WPS5241 Policy Research Working Paper 5241 Till Geography Do Us Part? Prolegomena to an Economic and Monetary Union between the Dominican Republic and Haiti Emmanuel Pinto Moreira The World Bank Latin American and the Caribbean Region Poverty Reduction and Economic Management March 2010 Policy Research Working Paper 5241 Abstract This paper offers a preliminary assessment of the potential cycle synchronization, based on basic and extended benefits and costs of an economic and monetary union integrated vector auto-regression models with exogenous (EMU) between the Dominican Republic and Haiti-- variables (VARX), is developed next. Overall, the analysis two countries sharing the same island but whose history suggests that at this stage several economic criteria are is one of conflict and divergent economic prospects not satisfied for the two countries to fully benefit from in recent decades. After a brief review of the historical an economic and monetary union. At the same time, context, it examines the nature of these potential benefits however, the endogeneity of most of these criteria and costs. It then conducts a preliminary analysis (using (including the degree of business cycle synchronization) basic statistical techniques) of some key criteria for the militates in favor of an aggressive medium-term agenda formation of an economic and monetary union between for integration between them. the two countries. A more formal analysis of business This paper--a product of the Poverty Reduction and Economic Management, Latin American and the Caribbean Region--is part of a larger effort in the department to understand the issues associated with economic and monetary unions between developing countries in the region. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The author may be contacted at epintomoreira@worldbank.org or epintomoreira@imf.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Till Geography do us Part? Prolegomena to an Economic and Monetary Union between the Dominican Republic and Haiti Emmanuel Pinto Moreira* *Senior Economist, Latin America and Caribbean Region, World Bank and Middle-East Central Asia Department of the International Monetary Fund. I am grateful to Pierre-Richard Agénor (University of Manchester and Centre for Growth and Business Cycle Research) for many helpful discussions and comments on a preliminary draft, and to Issouf Samake (International Monetary Fund) for support. The views expressed in this paper are my own and do not necessarily represent those of the World Bank. Table of Contents I. Introduction II. Historical Background III. Benefits and Costs of an EMU between Haiti and the Dominican Republic 1. Nature of an Economic and Monetary Union 2. Benefits and Costs of an Economic and Monetary Union IV. Criteria for an EMU: Preliminary Data Analysis 1. Volatility of Bilateral Exchange Rates 2. Similarity of Inflation Rates 3. Degree of Price and Wage Flexibility 4. Degree of Labor Mobility 5. Scope for Discretionary Fiscal Policy V. Degree of Symmetry of Shocks: A VARX Analysis 1. VARX Methodology 2. A Basic VARX Model 3. An Expanded VARX Model VI. Summary and Policy Implications References Appendix A--Benefits and Costs Associated with a Currency Union Appendix B--Data Sources and Unit Root Tests 2 I. INTRODUCTION Haiti and the Dominican Republic opened bilateral trade only 20 years ago with the hope of creating economic integration. Trade and migration have increased significantly, and the two nations are now taking steps to establish systems and structures to manage this relationship, with support from various quarters. In July 2008 for instance, the U.S. government and the Organization of American States provided funding for a new program aimed at increasing economic growth and cooperation along the 193-mile border between Haiti and the Dominican Republic. The program, called "Fwontye Nou/Nuestra Frontera" in Creole and Spanish ("Our Border" in English), aims to provide training, technical assistance, and conditions to foster cross-border projects and productive investments for both countries. The European Union has also begun to finance joint projects along the border. The remainder of the paper is organized as follows. Section II provides a review of the historical context. Section III discusses the potential benefits and costs of an economic and monetary union between the two countries. Section IV offers a preliminary analysis of some key EMU criteria, using basic statistical techniques (unconditional correlations), as in some other contributions (see for instance Karras (2007) and Furceri and Karras (2008)). Section V develops a more formal analysis of business cycle synchronization based on two integrated (two-country) VARX models: A "basic" model involving output gaps and inflation (measured in terms of deviations with respect to US inflation) in the two countries, and the US output gap, and an extended model, which adds other "external" variables (the world price of oil and the US real interest rate) and domestic variables (bank credit to the private sector as a share of GDP and bilateral real exchange rates). The models are used to analyze the response of domestic output and inflation to external and domestic shocks. Section VI draws together the policy implication of the analysis and offers some concluding remarks. II. HISTORICAL BACKGROUND History and geography bind the Dominican Republic and Haiti. On December 6, 1492, Christopher Columbus discovered an island that he qualified as "marvelous" and named it Hispaniola. The formation of these two states and their accession to independence is a long story marked by wars and reconciliations, open or hidden hostility and alliances against common dangers. History shows many attempts of unification of these two states, though unsuccessful. On October 1st 1798, Toussaint Louverture took control of the whole island after ousting foreign troupes, both English and French1. On August 1st 1800, he became the only 1 On October 1st 1798, English troupes left the place "Mole Saint Nicolas", last area under their control. On October 22, 1798, Toussaint Louverture obliged French Governor General Hedouville to leave the island for France. 3 master of the French part of Santo Domingo after defeating his rival, General Rigaud. On January 26, 1801, he took control of the Spanish part of Santo Domingo and became the only commander in chief of the entire island. He legitimized his power by elaborating a constitution, approved on July 7, 1801, which gave him all the powers, legislative and executive and allow him to serve as Governor for his lifetime. But Toussaint Louverture's power was short lived. The French troups sent by Napoleon Bonaparte, under the commandment of his brother-in-law General Victoire- Emmanuel Leclerc, entered in a two-year war with Toussaint Louverture's army. The war ended with the death of General Victoire-Emmanuel Leclerc on November 2nd, 1802 and the captivity, deportation, and death in prison of Toussaint Louverture on April 7th 1803. These events were followed by the granting of independence to Haiti on January 1st, 1804. After a year of peaceful relationships, in 1805, General Jean-Jacques Dessalines, first President of Haiti after the country gained its independence, launched a new war against the East cost of the island (Santo Domingo). But the war was short-lived as General Dessalines feared retaliation by the French Government. In 1820, a new unification attempt was launched by President Jean-Pierre Boyer. At that time, Spanish people of Santo Domingo were divided between two groups on the choice of their political regime. One group was in the view that independence of the Spanish part of the island was illusory because of the small population. Some advocated a union with Colombia, which had just gained independence. Others called for a fusion with Haiti. In this context, President Boyer used his diplomatic skills to convince Dominicans to go with the unification of the island. On February 1822, his troups entered in Santo Domingo. However, his policy of "Haitianization" of the whole island disappointed many inhabitants of the island including Haitians themselves. In January 1843, a revolution called "Revolution Praslin" under the commandment of Riviere Herard took place. After being defeated, President Boyer left the country on March 13, 1843, after 25 years of power. Dominicans ultimately proclaimed their independence on February 1844, following Riviere Herard's decision to close all ports of the East Coast to external trading. After a war launched by Riviere Herard to constrain Dominicans to renounce to independence in 1845, and the invasion of the Dominican Republic in March 1849 by Faustin Soulouque (who succeeded President Jean Baptiste Riche), the Dominican Republic reinforced its independence by approving a national constitution. The following years were marked by attempts to sign agreements between the two countries. On November 9, 1874, the two countries signed the first treaty of peace, friendship, trade, and extradition of criminals between them. This treaty established the principle of a customs union between the two countries. It also confirmed the economic and trade dominance of Haiti over its East neighbor. This resulted in Dominicans using Haitian trade and port services for their external trade. The treaty also entailed a financial support of the Haitian Government to the Dominican Republic, with a view to avoid that DR search for financial assistant to compensate for land losses. 4 However, these trade union efforts were stopped by the US occupation of the island for 19 years on the Haitian side (July 28, 1915 to August 21st, 1934) and 8 years on the Dominican side (October 29, 1916 to October 21, 1924). Following the departure of the US from the DR and the election of President Horacio Vasques on March 15, 1924 and Louis Borno on April 10, 1922, the issue of borders was sorted out. A treaty was signed on January 21st 1929 and approved by Congress in the Dominican Republic on February 7, 1929, which establishes the borders between the two countries. The treaty was ratified by Haiti on February 15, 1929 and by the Dominican Republic on February 25, 1929. Despite being independent countries, DR and Haiti share common characteristics. They share the same island space (Quisqueya or Hispaniola 76 780 km2, the second largest in the Caribbean). They display similarities in natural resources endowments and exhibit clear common traits, both cultural and socio-economic. The two countries have always maintained intense commercial exchanges and cultural ties, whether formal or informal, legal or illegal. Haiti is the third largest export market for the Dominican Republic; conversely, the DR is the first export market for Haiti. Haiti is the primary source of migrants for the DR. For thousands of Haitians, migration provides the way to employment in the DR as well as an escape valve for the pressure created by overpopulation and land scarcity at home; for others, political instability at home (n addition to economic needs) has led to voluntary exile in the DR. During the 1990s, some 30,000 Haitians were entering the DR each year, and about half that number was repatriated. Some 150,000 Haitian who remained from the previous 20 years would make a total of 250,000-300,000 emigrés, and these would join the estimated 100,000 Dominicans born of Haitian parents. The presencia haitiana in Dominican society could therefore involve, by this reckoning, some 350,000-400,000 persons of Haitian descent, whether born in Haiti or in the DR. Cultural and intellectual exchanges between the two countries have always been intense, especially along the border as well as among the elites at the national level. The Dominican economy needs Haitian labor and its industry depends largely on this source of manpower. A scarcity of land on the Haitian side continues to push peasants toward the east and to the north. The frontier markets and contraband trade indicate that, despite the stereotypes and notwithstanding the barriers to immigration, a vigorous economic life has created borderland `communities", which offer a flexible pattern of culture more accommodating of otherness than that of the interior zones. There is a fluidity and seeming ease with social interactions, including exchange of language (Haitian Kreyol and Spanish). Another type of exchange involves money, the Haitian gourde being valued here as is the Dominican peso. Intermarriages occur frequently, mostly of Dominican men with Haitian women. The exchanges of the borderlands remind people of the two countries that they share a background that embraces the history of colonialism, the experience of underdevelopment, and the struggle to survive. 5 Despite similarities and common traits, however, the growth patterns and outcomes achieved by the countries in recent decades have been widely different. The DR has gone successfully through a transition to lower middle-income status with substantial economic growth. Haiti has lagged behind in all aspects of socio-economic development. However, articulations and interdependencies are already a fact. With such interdependencies in mind, Muñoz (1995) correctly states that the "superstructural frame" that integrates countries into multinational blocs is not nationalities, but rather markets. Although "national identity and consciousness, by their own nature, cannot be negotiable"- in the sense that they have their own dynamic coherence ­ they do have functions within the nation-state system articulated by economics and migration.2 In fact, it may be argued that the sheer differences between the two countries create mutually beneficial situations building on their individual comparative advantages. Hence, by acting jointly and/or in a coordinated fashion in key areas, individual growth outcomes can be enhanced. Interdependence and cultural exchange continue within the asymmetrical frame of the Haitian and Dominican Republics. The problem of their contiguous coexistence, together with the flow of migrants across the shared border, expresses the situation of two states whose populations must seek the ad hoc solutions to social ills by mean of emigration. Integration could be realized in a fair and efficient regulation of the commercial traffic between the two countries. Such regulation would facilitate trade while providing for the revenue needed to develop each country's infrastructure. Thriving commerce in the frontier towns already indicates ways in which a less encumbered cross-border trade can profit the two national economies. In the Haitian southeast, Haitians and Dominicans come to do business at the rural market of Fonds-Verettes, Situated at a crossroads of major access routes. The market draws farmers, hawkers, and retail dealers from towns like Perdernales, Saltrou, Neiba, and Barahona. As illustrated in the example of Fonds-Verettes, informal linkages of commerce and culture have blurred the division between nations. They also point to the fact that Haitian and Dominican "sister towns" have become interdependent. Such pairings include not only Dajabón and Ouanaminthe, but also Elías Piña and Belladères; Jimani and Fonds Parisien; and Pedernales and Anses-à-Pitres. By transcending the limits of sovereign states, the two countries could define a new, holistic mapping of Hispaniola which could serve as a new paradigm of cooperation, interdependence, and collaboration between nations. Its articulation of insular counterpoints can legitimate and support initiatives for joint plans and projects for expanding commerce between Haiti and the Dominican Republic. 2 Muñoz, Maria Elena. Las Relationes Dominico-Haitianas: Geopolitica y Migracion. Santo Domingo: 6 III. BENEFITS AND COSTS OF AN EMU BETWEEN THE DOMINICAN REPUBLIC AND HAITI During the past few decades, the economic performance of the Dominican Republican and Haiti has diverged markedly. As shown in Figure 1, using 1963 as a base year, by 2007 real output in the Dominican Republic had increased almost ninefold (more than twice the increase in the United States during the same period), compared to quasi stagnation in Haiti. In per capita terms, real GDP tripled to about US$2,500 between 1960 and 2005 in the Dominican Republic, whereas it halved to US$430 in Haiti (Jaramillo and Sancak (2007)). This divergence in economic performance has put the two countries at the opposite ends of the spectrum in Latin America and the Caribbean. Figure 1 Dominican Republic and Haiti: Real GDP Index, 1963­2007 ___ Dominican Republic ___ Haiti --- United States 1,000 900 800 700 600 500 400 300 200 100 0 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 2003 2007 Source: World Bank. Editora Alfa & Omega, 1995. 7 In the last three years, the governments of the Dominican Republic and Haiti have taken several steps to strengthen their commercial ties and set up a bilateral commission on trade. Indeed, there appears to be greater recognition now on both sides of the benefits of regional economic integration. After defining the nature of an economic and monetary union (EMU), this section offers a brief overview of the potential benefits and costs that such a union could bring two countries that are geographically as close as the Dominican Republic and Haiti. The next section will examine the various economic criteria that have been proposed to assess whether these countries should form an EMU. 1. Nature of an Economic and Monetary Union An EMU between two countries can be defined as a single market with a common currency, managed by a single monetary authority. In turn, the adoption of a common currency is generally considered to be the final stage of a gradual process of economic integration, which typically involves the following stages: a) the elaboration of a preferential trading area (PTA), which usually allows preferential access to select goods within the block via a reduction in tariffs, but not necessarily by abolishing them; b) a free trade area (FTA), which involves removing tariffs and quantitative restrictions for trading of most goods amongst members themselves, if production structures are largely complementary; c) a customs union, which involves the formation of a common trade area with a common external tariff policy (and possibly import quotas and a common competition policy); d) a common market, which is formed with the objective of removing barriers for the free movement of capital, goods, services and labor, while retaining a common external trade policy. Once a single market has adopted a common currency, it is termed an EMU.3 Put differently, an EMU differs from a mere currency union because it involves a single market. The adoption of a single currency requires free movement of capital across member states, and the creation of an independent and supranational central bank responsible for implementing a common monetary policy. An EMU also involves harmonizing taxation and 3 The final stage of economic integration, often referred to as "complete economic integration," leads to a near complete harmonization of fiscal and monetary policies among members. It also amounts to a loss of political independence of member nations as they are unable to use the monetary and fiscal policies independently. 8 technical standards, with the objective of bringing in greater efficiency in resource allocation and enhancing competition. 2. Benefits and Costs of an Economic and Monetary Union The fundamental reason for creating an EMU is to bring in greater efficiency in resource allocation and enhanced competition. This is a particularly important consideration for the Dominican Republic and Haiti, given the small size of these economies taken individually. For instance, the formation of a PTA or an FTA could allow firms in the two countries to spread the costs of research and development over a larger market, thus reducing unit costs and encouraging greater innovation and technical progress. This could, in turn, generate positive spillovers as successful innovations are applied more broadly. Integration could boost productivity growth by allowing increased specialization, whereas increased competition could increase efficiency gains--possibly reinforced by foreign direct investment between the two countries. These benefits are further increased with a customs union and especially a common market (through the free movements of factors), although the use of a common external tariff policy may reduce scope for discretionary policy. At the same time, it is well recognized that PTAs and FTAs can entail some costs (see Bhagwati and Panagariya (1996)). In particular, they may lead to trade creation, by replacing relatively high-cost domestic production with lower-cost imports from partner countries. They may also lead to trade diversion, which occurs when imports are switched from efficient nonmember suppliers to less efficient member countries benefiting from tariff preferences. Higher trade volumes between member countries that result from the agreement may also lead to greater, not smaller, losses to an individual member A who joins the agreement from a higher initial set of tariffs, because joining the agreement at a common lower tariff leads to a redistribution of tariff revenues from A to other member countries with initially lower tariffs. Nevertheless, it has been recognized that PTAs and FTAs are worth pursuing for countries whose ultimate goal is complete dismantling of barriers to labor and capital mobility, and complete economic integration with the formation of an EMU. The benefits and costs associated with the final stage of an EMU, the creation of a currency union, has been the focus of much debate. As discussed in more detail in Appendix A, the traditional literature has focused on the reduction in transactions costs and reduced exchange rate uncertainty as the main benefits of a monetary union. Such a union entails a reduction of transaction costs, a reduction in uncertainty about financial variables and macroeconomic policy, and promotes integration of both financial and goods markets. At the same time, however, the formation of an EMU has important implications for national economic policies. It entails the loss of the exchange rate as an adjustment mechanism to combat any deterioration in competitiveness. Members are therefore deprived of one instrument to respond to shocks. At the same time, monetary policy autonomy is 9 "surrendered" to the common Central Bank. Fiscal management and the degree of labor mobility become therefore key factors in the response to adverse shocks. IV. CRITERIA FOR AN EMU: PRELIMINARY DATA ANALYSIS The conditions under which two countries should form an EMU and reap the associated efficiency gains have led over the years to a number of broad economic criteria. For the last stage of an EMU, in particular, these criteria relate to whether these countries can or should form a Common Currency Area (CCA) or a currency union.4 These criteria include the volatility of bilateral exchange rates, the similarity of inflation rates, the degree of correlation between economic shocks across countries, the degree of price and wage flexibility, the degree of labor mobility, and the scope for discretionary fiscal policy.5 In what follows we examine some of the these criteria especially the behavior of some of these variables in the context of the Dominican Republic and Haiti, especially inflation, business cycle synchronization, fiscal policy and public debt, exchange rates, and bilateral trade. Of course, the examination of historical data may provide a misleading indication of suitability of membership in a CCA, given that some of the criteria listed above are interrelated--or endogenous, as discussed later. Nevertheless, this discussion still provides valuable insights and will be complemented in the next section by a more formal analysis. 1. Volatility of Bilateral Exchange Rates It is now well recognized that exchange-rate volatility can be as harmful for intraregional trade (if not more) than tariff barriers. By blurring signals associated with relative price changes, exchange rate volatility (whether nominal or real) may hamper the ability of producers to reallocate resources production. Thus, the formation of a CCA can help to foster trade integration.6 4 The criteria are generally grouped under the header "theory of optimum currency areas" (or OCAs). Essentially, the theory of OCAs--first presented by Mundell (1961), in the context of the debate between fixed and flexible exchange rates of the 1960s. Note that a currency area is typically defined as a group of countries that undertake to fix their exchange rates, at least within narrow bands. In contrast, a currency union is typically defined as an area where a single currency circulates. The bilateral exchange rates are fixed and cannot be changed without a country quitting the union and reintroducing its own currency. In practice, however, the difference between a currency union and a currency area is probably smaller than in principle, given that in both regimes capital mobility constrains monetary policy independence. 5 Ishiyama (1975) provides an early review of the literature. Subsequent discussions include Masson and Taylor (1992), Tavlas (1993), Mongelli (2002), and De Grauwe (2007). See Appendix A for a more detailed review of the costs and benefits of a monetary union. 6 As an example of this growing awareness, in January 2008 members of the East African Community (consisting of Kenya, Tanzania, Uganda, as well as Burundi and Rwanda since July 2007) announced their intention to bring forward, to 2012 from 2015, the formation of their planned monetary union, in an effort to 10 Figure 2 shows the evolution of the bilateral nominal exchange rate between the Dominican Republic and Haiti during the period January 1979-September 2007, using monthly data. It suggests that the exchange rate between the two countries has been quite volatile--especially after the countries switched to a flexible exchange rate (19- for the Dominican Republic and 19- for Haiti) and during periods of financial crisis.7 Thus, the reduction in exchange rate volatility and transactions costs in foreign exchange that a CCA could bring would entail substantial benefits for the two countries. 2. Similarity of Inflation Rates The similarity of inflation rates (at levels consistent with overall macroeconomic stability) is one of the most common criteria used to evaluate the viability of a CCA. If inflation rates diverge too much between two potential members of a union, the ability to implement a common monetary policy (which follows from the adoption of a common currency and the shared goal of price stability) may be significantly hampered. bring about greater exchange rate stability among themselves and boost intraregional trade. 7 With higher frequency data (weekly or daily), the data would show even higher volatility during certain periods. 11 Figure 2 Dominican Republic and Haiti: Bilateral Exchange Rate, 1979­2007 (Dominican pesos per gourde) 2.50 2.00 1.50 1.00 0.50 0.00 1990M1 1991M5 1992M9 1994M1 1995M5 1996M9 1998M1 1999M5 2000M9 2002M1 2003M5 2004M9 2006M1 2007M5 2008M9 Source: International Monetary Fund. Figure 3 shows the behavior of annual inflation--measured in terms of the consumer price index--in the Dominican Republic and Haiti during the period 1963-2007, measured as deviations from US inflation. The figure shows that inflation has displayed significant spikes and considerable volatility at times in both countries--for the Dominican Republic for instance, most recently following the financial crisis and sharp exchange rate depreciation in 2003-04. The correlation between the two series is also quite high at times. These results suggest that although both countries could gain from a CCA in terms of price stability, their medium- and long-term performance in that regard does not appear to have been sufficiently effective and comparable to envisage a rapid transition toward such an arrangement. At the same time, however, it is important to note that convergence in inflation does not need to be complete (in the sense of very narrow differentials among members) before adopting a CCA; as discussed later, the very process of union formation may bring about greater similarity in inflation between member countries--in part through greater integration in goods markets. 12 Figure 3 Dominican Republic and Haiti: Consumer Price Inflation Deviations from US Inflation, 1964-2007 ___ Dominican Republic ___ Haiti 0.8 0.6 0.4 0.2 0 -0.2 -0.4 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 20 Source: World Bank. 3. Degree of Symmetry of Shocks As noted earlier, when a country joins a CCA or a currency union in the last stage of an EMU, the exchange rate becomes unavailable to act as a buffer when shocks hit the economy. A key issue is therefore the extent to which the exchange rate is, to begin with, a "shock absorber." In general, the larger the degree of asymmetry of shocks across a potential currency union the greater the need to absorb shocks at the country level. Put differently, if the potential partners to a CCA have a common business cycle (in the sense that they tend to be affected by shocks in similar ways) rather than divergent business cycles, the costs of fixing the exchange rate across the union will be smaller. By contrast, if members of a currency union are often at different stages of the business cycle, then a common monetary 13 policy for the entire area would not be appropriate to all of them and could be very costly for some of them.8 A reason why countries may face asymmetric (or idiosyncratic) shocks is that they have different commodity export baskets. The likelihood of asymmetric shocks tends to be smaller if a potential union partner is, to begin with, a key bilateral trading partner. This is because a boom in the union partner B will tend to increase demand for country A's exports, while rising demand in country B will also tend to reduce their exports to country A. More generally, a high ratio of bilateral trade to total trade implies that the exchange rate is a relatively ineffective means of securing macroeconomic adjustment, so that foregoing its use (as required in a CCA) is relatively less costly. Put differently, if two countries trade a lot with each other, they are likely to benefit from low transaction costs and the elimination of exchange rate risk. Thus, the extent of bilateral trade is an important consideration in any assessment of the net benefits associated with a CCA. To examine the degree of symmetry of shocks between the Dominican Republic and Haiti, a preliminary approach is therefore to look at the correlations of output gaps in the two countries, the extent of bilateral trade between them, and the behavior of their terms of trade. Figure 4 shows the behavior of the output gap in the two countries during the period 1963- 2007. The output gap is measured by the logarithm of the ratio of real GDP and its trend value, measured itself by the modified Baxter-King filter (see Christiano and Fitzgerald (2003)). The figure does not suggest much correlation between the output gap in the two countries; the actual correlation coefficient is only 0.03. In fact, the correlation between output gaps in the Dominican Republic and the United States is in fact higher, reflecting a greater degree of synchronization between them. 8 The idea that different regions within a currency union will sometimes be subject to shocks that cause their business cycles to diverge is hard to dispute; the key issue, however, is how often divergences occur and how significant they are. This is an empirical issue, which is discussed further later, in the context of Latin America and the Caribbean. 14 Figure 4 Dominican Republic and Haiti: Output Gap, 1963­2007 ___ Dominican Republic ___ Haiti --- United States 0.15 0.1 0.05 0 -0.05 -0.1 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 Source: Author's calculations based on World Bank data. See Appendix B for calculation method. The evolution of bilateral trade between the Dominican Republic and Haiti, using the IMF's Directions of Trade statistics, suggests that measured trade between the two countries remains very low. There are a number of factors that may help to account for this. The first is that any assessment of the extent of trade between Haiti and DR faces a major problem--the coverage of the data. Frontiers between the two countries are highly permeable and informal trade is of great significance, being equivalent according to some estimates to as much as total recorded trade. Nevertheless, even allowing for such effects, the consensus remains that trade between the two countries accounts for only a small fraction of their total trade. One reason why this may be so is related to the non-complementary production structures of these economies, with exports consisting of goods and services (primary products, tourism in the case of the Dominican Republic) heavily in demand by the industrial countries, whereas imports consist mainly of raw materials and finished investment and consumer goods that are not produced domestically. Alternatively, the low level of trade may be explained by the relatively poor transportation and communications networks between the 15 two countries; these networks are mainly geared toward maintaining links with each country's main partner, the United States, rather than between them. There are also historical reasons, as well as language and cultural barriers that may be relevant. However, while the level of trade between the two countries may be undoubtedly lower than one would expect based on their degree of proximity, from a normative standpoint it does not follow that it is somehow too low. Clearly, to make such a judgment requires an assessment of the optimal level of bilateral trade, based on the relevant structural and economic characteristics of the two countries. Although there are no formal studies focusing on this issue, standard trade models would suggest that in fact the level of trade between the Dominican Republic and Haiti is no lower than would be predicted given their structural characteristics. Indeed, it is probable that the low level of bilateral trade is well explained by the low trade potential of these countries, which in turn is a reflection of their very small economic size.9 Figure 5 shows the behavior over the past four decades of the terms of trade for the two countries. Clearly, there does not appear to be any sizable degree of correlation between the behavior of these variables. This reflects to a large extent differences in the composition of exports of the two countries, rather than the structure of their imports. 9 It should also be kept in mind that the overall degree of openness (as measured by total trade ratios) also matters from the perspective of a CCA. The reason is that the more open an economy, the more devaluation becomes a source of imported inflation; this may reduce significantly the benefits associated with an adjustment via currency changes because domestic inflation will then mitigate the initial effect of a nominal depreciation on the real exchange rate. 16 Figure 5 Dominican Republic and Haiti: Terms of Trade Index (2000=100), 1963­2007 ___ Dominican Republic ___ Haiti --- United States 400 1400 350 1200 300 1000 250 800 200 600 150 400 100 200 50 0 0 63 67 71 75 79 83 87 91 95 99 03 07 19 19 19 19 19 19 19 19 19 19 20 20 Source: IMF and World Bank. All three indicators suggest therefore that the Dominican Republic and Haiti are in general affected by asymmetric shocks. By itself, this would suggest that the formation of a CCA between these countries would not be advisable. However, before reaching this conclusion, there are two issues worth highlighting. First, in general two countries might exhibit low business cycle synchronicity not because using different currencies creates underlying structural asymmetries, but simply because the two monetary authorities are following different policies. Conversely, two countries might exhibit a relatively high level of synchronicity only because they are subject to common external shocks. What is needed therefore is to look at conditional business cycle correlations, instead of simple correlation coefficients; this is what is done in the next section. Second, the existence of asymmetric shocks (which are likely to remain, even in a well integrated union) is not sufficient to establish the case for retaining a separate currency. A CCA may still be viable in the presence of asymmetric shocks as long as there are alternative adjustment mechanisms available to deal with them. These mechanisms include most notably the degree of wage and price flexibility, the degree of labor mobility, and the scope for discretionary fiscal policy--as discussed next. 17 4. Degree of Price and Wage Flexibility Adjustment to asymmetric shocks requires in general an adjustment of the real exchange rate. If, as is the case in a currency union, the nominal exchange rate cannot adjust to cope with these shocks, a real exchange rate adjustment may come through movements in prices and wages. If wages are perfectly flexible, a reduction in nominal wages is essential similar to a nominal depreciation in its adjustment role. By contrast, when wages are rigid downward (as a result, for instance, of government interference through minimum wages, or the existence of trade unions), adjustment through wage reductions is much slower and more costly, to the extent that it may be accompanied by periods of high unemployment. Thus, a high degree of wage rigidity makes preserving nominal exchange rate flexibility and monetary independence more desirable. Price flexibility may also bring about an adjustment in the real exchange rate. However, in practice the degree of price flexibility is often limited. The slow response of prices to shocks may be caused by, for instance, price-setting behavior by firms and backward-looking expectations. More generally, it is often difficult in practice to judge whether a country has a sufficient degree of price and wage flexibility to facilitate smooth adjustment to idiosyncratic shocks. In the case of the Dominican Republic and Haiti, there are few studies documenting directly the degree of flexibility. In the case of the Dominican Republic for instance, estimation results by Hernández (2008) suggest that there is some significant degree of nominal inertia in price setting. At the same time however, both countries have a sizable informal sector; According to estimates by Gasparini and Tornarolli (2007, Table 3.2) for instance, informal employment as a share of total employment represented 51.3 in the Dominican Republic in 2004, 82.1 percent in Haiti in 2001. This suggests a high degree of nominal wage flexibility, at least for unskilled workers. 5. Degree of Labor Mobility When labor is mobile across union borders, asymmetric shocks to any individual member can in principle be absorbed through migration, without requiring relative price changes. Labor mobility may thus be particularly important in the presence of wage rigidities. For instance, if the real exchange rate does not adjust following a shift in relative demand across countries, unemployment may increase in the country where demand has been reduced; migration toward the other country where demand has increased may mitigate the problem.10 10 In the United States, labor mobility has been identified as the key regional adjustment mechanism for adjusting to regional unemployment, although migration does not seem to occur only in response to wage differentials. In contrast, labor movements across Europe are very limited compared to those across United 18 In the case of the Dominican Republic and Haiti, there has always been quite a significant degree of labor mobility at the border--although mostly illegal. In addition, mobility has taken the form of mostly unskilled labor flows. However, beyond that mobility has been limited, to a large extent for cultural and historical reasons. There is also evidence that the formal labor market in these countries suffers from a number of imperfections and distortions--including labor regulations that raise the cost of hiring and firing workers (see Inter-American Development Bank (2003)). Thee are also challenges related to youth unemployment, the mismatch between the educational system and the needs of the labor market, the creation of jobs, low levels of productivity coupled with relatively high wages, and sustained emigration flows of skilled labor from the region. This "brain drain" has helped to strengthen the economic links between each country and the United States. In addition, language is likely to remain a persistent barrier between the two countries. Labor mobility is therefore unlikely to provide much scope for absorbing shocks, thereby making wage flexibility and fiscal policy all the more important. 6. Scope for Discretionary Fiscal Policy As noted earlier, countries participating in a CCA must give up the use of monetary policy and the exchange rate. But when a country is subjected to an asymmetric shock, automatic fiscal stabilizers "kick in" and may ease the burden. In addition, discretionary fiscal policy (involving borrowing or lending) within nations could be as effective as centralized institutions in cushioning external shocks. Figure 6 shows the evolution of the overall fiscal balance in the Dominican Republic and Haiti over the period 1960-2007. The figure suggests that neither one of these countries has managed to maintain surpluses for long; on the contrary, both countries have experienced persistent periods of high deficits, which have fueled the accumulation of public debt. This discrepancy between fiscal outcomes makes it difficult to envisage a CCA between the two countries; limits are typically built in CCAs on debt levels and deficits. For instance, the Maastricht Treaty criteria required a 3 percent maximum fiscal deficit and a ratio of public debt to GDP inferior to 60 percent. There are also divergent domestic medium-term considerations that may guide fiscal policy in both countries, such as the need to engage in fiscal consolidation in order to reduce constraints on the financing of private investment, or the need to satisfy qualitative criteria, such as improving tax collection (a key issue in both countries) and/or implementing a better arbitrage between current and investment expenditure. These considerations may severely reduce the scope for discretionary policy actions in response to shocks. States regions. 19 Figure 6 Dominican Republic and Haiti: Overall Fiscal Balance (percent of GDP), 1960-2007 ___ Dominican Republic ___ Haiti 8 6 4 2 0 -2 -4 -6 -8 -10 -12 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Source: IMF. The foregoing discussion provides only mixed evidence (at best) in support of a CCA between the Dominican Republic and Haiti. However, it must be kept in mind, as noted earlier, that simple (unconditional) correlations can be misleading for assessing the degree of comovements between variables. In the short run, unconditional correlation coefficients can vary in response to a wide variety of macroeconomic factors. The analysis performed in the next section (which introduces a number of control variables in a multivariate framework) provides more accurate measures of business cycle synchronicity. More importantly, the criteria used may be endogenous, as a result of the very existence, and induced effects, of a CCA. For instance, it has been argued that similarity of inflation rates may be promoted by participating in a currency union. Hoffman and Remsperger (2005) for instance have found that, for the Euro area, the degree of persistence in inflation differentials fell significantly following the adoption of the common currency in 1999. Similarly, once two countries choose to form a CCA, their business cycle may become more synchronized, as the elimination of currency fluctuations may spur trade between them. There is also evidence suggesting that the 20 formation of a monetary union may foster macroeconomic integration.11 We will return to this issue in the concluding section of the paper. V. DEGREE OF SYMMETRY OF SHOCKS: A VARX ANALYSIS The foregoing discussion has provided a conceptual and preliminary statistical analysis of the various criteria that may be relevant in considering whether the Dominican Republic and Haiti could form an EMU. We now turn to a more formal analysis, based on vector autoregression (VAR) models, with a focus on understanding how the two countries respond to external and domestic shocks. We begin with ha description of the VARX methodology. We then present and analyze two VARX models, which differ by their degree of complexity. 1. VAR and VARX Methodology In general, a standard linear VAR model can be written in the structural form12 (1) where is a vector of n endogenous variables; is a vector of p exogenous variables (the first of which is a constant term); 11 See Fielding and Shields (2005) for the CFA franc Zone and Gil-Pareja, Llorca-Vivero, and Martínez- Serrano (2008) for the Euro Zone. 21 is a vector of n random shocks that are uncorrelated over time and mutually uncorrelated, with a (diagonal) variance matrix . To simplify, we will assume that all variables in this system are stationary, and that the exogenous variables only enter contemporaneously (that is, they are dated at period t). The elements of matrix A, of order n×n (respectively C, of order n×p) are the structural parameters related to endogenous variables (respectively, exogenous) and B(L) is a matrix polynomial which depends on the lag operator L, defined so that Lkzt = zt-k, that is where m+1 is the number of lags and the matrices Bi are of order n×n. Matrix A captures instantaneous causality links (or, more generally, contemporaneous interactions) between endogenous variables. A reduced form of the above system is given by (2) where The matrices D, of order n×n, and E, of order n×p, are nonlinear functions of the structural parameters on the contemporaneous endogenous variables, A, and the contemporaneous response of endogenous variables to the exogenous variables, C. Let denote the variance-covariance matrix of the reduced-form shocks t; exact identification of the parameters of the structural form equations from the estimated parameters of the reduced- 12 See for instance Lütkepohl (2006) for a detailed presentation of standard VAR models and their properties. 22 form equation requires that the number of "free" parameters in A and be equal to the number of independent parameters in . It can be shown that, if the diagonal elements of A are normalized to unity, n(n-1)/2 restrictions on the coefficients of A are needed to achieve exact identification (see for instance Lütkepohl (2006)). Consider for instance the case where n = 3, p = 2, and m = 0; the reduced-form VAR is therefore given by (3) If there are no restrictions on the coefficients dij and eij in matrices D and E, we must estimate 32 + 3×2 = 15 parameters. The problem this number rises rapidly with the number of variables and the number of lags, thereby leading quickly to problems of degrees of freedom and eventually preventing estimation. For instance, with two lags on all endogenous variables (m = 1), 24 parameters should be estimated, and with 3 lags, 33 parameters. Thus, even a parsimonious VAR may quickly exhaust the number of degrees of freedom available in a typical time-series application. A VARX model is an attempt to alleviate this problem. It involves imposing restrictions on the interactions, static and dynamic, between variables in a VAR, on the basis of a priori economic reasoning. Suppose that the model builder "knows" that variables y1t and y2t can interact between themselves (as well as possibly with other variables), but that they do not have any effect on variable y3t, which is itself determined only by its past values (here, the one-period lag), without a constant term. There is therefore an asymmetry between the model's endogenous variables; causality is unidirectional. A VARX system allows the modeler to exploit this asymmetry (or ex ante information) in order to reduce the number of parameters to be estimated, compared to a standard VAR; the resulting increase in the number of degrees of freedom may not only make estimation feasible, but it will also lead to improved efficiency of those parameters that are estimated. In the example being considered, the reduced form can now be written as 23 (4) The system becomes therefore recursive; the four restrictions that have been imposed (d31 = d32 = 0, and e31 = e32 = 0) imply that only 3²-2+2×2 = 11 parameters must be estimated, instead of 15 with the standard VAR. The gain in terms of degrees of freedom is thus quite significant. Equally important, the VARX specification allows the modeler to calculate impulse response functions associated with innovations on those variables that are "partially" endogenous, such as y3t here. By treating y3t as a purely exogenous variable (such as those included in the vector Xt), such calculations would not be feasible. A similar argument holds for variance decompositions. In general, restrictions imposed on the structural form are not always sufficient to guarantee that all structural parameters are identified. Thus, a VARX model is not synonymous to a structural VAR, which requires a specific number of exclusion restrictions (including those associated with the normalization of the system) to achieve exact identification, as discussed earlier. At the same time, the exclusion restrictions typically imposed in a VARX (that domestic variables do not affect foreign variables, for instance) are often used in structural VAR models, given their plausibility.13 Structural VARs have been used by a number of authors to discuss convergence issues in CCAs; among developing countries, these studies (which are based on the Blanchard-Quah technique or some variant of it) include Ahumada and Martinera (2001), Fielding and Shields (2001, 2005), Buigut and Valev (2005), and Saxena (2005), and Allegret and Sanc-Zantman (2009).14 A common criticism of structural VARs, however, is that the restrictions used are quite arbitrary. In this case, applying the Blanchard-Quah technique is particularly problematic: it assumes that aggregate demand shocks have a temporary impact on output, 13 Note that, in the formulation (4), restrictions are imposed directly on the reduced form; equivalently, they could be imposed on the structural form. In the present example, matrices A and B0 (given that m = 0) would become, using the normalization aii = 1, 14 As an alternative approach, Houssa (2008) uses dynamic factor models. 24 whereas aggregate supply shocks are have a permanent effect. But casual observation suggests that many of the shocks that affect the two countries here are temporary supply disturbances, related for instance to climatic shocks (hurricanes, etc.) or terms-of-trade shocks. We therefore do not attempt to identify the characteristics of structural shocks to these economies and focus instead on a reduced-form VARX approach. 2. A Basic VARX Model We consider first a basic VARX model consisting of five variables, which were defined in the previous section: output gaps in the Dominican Republic and Haiti, with trend values measured by the modified Baxter-King filter; inflation rates (measured as deviations from US inflation) in the two countries; and the US output gap, which is not only an indicator of activity abroad and demand for domestic exports for both countries, but also an indicator of capital flows associated with remittances from nationals abroad; indeed both countries have sizable diaspora outside the national territory and benefit from sizable unrequited transfers. By using an integrated model instead of separate models, we are able to analyze directly how domestic variables in each country respond to shocks to variables in the other country, while at the same time accounting for possible interactions between variables in both countries induced by external shocks (in this case, shocks to the US output gap). All five variables are stationary at a standard significance level of 5 percent, as documented by the unit root tests reported in Appendix B. Thus, we can exclude any cointegration relationship between them and, instead of a VECM, we can proceed but a VAR in levels. The first four variables are considered fully interactive in the model (in the sense that the associated parameters are not subject to restrictions), whereas the US output gap is treated as a partially interactive endogenous variable, which is subject to restrictions on its instantaneous causality and interdependence links. The constant term is the only strictly exogenous variable. Suppose that the five variables defined earlier are numbered according to the order in which they were presented, that is, variables 1 to 4 are the fully interactive variables (y1t and y2t being the output gaps in Haiti and the Dominican Republic, and y3t and y4t inflation rates in both countries) and 5 is the partially interactive variable (the US output gap). It is totally plausible to assume that output fluctuations in the U.S. affects output gaps and inflation in the Dominican Republic and Haiti, but it is unlikely that activity and inflation in these two countries have much effect on fluctuations in U.S. output. This variable is taken to depend solely on its past value(s). The strictly exogenous variable (the constant term) is numbered 1. In addition, suppose that the model considers uniformly only one lag, so that m = 0. In matrix form, the VARX model can then be written as 25 where , , and e1, d11 and d12 are submatrices defined as We can infer from this system that the gain in terms of degrees of freedom associated with the VARX relative to a standard VAR with no restrictions of any sort is given by 5*(5+1) ­ 4*(4+1+1) - 1*(1+1) = 4. If, instead, the model has uniformly two lags, so that m = 1, the gain is 5*(10+1) ­ 4*(10+1) - 1*(2+1) = 8. The variance decompositions obtained with two lags are shown in Table 1, using the Cholesky ordering: the US output gap, the output gap in the Dominican Republic, the output gap in Haiti, inflation deviation for the Dominican Republic, and the inflation deviation for Haiti. We therefore take the US output gap to be the most "exogenous" variable, followed by the variables for the Dominican Republic (the bigger of the two island countries), and last the two variables for Haiti.15 In the context of our analysis, variance decomposition tells us the extent to which variability in real output and inflation in each of the two countries is influenced by shocks to common factors--in this basic case, shocks to US output. This is important because differences in the cause of variability in the countries could be indicative of underlying differences in the transmission mechanism and the policy strategies in both countries, which could be an obstacle to monetary integration. 15 The order in which the variables for the Dominican Republic and Haiti was also inverted, to check for robustness; this did not affect significantly the results. 26 Table 1 Basic VARX Model: Variance Decompositions Variance Decomposition of GAP_HT_BK: Period S.E. GAP_HT_BK GAP_DR_BK DLINFL_DUS_HT DLINFL_DUS_DR GAP_US_BK 1 0.024 100.000 0.000 0.000 0.000 0.000 2 0.025 96.484 1.521 1.080 0.647 0.269 3 0.028 79.095 3.499 11.022 4.016 2.368 4 0.028 75.700 3.612 13.209 5.020 2.459 5 0.029 72.533 3.580 15.659 5.053 3.174 6 0.029 71.021 3.491 16.930 4.952 3.606 7 0.030 70.694 3.479 17.256 4.982 3.589 8 0.030 70.446 3.471 17.374 5.033 3.676 9 0.030 70.300 3.469 17.487 5.039 3.704 10 0.030 70.224 3.470 17.569 5.037 3.700 Variance Decomposition of GAP_DR_BK: 1 0.025 0.000 100.000 0.000 0.000 0.000 2 0.028 3.934 94.649 0.038 0.371 1.008 3 0.028 3.792 92.035 0.969 2.156 1.048 4 0.030 3.791 87.786 3.772 2.196 2.454 5 0.031 3.778 85.400 4.859 2.128 3.835 6 0.031 3.840 85.031 5.106 2.186 3.837 7 0.031 3.823 84.319 5.089 2.157 4.612 8 0.031 3.816 84.070 5.080 2.280 4.754 9 0.031 3.833 83.818 5.064 2.407 4.878 10 0.031 3.825 83.579 5.061 2.413 5.121 Variance Decomposition of DLINFL_DUS_HT: 1 0.083 0.697 4.207 95.097 0.000 0.000 2 0.095 1.214 5.121 93.185 0.416 0.064 3 0.107 2.522 5.431 91.126 0.586 0.334 4 0.112 2.796 5.514 90.544 0.697 0.449 5 0.115 2.886 5.341 90.431 0.911 0.431 6 0.116 2.887 5.240 90.232 1.200 0.441 7 0.117 2.858 5.213 89.971 1.517 0.441 8 0.117 2.837 5.238 89.691 1.728 0.505 9 0.117 2.826 5.255 89.545 1.816 0.558 10 0.117 2.823 5.255 89.509 1.850 0.564 Variance Decomposition of DLINFL_DUS_DR: 1 0.100 0.108 5.524 5.732 88.637 0.000 2 0.119 0.573 5.704 6.318 87.265 0.140 3 0.138 2.218 11.025 4.735 79.468 2.553 4 0.145 2.376 11.664 4.266 75.545 6.149 27 5 0.146 2.470 11.492 4.210 74.999 6.828 6 0.147 2.481 11.591 4.243 74.776 6.909 7 0.147 2.479 11.560 4.278 74.575 7.108 8 0.147 2.518 11.550 4.315 74.520 7.096 9 0.148 2.549 11.556 4.325 74.306 7.265 10 0.148 2.550 11.551 4.327 74.253 7.319 Source: Author's calculations Notes: GAP_DR_BK is the output gap for the Dominican Republic, GAP_HT_BK the output gap for Haiti, DLINFL_DUS_DR the inflation differential with respect to the US for the Dominican Republic, DLINFL_DUS_HT the inflation differential with respect to the US for Haiti, and GAP_US_BK the output gap for the United States. The Choleski ordering of the variables is as described in the text. The results indicate that at a horizon of 10 periods, fluctuations in the output gap in the Dominican Republic are explained mostly by its own innovations (more than 80 percent). This is less so for Haiti, with own innovations accounting for about 70 percent, and innovations in domestic inflation accounting for about 21 percent of output gap fluctuations. In both countries fluctuations in inflation are accounted for mostly by their own innovations (about 80 percent), with a somewhat more significant effect of the domestic output gap in the case of the Dominican Republic. In neither case does the US output gap play a significant role in domestic fluctuations. There are also no significant cross-country effect between the Dominican Republic and Haiti. This could reflect the fact that economic fluctuations in both countries are subject to large temporary supply shocks; although these shocks can be correlated (as is the case with hurricanes), this is not always the case. We examine next the impulse response function to an innovation in the US output gap and the response functions of domestic variables in response to shocks in the other country. The objective is to examine if and how each country's domestic macroeconomic variables respond to the same external shock, and if and how they respond to shocks in the other country. To the extent that both countries respond in a similar way to the first shock (that is, by displaying a high degree of symmetry), are better potential candidates for a CCA. Moreover, the speed of response matters also; in general, the slower is the adjustment after disturbances, the larger will be the cost of maintaining a single currency. 28 Figure 7 Basic VARX: Impulse Response Functions Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_DR_BK Response of GAP_DR_BK to GAP_HT_BK Response of GAP_DR_BK to DLINFL_DUS_DR Response of GAP_DR_BK to DLINFL_DUS_HT Response of GAP_DR_BK to GAP_US_BK .04 .04 .04 .04 .04 .03 .03 .03 .03 .03 .02 .02 .02 .02 .02 .01 .01 .01 .01 .01 .00 .00 .00 .00 .00 -.01 -.01 -.01 -.01 -.01 -.02 -.02 -.02 -.02 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of GAP_HT_BK to GAP_DR_BK Response of GAP_HT_BK to GAP_HT_BK Response of GAP_HT_BK to DLINFL_DUS_DR Response of GAP_HT_BK to DLINFL_DUS_HT Response of GAP_HT_BK to GAP_US_BK .04 .04 .04 .04 .04 .03 .03 .03 .03 .03 .02 .02 .02 .02 .02 .01 .01 .01 .01 .01 .00 .00 .00 .00 .00 -.01 -.01 -.01 -.01 -.01 -.02 -.02 -.02 -.02 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_DR_BK Response of DLINFL_DUS_DR to GAP_HT_BK Response of DLINFL_DUS_DR to DLINFL_DUS_DR Response of DLINFL_DUS_DR to DLINFL_DUS_HT Response of DLINFL_DUS_DR to GAP_US_BK .16 .16 .16 .16 .16 .12 .12 .12 .12 .12 .08 .08 .08 .08 .08 .04 .04 .04 .04 .04 .00 .00 .00 .00 .00 -.04 -.04 -.04 -.04 -.04 -.08 -.08 -.08 -.08 -.08 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_HT to GAP_DR_BK Response of DLINFL_DUS_HT to GAP_HT_BK Response of DLINFL_DUS_HT to DLINFL_DUS_DR Response of DLINFL_DUS_HT to DLINFL_DUS_HT Response of DLINFL_DUS_HT to GAP_US_BK .08 .08 .08 .08 .08 .04 .04 .04 .04 .04 .00 .00 .00 .00 .00 -.04 -.04 -.04 -.04 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Source : Author's calculations. Figure 7 shows the response of domestic variables in the two countries to an innovation in all the variables at a horizon of 10 periods and using the ordering described earlier. The solid lines (in blue) in the figures represent the impulse responses themselves, whereas the dotted lines (in red) are the associated 95 percent upper and lower confidence bands.16 The results suggest that domestic variables do not respond significantly to shocks to the US output gap. In addition, domestic variables in one country do not respond either to shocks to domestic variables in the other countries. Thus, these results appear to corroborate the preliminary analysis, based on unconditional correlations, which suggested limited business cycle synchronization between the two countries. At the same time, however, it is possible that the small number of variables included in the VARX does not allow one to properly capture the transmission process of these shocks. To investigate further this potential misspecification problem, we now turn to a more extended model. 16 The impulse responses and their associated confidence intervals are computed using Monte Carlo simulations employing 1,000 draws. Eviews 3.1 was used to perform all computations. 29 3. An Extended VARX Model We now extend the basic VARX model to consider, in addition to the five variables above (the output gaps and inflation deviations in the two countries, and the US output gap), the following variables: a) the ratio of credit to GDP in the two countries; b) the rate of change of the real bilateral exchange rate (with respect to the US dollar) in both countries;17 c) the rate of change of the price of oil in US dollars; d) the real US Treasury bill rate, given by the nominal rate deflated by the US inflation rate in consumer prices. Given what is known about the importance of banks in financing domestic activity in developing countries (see for instance Agénor and Montiel (2008)), one would expect the ratio of credit to the private sector to GDP to play an important role in the transmission of a variety of shocks to output and inflation in the two countries: it can be viewed as measuring the availability of loanable funds, and thus possibly a constraining factor on economic activity. As shown in Figure 8, these ratios have fluctuated significantly during the sample period. Similarly, the real exchange rate, being a key relative price, is also expected to play an important role in the transmission mechanism of both domestic and international shocks. The behavior of these rates is shown in Figure 9, with a rise denoting a depreciation.18 17 We constructed a bilateral real exchange rate index because standard measures, such as the IMF's real effective exchange rate, were not available for the whole sample period. Details of our calculations are provided in Appendix B. 18 In estimating the model, we introduced two dummy variables, taking the value 1 in 1985 and 2003 for the Dominican Republic, and 1992 for Haiti, to account for the financial crises (and large nominal depreciations) that were recorded in both countries during these periods. These "spikes" can be clearly identified in Figure 9. 30 Figure 8 Dominican Republic and Haiti: Credit to the Private Sector, 1963-2007 ___ Dominican Republic _ _ _ Haiti 0.6 0.6 Credit to the Private Sector (Unit to GDP) DR 0.5 0.5 0.4 0.4 0.3 0.3 HT 0.2 0.2 0.1 0.1 0 0 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 Source: IMF. 31 Figure 9 Dominican Republic and Haiti: Bilateral Real Exchange Rates, 1963-2007 ___ Dominican Republic _ _ _ Haiti 1.4 1.4 Bilateral Real Exchange Rate 1.2 1.2 RER_DR 1 1 0.8 0.8 0.6 0.6 RER_HT 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 Source: Author's calculations based on IMF data (see Appendix B). Given that the Dominican Republic and Haiti are both oil importers, accounting for changes in the price of oil, and studying its potential impact on both countries, is also important. In the absence of sufficiently long time-series on the domestic-currency retail price for each country (which would have the advantage of taking into account changes in subsidies and taxes), we use a common price index in foreign prices.19 While the US output gap captures (as before) "trade" and "remittance" channels, the US interest rate captures a "financial" channel, that is, a measure of borrowing conditions on world capital markets. This is more appropriate for the Dominican Republic (given its ability to borrow on world capital markets in normal times) as opposed to Haiti, which cannot borrow on these markets.20 19 See Appendix B for an exact definition of the index. 20 Ideally, what would be more appropriate to measure financial shocks for the Dominican Republic would be a country-specific measure of borrowing costs on world capital markets, which would account for the country-specific spread above and over the risk-free US Treasury rate. Unfortunately, such a variable is not available for the whole period under study. 32 As shown in Appendix B, all of these new variables are also stationary at a significance level of 5 percent, except for the US real interest rate, which is significant at a 10 percent significance level; we therefore specify again the VARX in levels. For convenience, we rename the US output gap variable as y11t. The degree of endogeneity of the new variables, numbered now 5 to 10, is as follows. Variables y5t and y6t, the real exchange rates in Haiti and the Dominican Republic, respectively, are taken to be fully interactive, just like the output gaps and inflation rates. All the other variables are taken to be partially interactive. Variables y7t and y8t,, the private sector credit variables in Haiti and the Dominican Republic, respectively, are taken to affect directly only the output gap in the respective country; they have no direct effect on the other fully interactive variables, inflation and the real exchange rates. In addition, these variables are also taken to depend only on their own lagged value and the lagged output gap, to capture possible bidirectional causality between credit and economic activity. Variable y9t, the rate of change of the price of oil, depends only on its past value; it can affect all domestic variables in the fully interactive block. Variable y10t, the US real interest rate, is assumed to depend not only on its past value but also depend on lagged value of the US output gap, which is now y11t. Conversely, the US output gap is also assumed to depend now not only on its lagged value but also on the lagged value(s) of the real US interest rate and oil prices. Thus, the model now accounts for interactions between the US variables themselves, and between oil prices and US activity--while maintaining the unidirectional causality between US variables and domestic variables in Haiti and the Dominican Republic (y10t and y11t can affect all domestic variables). Just as before, all of the partially endogenous variables can affect the fully interactive variables. Given the increase in the number of variables, to preserve degrees of freedom, the number of lags is restricted uniformly to 1 (m = 0). In matrix form, the extended VARX model can then be written as where now , , , , , 33 and the submatrices dij, based on the restrictions defined earlier, are given by21 21 The dependence of the credit-to-GDP ratio on the (lagged) output gap in the respective country, as noted earlier, explains why the submatrix d21 does not contain only zeroes but instead is specified as having d71 34 and Based on these matrices, we can calculate as before the gain in terms of degrees of freedom that the extended VARX model provides, relative to a standard VAR model with no restrictions of any sort. From the above matrices, this gain is now given by 11*(11+1) ­ (6*6 + 20 + 2 + 8) = 121 ­ 66 = 55.22 The variance decompositions are shown in Table 2, using the following Cholesky ordering: the world price of oil, the US real interest rate, the US output gap, the credit variables in the Dominican Republic and Haiti, the output gap in the Dominican Republic and Haiti, inflation deviation for the Dominican Republic and Haiti, and the real exchange rates in the two countries. We therefore take the world price of oil as the most "exogenous" variable, followed by the US variables; the rest of the ordering is as for the basic VARX model.23 For the purpose at hand, the important issue is the extent to which variability in the "core" macro variables (real output, inflation, and the real exchange rate) in each of the two countries is influenced by shocks to common external factors--captured now by shocks to two US variables and world oil prices. The results indicate that at a horizon of 10 periods, own innovations play a much less significant role for output gaps and inflation rates--in contrast to the results obtained with the 0 and d82 0. 22 In an unpublished Appendix (Appendix C, available upon request), we also estimated an alternative extended VARX model, in which the real exchange rate variables were treated only as partially interactive. Although the gain in terms of degrees of freedom implied by the additional restrictions proved substantial (equal to 121 ­ 46 = 75, with only 46 parameters to estimate, instead of 66), the results obtained were very similar to those reported here. 23 Again, the order in which the variables for the Dominican Republic and Haiti was inverted to check for robustness, but this did not have a significant impact on the results. 35 basic VARX model. For the output gap, the share of all three variables is about 7 percent for the Dominican Republic and 10 percent for Haiti. For inflation, the shares are 12 percent and 10 percent, respectively, and for the real exchange rate 7 percent and 8 percent, respectively. Innovations in credit ratios play a significant role (compared to the real exchange rate) in explaining fluctuations in the output gap in both countries; at the same time, innovations in output gaps and inflation rates have a sizable effect on fluctuations in the real exchange rate in both countries. Country effects are a bit more significant as well. Nevertheless, fluctuations in the external factors account for a relatively small share of the fluctuations in the domestic variables. Thus, the extended VARX model show some significant differences in differences in the cause of variability in the countries could be indicative of underlying differences in the transmission mechanism and the policy strategies in both countries, which could be an obstacle to monetary integration. 36 Table 2 Extended VARX Model: Variance Decompositions Source: Author's calculations. Notes: OIL is the rate of change of the world price of oil, RATE_US is the real US Treasury rate, CRED_DR (CRED_HT) is the ratio of private sector credit to GDP in the Dominican Republic (Haiti), and RER_DR (RER_HT) is the real exchange rate for the Dominican Republic (Haiti). For the other variables, see the note to Table 1. The Choleski ordering of the variables is as described in the text. 37 We examine next the impulse response functions associated with both domestic and external shocks. These functions, which are constructed as described earlier, are shown in Figures 10 to 16. For clarity, we report only those related to the output gap, inflation, and the real exchange rate in both countries. Without getting into all the details of these results, some broad features can be identified. In general, and contrary to what one would have expected for some of them (e.g., a shock to the US output gap, given the importance of the United States as a trade partner and a source of remittances for both countries), external shocks do not have large effects on domestic variables--as can be inferred from an examination of Figures 10 to 12. For instance, in Figure 10, which illustrates the impact of an oil price shock, the only significant responses are a slight increase in the inflation differential (on impact in the Dominican Republic, in the second period in Haiti), and an initial real appreciation in the Dominican Republic; the response of the other variables are not statistically significant, and therefore not much can be inferred from their pattern. By contrast, for domestic shocks, responses vary depending on the shock. A positive shock to the output gap in the Dominican Republic has no statistically significant effects on other domestic variables or variables in Haiti (see Figure 13), whereas a positive shock to the output gap in Haiti generates a small increase in the rate of real depreciation (brought about by a nominal exchange rate depreciation, given that the inflation differential for the country does not drop by much) and no effect on variables in the Dominican Republic (see Figure 14). More interestingly, although a positive shock to the inflation differential for Haiti affects only the country's real exchange rate (it leads to a significant and sustained real appreciation, as could be expected, see Figure 16) and no foreign variables, a positive shock to the inflation differential for the Dominican Republic leads not only to a higher rate of inflation domestically but also to higher inflation in Haiti (see Figure 15). Although the impact on inflation in Haiti is less persistent (two years instead of three for the Dominican Republic) it is highly significant statistically. 38 Figure 10 Impulse Response functions: Oil Price Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to OIL Response of GAP_HT_BK to OIL .015 .015 .010 .010 .005 .005 .000 .000 -.005 -.005 -.010 -.010 -.015 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to OIL Response of DLINFL_DUS_HT to OIL .06 .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.06 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to OIL Response of RER_HT to OIL .10 .06 .04 .05 .02 .00 .00 -.05 -.02 -.10 -.04 -.15 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 39 Figure 11 Impulse Response functions: US Real Interest Rate Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to RATE_US Response of GAP_HT_BK to RATE_US .010 .010 .005 .005 .000 .000 -.005 -.005 -.010 -.010 -.015 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to RATE_US Response of DLINFL_DUS_HT to RATE_US .04 .04 .03 .02 .02 .01 .00 .00 -.01 -.02 -.02 -.04 -.03 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to RATE_US Response of RER_HT to RATE_US .04 .08 .02 .00 .04 -.02 .00 -.04 -.04 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 40 Figure 12 Impulse Response functions: US Output Gap Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_US_BK Response of GAP_HT_BK to GAP_US_BK .010 .015 .005 .010 .000 .005 -.005 .000 -.010 -.005 -.015 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_US_BK Response of DLINFL_DUS_HT to GAP_US_BK .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.06 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_US_BK Response of RER_HT to GAP_US_BK .12 .06 .04 .08 .02 .04 .00 .00 -.02 -.04 -.04 -.08 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 41 Figure 13 Impulse Response functions: DR Output Gap Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_DR_BK Response of GAP_HT_BK to GAP_DR_BK .03 .010 .02 .005 .01 .000 .00 -.005 -.01 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_DR_BK Response of DLINFL_DUS_HT to GAP_DR_BK .06 .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_DR_BK Response of RER_HT to GAP_DR_BK .08 .04 .02 .04 .00 .00 -.02 -.04 -.04 -.08 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 42 Figure 14 Impulse Response functions: Haiti Output Gap Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_HT_BK Response of GAP_HT_BK to GAP_HT_BK .020 .03 .015 .02 .010 .01 .005 .00 .000 -.01 -.005 -.010 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_HT_BK Response of DLINFL_DUS_HT to GAP_HT_BK .06 .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.06 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_HT_BK Response of RER_HT to GAP_HT_BK .10 .08 .06 .05 .04 .00 .02 -.05 .00 -.10 -.02 -.15 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 43 Figure 15 Impulse Response functions: DR-US Inflation Differential Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to DLINFL_DUS_DR Response of GAP_HT_BK to DLINFL_DUS_DR .008 .015 .004 .010 .000 .005 -.004 .000 -.008 -.005 -.012 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to DLINFL_DUS_DR Response of DLINFL_DUS_HT to DLINFL_DUS_DR .12 .06 .08 .04 .04 .02 .00 .00 -.04 -.02 -.08 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to DLINFL_DUS_DR Response of RER_HT to DLINFL_DUS_DR .04 .06 .02 .04 .00 .02 -.02 .00 -.04 -.02 -.06 -.04 -.08 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 44 Figure 16 Impulse Response functions: Haiti-US Inflation Differential Shock Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to DLINFL_DUS_HT Response of GAP_HT_BK to DLINFL_DUS_HT .004 .004 .000 .000 -.004 -.004 -.008 -.008 -.012 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to DLINFL_DUS_HT Response of DLINFL_DUS_HT to DLINFL_DUS_HT .03 .08 .02 .06 .01 .04 .00 .02 -.01 .00 -.02 -.03 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to DLINFL_DUS_HT Response of RER_HT to DLINFL_DUS_HT .100 .02 .075 .01 .00 .050 -.01 .025 -.02 .000 -.03 -.025 -.04 -.050 -.05 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 45 VI. SUMMARY AND POLICY IMPLICATIONS The purpose of this paper was to offer a preliminary assessment of the potential benefits and costs of an economic and monetary union (EMU) between the Dominican Republic and Haiti--two countries sharing the same island but whose history is one of conflict and divergent economic prospects in recent decades. After a brief review of the historical context, it examined the nature of these potential benefits and costs. It then conducted a preliminary analysis (using basic statistical techniques) of some key criteria for the formation of an EMU between the two countries. The third part conducted a more formal analysis of business cycle synchronization, based on integrated (two-country) VARX models as opposed to standard VARs, to economize on degrees of freedom and permit estimation using annual time series. The goal of the analysis was to examine if and how each country's domestic macroeconomic variables respond to the same external shock, and if and how they respond to shocks in the other country. To the extent that both countries respond in a similar way to external shocks (that is, by displaying a high degree of symmetry), they are better potential candidates for a CCA. Moreover, the speed of response matters also; in general, the slower is the adjustment after disturbances, the larger will be the cost of maintaining a single currency. A basic VARX model (involving only inflation rate differentials in the two countries relative to US inflation, and output gaps, with the US output as the common external shock) and an extended model (involving, in addition to the previous variables, the world oil price, the US real interest rate, bilateral real exchange rates, and private sector credit-to-GDP ratios) were formulated, with explicit and plausible restrictions on the degree of interactions between these variables, and estimated using standard techniques. In general, impulse response functions indicated that innovations in external variables do not seem to generate clear and common patterns in domestic variables, thereby suggesting limited business cycle synchronization between the two countries in response to the occurrence of common external shocks. This is also the case for domestic shocks--there appear to be limited "spillovers effects" from one country to another--except for positive shocks to inflation in the Dominican Republic, which appear to translate quickly into inflation in Haiti. Overall, both the preliminary analysis and the more formal VARX models suggest that at this stage several economic criteria are not satisfied for the two countries to fully benefit from an EMU. For instance, the lack of similarity in the response and adjustment process to shocks, both external and domestic, may reflect very different policy responses between the two countries; indeed, if shocks are not immediately offset by a policy response, then their effect will vary substantially between the two countries, with no obvious common policy response appropriate to both of them. In turn, this suggests potentially severe coordination problems. 46 At the same time, however, it is important to keep in mind that most of the criteria for forming an EMU (including the degree of business cycle synchronization) are endogenous and may change in response to an explicit and credible commitment to move toward greater integration. This militates in favor of an aggressive medium-term agenda for integration and partnership between the two countries--provided that the political will to do so goes beyond the recent rhetoric. Indeed, although the focus of this paper has been on the economic criteria for the formation of an EMU, the lesson of history is that the political commitment toward forming and ensuring the viability of a union by all members may be just as important. 47 Appendix A Benefits and Costs Associated with a Currency Union Currency unions entail various benefits and costs. The traditional literature has focused on the reduction in transactions costs and reduced exchange rate uncertainty as the main benefits of monetary union, whereas more recent work has emphasized potential gains in terms of financial integration and credibility. One of the key disadvantages is the loss of monetary independence--because it deprives members of the ability to alter the nominal exchange rate of their currency in response to a shock. This Appendix provides a more detailed overview of these benefits and costs. A.1 Benefits Associated with Currency Unions A.1.1 Reduced Transactions Costs A common and immediate benefit from currency union is the elimination of transactions costs; there is no longer a need to convert currencies when trading with other countries within the union. These benefits can be particularly large for members for which weaker financial markets and technology makes transacting in foreign currencies more costly, and trade is denominated mostly in foreign currencies (as is the case for most developing countries). If, in addition to currency conversion, trade with the rest of the world involves additional financial costs (in the form, for instance, of letters of credit), a currency union may generate cost savings in those areas as well. A.1.2 Reduced Exchange Rate Uncertainty An immediate benefit of a common currency is a reduction in exchange rate risk (that is, the uncertainty associated with exchange rate movements) involved in trade and financial transactions between countries. For instance, a contract in the seller's currency can leave buyers uncertain as to the amount (in their own currency) that they will actually have to pay upon delivery. While there are ways to hedge against this risk (by using forward contracts and options, for instance), doing so may be costly or impossible, especially when a developing country currency is involved.24 Moreover, if a currency union reduces exchange rate variability relative to union partners, and leads to a decrease in the risk premium built into domestic interest rates, it may stimulate investment, by making attractive "marginal" projects that are not currently undertaken due to the relatively high cost of capital; there is therefore an "investment creation" effect, as discussed by Agénor and Aizenman (2008). Thus, the larger the trade and financial flows between member countries, the greater the gain from a reduction in exchange rate volatility.25 24 For instance, contracts that would allow domestic borrowers to hedge foreign currency-denominated obligations into domestic currency obligations may not be readily available for a sufficiently long horizon, corresponding to the term of the transaction. 25 An adverse effect of exchange rate volatility on trade in developing countries may be due to the fact that, in these countries, hedging markets do not exist, or may be very illiquid. Parsley and Popper (2006) for 48 However, the evidence about the effects of exchange rate variability on exports, capital flows, and investment is largely inconclusive, suggesting that the magnitude of saving may not be very high. For instance, using a broad sample of eighty seven countries (both industrial and developing) from 1970 to 1997, and after accounting for various sources of bias, Tenreyro (2007) found that exchange-rate variability had no significant impact on trade. This finding suggests that the availability of forward contracts, currency options, and other alternatives for risk diversification and management may provide sufficient hedging to reduce the potential drawbacks of exchange-rate variability on trade. From that perspective alone, then, the benefit of a currency union may be mitigated. At the same time, however, there is other evidence suggesting that confirms that exchange rate volatility may still be an important discouraging factor when it comes to trade flows involving developing countries among themselves, or between themselves and developed countries (see for instance Inter-American Development Bank (2006)). In such conditions, the gains highlighted above would remain quite substantial. A.1.3 Increased Financial Integration A currency union may (all else equal) improve welfare of all members by increasing financial diversification in two ways. First, by reducing currency risk, as noted earlier, it allows national financial intermediaries to seek greater exposure on their liability side. Second, it offers greater scope to these intermediaries to diversify their lending portfolios across member countries. From the perspective of individuals, a single currency area may also offer risk-sharing benefits when domestic capital markets are limited in their ability to facilitate consumption insurance (Ching and Devereux (2003)). As discussed later, however, to the extent that credit market imperfections continue to persist after the formation of the union (because of difficulties of harmonizing bankruptcy legislation among members with very different legal systems to begin with, for instance), these gains may be slow to materialize. A.1.4 Improved Credibility Joining a currency union provides a credible constraint for a country wishing to "tie its hands" and prevent discretionary changes in short-run monetary policy that are inconsistent with the long-run objective of low and stable inflation (Guillaume and David Stasavage (1999)). The decision to surrender the power to alter the exchange rate and manipulate interest rates is indeed a possible way to signal a country's commitment to low inflation and influence inflation expectations.26 In a sense, membership in a monetary union plays a role instance found that Asia-Pacific firms remain significantly exposed to fluctuations in major currencies and that their the degree of foreign exchange exposure has not diminished over time--suggesting that hedging options remain limited. 26 See Agénor (1991) for a more detailed discussion. As discussed in that paper, for a policymaker concerned with both inflation and competitiveness, the desirability of "tying one's hands" as a solution to the time inconsistency problem depends on what one's hands are tied to. When union members have stable, low inflation rates, precommitment to a fixed exchange rate may help demonstrate domestic resolve to maintaining 49 equivalent to appointing a "conservative" central banker, highly averse to inflation (see Rogoff (1985)). The weaker the political institutions are, the more desirable it is for the authorities to adopt an institutional arrangement that imposes large (political or otherwise) costs on reneging on it. Put differently, because joining a monetary union imposes irrevocably fixed exchange rates (or at least rates that are very difficult to change), it may provide a "quick gain" in credibility. A.2 Costs Associated with Currency Unions Costs associated with joining a currency union relate essentially to the use of the exchange rate as a policy instrument, particularly in the presence of large external shocks. Depending on the nature of these costs, a country's commitment to the "rules of the game" of a monetary union may lack credibility--thereby hampering the extent to which membership in a union can overcome time inconsistency problems. A.2.1 Loss of Monetary Independence From the standpoint of an individual country, the most important cost of a currency union is the loss of monetary independence. When effective, an independent monetary policy can be used as a stabilization tool, in order to dampen cyclical fluctuations. As discussed later, the greater the degree of asymmetry of shocks across countries, and the more limited alternative adjustment mechanisms are to these shocks, the greater the cost associated with the loss of monetary independence. However, as argued by Grubel (2005), the loss of monetary sovereignty that joining a currency union entails may be highly beneficial in some countries, to the extent that it also ensures greater independence of monetary policy from political influence. And the more open an economy, the more devaluation becomes a source of imported inflation; this may reduce significantly the benefits associated with an adjustment via currency changes because domestic inflation will mitigate the initial effect of a nominal depreciation on the real exchange rate. Moreover, if the economy of the potential union member is highly dollarized, the cost of losing monetary independence may not be as high as it commonly thought. Indeed, if debts are denominated largely in foreign currency, and the pass-through from exchange rates to prices and foreign-currency liabilities occurs promptly, a nominal depreciation could lead to bankruptcies and depression, through large balance sheet effects. Thus, in a dollarized economy, adjustment through currency devaluation loses an important advantage over a deflationary adjustment, namely, the capacity to dilute the real value of the debt. By implication, this suggests that the higher the initial degree of liability dollarization, the less the cost of foregoing monetary independence. financial discipline. But when the economy is subject to large nominal shocks, the credibility gain may be outweighed by the cost of lost autonomy. 50 A.2.1 Flexibility provided by an Alternative Nominal Anchor Irrevocably fixing a country's nominal exchange in the context of a currency union is not the only way to provide a strong nominal anchor to the economy and constrain monetary policy. Countries can impose such constraints by adopting an alternative credible nominal anchor, namely the inflation rate itself, as several developing countries have done in recent years with relative success (see Gonçalves and Salles (2008)). At the same time, an inflation targeting regime allows some degree of monetary independence, thereby providing some discretion to respond to (temporary) shocks. However, it is also true that institutional requirements for the adoption of inflation targeting are quite demanding, and that the credibility that it confers is not automatic--it requires sustained success in achieving the target. 51 Appendix B Data Sources and Unit Root Tests The data underlying the analysis presented in this paper are taken essentially from the IMF's International Financial Statistics and the World Bank. The output gap is measured as the log deviation of real GDP from its trend, with the trend value calculated using the modified Baxter-King filter (see Christiano and Fitzgerald (2003)). The output gap for the United States is calculated in a similar fashion. The inflation rate differential is measured by the difference between domestic inflation and inflation in the US; each inflation rate (at an annual rate) is measured in terms of the consumer price index. The US real interest rate is measured as the nominal three-month T-bill rate minus the US inflation rate (in terms of consumer prices). The world price of oil is the price of Brent crude in US dollars. The bilateral real exchange rate is constructed for each country by taking the nominal exchange between the national currency and the US dollar (average over the period, converted into an index), multiplied by the US consumer price index, and divided by the domestic consumer price index. The ratio of credit to the private sector is measured as claims of commercial banks on the private sector, divided by nominal GDP. The unit root tests (based on standard ADF tests) for all the variables used in the basic and extended VARX models are reported in Table B1. As can be inferred from the results, the null hypothesis of nonstationarity is rejected at 1 percent (critical value) for most of the variables, at 5 percent for all but one (the realinterest rate in the US), and at 10 percent for all of them. 52 Table B1 ADF Tests in Level (With a non zero intercept and no trend) OIL -6.164*** RATE_US -2.837* GAP_US_BK -6.146*** CRED_DR -6.958*** CRED_HT -4.984*** GAP_DR_BK -5.124*** GAP_HT_BK -5.502*** DLINFL_DUS_DR -3.514** DLINFL_DUS_HT -3.043** RER_DR -7.747*** RER_HT -4.734*** ADF is the Augmented Dickey Fuller t-statistic. *** denotes significance at 1 percent; ** 5 percent; and * at 10 percent. The critical values are respectively (-3.589, -2.930, and -2.603) Lag selection was via SIC with a maximum of 9 lags 53 Appendix C Alternative Extended VARX Model With partially Endogenous Bilateral Real Exchange Rates We extend the basic VARX model of the paper to consider, in addition to the five variables above (the output gaps and inflation deviations in the two countries, and the US output gap), the following variables: a) the ratio of credit to GDP in the two countries (which is a measure of the availability of loanable funds); b) the rate of change of the real bilateral exchange rate vis-à-vis the US dollar in both countries; c) the rate of change of a world index of commodity prices (including oil, given that both countries are oil importers); d) the real US 3-month Treasury bill rate (given by the nominal rate deflated by the US inflation rate in consumer prices). For convenience, we rename the US output gap variable as y11t. The degree of endogeneity of the new variables, numbered now 5 to 10, is as follows. Variables y5t and y6t, the real exchange rates in Haiti and the Dominican Republic, respectively, are taken to be partially interactive; they affect directly only the output gap in the respective county. It depends only on its lagged value and the output gap. Similarly, variables y7t and y8t,, the credit variables in Haiti and the Dominican Republic, respectively, only affect the output gap in the respective country, with no direct effect on inflation or the real exchange rate. In addition, it also depends only on its lagged value and the output gap, to capture bidirectional causality between credit and economic activity. Variable y9t, the rate of change in world commodity prices, depends only on its past value(s); it can affect all domestic variables. Variable y10t, The US real interest rate, is assumed to depend not only on its past value(s) but also on lagged value(s) of the US output gap, which is now y11t. Conversely, the US output gap is also assumed to depend now not only on its lagged value but also on the lagged value(s) of the real US interest rate and world commodity prices. Thus, the model now accounts for interactions between the US variables themselves, and between world commodity prices (including oil) and US activity--while maintaining the unidirectional causality between US variables and domestic variables in Haiti and the Dominican Republic (y10t and y11t can affect all domestic variables). In the same vein as before, all of the partially endogenous variables can affect the fully interactive variables. 54 Given the increase in the number of variables, to preserve degrees of freedom, the number of lags is restricted uniformly to 1 (m = 0). In matrix form, the extended VARX model can then be written as where now and the submatrices dij, based on the restrictions defined earlier, are given by27 27 The dependence of the credit-to-GDP ratio on the (lagged) output gap in the respective country, as noted earlier, explains why the submatrix d21 does not contain only zeroes but instead is specified as having d71 0 and d82 0. 55 and 56 Based on these matrices, we can calculate as before the gain in terms of degrees of freedom that the extended VARX model provides relative to a standard VAR model with no restrictions of any sort. This gain is now given by 11*(11+1) ­ (4*4 + 16 + 4 + 10) = 121 ­ 46 = 75. The following figures show the impulse response functions, which are broadly similar to those reported in the text. 57 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to OIL Response of GAP_HT_BK to OIL .016 .015 .012 .010 .008 .005 .004 .000 .000 -.005 -.004 -.008 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to OIL Response of DLINFL_DUS_HT to OIL .04 .06 .02 .04 .00 .02 -.02 .00 -.04 -.02 -.06 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to OIL Response of RER_HT to OIL .10 .06 .04 .05 .02 .00 .00 -.05 -.02 -.10 -.04 -.15 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 58 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to RATE_US Response of GAP_HT_BK to RATE_US .012 .010 .008 .005 .004 .000 .000 -.005 -.004 -.010 -.008 -.012 -.015 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to RATE_US Response of DLINFL_DUS_HT to RATE_US .04 .04 .03 .02 .02 .01 .00 .00 -.01 -.02 -.02 -.04 -.03 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to RATE_US Response of RER_HT to RATE_US .04 .08 .02 .00 .04 -.02 .00 -.04 -.04 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 59 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_US_BK Response of GAP_HT_BK to GAP_US_BK .010 .015 .005 .010 .000 .005 -.005 .000 -.010 -.005 -.015 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_US_BK Response of DLINFL_DUS_HT to GAP_US_BK .06 .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.06 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_US_BK Response of RER_HT to GAP_US_BK .12 .04 .08 .02 .04 .00 .00 -.02 -.04 -.08 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 60 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_DR_BK Response of GAP_HT_BK to GAP_DR_BK .03 .010 .02 .005 .01 .000 .00 -.005 -.01 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_DR_BK Response of DLINFL_DUS_HT to GAP_DR_BK .06 .06 .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_DR_BK Response of RER_HT to GAP_DR_BK .08 .04 .02 .04 .00 .00 -.02 -.04 -.04 -.08 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 61 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to GAP_HT_BK Response of GAP_HT_BK to GAP_HT_BK .015 .03 .010 .02 .005 .01 .000 .00 -.005 -.01 -.010 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to GAP_HT_BK Response of DLINFL_DUS_HT to GAP_HT_BK .06 .04 .04 .03 .02 .02 .01 .00 .00 -.02 -.01 -.04 -.02 -.06 -.03 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to GAP_HT_BK Response of RER_HT to GAP_HT_BK .10 .08 .06 .05 .04 .00 .02 -.05 .00 -.10 -.02 -.15 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 62 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to DLINFL_DUS_DR Response of GAP_HT_BK to DLINFL_DUS_DR .0050 .012 .0025 .008 .0000 .004 -.0025 .000 -.0050 -.004 -.0075 -.0100 -.008 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to DLINFL_DUS_DR Response of DLINFL_DUS_HT to DLINFL_DUS_DR .08 .06 .04 .04 .02 .00 .00 -.04 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to DLINFL_DUS_DR Response of RER_HT to DLINFL_DUS_DR .04 .04 .02 .02 .00 .00 -.02 -.02 -.04 -.04 -.06 -.08 -.06 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 63 Response to Cholesky One S.D. Innovations ± 2 S.E. Response of GAP_DR_BK to DLINFL_DUS_HT Response of GAP_HT_BK to DLINFL_DUS_HT .004 .002 .002 .000 .000 -.002 -.002 -.004 -.004 -.006 -.006 -.008 -.008 -.010 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of DLINFL_DUS_DR to DLINFL_DUS_HT Response of DLINFL_DUS_HT to DLINFL_DUS_HT .03 .08 .02 .06 .01 .04 .00 .02 -.01 .00 -.02 -.03 -.02 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Response of RER_DR to DLINFL_DUS_HT Response of RER_HT to DLINFL_DUS_HT .100 .02 .075 .01 .050 .00 .025 -.01 .000 -.02 -.025 -.03 -.050 -.04 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 64 References Agénor, Pierre-Richard, "Credibility and Exchange Rate Management in Developing Countries," Journal of Development Economics, 45 (August 1991), 1-16. Agénor, Pierre-Richard, and Joshua Aizenman, "Capital Market Imperfections and the Theory of Optimum Currency Areas," Working Paper No. 14088, National Bureau of Economic Research (June 2008). Agénor, Pierre-Richard, and Peter J. Montiel, Development Macroeconomics, 3rd ed., Princeton University Press (Princeton, New Jersey: 2008b). Agbeyegbe, Terence D., "On the Feasibility of a Monetary Union in the Southern Africa Development Community," unpublished, City University of New York (September 2003). Alesina, Alberto, and Robert J. Barro, "Currency Unions," Quarterly Journal of Economics, 117 (May 2002), 409-36. Alesina, Alberto, Robert J. Barro, and Silvana Tenreyro, "Optimal Currency Areas," in NBER Macroeconomics Annual, ed. by Mark Gertler and Keneth Rogoff, MIT Press (Cambridge, Mass.: 2002). Allegret, Jean-Pierre, and Alain Sand-Zantman, "Does a Monetary Union Protect against External Shocks? An assessment of Latin American integration," Journal of Policy Modeling, 31 (January 2009), 102-18. Bacha, Obiyathulla I., "A Common Currency Area for ASEAN? Issues and Feasibility," Applied Economics, 40 (- 2008), 515-29. Barro, Robert J., and Silvana Tenreyro, "Economic Effects of Currency Unions," Economic Inquiry, 45 (January 2007), 1-23. Bayoumi, Tamim, and Jonathan D. Ostry, "Macroeconomic Shocks and Trade Flows within Sub-Saharan Africa: Implications for Optimum Currency Arrangements," Journal of African Economies, 6 (- 1997), 412-44. Belke, Ansgar, and Daniel Gros, "Monetary Integration in the Southern Cone," North American Journal of Economics and Finance, 13 (- 2002), 323-49. Bhagwati, Jagdish, and Arvind Panagariya, "Preferential Trading Areas and Multilateralism: Strangers, Friends or Foes?," in The Economics of Preferential Trade Arrangements, ed. by Jagdish Bhagwati and Arvind Panagariya, the American Enterprise Institute Press (Washington DC: 1996). 65 Buigut, Steven K., and Neven T. Valev, "Eastern and Southern Africa Monetary Integration: A Structural Vector Autoregression Analysis," unpublished, Georgia State University (June 2005). Christiano, Lawrence J., and Terry J. Fitzgerald, "The Band Pass Filter," International Economic Review, 44 (May 2003), 435-65. De Grauwe, Paul, Economics of Monetary Union, 7th ed., Oxford University Press (Oxford: 2007). De Grauwe, Paul, and Francesco P. Mongelli, "Endogeneities of Optimum Currency Areas: What Brings Countries Sharing a Single Currency Closer Together?," Working Paper No. 395, European Central Bank (April 2005). Fielding, David, and Kalfinder Shields, "Modeling Macro-economic Shocks in the CFA Franc Zone," Journal of Development Economics, 66 (- 2001), 199-223. -----, "The Impact of Monetary Union on Macroeconomic Integration: Evidence from West Africa," Economica, 72 (- 2005), 683-704. Frankel, Jeffrey A., and Andrew K. Rose, "The Endogeneity of the Optimum Currency Area Criteria," Economic Journal, 108 (July 1998), 1009-25. Furceri, David, and Georgios Karras, "Is the Middle East an Optimum Currency Area? A Comparison of Costs and Benefits," Open Economies Review, 19 (September 2008), 479- 91. Gasparini, Leonardo, and Leopoldo Tornarolli, "Labor Informality in Latin America and the Caribbean: Patterns and Trends from Household Survey Microdata," unpublished, CEDLAS (February 2007). Glick, Reuven, and Andrew Rose, "Does a Currency Union Affect Trade? The Time- Series Evidence," European Economic Review, 46 (June 2002), 1125-51. Gil-Pareja, Salvador, Rafael Llorca-Vivero, and José Antonio Martínez-Serrano, "Trade Effects of Monetary Agreements: Evidence for OECD Countries," European Economic Review, 52 (May 2008), 733-55. Gonçalves, Carlos Eduardo S., and João M. Salles, "Inflation Targeting in Emerging Economies: What Do the Data Say?," Journal of Development Economics, 85 (February 2008), 312-18. Hernández, Raúl, "Fiscal Policy, Phillips Curve and Nonlinearities in Monetary Policy: Theories and Evidence for the Dominican Republic," unpublished, Rutgers University (October 2008). 66 Hoffman, Boris, and Hermann Remsperger, "Inflation Differentials among the Euro Area Countries: Potential Causes and Consequences," Journal of Asian Economics, 16 (June 2005), 403-419. Houssa, Romain, "Monetary Union in West Africa and Asymmetric Shocks: A Dynamic Structural Factor Model Approach," Journal of Development Economics, 85 (February 2008), 319-47. Inter-American Development Bank, Good jobs wanted, Inter-American Development Bank (Washington DC: 2003). Ishiyama, Yoshihide, "The Theory of Optimum Currency Areas: A Survey," IMF Staff Papers, 22 (July 1975), 344-83. Jaramillo, Laura, and Cemile Sancak, "Growth in the Dominican Republic and Haiti: Why has the Grass Been Greener on One Side of Hispaniola," Working Paper No. 07/63, International Monetary Fund (March 2007). Karras, Georgios, "Is Africa an Optimum Currency Area? A Comparison of Macroeconomic Costs and Benefits," Journal of African Economies, 16 (March 2007), 234- 58. Lafrance, Robert, and Pierre St-Amant, "Optimal Currency Areas: A Review of the Recent Literature," Working Paper No. 99-16, Bank of Canada (Octoer 1999). Masson, Paul R., and Mark P. Taylor, "Currency Unions: A Survey of the Issues," in Policy Issues in the Operation of Currency Unions, ed. by Paul R. Masson and Mark P. Taylor, Cambridge University Press (Cambridge: 1993). Mongelli, Francesco P., "`New' Views on the Optimum Currency Area Theory: What is EMU Telling Us?," Working Paper No. 138, European Central Bank (April 2002). Montiel, Peter J., "Exchange Rate Policy and Growth in the Dominican Republic," unpublished, Williams College (November 2006). Mundell, Robert A., "A Theory of Optimum Currency Areas," American Economic Review, 51 (September 1961), 657-65. ------, "Uncommon Arguments for Common Currencies," in The Economics of Common Currencies, ed. by Harry G. Johnson and Alexander K. Swoboda, G. Allen and Unwin (London: 1973). Neves, J. Anchieta, Leandro Stocco, and Sergio Da Silva, "Is Mercosur an Optimum Currency Area,?" MPRA Paper No. 2758 (November 2007). 67 Saxena, Sweta C., "Can South Asia Adopt a Common Currency?," Journal of Asian Economics, (August 2005), 635-62. Tavlas, George S., "The `New' Theory of Optimum Currency Areas," World Economy, 16 (November 1993), 663-85. Tenreyro, Silvana, "On the Trade Impact of Nominal Exchange Rate Volatility," Journal of Development Economics 82 (March 2007), 485-508. World Bank, Dominican Republic: Reforming Institutions for a more Efficient Public Expenditure Management, Public Expenditure Review, Report No. 23852-DO, World Bank (Washington DC: March 2004). -----, Dominican Republic: The Foundations of Competitiveness, Country Eco-nomic Memorandum, Report No. 35731-DO, World Bank (Washington DC: September 2006). -----, Caribbean Trade Report, LAC Region (September 2008). 68