Exchange rate regimes and inflation in Sub-Saharan Africa

The main argument in favor of a fixed exchange rate regime (ERR) is its ability to maintain lower inflation in the long run compared to a flexible ERR (Mundell, 1963; Fleming, 1962). This paper empirically tests whether the fixed ERR of the CFA franc currency union provides lower inflation to its members relative to inflation in the non-CFA Sub-Saharan African (SSA) countries. SSA countries are grouped by their exchange rate regimes using the International Monetary Fund (IMFs) de facto classification to analyze the dynamics of inflation within the groups of fixed ERR in comparison to the non-fixed ERR groups. The empirical results support the inflation-growth trade-off in the CFA zones. While the CFA countries experience a relatively lower inflation in the short and long run, they suffer from a pronounced output loss relative to all other non-CFA countries in general and relative to the non-CFA countries with pegged ERR in particular. As individuals’ welfare depends on the change in their consumption of goods and services rather than the growth level of inflation (Aiyagari, 1990), the finding of this paper suggests that the CFA countries’ fixed ERR compounded with an alignment to a common currency undermines their economic performances.


INTRODUCTION
The impact of exchange rate regime (ERR) on economic performance is one of the hotly debated issues in the field of international finance.This correlation gained more importance in the face of financial crises as international capital flows become increasingly unstable.The critical role of ERR in economic performance in our globalizing world has induced many countries in recent years to switch from one regime to another.If, following the demise of the Bretton Woods system, the choice of an ERR was important for stabilization outcomes, then, nowadays the choice of ERR may have important policy implications -particularly, for policy aimed at tackling external shocks and speculative attacks (Eichengreen, 2008).
Across the globe, different types of ERR ranging from hard peg to free floating regimes exist.There is no consensus on which type of regime better enhances economic performance.Alternative ERRs have some E-mail: jlohi@worldbank.orgTel: 3013647128 (JEL) classification: E31; F31; F43.
Author agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License strengths and weaknesses regarding economic outcomes in the country when they are at work.There are some arguments in favor of and against each type of the ERR.For instance, Mundell (1963) and Fleming (1962) argue that under a fixed ERR, trade and investment are more certain.The trade and investment advantages of the fixed ERR stem from the reduction of transaction costs and lower inflation expectations.These advantages are what led the European Economic Community (EEC) to adopt the fixed ERR to achieve their single market program.The lower inflation associated with the fixed ERR has been an important incentive that enticed Great Britain to return to the gold standard in 1925 after abandoning it in the wake of world war I in 1914 (Capie et al., 1986a).By fixing a currency to a foreign anchor, the domestic country imports the monetary policy of the anchor country.Such import is associated with political commitment and disciplinary monetary policy (that is, alignment to the anchor country's monetary policies) for anti-inflationary outcomes.However, the fixed ERR is criticized for poorly insulating the economy against external shocks (Obstfeld and Rogoff, 1995).
Supporters of the flexible ERR argue that it confers more independence of monetary policy through the flexibility of the nominal exchange rate.By changing the nominal exchange rate, the country gains control over the impact of disruptive economic shocks.Friedman (1953) pointed out that the speed at which a country adjusts relative prices when hit by a real shock depends on the ERR at work in that country.Friedman (1953) argues that in a world of sticky prices, the flexible ERR absorbs the effects of external shocks more effectively than the fixed ERR.Indeed, under a flexible regime, in the presence of shocks, the nominal exchange rate adjusts immediately, allowing relative prices to change.This mechanism reduces the effects of shocks on macro variables, especially on output.Previous empirical work has found support for Friedman's hypothesis.For example, Broda (2004) tests Friedman's hypothesis on the terms of trade shocks and finds that the response of real GDP to a terms of trade shock is much smaller under a flexible regime than under a fixed regime.Broda notes that in response to a 10 percent negative shock to the terms of trade, the real exchange rate depreciated faster under the floating system while the depreciation was slower in the pegged regime.As a result, real GDP fell by 1.9 percent under fixed regime and only by 0.2 percent in the flexible exchange rate regime.
The advantage of the floating ERR is that it insulates the economy from external shocks and eventual speculative attacks.However, floating regimes are expected to exhibit high volatility in exchange rates and high inflation.Mussa (1986) underlined that real exchange rates fluctuate a lot more in the short run in countries with flexible ERR than in countries with fixed ERR.This is so because nominal exchange rates are very volatile under flexible regimes.Similarly, Shambaugh (2004), Klein (2005), and Klein and Shambaugh (2008) show that exchange rates are more volatile under a floating ERR than under a fixed ERR.Specifically, Klein and Shambaugh (2008) find that in magnitude, pegged (fixed) ERRs have about 16 percent less volatility in the nominal exchange rate than a floating ERR.After classifying counties by the de facto behavior of the country's monetary authorities, Levy et al. (2001) show that a flexible ERR exhibits higher exchange rate volatility with lower volatility in international reserves while the opposite holds under a pegged ERR.
Whether in the aftermath one type of ERR outperforms the others in terms of the economic outcome; ERRs are crucial determinants of economic performance.Rose (2011) states, "exchange rate is an important asset price, perhaps the most important asset price".This implies that the regime monitoring the exchange rates is important to assets' prices and therefore economic outcomes.An ERR can impact the economy through different macroeconomic channels.For instance, flexible ERR can expose the economy through the inflationary channel; the fixed regime by retarding the adjustment of prices in the face of external shocks allows large fluctuations in output.Inflation expectations can lead to higher or lower interest rates in the country and thereby affect trade and investment incentives.In short, an ERR is a crucial determinant of economic outcomes.
Despite the prominent connection between ERR and economic performance, the literature is limited in addressing how the type of ERR implemented is retarding the economic take-off of some developing countries.This paper attempts to fill this gap for some Sub-Saharan African (SSA) countries which are still lagging behind economically.Specifically, this paper focuses on the inflation dynamics between three groups of countries with distinct types of ERR in SSA: franc with a conventional peg before their indepen-dence from France and they are still using it up to date.The CFA franc was pegged to the French Franc from December 26, 1945 -the date of its creation-to January, 1999.Since January, 1999 up to date the CFA franc is pegged to the Euro.From this fact, it is hard to assess how these countries would have performed economically under an alternative ERR and/or without belonging to a currency union (owning their personal central banks).
Secondly, SSA countries exhibit many commonalities in terms of their history and economic characteristics (market access issues, dependence on the export of few primary commodities, financial markets develop-ment, geography, level of industrialization, government efficiency, etc.).Therefore, it is appropriate to compare the economic performances of the CFA countries to that of the non-CFA SSA countries with alternative ERR.Discussing the correlation between economic outcomes and ERR for the CFA countries and distinguishing between the CFA and non-CFA countries of SSA can provide important policy prescriptions -for both exchange rate regime and monetary policy reforms -to aid in solving the countries' delay in economic take-off.This paper focuses essentially on inflation dynamics and compares SSA's CFA franc currency union to its Non-CFA-currency-union countries, because a key purpose of participating in a currency union is to benefit from lower inflation.
The goal carried out in this paper is so far an uncovered topic, especially distinguishing between the CFA and Non-CFA countries of SSA on inflation dynamics.The remainder of this paper is organized as follows: Section 2 states the stylized facts about the fixed and flexible ERR.Section 3 describes exchange rate regimes in general and provides the classification of SSA countries by exchange rate regime and by monetary policy framework.Section 4 explains the methodology, presents the models and the data, and frames the hypothesis and the discussion points.The results tables and their interpretations are in section 5. Section 6 concludes the paper.The figures are stored in the appendix I.

The stylized facts about fixed and flexible exchange rate regimes
Across the literature there are three key stylized facts about the ERR.The first is the inconsistency between the de facto and the de jure ERR (i.e., countries that officially claim to float heavily intervene in the exchange market to regulate the rate of exchange of their currencies).The second is that many countries have shifted to a flexible ERR since the demise of the Bretton Wood System.The third fact is what Eichengreen (1994) named the "hollowing-out hypothesis" and Fischer (2001) refers to as a "bipolar view".The "hollowing-out hypothesis" or the "bipolar view" stipulates that intermediate regimes including conventional pegs are incompatible with capital flows.Only the two extremes: hard peg or free floating are sustainable in the face of high capital flows.Some recent facts across the world support the vulnerability of the pegs in the face of capital mobility.More or less, countries involved in crises in the 1990s were associated with the fixed (pegged) ERR.The 1994 Tequila crisis of Mexico, the 1998 exchange rate crises of Russia and Brazil, and those of Turkey and Argentina in 2000, are few examples.Fischer (2001) mentions that in contrast to the emerging countries with pegged currencies who experienced the exchange rate crisis, other emerging countries with more flexible rates (South Africa, Israel) avoided crises of this type.The implication of this third fact is that fixed ERRs are less efficient in insulating economies from external shocks.
With a fixed ERR the country sacrifices its ability to stabilize the economy against attacks in return for credibility gains through commitments (Klein and Shambaugh, 2010).Under a fixed ERR, the slow responsiveness of the nominal exchange rate to adjust relative prices in the face of external shocks allows disturbances in real GDP (Friedman, 1953;Levy et al., 2001;Caballero, 2002;Broda, 2004, Edwards andYeyati, 2005).Therefore, fixed regimes would exhibit more loss in their per capita outputs while the reverse is expected for floating regimes.Nevertheless, for the fixed ERR, the loss in output is expected to be compensated by lower inflation through credibility and disciplinary monetary policies associated to the commitment of pegging the domestic currency to a foreign currency that plays the role of an anchor.
Note that pegging a currency is associated with various political commitments which allow importing the anchor country's monetary policies (disciplinary mone-tary policy), reduce inflationary policies, increase the credibility of the domestic monetary authorities, reduce inflation expectations and stabilize the economy.The high political cost of fixing the exchange rate is what forces policy makers to adopt certain monetary and fiscal policies to avoid the demise of the regime.This constraint confers credibility and discipline to the fixed ERR (Meltzer, 1986;Ghosh et al. 1997;Yagci, 2001;Levy et al., 2001).On the other hand, under a flexible ERR the management of the nominal exchange rate to facilitate the quick adjustment of relative prices is associated with higher inflation expectations.In fact, the flexibility of the nominal exchange rate makes the relative price less predictable.The correlation between inflation and the exchange rate regime is well described in the literature.The investigation of inflation persistence shows three main findings: 1) inflation rates vary over time and across countries due to the monetary policy framework; 2) the speed of this variation differs over time; 3) there is an inflation-output trade off associated with inflation adjustment (Fuhrer and Moore, 1995;Sargent, 2001;Cecchetti and Debelle, 2006).Although not all papers directly relate inflation to the ERR, three main models are used in the literature to study inflation: the flexible and sticky price models and the sticky information model.The flexible price model argues that inflation evolves over time due to the monetary authorities' action of adjusting monetary policy very frequently.The expansionary policy of the policymakers leads to inflationary outcomes (Barro and Gordon, 1983).The pioneers of the sticky price model use the wage contract in explaining inflation (Taylor, 1979;Calvo, 1983).However, the sticky price model falls short in explaining inflation after introducing the real wage (Fuhrer and Moore, 1995).Lastly, the sticky information model developed by Mankiw and Reis (2002) shows that rather than sticky wages, prices adjust slowly because the cost of information prevents economic agents from frequently updating prices according to current macroeconomic conditions.The flexible price model has some incarnation of the ERR.In fact, the frequent price adjustment of the monetary authorities reflects the flexibility of the nominal exchange which is the foundation for the flexible ex-change rate.Conclusively, a flexible exchange rate is associated with higher inflation.
Recently, Obstfeld and Rogoff (1995) described inflation dynamic in the contest of exchange rate regimes.The theoretical framework developed by the authors focused on the cost and benefits of the fixed exchange rate regime.According to Obstfeld and Rogoff (1995), there are three main reason why countries fix (peg) their currency's foreign value.The first reason is to avoid exchange rate volatility like the one under the floating ERR.Exchange rate volatility creates uncertainty about future assets' prices and reduces trade and investment (Mundell, 1963;Fleming, 1962).The second reason is to import the anchor country's inflation rate.Fixing the domestic currency to a foreign one with lower inflation allows the domestic country to experience lower inflation due to the credibility by committing to disciplinary policies.The third reason, closely related to the second is the disinflationary objective.Some countries adopt the fixed ERR after they have experienced higher degrees of inflation.
Fixing (pegging) the currency in this case becomes an objective solution to reducing inflation.Among all, the main purpose and the theoretical benefit of fixing a currency's foreign value is to have price volatility under control.Obstfeld and Rogoff (1995) also point on one inconvenience of fixing the exchange rate: the forgone control over domestic money supply that would have been used for stabilization purposes.Theoretically and practically, in the face of external shocks such as the drop in demand for exports goods, the country would adjust import and export prices by depreciating the real exchange rate.That is the monetary authorities can reduce the domestic interest rate.The reduction of the home interest rate puts demand pressure on foreign assets with a relatively higher interest rate.Therefore, the domestic currency depreciates and stimulates the short run demand for domestic goods.If quick, this adjustment reduces the impacts of the shock.But, if prices and nominal exchange rates are rigid in the short run like under the fixed exchange rate regime, firms will have to hire less or fire some workers to reduce output in the face of the lower demand for their products.In this situation, as the domestic interest rate is determined by the foreign rate, the domestic monetary authorities have no power to change it.Thus, domestic attempts to chan-ge the money supply have no effects.Indeed, under the fixed exchange rate, the money supply is out of the control of the monetary authorities.
Given these stylized facts and the theoretical frameworks about flexible and fixed exchange rate regimes, this paper tests whether the CFA franc currency union countries-whose common currency is pegged to a foreign anchor experience lower inflation rates in the short and long run compared to the others, the non-CFA countries of SSA as their benefit for scarifying output in the face of shocks.

Exchange Rate Regimes in Sub-Saharan Africa
The official classification of countries by their exchange rate regime (ERR) has been traditionally provided by the International Monetary Fund (IMF).But Calvo and Reinhart (2000) show how some countries that officially claim to have a floating regime intervene in the foreign exchange market.The mismatch between the de jure and the de facto classifications of countries has led economists in the field of international finance to make a clear distinction and reclassify countries based on their de facto regimes.The most known alternative classifications of countries based on the de facto approach are those of Levy et al. (2000aLevy et al. ( , 2003)), Reinhart and Rogoff (2004), Shambaugh (2004), and Ilzetzki et al. (2008).Each of them uses different methodologies.
4 However, the classifications of SSA countries from any of the above cited classification sources match those of the IMF.For this reason, the recent IMF's de facto classification of countries provided in the AREAER The IMF's annual report on exchange rate arrangement and monetary policy frameworks classifies exchange rate arrangements based on the degree to which the exchange rate is determined by the market.
Ten key types of ERR are listed from across the world.
(1) No separate legal tender (hard pegs), ( 2 exchange rate regimes employ monetary aggregate target and inflation targeting as their monetary policy frameworks.
All of the above types of ERR associated with different monetary policy frame-works are at work in different countries in SSA.WAEMU and CEMAC that make up the two CFA franc zones, use the conventional peg as their exchange rate regime.
The CFA franc of CEMAC and that of WAEMU have the same rate of exchange to the euro to which they are pegged.The monetary policy framework at work in the CFA zones is the exchange rate anchor.Eritrea, Cape Verde, Comoros, Sao Tome and Principles, Lesotho, Namibia and Swaziland also use the conventional peg as their exchange rate regime.The difference between the CFA franc zones and these countries is that the zones form a currency union (the CFA franc zone countries are linked to one central bank in each zone and use a common currency: the CFA franc) while the other countries have their own central banks.Zimbabwe uses the no-separate-legaltender regime and is pegged to the U.S. dollar.
Burundi and Rwanda use a stabilized arrangement ERR with a monetary aggregate target as their monetary policy framework.Botswana has the crawling peg regime with the currency compositely pegged.Ethiopia is pegged to the U.S. dollar under the crawl-like-arrangement exchange rate regime.Angola, Liberia, Guinea, Malawi, and Nigeria exhibit other managed arrangement regimes.Angola and Liberia are pegged to the 6 The peg regimes use an exchange rate anchor as the monetary policy framework.Under the exchange rate anchor, the monetary authority buys or sells foreign exchange to maintain domestic currency's rate of exchange at the targeted rate or within a range.The exchange rate represents the nominal anchor or intermediate target to monetary policy for these regimes (see IMF's AREAER, 2010). 7Pegged exchange rates within horizontal bands also use the exchange rage anchor framework. 8Managed arrangement regimes use exchange rate anchor, monetary aggregate target and inflation targeting as monetary policy frameworks.Under the monetary aggregate framework the targeted aggregate serves as the anchor to monetary policy.For inflation targeting framework, monetary policy decision depends on inflation forecasting and how the forecasted inflation deviates from the targeted one.Thus inflation forecast is the nominal anchor to monetary policy. 9Floating and free floating regimes employ monetary aggregate target and inflation targeting as their monetary policy frameworks.
Lohi 177 U.S. dollar with an exchange rate anchor while Guinea, Malawi, and Nigeria use a monetary aggregate target framework.Twelve SSA countries operate under the floating exchange rate regime.Ten 10 of these target a monetary aggregate while the other two -Ghana and South Africa, have inflation targeting as their monetary policy framework.Mauritius is the only SSA country where a free floating exchange rate regime is at work.In this paper the SSA countries will be grouped as CFA and non-CFA zones, where the non-CFA zone combines pegged, floating and some intermediate ERR's.

METHODOLOGY AND MODEL
Studies of inflation have frequently used augmented Phillips curve models in which the policy preferences of the natural rate of unemployment and the expected supply of expansionary policy are incorporated.Although these models well suit inflation persistence, there is no reliable record on employment for many SSA countries; making it difficult to use such models to empirically test inflation in SSA.Other models have been used to examine the inflation effects of exchange rate regimes in many developing countries.For instance, Levy and Sturzenegger ( 2001) developed an inflation model in which inflation is related to the changes in money supply growth, the change in GDP growth, the real interest rate and the change in money velocity.
However, their model appears as an identity 11 and thus, gives less opportunity to conduct the comparative analysis on inflation across the SSA countries.Kamin (1997) studies the linkage between inflation and the ERR for Asian, industrialized and Latin American countries.But the model does not distinguish between the short and the long run.Kamin's model is explained in Appendix II.
In this paper, a simple cross-groups comparative analysis methodology is adopted using a dummy variable technique while building on the theories about Fixed and Flexible Exchange Rate Regimes.To capture the short and the long run inflation differences between t h e sub-samples while avoiding estimating an identity model, a model isconstructed where inflation depends on trade openness 12 external shocks (terms-of-trade) and the lagged inflation.In this model, the CFA dummy 13 is added (Equation 1).It is important to control for the terms-of-trade shocks and the trade openness in the model, as they are potential inflationary channels.The rationale behind the inclusion of the lagged inflation rests on the potential serial correlation that can exist between current and past inflation (the inertia problem).Augmenting the model with the CFA dummy allows identifying the extent to which fixed ERR succeeds in maintaining lower inflation relative to the flexible exchange rate regime in the short and in the long run as a result of enhanced credibility and disciplined monetary policies.Sub-Saharan African (SSA) countries are subdivided into four sub-samples (S1, S2, S3, and S4) based on their exchange rate regime, and estimated the equations using OLS and the robust regression methodologies on each sub-sample (group) with inclusion of the CFA dummy.The robust regression methodology makes one to control the heteroscedasticity to avoid biased parameter estimates.

The model
(1), where, πit is inflation rate in country i at time t.C F Ai , Openit , T Tit , and πit−1 represent respectively the dummy for the CFA zone, trade openness, terms of trade, and lagged inflation.The CFA dummy takes the value of 1 if country i belongs to the CFA franc currency union or zero otherwise.
For the short run, equation ( 1) is estimated on the full sample (sample S1 comprising CFA and all non-CFA countries); the first reduced sample (the S2-sample without pegging non-CFA countries); the second reduced sample (the S3-sample with only the CFA and floating ERR non-CFA countries); and the third reduced sample (the S4-sample with only the CFA and pegged ERR non-CFA countries) using pooled OLS and fixed effects estimation.For the long run inflation regression, the average inflation (πi) over the data period is estimated with inclusion of the CFA dummy (Equation 2).By the theory, if the CFA countries have lower inflation relative to the other countries, it would be a result of a lower money growth.To examine the extent of money supply growth in the CFA zones relative to other SSA countries, equation ( 3) is estimated below.Equation ( 3) is similar to equation (2); however, the average growth of money supply (growthM 2i ) is used as the dependent variable in Equation 3. (2), (3),

Data description
The data used in this investigation are retrieved from the World Bank's World Development Indicators (WDI) database.The variables employed include: Consumer price index inflation (CPI), money supply (M2), real GDP, population, the terms of trade (computed as the ratio of exports to imports prices), and the trade openness (calculated as the ratio of the sum of exports and imports to GDP).The data cover a panel of 36 SSA countries over the period from 1980 to 2007.14 Countries included in the examination are those with valid data on all variables of interest over the entire period.Countries with hyperinflation (inflation rate exceeding 50 percent and persistent over many years during the examination period) are excluded.Countries like Zimbabwe, the Democratic Republic of Congo, and Angola are not included for either the hyperinflation issue or the lack of data or both reasons.Guinea Bissau, a current member of WAEMU is excluded from the analysis, because it joined the union in 1998.So, this country has been a member for less than 20 years according to the data period.Benin and many other countries were not included for lack of data on key variables over many years.

The hypotheses
Based on the theory and the stylized facts discussed above, if a 14 The data is linearized to capture the long run effects.
fixed exchange rate regime provides lower inflation, then inflation rates should be lower within the CFA franc groups both, the short and the long run compared to those in the non-fixed exchange rate regime countries.Thus, in equation ( 1) where CFA reflects the CFA franc group dummy, one would expect β4 to be negative and statistically significant, and in equation ( 2), β1 to be negative and statistically significant.The negative signs of these coefficients would imply that the CFA franc countries exhibit lower inflation relative to the non-CFA countries.Theoretically, there is no incentive to increase the money supply in an attempt to lower the nominal interest rate under a fixed exchange rate regime.Under a fixed ERR, inflation can be reduced by maintaining lower money supply growth.From this point of view, one would expect β1 in equation ( 3) to be negative and statistically significant.In addition, if the main source of lowering inflation is the extent of money suppy growth, then the magnitude of β1 in equation ( 3) would match the size of β1 in equation ( 2).However, if there is any mismatch between β1 of equation ( 2) and that of equation ( 3), then other sources might be influencing the inflation rates.These sources could be the extent of the growth of the real GDP per capita and/or the growth of money velocity.
To see how the growth of the real GDP per capita and money velocity influence inflation in the CFA zones, equation 3.1 and 3.2 are estimated below.
(3.1), (3.2), where, i cap RGDP / and i  are respectively, the average of the growth of real GDP per capita and that of the growth of money velocity over the data period.C F Ai is the CFA dummy.Theoretically, we would expect a nega-tive correlation between inflation and money velocity growth.Also, a higher growth of output would contribute in lowering inflation.If there is a mismatch between the sizes of β1 in equations 2 and 3, then the coefficient restriction on β1 in equation 3.1 and 3.2 will depend on the type of the mismatch.

Dıscussıon Case 1: |β1 | in equation (3) > |β1| in equation (2)
If the size of β1 in equation ( 3) is larger than that of β1 in equation (2) in absolute value, this would imply that the extent of the Real GDP per capita is rather resisting to the reduction of inflation.This resistance will be reflected in a negative coefficient of β1 in equation (3.1).Nevertheless, a negative β1 in equation (3.1) would mean the CFA countries face a loss in their output per capita, which could be detrimental to welfare.Moreover, if the sum of β1 in equation ( 3) and β1 in equation (3.1) still mismatches the size of β1 in equation ( 2), then the growth of the money velocity should have some influences on the inflation rates.
In the case where the sum of β1 in equation ( 3) and in equation (3.1) is larger in absolute value than the magnitude of β1 in equation ( 2), that would imply the presence of a positive growth of money velocity in the CFA zones relative to that in other non-CFA countries.Whereas, a positive growth of the money supply velocity leads to higher real interest rate.Real interest rates being the costs of borrowing, the economic consequences of having higher real interest rates can be in two folds.Higher real interest rates induce banks and consumers to avoid keeping money.
15 15 Consumers would like to put money in a saving to benefit from the higher interest rates rather than investing in activities that would provide outputs; on the banks' side, they would like to lend at the higher rates.
Thus, higher interest rates reduce domestic money demand and allow faster circulation of money (that is higher growth of money velocity).In either case, higher interest rates (implying higher growth of money velocity) and/or output loss would decay the economic performance of the CFA countries.Higher interest rates can reduce investment per capita and output.Moreover, the welfare of individuals depends on how much goods and services they can consume.Hence, output loss directly reduces people's welfare.
Case 2: |β1 | in equation (3) < |β1| in equation ( 2) If the size of β1 in equation ( 3) is smaller than that of β1 in equation ( 2) in absolute value, we would expect the reverse scenario from case 1.Note also that for the comparison between the CFA and the pegged non-CFA (sample S4), if the CFA countries perform better, this would be the currency union effect as both parties in the sample have the pegged ERR.The only difference is that the CFA zones are currency unions while each country of the non-CFA group with pegged ERR has its own central bank.

Keeping inflation lower
The focus on Sub-Saharan Africa (SSA) and the distinction between CFA and non-CFA countries in this examination reveals important information on the exchange rate regimes (ERR)' influences on economic performances across SSA countries.Tables 1, 2, 3, and 4 provide respectively the long run inflation, money growth, output per capita growth, and money velocity growth in the CFA countries relatively to: 1) all other non-CFA countries, 2) non-CFA-non-pegged ERR countries, 3) the non-CFA with floating regimes, and the non-CFA pegged ERR countries.Estimations are performed separately for each sample (S1, S2, S3, and S4) with inclusion of the CFA dummy in the model.Moreover, for each sample, the OLS and the robust estimations are performed respectively.
The OLS estimation of the long run inflation shows that the CFA countries have respectively 11, 13, 16, and 3 percent less inflation relative to all other non-CFA, the non-CFA-non pegged, the non-CFA-floating countries, and the non-CFA pegged ERR countries (all the results are statistically significant at a 99 percent level of confidence) (see OLS estimations -column 1 of each sample in Table 1).
It is good to note that the OLS estimation does not correct the heteroscadasticity problems, while over the period considered, in some Sub-Saharan African (SSA) countries there has been some temporally inflation peaks causing the heteroscadasticity problems within Lohi 179 While domestic borrowers might not be willing to borrow at high costs, foreigners borrowers might.
Lohi 179 the inflation data.In fact, at different occasions over the data period, there have been temporary hyperinflations in some non-CFA countries.However, after the peaks, the countries quickly recovered and inflation rate became as usual.For instance, in the early 1990s, Zambia experienced an occasional hyperinflation mounting up to 183 percent.But after this period, the inflation rate declined back to its common rate around 25 percent.Similarly, Uganda experienced some brief, but severe hyperinflation in the late 1980s; Ghana was subject to an inflation of 122 percent in 1983.In 1995, the inflation in Nigeria reached 72 percent; Uganda had its highest inflation (56 percent) in 1986 and Mozambique's inflation peak of 63 percent occurred in 1994.None of these countries have a peg regime.Rather, they are all floating regime countries; that is why there is no big difference between the long run results of the sample S2 (using non-CFA without the pegs) and that of the sample S3 (using the non-CFA floating regimes only).
As the occasional inflation peaks highlighted above do not reflect the usual average inflation rate of these non-CFA floating regime countries, these inflation peaks by creating the heteroscedasticity (outliers) problem in the data pump up the period average inflation rates of the non-CFA countries as a whole and make as if the CFA countries have relatively very lower inflation rates in the long run.
Given that the classical estimation methods such as the ordinary least square (OLS) are outlier sensitive, the presence of outliers causes the OLS estimation to be inefficient; leading to inflated and bias estimates of the residuals (Mia et al., 2008).To correct this mitigation, the robust regression methodology is used.The robust regression is the estimation methodology that aims to control heteroscedasticity in the data to avoid biased parameter estimates.There is an extensive literature on how outliers occur and how to limit their effects on the parameter estimates.Across literature, the most used method to correct heteroscedasticity in the data is the robust estimation methodology (Fellner, 1986).
After accounting for heteroscadasticity in the model, the inflation rates of the CFA countries turn out to be in average 5 percent less than that of other countries samples, and only 0.4 percent less than that of other non-CFA countries with pegged ERR (see Robust estimation-second columns for each sample in Table 1).As a result of the robust estimation, the CFA countries exhibit lower inflation relative to their SSA counterparts, even though the inflation gap between the CFA and others is not too large.This out performance of the CFAs on keeping inflation lower over other non-pegged ERR can be attributed to fixed ERR effect.Noticeably, the difference in inflation between the CFAs (pegged under currency union) and the non-CFAs with pegged ERR (but non currency union members-having each their own central bank) is very small (0.4 percent).Even though negligible, this small difference in inflation    between the CFAs and the non-CFA pegged ERR group can be considered as the pay-off of the currency union membership.Basically, there is a lower inflation advantage to the CFA countries from having a fixed ERR and being a currency union.

Keeping money growth under control
Theoretically, lowering inflation should be handled by maintaining lower money growth.The theory holds for the case of the CFAs as they seem to maintain lower rates of money growth relative to their counterparts as their tool of lowering inflation.Table 2 provides the estimation results on money growth in the CFAs relative to the other group.Considering the results of the robust regression (second columns for each sub-sample in Table 2), the CFA countries have respectively 7.6, 9, 11, and 7 percent less growth in their money supply relative to all other non-CFA, the non-CFA-non-pegged, the non-CFA-floating, and the non-CFA countries with pegged ERR.The CFA countries seem to have a restraint growth of money supply.However, the aftermath in comparing the extent of the money supply growth in the CFA countries to their lower inflation level reveals a mismatch.The CFA countries do not have as much lower inflation as they would in accordance to their restraint money growth.Across each sub-sample, the magnitude of the negative money growth is larger than that of inflation (money growth: -7.6 percent versus -4.7 percent in inflation in S1; money growth: -9.5 percent versus -5 percent in inflation in S2; money growth: -11 percent versus -7 percent in inflation in S3; and money growth: -7 percent versus -0.4 percent in inflation in S4).

Output loss
The observed mismatch might be caused by two phenomena within the CFA zones.First, it could be that the fixed ERR is having negative effects on the growth of output per capita and secondly, the growth of money velocity within the CFA zones might be higher than needed.As underlined in the hypotheses section, lowering inflation can be achieved through higher growth of output.However, this cannot be the case in the CFA zones since the CFA countries exhibit negative output per capita growth relative to the non-CFA states (Table 3).
The results of the robust regression in Table 3 show that the real GDP growth within the CFA zones is lower than in any other comparative group in SSA.The real GDP growth within the CFA zones is about 1 percent lower compared to that in all other SSA countries together with the non-CFA non-pegged, and the non-CFA flotting ERR countries.But, the gap between the CFAs and the non-CFA group with pegged ERR is much larger on output growth: the non-CFA group with a pegged ERR outperforms the CFAs on output growth by about 2 percent.
The negative coefficient of the RGDP per capita growth indicates that fixed ERR reduces the extent of output growth in the CFA zones relative to other countries.Most importantly, the larger gap in output growth between the CFAs and the non-CFA group with pegged ERR reveals that those countries with a pegged ERR which have their own national central banks perform much better on output growth than others in general and than the currency union, in particular.
This result implies that the CFA economies are hurt not only through the fixed ERR effects, but also for being locked in a currency union (a common currencya common central bank for all countries in each CFA zones).The direct consequence of the lower output growth in the CFAs is the weak impact of money growth on inflation; in other words, the lower output growth diminishes the effects of money growth on inflation.Doing the math from Tables 2 and 3, in S1, S2, and S3, about 1 percent of the lower money growth compen-sates for output loss.In S4, 2 percent of the lower money growth compensates for the output loss.

Money velocity growth
Recall the discussion point in "case 1", and it appears that the sum of β1 in equation ( 3) and in equation (3.1) is larger in absolute value than the magnitude of β1 in equation (2) (doing the sums of β1 in Tables 2 and 3 across the samples, especially in the robust estimation).This result means the CFA zones have positive growth of money velocity (or higher growth in money velocity) relative to that in other non-CFA countries.The results on money velocity support the assumption and imply that the impact of the restrained money growth on inflation in the CFA countries is also reduced by the growth of money velocity (Table 4).
From Table 4 and under the robust estimation, the CFA countries have respectively 1.6, 2 and 1 percent higher growth of money velocity relative to all the non-CFA, the non-CFA-non-pegged, and the non-CFAfloating ERR countries.So, the impact of the restrained money growth on inflation in the CFA countries is reduced respectively by these amounts of money velocity growth across the sub-samples.The higher growth of the money velocity in the CFA zones might mainly stem from the higher interest rates in these countries as demonstrated in Figure 2 of Appendix I.
After subtracting the sizes of the sum of β1 for output per capita growth in Table 3 and β1 for money velocity growth in Table 4, from β1 for money growth in Table 2, the difference is about the magnitude of the extent of the lower inflation in the CFA countries (for S1, S2, and S3).However, the mismatch persists for S4 where the CFAs are compared to the non-CFA countries with pegged ERR.The CFAs have 7 percent lower money supply growth compared to the non-CFA group with pegged ERR, but the difference in terms of inflation between the two groups is only 0.4 percent.In addition, the difference in money velocity growth between the CFAs and the non-CFA countries with pegged ERR is insignificant and the CFAs have up to 2 percent loss in output against the non-CFA countries with pegged ERR.The sub-sample S4 reveals that the non-CFA countries with pegged ERR have the ability to maintain lower inflation rate (closely as much as in the CFA countries), exhibit similar rates of money velocity growth, and have a higher growth of their output per capita relative to the CFA countries.
The size of the output loss in the CFA being larger than the CFAs' extent of lower inflation in comparison to the scenario within the group of pegged ERR non-CFA indicates that the CFA franc currency union membership could be economically harmful.Gurtner (1999) warns that the CFA zones do not meet the required conditions for an optimum currency area (OCA).According to Gurtner (1999), the CFA countries follow different supply cycles.Thus, they face growth barriers at different points of time.The paths of the GDPs of the CFA countries depend on the fluctuation of the prices of the primary commodities that underline the economies of these countries.
Moreover, there is no trade intensive within the zones that necessitate the reduction of transaction cost by using a common currency.In addition labor mobility across countries within the zones is not intense (only the labor mobility in the informal sector seems to be fulfilled, according to Gurtner (1999).As the CFA union countries do not meet the requirements of the OCA, it is not surprising that locking countries with such heterogeneous cyclical patterns under a common currency could erode their economic performances.The reason of the pronounced difference between the CFA and the other pegged non-CFA in economic performance is the difference in interest rates between the two groups.As shown in Figure 4 of the Appendix I, the pegged non-CFA has more investment per capita relative to the CFA zones.This higher investment per capita could be a result of a lower cost of borrowing (lower interest rates).
In general, the difference in inflation between the CFA and non-CFA countries is decreasing while the gap in investment per capita between them is enlarging (Figure 1 versus Figure 4 in Appendix I).The inflation gap between the CFA and the non-CFA is decreasing over time.The inflation rates of the CFA countries are increasing while that of the non-CFA is diminishing on average.This convergence is due to the fact that output is growing faster in the non-CFA zone and slower in the CFA zones.
The inflation-growth trade-off is a major question in the discussion of the economic performance of developing countries like the CFA states.Is it worth sacrificing output for "lower" inflation?The lower extent of output in the CFA zones is a result of different macroeconomic problems which slow economic activities.In fact, having higher nominal interest rates, the CFA countries experience higher real interest rate which is the cost of borrowing.As shown in Figure 2, the real interest rate has been high and more volatile in the CFA countries than in the non-CFA states.As real interest rate reflects the cost of capital in the production process, facing high real interest rate can limit investments and output growth (Figure 4 in Appendix I supports the lower rate of investment under the CFA zones as a result of higher real interest rates).
The evidence from the hypothetical money model states three main benefits from keeping lower inflation.The first is the transaction cost reduction.The second is the reduction of the capital income tax and the third is the reduction of uncertainty.However, Aiyagari (1990) studies the benefits and the costs of maintaining lower inflation.He mainly shows that the costs of such policy outweighs it benefits.For instance, the reduction of transaction could be achieved by creating more forms of money useable in transaction to earn market rates of interest.Moreover, he argues that reducing money supply in an attempt of keeping lower levels of inflation might not systematically reduce the variability of inflation.Thus, the impact of a lowering-inflation policy on welfare could be marginal (Aiyagari, 1990).Therefore, instead of an inflation lowering policy, which is associated with higher costs, one could simply implement alternative policies, save in costs and reach the same benefits.In addition, a study like Hercowitz (1982) shows that supply shocks have stronger effects on relative prices than the changes in money supply (inflation), at least for the US data.Since wage contracts are usually not fully indexed to price level variability, as the real value of money increase due to inflation reduction, the money amount of the contract is less likely to fully change and compensate for the change in the price levels.Thus inflation reduction is associated to welfare loss (Okun, 1978;Fischer, 1984).Fischer (1984) estimates the sacrifice ratio 16 at 6 percent.Okun (1978), Fischer (1984), and Aiyagari (1990) argue that what matters in improving welfare is the variability of personal consumption of goods and services.Thus the CFA countries incur welfare loss through their alignment to a fixed ERR and a common currency as they experience higher output loss in the long run.Another major finding in this paper is the inflation persistence in the CFA countries.As shown by the coefficient of the lag inflation in Tables 5, 6, and 7, there is a remarkable inflation inertia in the CFA zones.The magnitude 0.7 (0.6 for the fixed effect estimation) of the coefficient of πit−1 indicates that 70 (or 60) percent of the current inflation is due to the past inflation history.The past output growth and money growth have less effect on current inflation compared to the past inflation itself (see the coefficients for the lag of money growth, lag of output growth, and the lag of money velocity growth in each short run estimation table).This finding on the inflation inertia in the CFA zones endorses Chopra (1985) and Loungani and Swagel (2001) who showed that inertial components are more influential in the inflation process in developing countries, especially those with fixed exchange rate regimes.
In summary, the CFA countries exhibit lower inflation relative to their SSA counterparts.Thus, the hypothesis by which fixed ERRs provide a lower inflation is proven for the CFA countries.Nevertheless, the Friedman hypothesis that predicts more loss of output under a fixed ERR relative to a flexible ERR holds also for the CFA countries (see estimation results in Table 3 and also Figure 3).These two theories and the findings in this paper indicate that the inflation-growth trade-off is at work in the CFA zones.However, the goal of an economic policy in any country is to increase the wellbeing of the citizens.Given the empirically proven fact that the costs associated to maintaining lower inflation outweigh its benefits on the country's welfare, one can conclude that fixing the foreign value of their currency is distortive to the CFA countries' economies.Hence, the monetary authorities of the CFA countries could have use alternative policies to improve the countries' welfare rather than lowering inflation.

Conclusion
This paper finds an empirical support for the inflationgrowth trade-off associated with a fixed exchange rate regime (ERR) in the case of the CFA franc currency union countries of Sub-Saharan Africa (SSA).Despite the relatively lower inflation in the CFA countries compared to the non-CFA countries of SSA, the CFA countries experience output losses through their alignment to a fixed ERR and belonging to a currency union.The CFA countries pay high costs in the form of output loss in return to a slightly lower inflation level compared to their SSA counterparts.As lowering inflation has less impact on welfare than the change in output, this trade-off is detrimental to the CFA economies.In fact, the economic objective of any individual is to improve her propensity to consume goods and services.In other words, economic policies of countries should be oriented to the improvement of their welfare.The CFA countries could therefore employ alternative policies to avoid the welfare loss associated with a fixed ERR, and their alignment to a single currency.The welfare loss relative to all other non-CFA countries in general and that relative to the pegged non-CFA countries in parti-cular lead to the conclusion that the CFA countries would have performed economically better under an alternative ERR, and/or by not belonging to the CFA franc currency union.
floating ERRs, and the Non-CFA countries with pegged ERR.Two important facts explain the choice of the country sample.First, the countries of the CFA currency unions 3 started using the common currency -the CFA 1 At the creation of the CFA franc currency, CFA was standing for French Colonies of Africa.Nowadays, the "CFA" of the West African Economic and Monetary Union stands for Communaute Financiere Africaine (African Financial Community) and the "CFA" of the Central African Monetary and Economic Union stands for Cooperation Financiere Africaine (Financial Cooperation in Central Africa). 4

Table 1 .
The long run inflation estimation in Sub-Saharan Africa.

Table 2 .
The Long Run Growth of Money in Sub-Saharan Africa

Table 3 .
The long run growth of the RGDP per capita in Sub-Saharan Africa. is the CFA versus all other non-CFA, S2 is the CFA versus the non-CFA-non-pegged ERR, S3 is the CFA versus the non-CFA-floating ERR, and S4 is the CFA versus non-CFA-Pegged ERR.The dependent variable is the average growth of RGDP per capita. S1

Table 4 .
Long run money velocity growth in Sub-Saharan Africa. is the CFA versus all other non-CFA, S2 is the CFA versus the non-CFA-non-pegged ERR, S3 is the CFA versus the non-CFA-floating ERR, and S4 is the CFA versus non-CFA-Pegged ERR.The dependent variable is the average growth of money velocity.Note: *** p < 0.01, ** p < 0.05, * p < 0.10 The heteroscedasticity-consistent standard errors are in parentheses. S1

Table 5 .
Short run inflation estimation: CFA vs.All Non-CFA countries.