{Dist}}}_{\mathit{ij}}+{\varphi }_6{\mathit{CFA}}_{\mathit{ij}}+{\varphi }_7{\mathit{Dum}}_{\mathit{ij}}\end{array}\right\}+{\mu }_i+{\gamma }_j+{\xi }_t+{\epsilon }_{\mathit{ijt}}$$E3

${\mathit{CFA}}_{\mathit{ij}}$ is the dummy variable equal to 1 if countries $i$ and $j$ belong simultaneously to the Franc Zone and its subregions (respectively ${\mathit{XAF}}_{\mathit{ij}}$ for CAEMC, ${\mathit{XOF}}_{\mathit{ij}}$ for WEAMU, and ${\mathit{CFA}}_{\mathit{ij}}$ for the consolidated zone), and 0 otherwise. ${\mathit{Dum}}_{\mathit{ij}}$ is the vector grouping the dummy variables related to the simultaneous openness to the sea (${\mathit{\text{Open}}}_{\mathit{ij}}$), the sharing of a common language (${\mathit{CL}}_{\mathit{ij}}$), the sharing of a common land border (${\mathit{CB}}_{\mathit{ij}}$), and the sharing of a common colonizer (${\mathit{CC}}_{\mathit{ij}}$). ${\mu }_i$, ${\gamma }_j$, and ${\xi }_t$ are, respectively, exporter fixed effects, importer fixed effects, and time fixed effects, considered as proxies for multilateral resistances [21]. To avoid perfect multicollinearity between the bilateral dummy variables (${\mathit{CFA}}_{\mathit{ij}}$ and ${\mathit{Dum}}_{\mathit{ij}}$) and time-invariant variables like distance ($D_{\mathit{ij}}$) with bilateral fixed effects, I decide to omit the latter by adopting the country and time fixed effects according to the specifications. ${\epsilon }_{\mathit{ijt}}$ is the random term.

3.2 Estimation technique, data, and sample

Gravity models can have two types of specifications, namely a linear specification typically using OLS and a nonlinear specification using multiple estimators. I will initially, for preliminary results, apply OLS and for robustness apply a nonlinear approach, namely the Poisson pseudo-maximum likelihood (PPML) developed by Santos Silva and Tenreyro [22].

The data used come from three main sources, namely: UNCTAD, WDI, and CEPII. These data are observed over the period 1995–2019. Given the number of member countries (6 in CAEMC, 8 in WEAMU, and 14 in AFZ), the number of pairs per country (2400 pairs) and the number of partner countries (see sample of partner countries in the appendix), the number of observations is 14,400, 19,200, and 33,600, respectively, for CAEMC, WEAMU, and AFZ. The main characteristics of these data are shown in Tables A1–A3 in the appendices.

The sample includes two types of countries (Table A4 in appendices). The first is the reporting countries (country $i$), which are the member countries of CAEMC, WAEMU, and the AFZ. These countries are linked to partner countries (country $j$) that are part of several regional blocs in Africa and the world, namely SADC, AMU, EAC, EU, ASEAN+, MERCOSUR, and NAFTA.

4. Main findings

The results show that in CAEMC (see Table 1), the overall and average effect of the CFA is significant¹. However, by adopting a dynamic analysis, several salient facts emerge. Indeed, between 1995 and 2010, although the effect associated with the sharing of the single currency by the member countries of this community space is positive overall, it remains insignificant. This result reflects the fact that the CFAF did not generate the economies of scale and dynamic gains expected. Between 2011 and 2019, there is a positive impact from the sharing of the single currency in CAEMC. This analysis shows that in CAEMC, the positive effect of the single currency was delayed, with countries experiencing dynamic losses for over more than a decade. These results are discussed in general by Mignamissi [3].

Dependent variable: lnXij
	Full sample	1995	2000	2005	2010	2015	2019
Ln(GDPi)	0.845***	0.988***	1.685***	0.934***	0.270	0.411	0.375
	(0.0552)	(0.247)	(0.334)	(0.346)	(0.334)	(0.274)	(0.312)
Ln(GDPj)	2.036***	1.729***	1.908***	1.956***	2.132***	2.094***	1.974***
	(0.0311)	(0.159)	(0.143)	(0.159)	(0.156)	(0.159)	(0.168)
Ln(Popi)	0.843***	0.944***	0.681***	0.786***	1.137***	1.183***	1.401***
	(0.0457)	(0.278)	(0.224)	(0.227)	(0.217)	(0.228)	(0.254)
Ln(Popj)	−0.0301	0.203	0.155	−0.00305	−0.100	0.00323	−0.147
	(0.0370)	(0.197)	(0.174)	(0.190)	(0.182)	(0.184)	(0.191)
Ln(Distij)	−2.300***	−2.590***	−1.384***	−2.409***	−3.154***	−2.483***	−1.887***
	(0.0794)	(0.427)	(0.385)	(0.411)	(0.386)	(0.390)	(0.411)
XAFij	1.515***	2.155	1.950	1.244	1.432	2.438*	2.572*
	(0.257)	(1.345)	(1.215)	(1.345)	(1.258)	(1.258)	(1.309)
Opennessij	2.834***	1.746***	1.228**	3.193***	4.063***	3.779***	3.483***
	(0.110)	(0.539)	(0.575)	(0.594)	(0.554)	(0.553)	(0.594)
CLij	1.266***	−0.309	1.257**	1.593***	1.238**	0.555	1.350**
	(0.111)	(0.603)	(0.529)	(0.577)	(0.541)	(0.541)	(0.564)
CBij	0.456	1.304	1.377	1.131	−0.715	−0.213	−0.934
	(0.281)	(1.460)	(1.325)	(1.465)	(1.371)	(1.372)	(1.435)
CCij	2.776***	3.388***	2.688***	2.451***	3.010***	2.704***	2.562***
	(0.127)	(0.664)	(0.600)	(0.663)	(0.621)	(0.622)	(0.646)
Constant	−56.52***	−54.24***	−79.49***	−55.48***	−42.29***	−52.62***	−55.17***
	(1.317)	(5.360)	(7.270)	(8.443)	(8.125)	(6.802)	(7.141)
${\mu }_i$	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\gamma }_j$	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\xi }_t$	Yes	No	No	No	No	No	No
Observations	14,232	552	564	576	576	570	570
R-squared	0.502	0.466	0.527	0.474	0.525	0.528	0.468

Table 1.

Baseline results of dynamic effects in the CAEMC.

^***

p < 0.01.

^**

p < 0.05.

^*

p < 0.1.

Robust standard errors in parentheses.

Source: Author.

In the WAEMU (see Table 2), on the other hand, not only is the average and overall effect perceptible but also the dynamic effects are significant. In other words, the sharing of the single currency has distributive effects over time, indicating the existence of dynamic gains in this region. Moreover, the results obtained in this region can be qualified, as the significant effect is derisory in view of the low share of intra-regional trade in this area, which has barely exceeded 10% for several decades.

Dependent variable: lnXij
	Full sample	1995	2000	2005	2010	2015	2019
Ln(GDPi)	2.208***	2.909***	2.582***	1.312**	1.330**	2.741***	3.777***
	(0.119)	(0.648)	(0.574)	(0.586)	(0.593)	(0.612)	(0.517)
Ln(GDPj)	1.254***	1.429***	1.344***	1.167***	1.120***	1.111***	1.152***
	(0.0232)	(0.129)	(0.107)	(0.112)	(0.115)	(0.119)	(0.108)
Ln(Popi)	0.910***	−0.614	0.249	1.883**	2.615***	0.577	−0.737
	(0.159)	(0.817)	(0.765)	(0.785)	(0.788)	(0.829)	(0.711)
Ln(Popj)	0.371***	0.225	0.539***	0.403***	0.454***	0.398***	0.452***
	(0.0290)	(0.166)	(0.137)	(0.140)	(0.141)	(0.143)	(0.129)
Ln(Distij)	−2.468***	−2.710***	−2.418***	−2.470***	−2.269***	−2.387***	−2.032***
	(0.0551)	(0.318)	(0.261)	(0.269)	(0.267)	(0.270)	(0.246)
XOFij	1.784***	1.938**	1.859**	2.587***	1.887**	1.495**	1.835***
	(0.154)	(0.867)	(0.724)	(0.756)	(0.749)	(0.753)	(0.682)
Opennessij	1.858***	1.299**	1.170**	2.611***	2.461***	1.932***	1.457***
	(0.0944)	(0.553)	(0.475)	(0.453)	(0.440)	(0.469)	(0.426)
CLij	1.091***	0.431	1.329**	1.093	0.814	1.251*	0.971
	(0.140)	(0.730)	(0.662)	(0.692)	(0.686)	(0.690)	(0.627)
CBij	0.699***	0.891	0.500	0.470	0.668	0.367	0.466
	(0.177)	(1.018)	(0.829)	(0.867)	(0.858)	(0.863)	(0.783)
CCij	2.100***	3.287***	1.829***	1.651**	2.291***	2.065***	1.907***
	(0.148)	(0.788)	(0.701)	(0.732)	(0.724)	(0.728)	(0.660)
Constant	−73.13***	−64.18***	−75.33***	−66.66***	−80.49***	−78.14***	−85.55***
	(0.884)	(5.492)	(4.326)	(4.422)	(4.437)	(4.445)	(4.012)
${\mu }_i$	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\gamma }_j$	Yes	Yes	Yes	Yes	Yes	Yes	Yes
${\xi }_t$	Yes	No	No	No	Yes	No	Yes
Observations	18,976	736	752	768	768	760	760
R-squared	0.534	0.462	0.565	0.517	0.549	0.539	0.593

Table 2.

Baseline results of dynamic effects in the WEAMU.

^***

p < 0.01.

^**

p < 0.05.

^*

p < 0.1.

Robust standard errors in parentheses.

Source: Author.

From these results, discussing the future of the AFZ presents itself as a highly interesting opportunity. Such a debate could be structured around two main arguments, one political and the other economic. On the political level, the pressures generated by the desire of the WEAMU countries to enter into a new monetary union with the other ECOWAS countries have shifted the positions. This desire also revived recurrent internal debates which founded the legitimate aspiration of the populations of the Franc zone to choose their own economic destiny in general. On the economic level, one could point out the questionable effects of the mechanisms of the Franc Zone. To this end, it is relevant to question deeply the relevance of the agreements and operating principles of the Franc Zone (guaranteed convertibility, fixed parities, free transferability, and centralization of foreign reserves), the rate of economic cycle’s synchronization of the member countries, and the convergence speed of nominal and real indicators, among others.

As for the control variables, I distinguish two cases, namely the case of traditional quantitative variables and the case of bilateral dummies. Our results show that GDP, population, and distance between two countries support the intuition of gravity modeling. While GDP and population act as attractors to bilateral trade, distance acts as a repellent to trade between two countries. In other words, GDP and population are proxies for market size, which is a factor driving bilateral trade when this size reaches a critical level in the partner countries. Moreover, the further apart two countries are, the less they trade, because of the multiplication of transaction costs, especially transport costs. This result is fundamental to all gravity models.

Finally, the sharing of certain historical (common colonizer), geographical (openness to the sea and land border), and cultural (language) characteristics is favorable to market integration. The coefficients associated with the dummies that capture them are in most specifications positive and significant. Indeed, these variables not only help to reduce transaction costs but also strengthen the social and historical ties between peoples, which would be favorable to the mixing of populations and the intensification of exchanges between them.

5. Sensitivity analysis: the effects of partners’ regions on bilateral trade

The idea of this test is to see if any of the factors limiting bilateral trade in the AFZ have external origins. To do so, I model through dummies the effect of the membership of partner countries outside the AFZ in their respective communities. I thus capture the effect of partner countries’ membership in the Southern African Development Community (SADC), in the Arab Maghreb Union (AMU), in the East African Community (EAC), in the European Union (EU), in the Association of Southeast Asian Nations (ASEAN), in the Mercado Común del Sur (MERCOSUR), and in the North American Free Trade Agreement (NAFTA). By augmenting the gravity model with dummies capturing membership in these integration spaces, two types of results emerge.

In the CAEMC, countries’ membership in the African regional economic communities (RECs) (SADC, UMA, and AEC) reduces the bilateral trade (Table 3). In other words, countries outside the AFZ contribute to the detour of trade flows within the AFZ member countries. This detour effect is more pronounced in the SADC and EAC. This result could be justified by the multiple memberships of Central African countries in the broader sense of ECCAS in several regional economic communities (RECs). The coexistence of several communities (CEN-SAD², CAEMC, ECCAS³, ECGLC⁴, and EAC) in the same integration space, sometimes with similar objectives, results in a waste of resources, which is a brake on integration through the market. Also, the dynamic analysis of these effects shows that they are decreasing for SADC and UMA, but increasing for EAC. This shows a desire for faster commercial integration with the first two mentioned. On the other hand, the membership of CAEMC partner countries in regional economic communities outside Africa (EU, MERCOSUR, ASEAN, and NAFTA) seems beneficial to their trade. Indeed, because of their low level of diversification and product sophistication, CAEMC countries have difficulty gaining consistent market shares in these communities, which tends to amount to an illusory increase in their bilateral trade. Here, the effect of creation of trade flows weakens for the EU and NAFTA but consolidates for ASEAN, in particular over the last 7 years.

	Dependent variable: lnXij
	Full sample	1995	2000	2005	2010	2015	2019
XAFij	1.174**	1.856	1.926	0.801	0.826	2.042	1.581
	(0.262)	(1.363)	(1.228)	(1.380)	(1.293)	(1.278)	(1.303)
SADCj	−1.360**	−0.894	−1.798*	−2.325**	−2.013**	−0.420	−0.761
	(0.154)	(0.791)	(0.718)	(0.811)	(0.763)	(0.764)	(0.781)
UMAj	−0.401*	1.685	−1.812*	−1.203	−1.297	−0.312	−0.745
	(0.194)	(1.082)	(0.893)	(1.007)	(0.945)	(0.943)	(0.967)
CAEj	−2.809**	−2.771*	−2.879*	−3.053*	−2.015	−2.621*	−4.510**
	(0.249)	(1.125)	(1.174)	(1.320)	(1.238)	(1.230)	(1.260)
UE28j	0.679**	0.866	1.640*	−0.367	−0.648	0.947	0.00635
	(0.161)	(0.824)	(0.740)	(0.852)	(0.794)	(0.835)	(0.875)
ASEANj	1.441**	2.011*	−0.316	−0.479	−0.223	2.853**	3.964**
	(0.194)	(1.021)	(0.913)	(1.006)	(0.944)	(0.987)	(1.013)
ALENAj	0.0889	1.721	−0.211	−0.937	0.0141	−0.782	−0.193
	(0.296)	(1.547)	(1.390)	(1.556)	(1.454)	(1.483)	(1.521)
Controls and resistances	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	14,232	552	564	576	576	570	570
R-squared	0.513	0.481	0.550	0.485	0.533	0.544	0.505

Table 3.

Dynamic effects in CAEMC (augmented gravity model).

^**

p < 0.01.

^*

p < 0.05.

Robust standard errors in parentheses.

Source: Author.

In WAEMU, only AMU member countries seem to create a diversion effect, the creation effects being globally observed with the rest of the regional economic communities (Table 4). In dynamic analysis, I note a high and permanent effect over the entire study period for ASEAN, which is the opposite in NAFTA. The EU has a weak creation effect, noted between 1999 and 2003 according to our estimates.

Dependent variable: lnXij
	Full sample	1995	2000	2005	2006	2010	2015	2019
XOFij	1.232***	1.081	1.455*	2.241***	1.901**	1.526**	0.999	1.380**
	(0.158)	(0.895)	(0.746)	(0.778)	(0.860)	(0.764)	(0.774)	(0.698)
SADCj	1.014***	0.363	1.311**	1.855***	0.556	1.593**	1.614**	1.592***
	(0.128)	(0.721)	(0.606)	(0.630)	(0.696)	(0.619)	(0.637)	(0.575)
UMAj	−0.907***	−1.089	−1.546**	−0.706	−1.355	−0.449	−0.723	−0.256
	(0.155)	(0.932)	(0.723)	(0.756)	(0.836)	(0.741)	(0.754)	(0.684)
CAEj	1.634***	1.168	2.597***	2.728***	1.211	0.145	0.527	1.795*
	(0.209)	(1.168)	(0.982)	(1.028)	(1.135)	(1.010)	(1.029)	(0.929)
UE28j	0.420***	0.669	1.011*	0.787	1.022	−0.751	0.199	0.766
	(0.127)	(0.700)	(0.591)	(0.625)	(0.693)	(0.610)	(0.654)	(0.609)
ASEANj	3.047***	3.717***	1.717**	3.049***	2.115**	3.223***	3.252***	3.708***
	(0.165)	(0.928)	(0.781)	(0.807)	(0.891)	(0.791)	(0.835)	(0.756)
ALENAj	0.0360	0.462	−0.256	1.007	0.802	−0.0202	0.609	0.655
	(0.233)	(1.301)	(1.100)	(1.149)	(1.270)	(1.122)	(1.165)	(1.059)
Controls and resistances	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	18,976	736	752	768	768	768	760	760
R-squared	0.547	0.479	0.578	0.531	0.515	0.569	0.554	0.610

Table 4.

Dynamic effects in WEAMU (augmented gravity model).

^***

p < 0.01.

^**

p < 0.05.

^*

p < 0.1.

Robust standard errors in parentheses.

Source: Author.

These controversial results call for debate on the nature of monetary integration in the AFZ, as well as its potential economic effects. Mignamissi's [3] contribution to this debate is more empirical than analytical. The author evaluates the monetary costs/benefits of market integration in the Franc Zone. He defines, starting from the status quo, four scenarios (Cooperation, Aggregation, Consolidation, and Enlargement). Based on an augmented gravity model, he identifies costs and shows that the best scenario for CAEMC is consolidation and for WAEMU is cooperation. In general, this analysis corroborates those of Allechi and Niamkey [23], Masson and Pattillo [24], Beetsma and Giuliodori [25], and Carrere [26].

6. Robustness

Santos Silva and Tenreyro [22, 27] show that the specification in log-linear form of the gravity model yields biased estimators due to the heteroscedasticity of trade levels. The authors also show that the PPML estimator is more efficient than the nonlinear least squares estimator when trade is specified in levels. They also point out that to ensure the consistency of the Poisson estimator of the PPML, the data do not necessarily have to follow a Poisson distribution.⁵ This estimator corrects three main biases: (i) a bias induced by the log transformation, (ii) a bias due to heteroscedasticity, and (iii) a bias due to the presence of zeros in the dependent variable.

Applying the Poisson specification to the gravity model [28], I obtain

$$\mathit{Pr}\left(T_{\mathit{ij}}=C|x_{\mathit{ij}}\right)=\frac{e^{-\mu \left(x_{\mathit{ij}}\beta \right)}\mu {\left(x_{\mathit{ij}}\beta \right)}^{{\mathit{Com}}_{\mathit{ij}}}}{\left(T_{\mathit{ij}}\right)!}$$E4

where $T_{\mathit{ij}}=0,1,2,$… with $T_{\mathit{ij}}$! the factorial of bilateral trade. The Poisson model stipulates an egalitarian dispersion and the conditional variance of $T_{\mathit{ij}}$ is equal to its mean $\mu \left(x_{\mathit{ij}}\beta \right)$.

Solving the first-order conditions of the log-likelihood of the above expression, I obtain

$${\widehat{\beta }}_{\mathit{\text{Poisson}}}={\mathit{\text{ArgMax}}}_{\beta }\left\{\sum _{i=1}^N\sum _{j=1}^N\left[-e^{\left(x_{\mathit{ij}}\beta \right)}+{\mathit{Com}}_{\mathit{ij}}\left(x_{\mathit{ij}}\beta \right)-\mathit{Log}\left({\mathit{Com}}_{\mathit{ij}}\right)!\right]\right\}$$E5

In the presence of heteroscedasticity, this estimator is consistent and more efficient than the previously developed estimators of the gravity model. Finally, because of its multiplicative form, the Poisson estimator offers a natural technique for handling zeros in the dependent variable.

The estimable form of the model is as follows:

$$X_{\mathit{ijt}}=\mathit{Exp}\left\{\begin{array}{c}\mathit{ln}{\varphi }_0+{\varphi }_1\mathit{ln}{Y}_{\mathit{it}}+{\varphi }_2\mathit{ln}{Y}_{\mathit{jt}}+{\varphi }_3\mathit{ln}{\mathit{Pop}}_{\mathit{it}}\\ +{\varphi }_4\mathit{ln}{\mathit{Pop}}_{\mathit{jt}}+{\varphi }_5\mathit{ln}{\mathit{\text{Dist}}}_{\mathit{ij}}+{\varphi }_6{\mathit{CFA}}_{\mathit{ij}}+{\varphi }_7{\mathit{Dum}}_{\mathit{ij}}\end{array}\right\}\mathit{Exp}\left({\mu }_i+{\gamma }_j+{\xi }_t+{\epsilon }_{\mathit{ijt}}\right)$$E6

The results of this robustness test confirm the previously established results (See Tables 5 and 6). While in the CAEMC the dynamic effects of the CFA franc on bilateral trade between member countries are delayed, in the WAEMU the dynamic effect is positive, permanent, and significant over the entire study period. Moreover, the analysis confirms that in comparative statistics, the effect of certain years is greater than the overall effect. In other words, the overall average cumulative effect suffers from the specific economic conditions associated with the various events and crises that the region has experienced. This analysis reflects the fact that dynamic gains, when they exist, are not uniformly distributed over time.

Dependent variable: Xijt
	Full sample	1995	2000	2005	2010	2015	2019
XAFij	0.244***	0.377*	0.251	0.244	0.265*	0.328**	0.305**
	(0.0358)	(0.209)	(0.208)	(0.176)	(0.158)	(0.140)	(0.143)
Controls and resistances	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	14,232	552	564	576	576	570	570
R-squared	0.428	0.397	0.437	0.413	0.444	0.445	0.402

Table 5.

PPML estimates in CAEMC.

^***

p < 0.01.

^**

p < 0.05.

^*

p < 0.1.

Robust standard errors in parentheses.

Source: Author.

Dependent variable: Xijt
	Full sample	1995	2000	2005	2010	2015	2019
XOFij	0.163**	0.208*	0.176**	0.235**	0.164**	0.121*	0.146*
	(0.0137)	(0.0899)	(0.0657)	(0.0636)	(0.0624)	(0.0618)	(0.0589)
Controls and resistances	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	18,976	736	752	768	768	760	760
R-squared	0.472	0.399	0.492	0.463	0.502	0.488	0.523

Table 6.

PPML estimates in WEAMU.

^**

p < 0.01.

^*

p < 0.05.

Robust standard errors in parentheses.

Source: Author.

7. Concluding remarks

The AFZ has historically been considered one of the oldest currency areas in the world. However, the external economic effects of this currency grouping have rarely been shown to be conclusive. First, the operating principles of the AFZ, such as the fixed exchange rate regime and the centralization of foreign exchange reserves in the French Treasury, are widely debated. Second, the sharing of the CFA franc has not always ensured the synchronization of member countries’ cycles, as countries have experienced asymmetric reactions to exogenous shocks (financial crisis, hunger crisis, oil crisis, and COVID 19). This asymmetry in reaction to exogenous shocks slows down the speed of convergence of nominal and real macroeconomic indicators.

Specifically, the macroeconomic effects of the CFA franc on the bilateral trade of AFZ member countries have not been well documented. Thus, the few recent studies [2, 3] take a global approach by estimating an average effect over the study period. The specificity of this chapter is to adopt a dynamic approach by estimating for each year the marginal effect of the CFA franc between 1995 and 2019. Using ordinary least squares and the Poisson pseudo-maximum likelihood estimator, I obtain different results in the two currency unions of the AFZ. In the CAEMC, the dynamic effects of sharing the CFA franc are delayed and begin to be noticeable from the early 2010s. In contrast, the dynamic effects of the CFA franc in the WAEMU are permanent over the entire study period. However, these different effects must be contrasted with the low effective share of intra-zone trade.

Some lessons can be learned from the results obtained. While there is no perfect monetary structure or exchange rate that is good all the time for a country, a profound reflection on the future of the franc zone must be conducted. Moreover, decisive steps have already been taken within WAEMU with the project to create a single currency in ECOWAS, although this has been delayed compared to the initial deadlines. In CAEMC, the project to rationalize with ECCAS is already in place. A High Monetary Authority responsible for setting up a Monetary Union and a Central Bank in the Union/Community through the monitoring of macroeconomic convergence as well as the harmonization of monetary, banking, and financial policies has recently been created. These dynamics are in line with the African Union’s desire to eventually move to a single currency on the scale of the continent from the various subregional currencies. This desire is affirmed by the implementation of the African continental free trade area, which is one of the preliminary steps toward the feasibility of a single currency with proven dynamic potential effects, enhanced tenfold by a large-scale integration zone.

A.

	CAEMC					WEAMU					Source
	N	Mean	S.D.	Min	Max	N	Mean	S.D.	Min	Max
lnXij^*	14,352	8.752	6.965	0	22.39	19,136	9.511	6.692	0	21.18	UNCTAD
lnYi^**	14,400	22.85	0.960	19.90	24.40	19,200	22.65	1.015	20.25	24.52	WDI
lnYj^**	14,280	24.59	2.326	18.62	30.54	19,040	24.59	2.324	18.62	30.54	WDI
lnPopi^**	14,400	15.19	1.088	13.12	17.07	19,200	16.02	0.806	13.90	17.06	WDI
lnPopj^**	14,400	16.16	1.717	11.23	21.06	19,200	16.15	1.721	11.23	21.06	WDI
lnDistij^***	14,400	8.305	0.757	2.349	9.512	19,200	8.312	0.820	4.656	9.576	CEPII
Openessij^****	14,400	0.542	0.498	0	1	19,200	0.508	0.500	0	1	Author
CLij^****	14,400	0.382	0.486	0	1	19,200	0.227	0.419	0	1	Author
CBij^****	14,400	0.0435	0.204	0	1	19,200	0.0533	0.225	0	1	Author
CCij^****	14,400	0.224	0.417	0	1	19,200	0.199	0.399	0	1	Author
CFAij^****	14,400	0.0521	0.222	0	1	19,200	0.0729	0.260	0	1	Author
SADCj^****	14,400	0.135	0.342	0	1	19,200	0.135	0.342	0	1	Author
AMUj^****	14,400	0.0625	0.242	0	1	19,200	0.0625	0.242	0	1	Author
CAEj^****	14,400	0.0317	0.175	0	1	19,200	0.0313	0.174	0	1	Author
EUj^****	14,400	0.292	0.455	0	1	19,200	0.292	0.455	0	1	Author
ASEANj^****	14,400	0.135	0.342	0	1	19,200	0.135	0.342	0	1	Author
MERCOSURj^****	14,400	0.0521	0.222	0	1	19,200	0.0525	0.223	0	1	Author
NAFTAj^****	14,400	0.0313	0.174	0	1	19,200	0.0313	0.174	0	1	Author

Table A1.

Descriptive statistics.

Source: Author.

	lnXij	lnYi	lnYj	lnPopi	lnPopj	lndistij	XAFij
lnXij	1.0000
lnYi	0.2913	1.0000
lnYj	0.5196	0.0343	1.0000
lnPopi	0.1814	0.2453	0.0197	1.0000
lnPopj	0.4222	0.0191	0.7099	0.0046	1.0000
lnDistij	−0.0002	−0.0204	0.4789	0.0008	0.2117	1.0000
XAFij	0.0981	−0.0011	−0.1749	−0.0005	−0.1339	−0.4782	1.0000

Table A2.

Full correlation matrix (CAEMC).

Source: Author.

	lnXij	lnYi	lnYj	lnPopi	lnPopj	lndistij	XOFij
lnXij	1.0000
lnYi	0.4625	1.0000
lnYj	0.3477	0.0252	1.0000
lnPopi	0.4193	0.9327	0.0224	1.0000
lnPopj	0.3490	0.0118	0.7111	0.0096	1.0000
lnDistij	−0.2222	−0.0101	0.4011	−0.0233	0.1896	1.0000
XAFij	0.1803	−0.0007	−0.2351	−0.0007	−0.0214	−0.4822	1.0000

Table A3.

Full correlation matrix (WEAMU).

Source: Author.

CAEMC	Cameroon, Central African Republic, Chad, Congo, Equatorial Guinea, Gabon
WEAMU	Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal, Togo
SADC without Tanzania	Botswana, Eswatini, Lesotho, Madagascar, Malawi, Mauritius, Mozambique, Namibia, Seychelles, South Africa, Zambia, Zimbabwe
AMU + Egypt	Algeria, Libya, Mauritania, Morocco, Tunisia, Egypt
EAC without Burundi	Kenya, Uganda, Tanzania
EU of 27 + UK	Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, UK
ASEAN + Japan, China, South Korea	Brunei Darussalam, Cambodia, China, Indonesia, Japan, Lao PDR, Malaysia, Myanmar, Philippines, Singapore, South Korea, Thailand, Vietnam
MERCOSUR	Argentina, Brazil, Paraguay, Uruguay, Venezuela
ALENA	Canada, Mexico, USA

Table A4.

Sample.

Source: Author.