Chapter 8. Exchange Rate Arrangements in the Transition to East African Monetary Union

Paulo Drummond, S. Wajid, and Oral Williams
Published Date:
January 2015
  • ShareShare
Show Summary Details
Christopher S. Adam, Pantaleo Kessy, Camillus Kombe and Stephen A. O’connell 

This chapter considers alternative exchange rate arrangements for the countries of the East African Community (EAC) in the transition to monetary union. Three main considerations shape our analysis. First, while existing exchange rate policies differ in some important ways across the EAC, the partner states wish to achieve a common exchange rate policy during the transition to union. Second, since the transition may well occupy several years or more, the arrangements adopted during this period should be consistent with macroeconomic stability and financial development on a country-by-country basis. Third, transition should be designed to avoid real exchange rate misalignment in the first few years of union.

In focusing on how to move to union, we are intentionally leaving aside the deeper question of whether monetary union is in fact desirable for these countries. Growing empirical literature takes a largely skeptical view, based on the lack of an asymmetry in the impact of economic shocks across the community and the limitations of cross-border labor mobility and other substitutes for intra-community adjustments in monetary policy and exchange rates (European Central Bank [ECB], 2010; Kigabo, Masson, and Masson, 2012).

Debrun, Masson, and Pattillo (2010) add that although a reduction in inflation bias is one of the potential benefits of monetary union, central banks in the EAC may be able to accomplish this unilaterally, through institutional measures that build on their relatively strong inflation performance over the past decade, without ceding control to a supranational institution. But these reservations are not decisive. Global evidence suggests, for example, that entering a monetary union tends endogenously to reduce some forms of macroeconomic asymmetry across members (De Grauwe, 2009). And while asymmetries will surely persist, the transition process itself represents an opportunity to reduce the union’s exposure to asymmetric shocks—including financial and fiscal spillover that may be exacerbated rather than reduced by union—and to strengthen alternative mechanisms for macroeconomic adjustment. The issue of how to move to union therefore remains of considerable interest, both for the EAC and for other regions where political or economic considerations may favor deeper economic integration.

Exchange rate commitments played a central role in Europe’s transition to monetary union, and the main question we address in this chapter is whether the EAC countries should follow suit by adopting a grid-plus-band system in the run up to union. Our answer, on balance, is no. There is a straightforward trade-off in the EAC case between the costs of tight exchange rate commitments during the convergence period and their potential value in acclimatizing the partner states to the loss of intraunion exchange rate flexibility once union is accomplished. The ECB, in its comprehensive report on the transition to union in the EAC, resolves this trade-off in favor of an exchange rate grid (ECB, 2010). We resolve it in favor of managed floats with country-level inflation targets. More specifically, we distinguish two broad phases of the transition, as Figure 8.1 illustrates: a convergence phase of uncertain duration, during which the partner states work to achieve a set of preconditions that may or may not include exchange rate criteria, followed by a brief conversion phase that should be of fixed duration and that follows an irrevocable decision to move to union. Our view is that tight exchange rate commitments are appropriate only in the final stage, a stage we view as some way off and that we argue should probably be no more than a year in duration.

Figure 8.1The Transition to Monetary Union

We begin the chapter by setting out the initial conditions we consider important for guiding the choice of exchange rate arrangements during the transition. Then, we review the evolution of exchange rate policy in the EAC, emphasizing the increasingly uniform commitment of member states to an open capital account. Next, we assess the degree to which international capital mobility constrains monetary policy in the EAC. While direct measures suggest a limited degree of financial integration with global markets, an indirect, regression-based approach delivers a distinctly different impression. Particularly in Kenya and Uganda, we find that foreign exchange market pressures generated by discretionary changes in monetary policy are consistent with a substantial degree of short-term capital mobility. These findings suggest that tight, but adjustable exchange rate commitments could be subject to speculative attack, particularly if made over any considerable period.

Following that, we address the appropriate exchange rate regime for the transition. We characterize alternative regimes according to the approach they adopt to anchoring inflation at the community-wide level and the scope they leave, if any, for monetary autonomy at the country level. We defend our preference for a managed float in terms of the structure and history of the EAC and with reference to our earlier evidence on capital mobility.

Then, we focus on the conversion phase. We discuss the introduction of an East African Currency Unit (EACU) and the choice of conversion parities. We argue that the method for choosing conversion parities should be subject to advance agreement by the partner states, both to avoid damaging misalignments at the outset of union and to discourage competitive depreciations during the convergence phase. The costs of initial misalignment will of course be relatively minor if nominal wages and prices can adjust rapidly to real exchange rate disequilibrium. A straightforward econometric approach suggests, however, that wage and price adjustment is not markedly more rapid in the EAC countries than in emerging market or advanced economies with flexible exchange rates. This finding underscores the relevance of adequate exchange rate assessments at the outset of the convergence phase.

We conclude our analysis by briefly considering the postunion impact of monetary union on real exchange rates and competitiveness. Using evidence from the CFA franc zone, we argue that one of the main impacts of union is likely to be a convergence of equilibrium real exchange rates across partner countries. Thus, for example, equilibrium real exchange rates will depreciate in resource-abundant members of the monetary union, moderating some portion of their Dutch disease, whereas resource-poor partners will experience a real appreciation and a shift in the composition of their exports toward the region and away from the rest of the world.1 Finally, we close the chapter with a summary of our contributions and a pointer to areas for further research.

Current Exchange Rate Policies

The three large countries of the EAC currently operate managed floats, with fully open capital accounts in Kenya and Uganda and a liberalization under way that will remove most of the remaining capital account restrictions in Tanzania. The context for these choices predates the reconstitution of the three-member community in 2001 and the accession of Burundi and Rwanda in 2007. Operating independently in the first half of the 1990s, each of the big three (Kenya, Tanzania, Uganda) implemented a set of reforms designed to eliminate a distorting system of foreign exchange rationing, develop an interbank market for foreign exchange, and establish a unified and market-determined exchange rate. Uganda and Kenya liberalized their exchange controls very substantially in the mid-1990s, dismantling controls on the capital account as well as the current account. Tanzania liberalized the current account fully, but has only recently begun to liberalize its capital controls as part of harmonization efforts within the EAC. It is highly likely that the exchange rate policy framework the union-wide central bank adopts will be close in outline to the framework currently operated in the big three countries, with a flexible, market-determined exchange rate and a largely open capital account. Burundi and Rwanda appear to be on a gradual path to such a regime, and would be likely to continue on this path even in the absence of monetary union.

In Adam and others (2012) we provide a summary of current foreign exchange market arrangements and operating procedures in the EAC. Four of the five countries are formally committed to exchange rate convertibility for current account purposes.2 There are differences across countries, however, in de facto exchange rate flexibility, stringency of capital controls, and sophistication of interbank foreign exchange markets. In its classification of de facto exchange rate regimes, the IMF characterizes the regime in Kenya, Tanzania, and Uganda as managed floating, Rwanda as a crawl-like regime, and Burundi as a stabilized regime (as of 2011). These differences in exchange rate flexibility are apparent, particularly over the past few years, in Figure 8.2, which tracks EAC exchange rates against the U.S. dollar over the past decade.

Figure 8.2East African Community Exchange Rates

Source: IMF, International Financial Statistics database.

Note: The figure shows log differences relative to 2005m12, multiplied by 100 to convert into approximate percentage differences.

In terms of operating procedures for exchange rate management, Kenya, Uganda, and Tanzania deploy comparable frameworks in which the authorities commit to a floating exchange rate whose value is determined in the interbank foreign exchange market, and structure their foreign exchange operations around reserve coverage and liquidity management objectives established in their reserve money programs. The reserve money programs in these countries are anchored, in turn, on explicit targets for inflation. Similar arrangements are not yet fully in place in Rwanda and Burundi, but both countries are working to develop interbank foreign exchange markets and are moving in the direction of greater exchange rate flexibility.

These differences in exchange rate management within the EAC are reflected in the structural characteristics of the foreign exchange markets. In Kenya, Uganda, and Tanzania, the markets are perceived to be broadly competitive, and the central banks are an important but not decisive player in the market. Central banks may seek to trim short-run volatility in the market, but would not normally expect to be able to decisively influence the underlying rate. In Rwanda and Burundi, where foreign aid flows to government account for around half of all foreign exchange inflows, and where the private financial sector is less developed, the central banks are the decisive players and, to a large degree, still the market makers. Residual controls on the capital account in Tanzania and the lower levels of financial sector and capital market development in Rwanda and Burundi would suggest that these countries continue to experience significant insulation from portfolio capital flows and the associated risk of speculative pressures on the currency. As we will see in the next section, these differences are apparent in indirect measures of capital mobility, although they do not show up strongly in direct comparisons of the volume of private capital flows.

Assessing Short-Term Capital Mobility

Portfolio behavior is a potentially powerful source not only of day-to-day pressures in foreign exchange markets, but also of speculative attacks that can test a central bank’s commitment to a fixed exchange rate. These attacks can be costly for the economy whether they succeed or fail. A successful attack produces a large devaluation, with its impact on inflation, external debt burden, and central bank credibility, and a failed attack may require an aggressive monetary policy response that damages the domestic economy through very high interest rates. The celebrated trilemma proposition in international macroeconomics states that in the presence of high capital mobility, a central bank must choose between exchange rate targets and domestic targets for monetary policy: it cannot sustain strong exchange rate commitments unless it is prepared to abandon domestic objectives when this is required to defend the exchange rate. This reasoning is one of the influences behind the move in Kenya and Uganda to exchange rate flexibility in the 1990s, when they were also choosing to open their capital accounts. Regardless of the other merits of flexible exchange rates (such as in supporting exchange rate unification, and allowing rapid adjustment of the real exchange rate to terms of trade shocks), a flexible rate was thought necessary to support each country’s commitment to an open capital account.

Concerns about the trilemma also apply, of course, to exchange rate commitments during the transition to monetary union. Once union is consummated, intraunion exchange rates are eliminated as potential sources of speculative capital flows. But during the transition, the advisability of tight exchange rate commitments depends on the degree to which these commitments may be exposed to speculative attack. This in turn depends on the degree of de facto capital mobility on a country-by-country basis, particularly in short-maturity claims, which are thought to be more responsive to transitory changes in market sentiment.

Capital mobility can be assessed either directly, by looking at the stringency of legal restrictions on capital movements (de jure openness) or the volume of cross-border flows (de facto openness), or indirectly, by assessing the impact of portfolio behavior on domestic interest rates and exchange rates. We employ both approaches in this section.

Figure 8.3 shows the Chinn and Ito (2008) measure of de jure financial openness, which is based on the prevalence of legal restrictions on capital account transactions as reported in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions. This measure varies quite substantially across the EAC. Even excluding the industrial countries, Burundi was at the bottom of the global distribution of de jure financial openness in 2010 (7th percentile), and Tanzania was in the 44th percentile along with South Africa. Rwanda was in the 45th, and Kenya the 68th. Uganda was in the most financially open group in the world in 2009, whether including industrial countries or not.3

Figure 8.3De Jure Capital Account Openness, 1990–2010

These differences in legal openness are not strongly reflected, however, in measures based on observed private capital flows. Table 8.1 reports a measure of de facto financial openness based on gross transactions in the capital account. The Lane and Milesi-Ferretti measure (2007) is the sum of total external financial assets and liabilities as a share of GDP; this is the stock analog, for the capital account, of a widely used measure of de facto openness to trade in goods and services.4 The table also reports an adjusted Lane and Milesi-Ferretti measure that corrects for the prevalence of what might best be characterized as nonmarket financial claims in low-income countries. Following Dhungana (2008), this measure deletes official reserves (excluding gold) on the asset side and concessional debt owed to official creditors on the liability side. We show medians rather than means because of the presence of extreme values in the sample.5

Table 8.1Stock Measures of De Facto Capital Account Openness(Financial assets + liabilities/gross domestic product [%], group and period medians)
Country or RegionLane and

Milesi-Ferretti Measure

Adjusted Lane and

Milesi-Ferretti Measure

East African Community8747
South Africa157148
Latin America and Caribbean155127
South and East Asia14495
Sources: Lane and Milesi-Ferretti (2007); IMF, World Development Indicators database and Global Development Finance database.Note: OECD = Organization for Economic Cooperation and Development.
Sources: Lane and Milesi-Ferretti (2007); IMF, World Development Indicators database and Global Development Finance database.Note: OECD = Organization for Economic Cooperation and Development.

As noted by Dhungana (2008), sub-Saharan Africa appears to be significantly less integrated into global capital markets once the nonmarket flows are excluded. The EAC is comparable to the rest of Africa in this respect, including Kenya, which is surprisingly low given its relatively advanced degree of financial development (see also O’Connell and others, 2010).6Figure 8.4 tracks recent developments by showing a flow version of the unadjusted Lane and Milesi-Ferretti measure—the ratio of gross financial flows to GDP for the past decade. While all EAC countries except Tanzania have experienced at least a modest trend increase in gross flows in recent years, these flows remain small by comparison with GDP, both in absolute terms and by the standards of other countries in sub-Saharan Africa. They also differ by less across the countries of the EAC than might be expected given their sharp differences in de jure openness.

Figure 8.4Gross Financial Account Flows

Source: IMF, International Financial Statistics database.

Note: Gross financial flows consist of foreign direct investment, portfolio flows, financial derivatives, and long-term debt flows (denoted “other investment” in the International Financial Statistics database). Our measure of long-term debt flows omits flows attributed to “general government” or “monetary authority.”

The direct measures therefore suggest that a substantial gap exists between the degree of legal openness in the EAC and the actual volume of international capital flows. What matters for monetary policy, however, is capital mobility at the margin. If this is increasing rapidly, or if short-term capital movements are poorly captured in the capital account, then these direct measures may underestimate the degree to which capital mobility constrains monetary policy in the EAC.

The remainder of this section assesses de facto short-term capital mobility in the three large countries of the EAC indirectly by adapting a time-honored methodology that dates from Porter (1972) and is widely used in the industrial country and emerging markets literature. The central idea is that an unanticipated monetary expansion, such as an increase in domestic credit to the government, reduces domestic interest rates and thereby encourages portfolio substitution in favor of foreign assets. If the monetary authority is committed to a fixed exchange rate, the induced capital outflow drains international reserves, producing a contraction in the monetary base that partially or fully offsets the original expansion. The offset coefficient is the share of the original monetary expansion that is offset through capital outflows. This can be measured in principle at each horizon (such as a month or a quarter), with the size of the offset coefficient indicating the degree to which de facto short-term capital mobility imposes a constraint on monetary policy. In Kenya, Tanzania, and Uganda, where the monetary policy framework and exchange rate regime have been reasonably stable for over a decade, we have sufficient data to apply a modified version of the offset coefficient approach to measure the degree to which portfolio behavior imposes a constraint on monetary policy.

Two adaptations are required, however, before this approach can be applied to countries on managed floats, rather than fixed exchange rates, and in which the central bank routinely engages in foreign exchange transactions with other parts of the public sector. To accommodate flexible exchange rates, we broaden the focus from reserve losses to exchange market pressure, defined as the sum of reserve losses and exchange rate depreciation.7 An unanticipated monetary expansion now creates a combination of reserve losses and exchange rate depreciation, with the size of the overall impact dependent on the degree of capital mobility and the mix between reserve losses and depreciation dependent on the central bank’s behavior.

The second adaptation arises from the prevalence of intragovernmental transactions in foreign exchange on the balance sheets of central banks in low-income countries. An aid inflow, for example, initially produces a simultaneous increase in central bank reserves and reduction in domestic credit, as the domestic-currency value of the inflow is credited to the relevant government ministry. The induced change in domestic credit, however, has nothing to do with monetary policy, and the directly offsetting change in international reserves has nothing to do with private capital flows. The reverse pattern occurs when a government ministry spends a portion of previously received budget support on imports. We control for these intra–public-sector transactions by allowing for an unobserved “aid” shock that is unrelated to monetary policy and that moves the current values of net domestic and net foreign assets in equal and opposite directions on the central bank’s balance sheet.

To identify the dynamic impact of monetary policy changes on exchange rates and reserves, we estimate monthly structural vector autoregressions in logs of net domestic assets (NDA), net foreign assets (NFA), and the nominal exchange rate(S) for January 2002 to May 2011.8 We proceed separately, country by country, measuring all three variables in first differences in order to guarantee stationarity of the residuals.9 Our interest is in the impulse responses of the exchange rate and reserves to changes in monetary policy. To estimate these responses, we need to identify a set of unanticipated changes in monetary policy in the sample. We do this by assuming that the reduced form shocks to net domestic assets, net foreign assets, and the exchange rate, all of which we can observe directly from the reduced form vector autoregression, are linearly related to a set of unobservable structural shocks to monetary policy (M), aid (A), and the private sector’s desired balance of payments (B). Denoting the latter shocks by εt, our identifying assumptions can be summarized by

along with the requirement that the structural shocks are mutually uncorrelated, with zero means and unit variances.

At the heart of this identification scheme is the relationship ut = Cεt, where C is an invertible matrix with positive terms on the diagonal.10 We need three restrictions to identify C, which we obtain in equation (8.1) by assuming (1) that the private sector’s structural balance-of-payments shock εtB does not contemporaneously affect the change in domestic credit; (2) that the aid shock εtA does not contemporaneously affect the exchange rate, and (3) that all contemporaneous and mutually offsetting movements in domestic credit and international reserves reflect internal transactions in foreign exchange between the central bank and the rest of the public sector.11

The expected signs of the elements of C follow those of the first matrix in equation (8.1). The diagonals and C31 are therefore expected to be positive, while c21 and c23 are expected to be negative. The variances of the structural shocks can be inferred from C using c11 = σM, c22 = σA and c33–23 = σBTable 8.2 shows our estimates of the elements of C. These estimates have the correct signs in all cases and are generally highly statistically significant.

Table 8.2Coefficient Estimates for the C Matrix
C112.15 (5.16”*)4.37 (5.9***)5.54 (6.05***)
C21−0.13 (-0.29)−0.96 (-1.4)−2.41 (-2.50*”)
C224.14 (14.90”*)6.74 (14.4*”)8.53 (14.42***)
C23−1.14 (-2.89***)−0.39 (-0.8)−2.08 (-4.18*”)
C310.92 (2.72”*)0.08 (0.31)1.45 (5.22***)
C331.53 (7.49”*)1.70 (14.40*”)1.43 (5.78***)
Source: Authors’ calculations.Note: t–statistics are in parentheses; * denotes significance at the 10 percent level, ** at the 5 percent level, and *** at the 1 percent level. All vector autoregressions are estimated from January 2002 to May 2011 (excluding lags) and include seven lags and monthly seasonal dummies.
Source: Authors’ calculations.Note: t–statistics are in parentheses; * denotes significance at the 10 percent level, ** at the 5 percent level, and *** at the 1 percent level. All vector autoregressions are estimated from January 2002 to May 2011 (excluding lags) and include seven lags and monthly seasonal dummies.

In Figure 8.5, we show the cumulative impulse responses of net foreign assets and the exchange rate to a one-time shock to domestic credit. These results should be treated with caution; the standard error bounds around our vector autoregression coefficients are large, and generally only the first month of impulse responses is statistically significant. The results nonetheless suggest that there is greater exposure to short-term capital mobility in the EAC than is indicated by the de facto direct approaches. In all three countries, an unanticipated monetary expansion produces a decline in reserves and a depreciation of the nominal exchange rate, consistent with a managed float in which the authorities “lean against the wind” to limit the short-run volatility of the exchange rate. These impacts are small in Tanzania, consistent with its relatively closed capital account as indicated by the de facto measures. Kenya and Uganda, by contrast, show substantial impacts, with roughly half of the monetary impulse offset through a drain in reserves within a six-month horizon, despite the willingness of their central banks to tolerate a weakening exchange rate.

Figure 8.5Cumulative Impulse Responses to a One-Unit Shock to Net Domestic Assets

Source: Authors’ calculations.

Note: Three-variable structural vector autoregression in the change in adjusted net domestic assets, the change in adjusted net foreign assets (dNFA) and the change in the log of the nominal exchange rate in local currency per U.S. dollar [d log(S)]. See Annex 8.1 for a definition of the variables. KEN = Kenya; TZA = Tanzania; UGA= Uganda.

The cumulative impact of a monetary expansion on exchange market pressure, defined as the sum of reserve losses and exchange rate depreciation,12 is roughly equal to the size of the initial monetary impulse in Kenya and Uganda, and is easily twice the impact observed in Tanzania. Kenya’s impacts emerge more gradually than Uganda’s, consistent with the greater openness of Uganda’s capital account; but in both countries, the results suggest that portfolio behavior exerts a substantial constraint on monetary policy. In Kenya’s case, in particular, the size of the cumulative impacts suggest that the de facto measures considered previously may seriously underestimate that country’s integration with global financial markets.

In interpreting these varied results, we place weight on the increasing sophistication of interbank markets in the EAC countries, and also on the recent liberalization of capital controls within the EAC itself. While the EAC countries are well short of the trilemma, their scope for maintaining tight exchange rate commitments while simultaneously pursuing domestic monetary policy objectives is limited. This is particularly true for Kenya and Uganda. But as capital movements become freer within the EAC, portfolio capital will be able to move between each member and global markets indirectly, via transit through partners whose capital accounts are more open to the rest of the world. As in a free trade area without a common external tariff, the de facto exposure of each participant to external markets will tend to converge to that of the most open partner.

Exchange Rate Options for the Convergence Phase

The convergence phase is of uncertain duration because it depends on political developments and on the pace at which the partner states can put convergence criteria in place. Exchange rate policies for this phase should therefore be consistent with the successful conduct of monetary policy for a potentially extended period. In analyzing the options, we place a particular emphasis on the transparency of alternative arrangements, their suitability for the EAC countries, and how they anchor inflation. We assume that, regardless of the option chosen, countries will work during the convergence phase to deepen market integration and harmonize financial market regulations.

Europe’s Experience

Should nominal exchange rate commitments, in particular, convergence criteria for intraunion nominal exchange rates play a central role in the lead-up to monetary union? The complete elimination of intraunion exchange rates is of course the single most obvious consequence of adopting a common currency, and the appropriateness of this commitment plays a central role in the theory of optimal currency areas. In Europe, moreover, where the convergence phase spanned nearly three decades, nominal exchange rate commitments played a prominent role throughout transition. The exchange rate mechanism specified a grid of bilateral central exchange rates between the potential partners and a set of country-specific bands within which exchange rates were allowed to fluctuate. The central rates could be adjusted by mutual agreement, but for the two years leading to entry, countries were not to devalue their central rates and were to manage their economies so as to remain within a ±2.5 percent band of the parities without experiencing severe pressure. The central rates ultimately served as conversion rates: in May 1998, the entering members announced that they would convert to the euro on January 1, 1999, at the prevailing central rates. At the time of the announcement, no member was more than ½ percent from its central parity. Subsequent movements gradually closed the gaps to zero in the weeks and days before conversion (De Grauwe, 2009).

Not surprisingly, an agreement to limit cross-country exchange rate movements during the transition period features prominently in the ECB study of the prospective East African monetary union (ECB, 2010). A mutual grid is also a feature of the transitional arrangement among the countries of the Gulf Cooperation Council. But the advisability of such a scheme for the EAC countries is not obvious. Unlike either the European countries or the Gulf Cooperation Council members, who were on pegged arrangements before entering the transition period, the three large countries of the EAC operate de facto flexible exchange rate regimes and have done so for some time. The European grid, moreover, was subject to multiple speculative attacks during the transition, and while the Gulf states have been largely free of such concerns, their success in maintaining fixed pegs to the dollar predates the transition period and may be due in large part to an unusual combination of labor market flexibility and fiscal flexibility. These conditions are not prevalent in the EAC.13 For these reasons, we consider a range of options for the convergence phase, including one in which exchange rate commitments are absent altogether.

Key Trade-Offs

Independently of their exchange rate histories, any set of countries planning to enter a monetary union faces the looming elimination of intraunion exchange rates as a macroeconomic adjustment mechanism. Here, we briefly review three potentially important arguments in favor of convergence criteria for nominal exchange rates during the transition, and two arguments against.

The first argument in favor of limiting the flexibility of intraunion exchange rates during the transition is that this may endogenously increase the flexibility of other national economic variables—wages and prices, labor mobility, fiscal policy—that will necessarily bear the brunt of addressing real exchange rate misalignments in the postunion period. The partners will therefore enter the union on a stronger footing for handling asymmetric shocks. This argument, however, accepts a set of well-known costs of limited flexibility, articulated in the following, for benefits that are highly uncertain and that may be secured, at least in part, through policy initiatives that directly target enhanced flexibility.

The second argument is that members may seek to depreciate their exchange rates during the transition in order to enter the union on a highly competitive footing. By preventing this behavior, nominal exchange rate criteria would avoid the erosion of national inflation anchors during the transition, while also limiting the extent of real exchange rate misalignment at the outset of union. We will argue here that the second of these concerns is best handled by setting conversion rates based on an empirical assessment of real exchange rate misalignment. We therefore focus here on the first.

The concept of local exchange rate competition plays a central role in the work of Debrun, Masson, and Pattillo (2005, 2010) on monetary unions in Africa. These authors argue that the formation of a union reduces the average degree of inflation bias across partners, by eliminating the ability of each partner to depreciate relative to the others. The empirical relevance of this argument, however, depends crucially on the strength of the intragroup externality from exchange rate depreciation. In Europe, this externality was widely thought to be important because the union itself was each member’s largest trading partner. At least currently, however, the EAC countries trade much less with each other than the European countries did. It is doubtful, therefore, that intraunion exchange rates, as opposed to exchange rates vis-à-vis global markets, are an important focus of national concerns about competitiveness.

If the intraunion externality is relevant for the EAC, therefore, it must operate either via large anticipated increases in local trade or via linkages to global markets. One such channel may be competition for foreign direct investment (FDI). If the partner states view themselves as competing for labor-using FDI that would serve either a growing EAC single market or global export markets, then intraunion depreciation may be viewed as a way of attracting such FDI by reducing national wage costs, measured in global currencies, relative to the wage costs of partners. Our view is that although some such motivation may be present,14 it can be adequately contained through a combination of convergence criteria on inflation and the option of adjusting conversion parities for severe misalignment.

The final argument in favor of exchange rate criteria is that a successful union requires a full understanding and acceptance by the partner states of their impending loss of monetary sovereignty. Exchange-rate commitments involve a reduction in monetary autonomy and may also involve explicit cooperation across partners. As convergence criteria, therefore, they may help to ensure that partner states confront the implications of a decision that may be costly to maintain as well as to reverse (Beetsma and Giuliodori, 2010). Europe’s experience suggests that this concern may be the strongest of the three we have mentioned. The current challenges of the euro area have emerged despite a process of political legitimization that spanned nearly 35 years and that allowed partners with serious reservations about monetary integration to remain outside the euro area, for an initial period for all European Union members, and indefinitely in at least one case (the United Kingdom). Our view, however, is that explicit discussions and ex ante agreements on fiscal union are the appropriate locus for handling this concern in the EAC case.

Any considerations in favor of exchange rate criteria have to be balanced against two potentially substantial costs associated with limited nominal exchange rate flexibility during the transition. The first is exposure to speculative attacks. We argued in the previous section that this cannot be ignored in the EAC, in the context of open capital accounts and an ongoing increase in de facto capital mobility.

The second is the loss of nominal exchange rate flexibility as a mechanism for avoiding intraunion real exchange rate misalignments during a potentially extended transition. The countries of the EAC will remain subject to asymmetric shocks during the transition period, including transitory changes in the terms of trade, aid, or financial flows and more permanent shocks associated with natural resource discoveries and, possibly, catch-up growth in Burundi and Rwanda.15 If domestic wages and prices are relatively inflexible, nominal exchange rate adjustments will have an important role to play in securing the requisite real exchange rate adjustments. The next section provides indirect and perhaps surprising evidence of the limited degree of short-run wage/price flexibility in the EAC. While we view this evidence as provisional, it suggests a potentially important ongoing role for intraunion flexibility in nominal exchange rates.

A Taxonomy of Alternatives

Table 8.3 summarizes the five leading options for exchange rate management during the convergence phase. The options differ on two main and related dimensions. The first is the strength of exchange rate commitments on a country-by-country basis and therefore the degree to which exchange rate commitments provide an anchor for inflation. The options here range from a managed float system in which countries make no exchange rate commitments, to an external grid system in which each country unilaterally pegs to a global currency. The second is the allocation of monetary policy autonomy across the union, defined in terms of the de facto locus of policy sovereignty and the scope for directing monetary policy to internal objectives like inflation or aggregate demand. Here the options range from a managed float system, in which sovereignty remains fully at the national level and flexible exchange rates allow considerable autonomy for monetary policy, to a collective anchor system, in which a supranational institution acquires policy authority in advance of formal union, or an external grid system, in which sovereignty remains at the national level but an exchange rate peg receives priority over internal objectives. In what follows, we first describe the five alternatives and then assess their relative merits in the context of the EAC.

Table 8.3Options for Exchange Rate Management During the Convergence Phase
OptionNominal anchorLocus of monetary policy autonomy
1—Managed float with inflation targetsNational money growth or inflation targetsNational
2—External gridNational exchange rate pegsNational (limited by peg)
3—Delegated anchor and internal gridAnchor country: money growth or inflation; Others: exchange rate peg to the anchor countryAnchor country: national; Others: national (limited by peg)
4—Internal grid with inflation targetsNational money growth or inflation targets, coexisting with limited flexibility of intra-EAC paritiesNational (limited by grid)
5—Collective anchorCollectively managed money growth and/or inflation targets, possibly coexisting with limited flexibility of intra-EAC paritiesShared (possibly limited by grid)
Source: Authors.Note: EAC = East African Community.
Source: Authors.Note: EAC = East African Community.

Option 1: A Managed Float

Our first option combines country-level managed floats with convergence criteria on inflation or, equivalently, jointly agreed inflation targets. This option is close to the status quo in the EAC (see the previous discussion), where all countries but Burundi maintain a fairly clear hierarchy in which reserve money growth is the dominant inflation anchor and exchange rate intervention is mainly limited to smoothing short-run volatility.16 In a managed float system, therefore, the national central banks would continue to operate their existing frameworks, basing their monetary programs on jointly agreed inflation targets that might in principle be country specific, but with minimal adjustments to existing national targets that could be set at a uniform 5 percent. Monetary autonomy would remain fully at the national level, and exchange rate flexibility would leave considerable room to direct policy to domestic ends (including prominently, the inflation target). The partner states would pursue fiscal convergence criteria as entry requirements and develop a set of union-wide fiscal safeguards to be operated postunion. They would work out the operational details of the postunion monetary framework and put structures in place to increase information sharing about the conduct of monetary policy. Explicit exchange rate commitments would be deferred until the final stage of the move to union.

Option 2: An External Grid

In an external grid system, each country specifies a central rate against a global currency, such as the U.S. dollar or the euro, and commits to keep its exchange rate within a predetermined band around this rate. Gulf Cooperation Council members operate such a system, in which each member has committed to maintaining its preexisting peg to the U.S. dollar during transition.

In its impact on exchange rates, an external grid system is the equivalent of a system of mutual pegging between the n member currencies, supplemented by an explicit link of any one of their currencies (or any basket) to global currencies. The external link is crucial to the anchoring properties of the system, because although a set of n external pegs and bands implies a grid of n – 1 bilateral cross-rates and bands within the community, the reverse is not true. Mutual fixity among n partners—what we will call an internal grid—determines only n – 1 exchange rates, leaving the relationship of the group to external currencies indeterminate (the so-called n1 currency problem).

By tying the system to external currencies, an external grid also provides an inflation anchor for individual countries and the system as a whole. Policy autonomy remains at the country level, although the pegs leave relatively limited scope for discretionary monetary policy.

Option 3: A Delegated Anchor

Neither of our first two options involves an explicit mechanism for stabilizing cross-rates among union partners. An external grid stabilizes these rates as a side effect of country-level links to the global currency, while a float leaves them tethered only by the commitment to a common inflation target. In a delegated anchor system, an internal grid emerges through individual pegs to the currency of a selected union partner. In this approach, the partners delegate one country from among them to maintain a strong domestic anchor. Each of the other countries then commits to keeping its own currency within a narrow (or narrowing) band of the delegated country’s currency, with an arrangement for coordinated intervention when a member’s rates go to the weak edge of the band. Within this system, the exchange rates of partner countries with respect to global currencies reflect those of the delegated currency.

While the European exchange rate mechanism is sometimes interpreted as a nonanchored internal grid system (such as by ECB, 2010), it in fact operated as a delegated internal anchor system. Germany, the largest economy of the community and the one with the strongest post–World War II inflation performance, provided the de facto delegated anchor.

Option 4: An Internal Grid

Option 4 combines a grid of central parities with inflation targets at the country level. This is the option the ECB report favors. Its exchange rate component mimics the internal grid system operated in Europe during the long transition to union. A centerpiece of the European transition was the establishment of a grid of central parities between national currencies and the European currency unit (ECU), the precursor to the euro. By committing to stay within a narrow band of the ECU, each country was in effect committing to keep its own exchange rate movements relative to external currencies close to those of its partners.

An internal grid system with country-level inflation targets is, in principle, overconstrained, in the sense of imposing 2n – 1 constraints on nominal variables (n inflation targets and n – 1 internal exchange rates) when only n are independently feasible for any extended period. For consistency, therefore, inflation targets would have to be derived from the exchange rate commitments, or vice versa. But even with consistency across targets, the system lacks transparency. Policy autonomy would rest at the national level, but without clarity about how national responsibilities should be allocated between anchoring inflation and maintaining exchange rate commitments.

Option 5: A Collective Anchor

For completeness, we include as a final option a collective anchor system that would vest the authority for anchoring union-wide inflation in a supranational body composed of representatives of the national central banks. In principle, such an authority could devote either weak or strong attention to stabilizing cross-rates among union partners; the key is that the structure as a whole would be anchored indirectly through allegiance to money growth and/or inflation targets union wide. A collective anchor system with intraunion exchange rate commitments would come closest to mimicking the monetary framework that is likely to prevail under the union, in which intraunion exchange rate changes will be absent and union-wide monetary policy will be formed through a collective process over which all members have some influence. A system without intraunion exchange rate commitments would be similar to our managed float option, but with significant policy sovereignty ceded immediately to a supranational agency. The task of the supranational body would be to coordinate national policies that continue to employ internal anchors (money growth rates or inflation forecasts). Given the monetary frameworks currently in use in EAC countries, this might be accomplished by creating a zone-wide financial program to which the national reserve money programs are subordinate in some well-defined sense.

Assessing the Alternatives

Two of our options, the external grid and the collective anchor system, appear to lie at unsuitable extremes for the EAC countries. The external grid gives up all the benefits of exchange rate flexibility for a potentially extended period, and faces potentially existential challenges from speculative attacks in the context of open capital accounts and growing de facto capital mobility. Unlike the Gulf Cooperation Council countries, the three large countries of the EAC definitively abandoned their adjustable peg regimes and have been operating flexible exchange rate regimes, in most cases, since the mid-1990s. The reasons for this are various and include unfavorable experiences with exchange controls, vulnerability to substantial current account shocks, a desire for some degree of monetary autonomy in the context of an open capital account, and a desire to spur the development of domestic markets for exchange risk. Any disadvantages of reverting to fixed but adjustable exchange rates are greater, moreover, in a context of uncertainty about the duration of the transition.

The collective anchor system fails for a very different reason: it transfers sovereignty prematurely and ambiguously in the context of unclear political commitments to union by the partner states. If the authority of the union-wide agency is seen as contingent on unresolved political decisions by national governments, the agency may find itself either unable or unwilling to constrain national policies during the transition. The system may then evolve into a de facto delegated anchor or a managed float system, but with substantial policy uncertainty in the meantime and with damage to the credibility of the union-wide central bank.

Among the remaining three options, the delegated anchor and internal grid systems both feature strong convergence criteria on nominal exchange rates. The delegated anchor system has the advantage of greater transparency but the near-fatal disadvantage of demanding a highly asymmetric sacrifice of policy autonomy across the partners. In effect, four of the five partners acquiesce to the monetary policy choices of the fifth. It may be possible to contemplate a political bargain in which Kenya—the obvious delegated anchor in terms of economic size and long-term inflation history—retains monetary policy autonomy during the transition while the other partners operate peg-plus-band systems vis-à-vis the Kenya shilling. However, the content of such a bargain is unclear, and in its absence it is extremely difficult to imagine the larger partners, particularly Tanzania, accepting a subordinate role. A delegated anchor system seems almost as unlikely to emerge endogenously from an internal grid system as it did in Europe.

The list of plausible ex ante choices for the convergence period, therefore, narrows to two: a managed float system or an internal grid, in both cases supported by country-level inflation targets. We favor the managed float option, which minimizes the disturbance to monetary frameworks that are currently in use in the EAC and have performed successfully, in the larger countries, for over a decade. A managed float is not only robust to a potentially protracted convergence period, but also supports the continued development of domestic financial markets and the continued refinement of existing policy frameworks, including moves toward inflation targeting. In light of the tradeoffs reviewed earlier, the managed float also retains nominal exchange rate flexibility, a reality that traders and financial markets have become accustomed to in the larger countries of the community and that has probably had some protective effect with respect to speculative movements against national currencies. As we argued previously, intraunion exchange rate flexibility also has a continuing role to play in facilitating adjustments to asymmetric shocks to trade, aid, and longer-term capital flows.

We acknowledged previously that a managed float system could be subject to manipulation by partners seeking to enter the union with relatively weak exchange rates. More generally, flexible exchange rates can be a source of short-run real exchange rate misalignment, due to thin market effects or other sources of volatility unrelated to the fundamentals. Our view, however, is that the combination of convergence criteria on inflation and adjustments for misalignment at the conversion step handle these concerns adequately.

A final and potentially serious concern about a managed float system is the flip side of its strongest advantage. By comparison with the other options, a managed float is a minimalist option for the EAC, not just in the sense of leaving existing anchors largely unchanged, but also in the deeper sense of requiring the smallest sacrifice of monetary policy autonomy during the convergence phase. In a managed float system, the national authorities agree on an inflation target but do not subordinate domestic objectives to exchange rate commitments and do not coordinate policy instruments explicitly during the convergence period. Such a system would preserve the strengths of existing national frameworks (especially for the large countries), and in that sense would be highly robust to uncertainties about the timing of the move to union. It would do little, however, to create, legitimize, and strengthen the union-wide institutions that will be fully responsible for policy from the conversion point forward. By the same token, it would do little to confront national stakeholders with the costs of union in terms of the inevitable loss of policy sovereignty. Because this option demands so little sacrifice of sovereignty, partner states must be more purposive in their investment in and commitment to EAC-wide institutions and the delegation of at least advisory functions to the proposed East African Monetary Institute (EAMI) during the convergence period.

The Conversion Phase

Once the requisite macroeconomic convergence criteria have been satisfied, the final steps in the consummation of monetary union are the transfer of sovereignty to a supranational institution and the replacement of national currencies with a union-wide currency. In Adam and others (2012), we argue that these steps should be preceded by a conversion phase of fixed and short duration (perhaps about two calendar quarters or so) during which the partner states commit to maintain their exchange rates within narrow bands of a set of preannounced conversion parities. This phase is not strictly necessary; an unannounced “overnight” transition at prevailing market exchange rates could work equally well if it were prepared logistically in advance by the partner states, immediately understood and embraced by market participants, and preceded by adequate internal surveillance to prevent competitive depreciation. The implausibility of these conditions is what recommends our approach, which draws heavily on the successful final transition in Europe. In this section, we briefly review the nature of the transition, including the role of a new regional currency and the choice of conversion parities.

Introducing the East African Currency Union

Despite the current travails of the euro area, the successful replacement of the individual currencies of its 11 founding members with the euro on January 1, 1999, was acclaimed in all quarters and offers a valuable template for managing the final stages of the transition to monetary union in East Africa. Based on this model, the conversion phase should commence with a major, union-wide announcement of (1) a date for the creation of full monetary union, (2) a commitment by the partner states to the irrevocable conversion of their national currencies into a new East African currency on that date, (3) a set of final basket weights for the EACU, and (4) a set of irrevocable central parities between national currencies and the EACU at which conversion to the new currency will occur.

At the heart of this process is the EACU, which would be defined well in advance of the conversion phase as a fixed composition basket (like the ECU or the special drawing right of the five regional currencies). Weights would reflect initial GDP and/or trade shares, and could be subject to minor adjustments when appropriate, including at the outset of the conversion phase.17 From its inception, the EACU would float vis-à-vis third-party (i.e., global) currencies, its value reflecting the third-party exchange rates of its constituent currencies. Figure 8.6 shows a hypothetical EACU, constructed using 2010 GDP weights and set equal to US$1 in December of 2010. Like any weighted average, the basket tends to be less variable against third-party currencies than its constituents, though its movements can be large when the exchange rates of the big three countries move together, as in the second half of 2011.

Figure 8.6A Sample East African Currency Unit

(2010 GDP weights at official exchange rates)

Source: IMF, International Financial Statistics database; and authors’ calculations.

Note: This EACU is composed of 25.22 Burundi francs, 32.33 Kenya shillings, 42.50 Rwandan francs, 426.23 Tanzania shillings, and 498.84 Uganda shillings. These imply initial basket weights of 2.05 percent, 39.90 percent, 7.15 percent, 29.29 percent, and 21.61 percent, which are the GDP shares of the five partners based on U.S. dollar figures at official exchange rates in 2010. EACU = East African Currency Union.

In Europe, the ECU served as a parallel unit of account during the transition period, and then, as the euro, the euro area’s exclusive legal unit of account from the outset of union. Euro notes and coin were not actually circulated until two to three years after the consummation of union. The ECU had been intended to serve as a currency of denomination during the long transition—not just for official intraunion clearing purposes, but also for private commercial transactions and financial contracts including bonds and bank deposits—but it never accounted for a significant share of transactions in the European Union, despite the advantages of denominating in a stable unit of account. Eichengreen (2006) attributes this outcome to network externalities in favor of incumbent currencies. The same outcome seems highly likely for any EACU introduced solely as a unit of account during the transition. This situation implies no loss to the process, provided that other aspects of the transition support the acceptability of the EACU once the union is established.

Masson (2012) and Kigabo, Masson, and Masson (2012), however, have advocated a more substantial role for the EACU based on the introduction of notes and coin at a much earlier stage (see also Agarwala, 2003, on South Asia, and Mori and others, 2002, on Association of Southeast Asian countries). Their argument is that an early introduction of EACU currency provides a way to demonstrate partner states’ political commitment to union and to involve the public in the process of monetary integration during a potentially protracted convergence period. The currency issue would be regulated by a currency board–type arrangement, in which a supranational body (such as the EAMI) would hold a 100 percent reserve of constituent currencies and would be committed to exchanging the EACU for these currencies on demand.

This proposal deserves serious consideration, although it raises concerns associated with competing media of exchange that were absent in the European case. The network externalities associated with exchange suggest that in the absence of strong legal restrictions there will typically be multiple locally stable equilibria with widely divergent market shares of the competing currencies (Dowd and Greenaway, 1993). These include equilibria like those observed in Europe, where the EACU fails to establish a foothold in actual usage—though in this case the stakes seem higher because they imply the outright rejection of EACU notes and coin. More appealing would be a set of equilibria in which the EACU establishes a relatively stable but nontrivial role in the local economies, perhaps even reducing the extent of transactions and liability dollarization in Tanzania and Uganda where it is currently substantial. More worrisome, however, are equilibria in which the weaker local currencies are driven out by the EACU, or in which a combination of credibility problems at the national currency and/or EACU level leads to an advance rather than a retreat of dollarization. These considerations, together with the potentially destabilizing effects of portfolio substitution on the effectiveness of national monetary policies, suggest that the advantages of the currency board approach should be carefully weighed against possible damages to the credibility of the overall process.

At the point of union, the EACU will convert at an exchange rate of 1:1 with the new East African currency unit. Following the European example, the new currency should initially be used for nonphysical transactions (i.e., electronic and interbank transfers), with national currency notes and coin continuing to circulate at their fixed parities within national jurisdictions for a period of time prior to the introduction of new notes and coin. From the moment of union, however, all public sector transactions, interbank payment system clearing activities, and new debt issues and rollovers would be obliged to be denominated in the new currency. Dual pricing in new and old local currency values will be both necessary and desirable during the changeover, for wholesale and retail goods and services including, for example, all invoicing, wage and salary notification, and bank statements. If the euro area blueprint is followed, the duration from formal union to the sole circulation of the new currency at the retail level could be in the order of two to three years (Adam and others, 2012).

Choosing Conversion Parities

The choice of conversion parities is subject to the familiar n–1 problem of what will anchor the EACU vis-à-vis global currencies and also to the challenge of avoiding severe misalignments. In Adam and others (2012), we addressed the first issue by arguing that a coordinated set of national inflation targets could anchor the internal exchange rate grid during the conversion phase, provided that some degree of de facto sovereignty had been transferred to the East African Central Bank or its institutional predecessor (such as the EAMI) in advance of the conversion phase. The exchange rate commitment would have to be backed up in practice by the near equivalent of a set of external pegs, supported by mutual intervention where necessary; this underscored the need for a functioning supranational institution in advance of the conversion phase. We argued that the ambiguities inherent in such a situation argued strongly in favor of a short duration for it, as did the danger of speculative attacks in a system of open capital accounts.

Here, we focus on the second question: the choice of conversion parities. Once this choice is made, the partner states give up any prospect of using intraunion exchange rate adjustments to address asymmetric real exchange rate misalignments.18 Countries that enter the union with relatively strong currencies will therefore be at a potentially persistent competitive disadvantage relative to those that enter with relatively weak currencies. The resulting macroeconomic strains may be substantial: the high-cost economies will tend to have lagging exports and employment, and the low-cost economies will tend to be strong exporters and may be favored destinations for FDI. Depending on national tax structures, these differences may affect relative fiscal performance as well.

These concerns suggest that the EAC partners should agree in advance that conversion parities will be chosen to avoid locking in severe real exchange rate asymmetries at the outset of union. In Adam and others (2012), we review alternative methodologies for real exchange rate assessment and argue that a combination of approaches should be employed, along the lines of the IMF’s current assessment methodology (Lee and others, 2008). It is worth emphasizing that if our managed float proposal is adopted for the convergence phase, the advance commitment to a real exchange rate assessment may be as important as any parity adjustments such an exercise actually delivers. Retrospectively, of course, the assessment will have the protective effect of allowing conversion parities to differ from internal spot rates at the outset of the conversion phase. But even if adjustments are not required, the knowledge that such an assessment will take place should have the separate impact of reducing the incentive for the partner states to implement competitive depreciations late in the convergence phase. Europe was explicitly protected from the latter by the internal exchange rate grid and an explicit rule against devaluation in the final two years before union. National inflation targets should limit the scope for competitive depreciation in the EAC case, but with uncertain efficacy, particularly if convergence criteria on inflation are nondemanding or adherence to them is weak.

Are Nominal Rigidities a Barrier to Real Exchange Rate Adjustment?

The remainder of this section focuses on an empirical question that informs both the costs of initial misalignments and the strength of national incentives for late-stage depreciation: how severe are wage and price rigidities in the EAC? If the large informal sectors and relatively uncomplicated supply chains that are characteristic of these economies mean that price rigidities are largely absent, then the likelihood of extended real exchange rate misalignments—and the urgency of setting appropriate conversion parities—may be relatively limited. We use monthly data for 1995 to 2011 to assess the relative importance of exchange rate and price adjustment in resolving real exchange rate misalignment for the five EAC countries. For comparison purposes, we undertake the same estimation for a sample of 19 countries that Reinhart and Rogoff (2004) classify as either “managed floating” or “freely floating” for the majority of the period.19

Observed reliance on nominal exchange rate adjustment may of course reflect a preference for letting the nominal exchange rate carry the burden of correcting misalignment, rather than the existence of structural rigidities to domestic price adjustment. Our analysis nonetheless suggests two main conclusions. The first is that the nominal exchange rate appears to play a decisive role in eliminating real exchange rate misalignments in the EAC. This effect is strongest in Kenya, Tanzania, and Uganda; domestic prices bear a larger share of the adjustment burden in Rwanda, and especially in Burundi. The second finding is that the EAC pattern is quite similar to what we observe in emerging market and advanced economies. In particular, we find only very weak evidence that domestic prices play a greater role in achieving real exchange rate adjustment in the EAC countries than they do in higher-income countries.

An Empirical Framework

The log of a trade-weighted real effective exchange rate, et, can be calculated as et=pt+(Stpt*), where pt is the log of the consumer price index in the home country and St and pt* are the logs of partner trade-weighted indexes of nominal exchange rates and partner country (wholesale) prices. Increases in et and St are appreciations. Denoting the log of the unobservable equilibrium real exchange rate as e˜t, the degree of misalignment of the real exchange rate is given by

Theories of the equilibrium real exchange rate imply that misalignment is a stationary variable.20 It follows that if the equilibrium real exchange rate is non-stationary, the real exchange rate itself must also be nonstationary and the two must be cointegrated. This in turn implies that one or both of St and (ptpt*)t must be not only nonstationary but also cointegrated with e˜t. All three elements of yt=[(ptpt*),St,et] are in fact I(1) for the EAC countries and for the majority of the comparator group.21 Stationarity of equation (8.2) therefore implies that these variables should be cointegrated, with an equilibrium error given by the degree of misalignment.

Cointegration implies, in turn, that at least one of the components of Δyt must respond in an equilibrating direction to the lagged equilibrium error, mt-1. This could in principle include the equilibrium rate itself if policymakers adjusted underlying policy determinants like government spending to reduce misalignments. But in what follows, we assume that e˜t is weakly exogenous (i.e., it does not adjust) so that adjustment is confined to the actual real effective exchange rate. The real depreciation required by a situation of temporary overvaluation (m˜t1>0) can then be accomplished either through nominal depreciation (ΔSt<0) or through disinflation relative to trading partners [(Δptpt*)<0]. Our analysis asks how the burden of adjustment is shared between these two channels.22 Notice that the EAC countries are small, open economies, implying that wholesale prices should be weakly exogenous and therefore that any adjustments in relative prices reflect movements in domestic prices.

To assess the relative importance of exchange rates and domestic wages and prices in real exchange rate adjustment, we embed both aspects of behavior within a partial vector error correction model of the form

where Δyt=[ΔSt,Δ(ptpt*)] and where α consists of the parameters of interest, measuring the adjustment respectively of the nominal exchange rate and relative prices to the realized real exchange rate misalignment, m˜t1. This adjustment can in principle be asymmetric, for example if wages and prices rise more readily than they fall. The vector Γi consists of the short-run parameters of the model, and ut is a white noise error term.

As we do not directly observe the equilibrium real exchange rate, we must estimate it in order to infer the degree of misalignment. A standard approach is to model the fundamental rate as a linear combination of observable fundamentals such as aid, the terms of trade, and government spending. Here we adopt the alternative approach of decomposing the actual real exchange rate, et, into its stationary and nonstationary components using the Hodrick-Prescott filter. The nonstationary trend component of this decomposition serves as our proxy for the equilibrium real exchange rate, and the stationary component for misalignment, so that m˜t=ete_hptwheree_hpt is the Hodrick-Prescott estimate of the nonstationary (equilibrium) component of the real effective exchange rate.23 Lagging once and substituting the resulting expression for m˜t1, we finally estimate equation (8.3).


Table 8.4 reports the estimated coefficient vectors α from equation (8.3) estimated for the EAC and our comparator country groups, along with a range of diagnostic tests. The model is estimated for January 1995 to August 2011.

Table 8.4Adjustment to Real Exchange Rate Shocks
Adjustment Coefficients

UnrestrictedPositive shocks2Negative shocks2
Mean EAC-5−0.27−0.06−0.34−0.04−0.18−0.06
Mean EAC-3−0.28−0.03−0.29−0.03−0.30−0.03
Joint F test3 (p-value):Test of symmetric response4 (p-values):
Difference in means5 versusKenya0.440.82
Source: Authors’ calculations.Notes: Real exchange misalignment is defined as the deviation of the real effective real exchange rate from its long-run trend, where the latter is estimated by a Hodrick-Prescott filter with smoothing parameter 14,400. The adjustment coefficients are the error-correction coefficients from a dynamic error-correction model. Asterisks denote significance levels of *10 percent, **5 percent and *** 1 percent. EAC-3 = Kenya, Tanzania, Uganda; EAC-5 = EAC-3 plus Burundi and Rwanda; NEER = nominal effective exchange rate; OECD = Organization for Economic Cooperation and Development; RELP = relative prices.

The null hypothesis is that the trade-weighted world price index is weakly exogenous; p-values correspond to the likelihood ratio test proposed by Johansen (1992).

Positive and negative misalignments indicate overvalued and undervalued real exchange rates, respectively.

The null hypothesis is that all error-correction coefficients are zero.

The null hypothesis is that the adjustment coefficients on positive and negative lagged misalignments are identical.

The subtable reports the difference in average adjustment coefficients between the East African Community countries and the OECD countries. Asterisk indicates significance levels for a test of equality of means.

Source: Authors’ calculations.Notes: Real exchange misalignment is defined as the deviation of the real effective real exchange rate from its long-run trend, where the latter is estimated by a Hodrick-Prescott filter with smoothing parameter 14,400. The adjustment coefficients are the error-correction coefficients from a dynamic error-correction model. Asterisks denote significance levels of *10 percent, **5 percent and *** 1 percent. EAC-3 = Kenya, Tanzania, Uganda; EAC-5 = EAC-3 plus Burundi and Rwanda; NEER = nominal effective exchange rate; OECD = Organization for Economic Cooperation and Development; RELP = relative prices.

The null hypothesis is that the trade-weighted world price index is weakly exogenous; p-values correspond to the likelihood ratio test proposed by Johansen (1992).

Positive and negative misalignments indicate overvalued and undervalued real exchange rates, respectively.

The null hypothesis is that all error-correction coefficients are zero.

The null hypothesis is that the adjustment coefficients on positive and negative lagged misalignments are identical.

The subtable reports the difference in average adjustment coefficients between the East African Community countries and the OECD countries. Asterisk indicates significance levels for a test of equality of means.

Column [1] reports Johansen’s (1992) weak exogeneity test of partner wholesale prices, pt, which indicate that for the large EAC countries and for the two sets of comparator countries, partner wholesale prices play no significant role in eliminating real exchange rate misalignment. This justifies our focusing exclusively on the respective adjustment burden carried by nominal exchange rate and domestic price adjustment. Only for Burundi and Rwanda do we reject the null of weak exogeneity, although this result is caused by the heavily managed nature of these two countries’ exchange rates over the estimation period.24

The central results of the analysis are contained in columns [2] to [7], which record the adjustment coefficients αs and α(p-p*) for the nominal effective exchange rate and relative prices, respectively. In each case, we report the coefficients for both symmetric and asymmetric adjustment. Columns [8] and [9] then report F-tests against the null that adjustment to real exchange rate shocks is symmetric, that is, that the speed of adjustment is the same when the real exchange rate is overvalued as when it is undervalued.

A number of features emerge from the analysis. First, taking the EAC group as a whole, adjustment to real exchange rate misalignment occurs principally through movements in the nominal exchange rate rather than through domestic prices. The nominal exchange rate adjustment coefficients are, in the main, statistically significant while those measuring the adjustment of relative prices are not. This pattern is stronger for the three big partner states, Kenya, Tanzania, and Uganda than for Rwanda and, especially, Burundi, where the burden of adjustment is shared more evenly between nominal exchange rate and price adjustments.

Second, for Kenya, Tanzania, and Uganda exchange rate and price adjustments are symmetric between temporary overvaluation and undervaluation of the real exchange rate. There is weak evidence of asymmetry in Rwanda and Burundi, where the nominal exchange rate responds somewhat more strongly to overvaluation than to undervaluation, but the difference is not statistically significant (the F-tests on symmetry are not rejected, but this reflects the low precision of the coefficient estimate on negative real exchange rate shocks).

Third, for the large EAC countries as a whole, the strength and pattern of adjustment is broadly comparable to that observed for our emerging markets comparator group. In both groups, the balance of adjustment is strongly skewed in favor of the exchange rate, and the adjustment pattern is broadly symmetrical with respect to over- and undervaluations of the real exchange rate. Within the arguably more homogenous Organization for Economic Cooperation and Development subsample, the balance of adjustment is even more decisively skewed in favor of movements in the exchange rate—to the point that for this group, we can comfortably accept the restriction that all the adjustment is through the nominal exchange rate. This is not the case for the emerging markets group, and is only just accepted for Kenya, Tanzania, and Uganda taken together.

In summary, our investigation strongly suggests that the larger EAC countries have tended to rely as heavily on exchange rate flexibility for eliminating real exchange rate misalignments as have other more developed floating-rate economies, and that wage and price flexibility has not shouldered a significantly larger share of the adjustment burden. It is not self-evident, therefore, that wage and price flexibility can be assumed to rescue the EAC countries from the oldest concern of the literature on optimum currency areas, which is that the loss of intraunion exchange rates increases the costs of asymmetric shocks. This underscores the value of an advance commitment to setting conversion parities that avoid severe initial misalignments.

Real Exchange Rates and Monetary Union

We close the chapter by briefly addressing an exchange rate issue that relates to the postunion period rather than the transition: what impact will union have on real exchange rates and competitiveness across the union? Both theory and the experience of other African monetary unions suggest that national price levels will be tied together much more closely in the postunion period than before. Two very different mechanisms bring this about. One is the adoption of a union-wide exchange rate and monetary policy, which increases the comovement in nominal variables, such as the prices of internationally traded goods, regardless of the degree of intraunion trade. But the choice of nominal anchor cannot matter for the real exchange rate in the long run, because relative prices are tied down eventually by real rather than nominal factors. The second factor, therefore, and the factor we focus on here, is the increase in cross-border price arbitrage that follows a reduction in trading costs. We argue that monetary union will produce a convergence of equilibrium real exchange rates by stimulating an increase in intraunion trade. We provide evidence from sub-Saharan Africa that is consistent with this effect.

The essence of this argument can be understood by considering a pair of countries that form a currency union starting from a nonunion equilibrium in which they may or may not be engaged in bilateral trade.25 The countries are identical, with the exception that one of them has a large oil export sector. The two countries face the same world prices for all traded goods and services, but the oil-rich country has a higher domestic price of nontraded goods and therefore a more appreciated real exchange rate. Reflecting the Dutch disease effect of its natural resource wealth, this country has a larger nontraded goods sector and a smaller manufacturing and cash crop sector than the resource-poor country.

Monetary union alters this situation by driving down the cost of intraunion trade. The size of this cost reduction is not known, but the trade-creating effects of monetary union appear to be large. Rose (2000) found that, all else equal, membership in a monetary union tripled intraunion trade. The subsequent empirical literature has reduced this substantially to an impact somewhere between 5 percent and 25 percent (De Grauwe, 2009; Dellas and Tavlas, 2009), but a consensus remains that regional trade will increase, and perhaps sharply. The impact on price arbitrage is of course more extensive than the impact on trade flows, because price competition can operate as long as trade is feasible on the margin, regardless of the volume of that trade. The evidence is very strong, in other words, that monetary union widens the scope for cross-border price arbitrage within the union.

Monetary union therefore widens the scope of intraunion price arbitrage. Previously nontraded goods become tradable: they flow from where their prices were initially lower, in the resource-poor country, to where they were initially higher, in the oil-exporting country. The real exchange rate differential between the two countries shrinks, and may disappear altogether. The oil exporter experiences a long-run real depreciation, which unwinds a portion of that country’s Dutch disease and produces an expansion of noncommodity exports to the world economy (or a contraction of imports: its trade balance with global markets increases). The resource-poor partner, in turn, experiences a real appreciation and a shift in the composition of its exports from globally traded goods to regional goods that were previously nontraded.

In Annex 8.1, we undertake a set of regression-based tests of the hypothesis that monetary union narrows real exchange rate differentials. Sub-Saharan Africa provides a unique empirical lens on this hypothesis, because of the relatively large number of countries in the CFA franc zone and the substantial variation in natural resource intensity across the continent.26 Using a difference-in-difference approach, we test our hypothesis by asking whether natural resource endowments have a smaller effect on equilibrium national price levels across the members of the CFA zone than across countries in sub-Saharan Africa with independent national currencies.

The empirical results for 1988–2008 and various subperiods are strongly consistent with our hypothesis. We find that the Dutch disease effects of national resource endowments are sharply lower among the CFA countries than among countries with independent currencies; in some specifications, in fact, we cannot reject the hypothesis that variations in resource rents have no impact at all on relative real exchange rates in the CFA zone.

These results suggest an answer to the question we posed at the outset of this section. Monetary union is likely to spur price arbitrage across the union, particularly with respect to goods that are not traded with global markets. Relatively resource-rich members will thereby tend to “export” some portion of their Dutch disease in global markets (the competitiveness problem faced by their non–commodity traded goods sectors) to relatively resource-poor members of the union. Holding overall trade balances constant, the trade balances of the latter group will tend to deteriorate vis-à-vis global partners, as resources previously devoted to globally traded goods are shifted over to serve the expanded common-currency market. The same logic applies, of course, where the “resource endowment” is not minerals or another natural resource, but other ongoing sources of financing for a country’s deficit in goods and services, such as large and persistent inflows of foreign aid or private capital.

It is important to emphasize that the welfare implications of these effects are far from obvious: for the non-oil or non-aid economy, for example, any Dutch disease impact of membership applies to the country’s preunion export sector, and not to its overall exports. The latter are likely to rise, as we have emphasized, as previously nontraded goods find a favorable single-currency market. Our analysis in this section, moreover, is only partial, based on the pure trade-creation effects of currency union. The full impact of union on relative competitiveness and growth may depend on a host of additional factors that lie outside of the scope of this chapter. These include dynamic effects on factor markets, on the location of investment, and, importantly, on how resource revenues are spent, particularly in the areas of infrastructure.


What exchange rate arrangements are appropriate for the EAC in the transition to monetary union? After reviewing the feasible options, we have argued against tight exchange rate commitments based our interpretation of the EAC’s policy history, its vulnerability to external shocks, and the openness of its capital account. The appropriate exchange rate regime during the transition to union, in our view, is a managed float in which national inflation targets—consistent with community-wide convergence criteria for inflation—serve as the nominal anchor. The advantages of a managed float are strengthened considerably by the uncertain horizon over which union will occur.

In developing this argument, we undertook a set of empirical exercises that may have applications beyond this chapter. The first is an adaptation of the time-honored literature on offset coefficients to low-income and potentially aid-dependent countries operating flexible exchange rates. For Kenya and Uganda, we find that domestic monetary expansions create a high degree of exchange market pressure, leading to a combination of exchange rate depreciation and reserve losses that are concentrated in the first three or four months in Uganda, and appear after a lag of about a quarter in Kenya. For these countries, exchange rate commitments would appear to harbor significant risks of speculative attack. Pressures emerge more slowly in Tanzania and are cumulatively much smaller, consistent with that country’s relatively tight capital controls. These conclusions could not be drawn with any confidence from direct measures of the scale of private capital flows.

The partner countries will of course eventually face the irrevocable loss of intraunion exchange rates. In a second empirical exercise, we generated indirect evidence of the potential costs of this loss by assessing the relative importance of nominal exchange rate movements and domestic price adjustments in addressing real exchange rate misalignments in the past. We found only very weak evidence that wages and prices were more flexible in this sense in the big three countries of the EAC than in emerging market and advanced economies operating flexible exchange rates. This finding provides strong support for our argument that the partner states should agree, in advance, on an approach to choosing conversion parities that avoids locking in severe misalignments at the outset of union.

Our final contribution focused on the likely impact of union on equilibrium real exchange rates and national competitiveness. Considerable empirical literature suggests that monetary union brings a substantial increase in intraunion trade. Cross-border price arbitrage will increase even more, since prices can be disciplined by the possibility of trade even when the volume of trade is low. We argued that union-wide price convergence on previously nontraded goods would moderate equilibrium real exchange rate differentials in the union, producing a situation in which the relatively resource-rich partners export some portion of their Dutch disease to the resource-poorer partners. Sub-Saharan Africa provides an unusually favorable environment for testing this hypothesis, given the large number of participants in the CFA zone and the substantial differences in resource endowment across the continent. We found strong confirmation of our hypothesis: differences in natural resource endowment have much smaller effects on long-run differentials in real exchange rates across the members of the CFA zone than across countries in sub-Saharan Africa operating independent national currencies.

Our analysis suggests a number of areas in which further work would strengthen the basis for a successful transition. One is the early development of a regular, union-wide real exchange rate surveillance exercise. Given the vagaries of existing empirical methods, only substantial misalignments—such as 15 percent or more by a combination of approaches—would justify a departure from current spot rates when conversion parities are set. An internal, union-wide exercise would nonetheless be of very substantial value if undertaken regularly and subject to high-level review during the convergence phase. It would not only facilitate agreement over conversion parities, but also minimize exchange rate surprises for market participants at the outset of the conversion phase.

The possibility of new entrants into the union raises a second topic for study. Accession by new members will undoubtedly be dictated by macroeconomic and institutional convergence requirements, and in these arenas the European Union’s recent enhancement of its fiscal convergence and coordination criteria deserve close study by EAC members.27

But what nominal exchange rate criteria should new entrants be required to satisfy, vis-à-vis the EACU, in advance of their entry into the union? Our taxonomy implies that there is a tension between rules that are robust enough to be applied uniformly to all potential entrants and rules that are optimal on a country-by-country basis. If practical considerations favor a uniform rule, however, the leading contenders seem likely to be the two that emerged from our discussion. A version of our managed float approach would leave new entrants free from exchange rate commitments vis-à-vis the EACU until a final, brief stage; the alternative would be a grid-based approach that would require them (as with new entrants to the euro area) to maintain a narrow band with respect to the EACU two or three years before entry. Open capital accounts and other arguments in favor of exchange rate flexibility favor the first arrangement. The broader context of ultimate fiscal or even political union, however—the relevance of which has been illustrated by the recent experience of the euro area—suggests that incumbent members may wish to seriously consider imposing a grid-based criterion on new entrants to enforce greater macroeconomic discipline during the accession process.

Annex 8.1.

Real Exchange Rates and Monetary Union in Sub-Saharan Africa

We argue in this chapter that monetary union will produce a convergence of equilibrium real exchange rates by stimulating an increase in intraunion trade. The structure of this argument appears in Figure 8.1.1, where rectangles denote the global market and circles denote countries. The left-hand rectangle shows a pair of potential partners before union. These countries are small relative to the global market, and they maintain identical trade policies and constant overall trade balances (financed by some sustainable level of aid, for example, or long-term capital inflows); for concreteness, we assume that they each run a small overall trade deficit. Their real exchange rates are defined as the ratio of the domestic price of nontraded goods to the domestic price of a composite basket of internationally traded goods. If the two countries were identical in preferences, factor endowments, net international asset positions, and other macroeconomic fundamentals, theory would predict identical real exchange rates in a situation of internal and external balance (i.e., identical equilibrium real exchange rates) regardless of the choice of exchange rate regime in the two countries or of whether they share a common currency.

Figure 8.1.1Trade and Real Exchange Rate Effects of Monetary Union

Source: Authors.

Note: R = natural resource export; T = traded goods; N = nontraded goods. Arrow lengths show the values of trade at international prices. Both countries have small overall trade deficits both pre- and postunion.

To induce a difference in preunion real exchange rates, we endow country 1 with a natural resource that is not used at home, but can be extracted at zero cost and sold on world markets. Trade patterns now differ, despite the fact that the two countries are identical in all other respects. The resource-rich economy exports its natural resource (denoted by R) to the global market to pay for (most of) its deficit in all other traded goods (denoted by T), while the non—resource-rich economy runs a small deficit in the latter. The degree to which the two countries trade with each other in this equilibrium is immaterial: both are price takers in the global market, and purchasing power parity holds for traded goods in equilibrium, so their bilateral balance in T goods has no macroeconomic impact in either country. Behind the scenes, of course, both countries also consume non-traded goods, which (by definition) they produce themselves. By virtue of its commodity rents, the first country has a more appreciated real exchange rate in the initial equilibrium, along with a higher absolute price of nontraded goods when measured in a common currency. Its nontraded goods sector is larger, and its traded-goods sector smaller (excluding the commodity export), than that of country 2. Cross-border trade costs are high enough in the initial equilibrium to prevent trade from arbitraging away the price gap in nontraded goods.

Monetary union alters this situation by driving down the cost of intraunion trade. The right-hand box in Figure 8.1.1 therefore interprets union as a widening of intraunion price arbitrage. As discussed in this chapter, previously nontraded goods (denoted by N) become tradable, and flow from the resource-poor to the resource-rich partner. The real exchange rate differential between the two countries shrinks and may disappear altogether. The resource-rich economy experiences a long-run real depreciation, which unwinds a portion of that country’s Dutch disease and produces an expansion of noncommodity exports to the world economy (or a contraction of imports; trade balance in T goods increases). The resource-poor partner, in turn, experiences a real appreciation and a shift in the composition of its exports from globally traded goods to regional goods that were previously nontraded.

Table 8.1.1 shows a set of regression-based tests of the hypothesis that monetary union narrows real exchange rate differentials. Our basic model is therefore a cross-sectional regression of the form

Table 8.1.1Differences-in-Differences: Monetary Union and Equilibrium Real Exchange Rates
NR= IndicatorNR= Rents
Number of countries36363636
Number of observations2511797272
R2 (adjusted)0.35370.35870.45520.3885
F test on the sum of the coefficients on NR and CFA*NR (p-value)
Source: Authors’ calculations.Note: The dependent variable is the log of the ratio of the official exchange rate in dollars per unit of local currency to the purchasing power parity exchange rate. Observations are nonoverlapping three-year averages. The sample includes all countries in sub-Saharan Africa with available data and population of at least one million in 2000, including South Africa but excluding other members of the Common Monetary Africa. All regressions include a full set of period dummy variables, and the earlier regressions include a dummy variable for Angola, due to a very large outlier observation. T-statistics based on robust standard errors are in parenthesis (asterisks denote significance at the *10 percent, **5 percent, and ***1 percent levels). CFA =African Financial Community, NR = natural resource intensity.
Source: Authors’ calculations.Note: The dependent variable is the log of the ratio of the official exchange rate in dollars per unit of local currency to the purchasing power parity exchange rate. Observations are nonoverlapping three-year averages. The sample includes all countries in sub-Saharan Africa with available data and population of at least one million in 2000, including South Africa but excluding other members of the Common Monetary Africa. All regressions include a full set of period dummy variables, and the earlier regressions include a dummy variable for Angola, due to a very large outlier observation. T-statistics based on robust standard errors are in parenthesis (asterisks denote significance at the *10 percent, **5 percent, and ***1 percent levels). CFA =African Financial Community, NR = natural resource intensity.

where pt is a measure of national prices converted to U.S. dollars for a comparable basket of goods across countries, CFAi is a dummy variable taking the value of 1 if country i is a member of a CFA franc zone, NRi is a measure of the natural resource intensity of country i, and CFAi · NRi is the interaction of the two effects. The coefficient on the interaction term provides a test of our hypothesis: a large, negative, and statistically significant value for α3 would constitute strong evidence that the impact of national macroeconomic fundamentals on equilibrium real exchange rates is “shared out” among the members of a monetary union, and is therefore muted relative to what would emerge in the absence of union. The peripheral members of the Common Monetary Area (Lesotho, Namibia, Swaziland) are excluded, so the regression compares CFA zone members with countries in sub-Saharan Africa operating independent currencies.28

To control for the differing compositions of national expenditure, we draw on the Penn World Tables to benchmark national price levels against international reference prices. Our dependent variable is the ratio of GDP measured at the official exchange rate to GDP measured at the purchasing power parity exchange rate.29 An increase in this series denotes an increase in domestic prices relative to the foreign prices of similar goods, and therefore a real appreciation.30 To focus on low-frequency movements, we take nonoverlapping, three-year averages of the data, for 1988 to 2008, and include a set of fixed time effects to capture developments in global markets. Because higher levels of real GDP per capita are associated with more appreciated real exchange rates across countries (the Harrod-Balassa-Samuelson effect), we control for the log of once-lagged real per capita income, log yi,t-1, in all regressions.

The sample consists of 36 countries, comprising all countries in sub-Saharan Africa with available data and with populations of at least one million in 2000, with the exception of the peripheral members of the Common Monetary Area. The MU dummy variable takes on the value “1” for members of the two CFA monetary unions—12 of these are in our sample, and we treat them as members of a single monetary union—and zero for the remaining 24 countries in the sample. Table 8.1.1 reports the results. In columns [1] through [3], we measure resource endowments using a time-invariant binary indicator for whether a country is a significant natural resource exporter; there are three such countries in the CFA group and eight in the non-CFA group.31 Column [2] focuses on the sub-period from 1988 to 2002, a favorable one for assessing the equilibrium implications of resource abundance given the relative tranquility in global commodity markets. Column [3] considers the more recent subperiod from 2003 to 2008. Resource abundance surely mattered a great deal during this period given the quintupling or so of nominal global fuel prices, but the empirical test is less clean than in the earlier period because rigid nominal exchange rates are likely to have generated larger short-term real exchange rate misalignments in the CFA countries than elsewhere in Africa. In the final column, we report a regression for the later period that measures natural resource dependence as the lagged value of natural resource rents relative to GDP.

The results in Table 8.1.1 are strongly consistent with our hypothesis. For the full period and the period since 2003, the familiar “resource curse” effect is apparent among the non-CFA countries, because α2> 0 implies that higher levels of resource wealth generate more appreciated currencies within this group. The point estimate remains positive for the relatively tranquil period before 2003, although it is smaller and statistically insignificant. The very strong results from the 2003–2008 period should of course be treated with caution, because this was a period of sharp cumulative increases in global fuel prices; the interaction term therefore provides a joint test of our hypothesis about equilibrium real exchange rates and a complementary hypothesis about greater short-run misalignments in the CFA zone due to the lack of nominal exchange rate adjustment. But our estimates of α3 are negative and highly statistically significant for the full period, and for both subperiods. Differences in natural resource endowment therefore have much smaller impacts on real exchange rate differentials across the CFA zone than across countries with independent national currencies. In columns [1]–[3], in fact, we cannot reject the hypothesis that variations in resource rents have no impact at all on relative real exchange rates in the CFA zone.


    AdamC.C.KombeP.Kessy and S.O’Connell2012Exchange Rates in the Transition to East African Monetary Union,International Growth Centre Working Paper 12/0458 (London: International Growth Center).

    • Search Google Scholar
    • Export Citation

    AgarwalaR.2003Road to a Single Currency for South AsiaRIS Policy Brief #9 (New Delhi: Research and Information System for the Non-Aligned and Other Developing CountriesDecember).

    • Search Google Scholar
    • Export Citation

    BeetsmaR. and M.Giuliodori2010The Macroeconomic Costs and Benefits of the EMU and Other Monetary Unions: An Overview of Recent Research,Journal of Economic Literature Vol. 48 No. 3 pp. 603641.

    • Search Google Scholar
    • Export Citation

    ChinnM. and H.Ito2008A New Measure of Financial Openness,Journal of Comparative Policy Analysis Vol. 10 No. 3 pp. 309322.

    • Search Google Scholar
    • Export Citation

    DebrunX.P.Masson and C.Pattillo2005Monetary Union in West Africa: Who Might Gain, Who Might Lose, and Why?Canadian Journal of Economics Vol. 38 No. 2 pp. 454481.

    • Search Google Scholar
    • Export Citation

    DebrunX.P.Masson and C.Pattillo2010Should African Monetary Unions Be Expanded? An Empirical Investigation of the Scope for Monetary Integration in Sub-Saharan Africa,Journal of African Economies Vol. 20 No. 2 pp. 104150.

    • Search Google Scholar
    • Export Citation

    De GrauweP.2009Economics of Monetary Union (Oxford: Oxford University Press, 8th ed.).

    DellasH. and G.S. Tavlas2009An Optimum-Currency-Area Odyssey,Journal of International Money and Finance Vol. 28 No. 7 pp. 11171137.

    • Search Google Scholar
    • Export Citation

    DhunganaS.2008Capital Account Liberalization and Growth Volatility,unpublished BA thesis (Williamstown, Massachusetts: Williams College).

    • Search Google Scholar
    • Export Citation

    DowdK. and D.Greenaway1993Currency Competition, Network Externalities and Switching Costs: Towards an Alternative View of Optimum Currency Areas,Economic Journal Vol 103 No. 20 pp. 11801189.

    • Search Google Scholar
    • Export Citation

    EichengreenB.2006The Parallel Currency Approach to Asian Monetary Integration,American Economic Review Vol. 96 No. 2 pp. 432436.

    • Search Google Scholar
    • Export Citation

    EichengreenB.A.K. Rose and C.Wyplosz1994Speculative Attacks on Pegged Exchange Rates: An Empirical Exploration with Special Reference to the European Monetary System,” in The New Transatlantic Economy ed. by M.CanzoneriP.Masson and V.Grilli (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation

    European Central Bank2010Study on the Establishment of a Monetary Union among the Partner States of the East African Community,ECB Staff Study (Frankfurt).

    • Search Google Scholar
    • Export Citation

    HinkleL. and P.Montiel1999Exchange Rate Misalignment: Concepts and Measurement for Developing Countries (Oxford: Oxford University Press).

    • Search Google Scholar
    • Export Citation

    JohansenJ.1992Testing Weak Exogeneity and the Order of Cointegration in UK Money Demand Data,Journal of Policy Modeling Vol. 14 No. 3. pp. 313334.

    • Search Google Scholar
    • Export Citation

    KigaboR.T.Masson and P.Masson2012Design and Implementation of a Common Currency Area in the East African Community,” unpublished.

    • Search Google Scholar
    • Export Citation

    LaneP. and G. M.Milesi-Ferretti2007The External Wealth of Nations Mark II,Journal of International Economics Vol. 73 No. 2 pp. 223250.

    • Search Google Scholar
    • Export Citation

    LeeJ.G. M.Milesi-FerrettiJ.OstryA.Prati and L. A.Ricci2008Exchange Rate Assessments: CGER MethodologiesIMF Occasional Paper No. 261 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    MassonP.2012A Simple Method of Introducing a Regional Currency,” unpublished (Kigali: National Bank of Rwanda).

    MoriJ.M.KinukawaH.Nukaya and M.Hashimoto2002Integration of the East Asian Economies and a Step by Step Approach towards a Currency Basket Regime,Institute for International Monetary Affairs Research Report No. 2 (Tokyo: Institute for International Monetary Affairs).

    • Search Google Scholar
    • Export Citation

    O’ConnellS.A.B.O.MaturuF. M.MwegaN. S.Ndung’u and W. R.Ngugi2010Capital Mobility, Monetary Policy and Exchange Rate Management in Kenya” in Kenya: Policies for Prosperity ed. by ChristopherAdamPaulCollier and NjugunaNdung’u (Oxford: Oxford University Press).

    • Search Google Scholar
    • Export Citation

    PorterM. G.1972Capital Flows as an Offset to Monetary Policy: The German Experience,IMF Staff Papers Vol. 19 pp. 395424.

    • Search Google Scholar
    • Export Citation

    ReinhartC. M. and K. S.Rogoff2004The Modern History of Exchange Rate Arrangements: A Reinterpretation,Quarterly Journal of Economics Vol. 119 No. 1 pp. 148.

    • Search Google Scholar
    • Export Citation

    RoseA.2000One Market, One Money: Estimating the Effect of Common Currencies on Trade,Economic Policy Vol. 30 No. 1 pp. 745.

    • Search Google Scholar
    • Export Citation

    WillettT. D.K. M.Al-Barwani and S. M.El Hag2009The Gulf Cooperation Council’s Fixed Exchange Rate: A Major Anomaly for OCA Analysis(School of Politics and Economics Claremont Graduate University California).

    • Search Google Scholar
    • Export Citation
1Dutch disease is the apparent relationship between the increase in exploitation of natural resources and a decline in competitiveness. The mechanism is that an increase in revenues from natural resources (or inflows of foreign aid) will make a given nation’s currency stronger compared to that of other nations (appreciation), resulting in the nation’s other exports becoming more expensive for other countries to buy, making the manufacturing sector less competitive.
2Dates of accession to Article VIII status in the IMF are as follows: Kenya and Uganda 1994, Tanzania 1996, and Rwanda 1998. Burundi has not yet ratified Article VIII.
3Capital account liberalization is ongoing in Rwanda and Tanzania, so the Chinn-Ito indexes may show increases for 2011 and 2012.
4The trade measure is the sum of imports and exports divided by GDP. Financial assets and liabilities include foreign direct investment, portfolio equity, debt, financial derivatives, and, on the asset side, foreign exchange reserves (including gold).
5These tend to be associated with small economies and economies in which GDP at official exchange rates may be significantly undermeasured. In sub-Saharan Africa, these include Liberia (adjusted openness measure = 863 percent), Seychelles (240 percent), Guinea-Bissau (220 percent), Lesotho (200 percent), the Gambia (140 percent), and São Tomé and Príncipe (140 percent).
6The unadjusted Lane and Milesi-Ferretti data suggest that Burundi’s capital account is much more open than that in the EAC’s larger and more developed economies. This anomaly is consistent with the evidence for small economies with doubtful-quality GDP data, and, in the case of Burundi, is driven in part by very high external official indebtedness. Much of the anomaly is eliminated when we adjust for nonmarket transactions.
7This variable was introduced by Eichengreen, Rose, and Wyplosz (1994); variants have been used extensively in the literature on balance of payments crises.
8Data limitations preclude extending the sample to earlier years. The vector autoregressions include seven lags and a full set of centered monthly seasonal dummies.
9We adjust net foreign and domestic assets for exchange rate valuation effects and measure both variables as shares of the lagged monetary base.
10Following the standard structural vector autoregression approach, ut = Cεt implies Ω = CC', where the covariance matrix of εt is a 3 x 3 identity matrix and where Ω = E[uu'] is the covariance matrix of reduced-form shocks. Ω can be estimated from the reduced-form vector autoregression, but because Ω is symmetric, this estimate only delivers six restrictions on the elements of C. Equation (8.1) shows the three additional restrictions we are imposing in order to identify the C matrix and recover the structural shocks.
11A final bit of structure in equation (8.1) comes from interpreting σB as the impact of the balance-of-payments shock on exchange-market pressure, defining the latter as the sum utsutF, of exchange-rate-depreciation and reserve losses. This restricts the difference between the (3,3) element and the (2,3) element of the first matrix to be 1. In the absence of an independent estimate of the standard deviation σB, however, this property does not impose an additional restriction on the elements of C.
12Eichengreen, Rose, and Wyplosz (1994) introduced this variable; variants have been used extensively in the literature on balance-of-payments crises.
13Willett, Al-Barwani, and El Hag (2009) point out that more than half of the labor force is composed of guest workers in a number of Gulf Cooperation Council countries. This affords an unusual degree of labor market flexibility. Oil wealth, in turn, affords an unusual degree of fiscal flexibility and obviates the need for active exchange rate management to maintain external balance.
14Concerns about the location of FDI were cited as among the acute strains within the three-member EAC in the early postindependence period.
15The last point refers to the Harrod-Balassa-Samuelson effect, whereby the poorer of two partners tends to experience a bilateral real appreciation as it converges to global productivity norms at a faster rate than its richer partner. With a purchasing power parity–adjusted real income that is easily two-thirds below that of the remainder of the EAC, Burundi in particular would be expected to experience a gradual equilibrium real appreciation within the EAC, regardless of the progress to monetary union. Many economists would expect monetary union to accelerate this process, for example, by enhancing Burundi’s access to international capital markets (as occurred in the euro area), or accelerating price convergence in regionally produced goods and services, or both.
16Uganda is implementing a formal inflation-targeting system, which will ultimately place significantly greater emphasis on inflation forecasts than on money growth rates as intermediate targets, and on policy interest rates rather than reserve money as operational instruments. But its exchange rate policy remains a managed float.
17The ECU’s weights, set initially in 1979, were adjusted in 1984 and 1998.
18They also give up the ability to reconcile divergent fiscal needs for devaluation, which may be associated either with different medium-term preferences for inflationary finance or with short-run fiscal solvency pressures that favor a de facto default on nonindexed domestic-currency liabilities of key actors (such as the public sector). With respect to misalignments, note that only asymmetric misalignments matter, because symmetric ones can be eliminated through movements in the union-wide currency. See Beetsma and Giuliodori (2010).
19Our sample consists of countries that have maintained a broadly floating exchange rate regime and for whom the IMF reports trade-weighted nominal and real effective exchange rate indices over the 1995–2011 period. The Organization for Economic Cooperation and Development subsample consists of Australia, Canada, Denmark, Japan, New Zealand, the United Kingdom, and the United States. The emerging market subsample consists of Algeria, Bolivia, Brazil, Chile, Ghana, Malaysia, Mexico, Nigeria, the Philippines, Singapore, South Africa, and Tunisia. Of these, only Australia, Japan, South Africa, and the United States are characterized as freely floating. Within the managed float category, Algeria, Bolivia, Brazil, Burundi, Denmark, Malaysia, Rwanda, and Singapore are classified as maintaining a de facto crawling band around their reference currency for at least some of the period.
20In IMF parlance, e˜ is typically referred to as the “fundamental equilibrium real exchange rate.” See Hinkle and Montiel (1999) for a survey of theory and evidence on equilibrium real exchange rates and misalignment.
21Details available on request.
22Notice that we assume that policymakers seek to adjust the nominal effective exchange rate. In practice, policymakers in the EAC have tended to focus particularly on the U.S. dollar exchange rate. Although we have not chosen to do so here, we could in principle reflect this in our empirical work by decomposing the nominal effective exchange rate into the bilateral rate versus the U.S. dollar and a “correction” term that depends only on the exchange rates of partner-country currencies versus the U.S. dollar.
23The results in Table 8.4 filter the real exchange rate using a smoothing parameter of λ = 14,400. A potential drawback to our approach is that univariate filtering methods tend to track the actual evolution of the real exchange rate, and may therefore attribute “too much” of its observed movement to changes in the underlying equilibrium real exchange rate. We find, however, that varying the smoothing parameter does not fundamentally alter our results. Full results for alternative values of the smoothing parameter, and for estimation under the Baxter-King band-pass filter, are available on request.
24Johansen’s likelihood ratio test entails testing the partial system restriction that αp* = 0 on the rank-one vector error correction model shown in equation (8.3), where Δyt=[Δst,Δpt,Δpt*] This restriction implies that the cointegrating vector is weakly exogenous with respect to Δpt*.
25Annex 8.1 develops this argument in greater detail.
26The Common Monetary Area is a third monetary union in sub-Saharan Africa, but given South Africa’s predominance in the union, it is best to think of South Africa as operating an independent national currency.
27The European Union’s Treaty on Stability, Coordination and Governance in the Economic and Monetary Union came into force in January 2013, superseding the Stability and Growth Pact that accompanied the launch of the euro in 1999.
28The differences-in-differences design compares members of individual monetary unions with countries with independent national currencies. It is critical that each group be of sufficient size and that it contain appreciable variation in natural resource rents. The Common Monetary Area group is too small in these respects.
29This underlying data series as reported in the World Bank World Development Indicators databse are: GDP, PPP (current international $) (NY.GDP.MKTP.PP.CD), and GDP current US$ (NY.GDP MKTPCD). This is equivalent to the inverse of the PPP Conversion Factor (GDP) to market exchange rate ratio reported by the World Bank as series (PA.NUS.PPPC.RF).
30Recall that purchasing power parity ties the prices of traded goods together in a common currency, in the long run. Lower overall national price levels therefore mean lower prices of nontraded goods measured in a common currency, and therefore a more depreciated real exchange rate.
31Countries’ natural resource intensity is based on total natural resource rents as a share of GDP (World Development Indicators series NY.GDP.TOTL.RT.ZS). The indicator variable NR takes the value “1” for countries with a share of natural resources in GDP exceeding 10 percent on average between 1986 and 2008.

    Other Resources Citing This Publication