ECCU Business Cycles
Impact of the U.S.

Contributor Notes

Authors’ E-Mail Addresses: ysun2@imf.org and wsamuel@imf.org

With a fixed peg to the U.S. dollar for more than three decades, the tourism-dependent Eastern Caribbean Currency Union (ECCU) countries share a close economic relationship with the U.S. This paper analyzes the impact of the United States on ECCU business cycles and identifies possible transmission channels. Using two different approaches (the common trends and common cycles approach of Vahid and Engle (1993) and the standard VAR analysis), it finds that the ECCU economies are very sensitive to both temporary and permanent movements in the U.S. economy and that such linkages have strengthened over time. There is, however, less clear-cut evidence on the transmission channels. United States monetary policy does not appear to be an important channel of influence, while tourism is important for only one ECCU country.

Abstract

With a fixed peg to the U.S. dollar for more than three decades, the tourism-dependent Eastern Caribbean Currency Union (ECCU) countries share a close economic relationship with the U.S. This paper analyzes the impact of the United States on ECCU business cycles and identifies possible transmission channels. Using two different approaches (the common trends and common cycles approach of Vahid and Engle (1993) and the standard VAR analysis), it finds that the ECCU economies are very sensitive to both temporary and permanent movements in the U.S. economy and that such linkages have strengthened over time. There is, however, less clear-cut evidence on the transmission channels. United States monetary policy does not appear to be an important channel of influence, while tourism is important for only one ECCU country.

I. Introduction

Given the close proximity between the United States and Caribbean countries, it is unsurprising that there are close economic relationships and strong linkages among these economies. The countries of the Eastern Caribbean Currency Union (ECCU) are no exception and a recognition of this fact resulted in the shifting of the Eastern Caribbean dollar peg from the pound sterling to the U.S. dollar in 1976. This shift has the potential for transmitting U.S. monetary policy to the ECCU countries and further strengthening the economic relationships between the two. By far the most important influence of the U.S. on the tourism-dependent ECCU countries is through tourism receipts, with trade in goods playing a much smaller role. About one third of the stayover tourists to the ECCU countries are from the U.S., the top tourism-source country. These economies are also heavily dependent on the U.S. for foreign direct investment mainly in tourism. In contrast, the U.S. accounts for less than 5 percent of ECCU exports. Furthermore, the flow of remittances is an important channel of influence, with a significant proportion of Caribbean migrants making their homes in the U.S.

The recent slowdown in the U.S. economy raises questions about the extent of the spillover effects of cyclical fluctuations to the ECCU economies. Thus the motivation of this paper is to quantify the effects of U.S. business cycles on ECCU economies and identify the channels of spillovers. This paper follows on previous work on the Caribbean to identify the characteristics of Caribbean business cycles, for example Cashin (2006) and Kandil (2009). Two empirical procedures are used in this paper. The first is based on the “common trends and common cycles” approach developed by Vahid and Engle (1993). The paper decomposes real GDP into trend and cycle for selected Caribbean economies treating the ECCU as a single country, and like Roache (2008) estimates the growth elasticities of the cycle and trend to U.S. growth. The paper also uses standard VAR analysis to estimate the magnitude of spillovers from the U.S. to the ECCU and identify different channels through which spillovers occur, along the lines of Bayoumi and Swiston (2008).

The analysis based on common trends and common cycles reveals that both the trend and cycle of the ECCU economies are highly sensitive to movements in the U.S., with a growth elasticity close to 1. This analysis also finds that reactions to U.S. economic movements, both trend and cycle, can vary significantly across Caribbean economies, with different directions and magnitudes. Furthermore, the VAR analysis reveals the strong and increasing impact of the U.S. economic movements on the ECCU. However, evidence on the channels for spillover is less clear-cut and requires further investigation. United States monetary policy does not appear to be an important channel of influence, while tourism is statistically important for only one ECCU country.

The paper is organized as follows: Section II reviews the literature on business cycles and spillovers. The methodology and data issues are discussed in Section III and the empirical results are analyzed in Section IV. The final section discusses the policy implications and gives some concluding remarks.

II. Business Cycles and Spillovers

A. Analysis of Business Cycles in the Caribbean

There is a dearth of literature on Caribbean business cycles. The Caribbean literature has focused more on trend growth rather than fluctuations, and data inadequacies, in particular the relatively short time series, are also a serious challenge to such analysis. Earlier analyses of Caribbean business cycles have tended to transform the data by differencing or filtering to ensure stationarity.2 Recent studies by Cashin (2006) and Kandil (2009) have used more efficient filters to analyze the characteristics of Caribbean business cycles.

Using an ideal band-pass frequency filter to extract the cyclical component from the real GDP series for the ECCU countries, Cashin (2006) observed that there is strong co-movement between Canadian and Caribbean classical business cycles. He found less synchronization with U.S. and United Kingdom (U.K.) classical cycles (defined as sequence of expansion in out followed by a sequence in contraction in output). The analysis also showed that:

  • Caribbean classical cycles are asymmetric with long periods of expansion and short sharp periods of contraction. Their growth cycles (defined as a sequence of output expansion above trend followed by below trend output growth), on the other hand are more symmetric in duration and amplitude.

  • The classical cycles are also longer than similar cycles in other middle-income developing countries, and developed countries.

  • ECCU growth cycles are also more synchronized with Canadian growth cycles than they are with the U.S. The close relationship with Canada was likely due to the flow of development assistance from Canada, the dominance of Canadian banks in the financial system and the flow of remittances from Canada.

Explaining the asymmetries of output growth between the expansion and contraction phases of Caribbean business cycles was the main focus of Kandil (2009). The analysis includes a larger sample of Caribbean countries, which encompasses the ECCU sample in Cashin 2006, and used annual real GDP for the period 1975–2006. After filtering the data to remove trend, regression analysis is used to discriminate between the responsiveness or output and inflation to demand shocks during contraction and expansion, controlling for natural disasters and oil price shocks. Kandil finds that output increases more slowly in the expansionary phase and contracts sharply in a recession. On the contrary prices rise more quickly in the expansion phase and deflation is less during contractions. This is consistent with a kinked supply curve related to rigidities in the economies.

B. Common Trend and Cycle Analysis

Common trend and common cycle analysis has not been used extensively in the literature on business cycles. Two exceptions are Hernandez (2004) and Roache (2008). Hernandez employed a two-country version of the Vahid-Engle methodology to estimate the sensitivity of Mexican business cycles to those of the U.S. He found that the response of short-term fluctuations of the Mexican economy to shocks in the U.S. appear to be stronger when the Vahid-Engle methodology in used. In particular he found that:

  • The short-term elasticity estimated from this method was larger than that estimated using other methods like the Hodrick-Prescott filter.

  • The short-term elasticity (response to temporary shocks) was much higher than the long-term elasticity (response to permanent shocks), which was less than 1.

  • The short-term fluctuations in the Mexican economy in response to temporary shocks in the U.S. are more pronounced (3.78 times) than in the U.S. itself, implying that “when the U.S. sneezes Mexico catches the cold.”

Roach (2008) uses a multicountry version of the Vahid-Engle methodology to analyze common cycles and common trends between Central America and the U.S. This allows him to estimate both regional and U.S. influences on Central American cycles and trends. Using data from 1950 to 2006 for six Central American countries he concludes that the cyclical linkages are stronger than previously thought. The U.S. and Central America share a common business cycle, however, the linkages between long-run growth shocks are weak because of the influence of military conflicts, common terms of trade shocks and poor policy responses in Central America. As a consequence, simple regression and other methods that average the short-run (cycle) and long-run (trend) growth elasticities would show a weaker relationship than that implied by cyclical fluctuations alone.

C. Transmission of U.S. Shocks to the Caribbean

Five channels have been proposed in the literature through which spillovers can be transmitted from the U.S. to the ECCU. These are trade, commodity prices, financial markets, remittances and official development assistance. For the Caribbean, the trade channel (trade in tourism services) would appear to be the most important means of transmitting U.S. growth shocks to ECCU countries, followed by remittances.

Trade. As noted earlier the proportion of ECCU goods exports to the U.S. is very small, while the bulk of imports comes from that source. The major influence on ECCU output however comes from the impact on tourism. The U.S. is the single largest source of tourist arrivals to the ECCU countries, accounting for almost one-third of stay-over tourist arrivals.

Commodity prices. The ECCU countries are net commodity importers. Moreover, their major commodity exports (bananas and sugar in the early years) are mainly to the U.K. Thus commodity prices are unlikely to be a significant channel of spillover from the U.S. On the import side, their major commodity imports are mainly from non-U.S. sources (Trinidad and Tobago for petroleum, Guyana for rice and sugar, and Canada for wheat).

Financial markets. Kluyev (2008) identifies indirect and direct channels through which U.S. financial markets can cause spillovers to the Canadian economy. In a similar vein, U.S. monetary conditions can influence economic developments in the ECCU. The exchange rate peg to the U.S. dollar transmits U.S. monetary policy to the ECCU.3 However, this influence is likely to be quite small given the low elasticity of investment and consumption to interest rate changes. Foreign direct investment to the ECCU countries has been significant, averaging about 21 percent of GDP over the last five years. A significant portion of foreign investment in the ECCU, which is concentrated in the tourism sector, comes from the U.S. Thus, to the extent that a tightening of domestic U.S. financial conditions can either cause a delay or outright cancellation of foreign direct investment by U.S. corporations, they serve as a direct transmission mechanism to the ECCU.

Remittances. The six Fund members of the ECCU are in the world’s top-twenty countries in terms of highest rates of migration, and a large proportion of these migrants live in the U.S. (Mishra 2006). Consequently, remittances could provide a strong link between the U.S. and ECCU economies. The sign of the relationship however, cannot be determined a priori. Rappoport and Doquier (2005) show that there are several motives for remittances, and the net effect depends on which motive dominates. For example, under the altruism and exchange motives a decline in income in the destination country would result in a decline in remittances, however, if the migrant’s home country is experiencing a recession at the same time, the insurance motive would suggest an increase in remittances. Empirically, it has been shown that remittances to developing countries are typically countercyclical.

Official Development Assistance (ODA). ODA to the ECCU countries has declined recently, but, in the past, accounted for a significant amount of foreign inflows. To the extent that budgetary resources in the donor countries have a cyclical component that translates into fluctuations in ODA, this could be another source of spillovers. However, the U.S. has not been a significant source of ODA for the ECCU countries hence it would likely play a minor role, if any, in transmitting shocks.

III. Econometric Methodology and Data

A. The Common Trends and Common Cycles Approach

The technique used in this paper is based on the “common trends and common cycles” approach illustrated by Vahid and Engle (1993). Essentially this approach estimates both long-run and short-run comovements (i.e., trends and cycles) in a set of time series. This section will briefly go over this approach, while details can be found in Vahid and Engle (1993).

The common trends and common cycles approach is built on the existence of cointegration relationships. Consider a VAR representation for a set yt of n variables:

yt=μ+A1yt1+A2yt2++Akytk+εt(1)

where yt is a vector of n endogenous variables, and εt is a vector of white noise. Equation (1) can be written in a VAR error correction form:

Δyt=μ+Ayt1+i=1k1ΓiΔyti+εt(2)

where A=i=1kAiI

and Γi=j=i+1pAj.

With no cointegration A will be a zero matrix. If the series of yt are cointegrated, the matrix A can be decomposed into A = βα′ where α is the matrix of cointegration vectors and β is the matrix of speed of adjustment. The existence of r cointegration vectors implies (n-r) common trends.

Δyt=μ+β(αyt1)+i=1k1ΓiΔyti+εt(3)

Once cointegration vectors are estimated, we can test the existence of common cycles using the canonical correlation procedure developed by Vahid and Engle (1993). The presence of common cycles implies that there exist s linear combinations of Δyt which cannot be forecasted and eliminate serial correlation. In other words, there are s cofeature vectors, implying (n-s) common cycles. Basically the squared canonical correlation between Δyt and its relevant history x =(αyt−1, Δyt−1 ,...., Δytk+1) are calculated. The number of cofeature vectors (s) equals the number of squared canonical correlations equal to zero. The squared canonical correlations are obtained by solving for the eigenvalues of the matrix constructed by Δyt and x.

The test statistic for the null hypothesis that there are at least s cofeature vectors is:

C(k,s)=(Tk1)j=1slog(1λj2)(4)

where λj2 ’s are the s smallest squared canonical correlations between Δyt and its relevant history x =(αyt−1, Δyt−1 ,...., Δytk+1). Under the null, the statistic C(k,s) has a chi-squared distribution with s(s + nk + r – n) degrees of freedom.

When the numbers of common cycles and common trends add up exactly to the total number of variables, i.e., r + s = n, Vahid and Engle (1993) suggested that each series of yt can be decomposed into permanent (trend) and transitory (cycle) components as follows:

yt=α˜α˜yt+ααyt=trend+cycle(5)

where α is the matrix of cointegration vectors and α˜ is the matrix of cofeature vectors.

An important special case is a first-order cointegration system where k = 1. In this case the error-correction representation equation (2) would have no lagged differences on the right-hand side:

Δyt=μ+Ayt1+εt(6)

If the matrix of A has rank of r, there will be r cointegration vectors and n – r common trends. We can see that the left null place of A is s = n – r. So there exists s cofeature vectors and r = n – s common cycles. Specifically, it can be shown that:

(IA1)Δyt=(IA1)(μ+Ayt1+εt)=(IA1)(μ+εt)=an innovation.(7)

In sum, all first-order cointegration systems have common cycles.

B. The VAR Analysis

Vector Autoregressions (VARs) have long been used to analyze spillover effects across countries. In this paper, we try to identify the magnitude by which shocks to U.S. growth affect the growth of individual ECCU countries by estimating impulse responses. The basic VAR equation can be written as:

(ΔyUS,tΔyi,t)=A1(ΔyUS,t1Δyi,t1)++Ak(ΔyUS,tkΔyi,tk)+(εUS,tεi,t)(8)

where ΔyUS,t is the real growth rate of the U.S. at time t and Δyi,t is the real growth rate of an ECCU country i at time t. Since we can safely assume the dominance of U.S growth shocks, the order of Cholesky decomposition is straightforward. An impulse response function can help us trace the effect of a one-time shock to the U.S. growth rate, one of the endogenous variables in the VAR, on the ECCU growth rate.

The procedure in Swiston and Bayoumi (2008) is followed to identify channels through which shocks to U.S. growth are transmitted to individual ECCU countries. Essentially, we augmented the basic VAR equation by adding each possible channel, for example, trade and financial, as an exogenous variable. The individual channel’s contribution to spillovers would equal to the difference between this response and the one from the basic VAR in equation (8):4

ci,j=riri,j.(9)

The impulse response from the VAR with only real growth is ri and ri,j is the impulse response of domestic growth to U.S. shocks from the VAR with channel j included.

C. The Data

The analysis uses annual real GDP data for 1963–2007, covering most CARICOM countries—Belize, Barbados, Guyana, Jamaica, six ECCU Fund member countries (Antigua and Barbuda, Dominica, Grenada, St. Kitts and Nevis, St. Lucia, and St. Vincent and the Grenadines), and Trinidad and Tobago—and the U.S. The data were obtained from the IMF’s World Economic Outlook (WEO) database. Unavailability of quarterly real GDP data for many of the Caribbean countries including the ECCU, the main focus of the study, prevents the use of quarterly data for business cycle analysis.

Summary statistics of real GDP and GDP growth are provided in Table 1. All real GDP series are found to be I(1) while their first differences are I(0).

Table 1.

Summary Statistics of Real GDP growth

article image
Source: Authors’ calculations.Notes: Sample moments were computed from log-difference of real GDP.

P-values from augmented DF unit root tests with lags selected using Aikake information criteria.

IV. Empirical Results

This section presents our findings from the two empirical approaches discussed above. First, results from the common trends and common cycles approach are presented. In this study, all six ECCU Fund-member countries are aggregated to represent the currency union as a whole. The sample, therefore, has 7 time series, including 6 Caribbean countries (Belize, Barbados, Guyana, Jamaica, ECCU, and Trinidad and Tobago) and the U.S. Second, results from standard VAR analysis are presented, focusing specifically on the impact of U.S. growth shocks on individual ECCU countries. The sample also covers 7 countries, including six ECCU Fund-member countries and the U.S.

A. Caribbean Common Trends and Common Cycles

Cointegration and cofeature vectors

The first and key step is to choose the lag order of the system, that is the number of k in equation (1). As indicated in Table 2, two of the four selection criteria indicate a lag order of one. Based on this, we proceed to test the existence of cointegration in a first-order system. Both eigenvalue and trace tests suggest 3 cointegration vectors, which imply 4 common trends among the seven GDP time series (Table 3).

Table 2.

VAR Lag Order Selection

article image
Source: Authors’ calculations.

indicates lag order selected by the criterion

LR: Sequential modified likelihood ratio test statistic (each test at 5% level)AIC: Akaike information criterionSC: Schwarz information criterionHQ: Hannan-Quinn information criterion
Table 3.

Tests for the Number of Cointegrating Vectors

article image
Source: Authors’ calculations.

As indicated above, all first-order cointegration systems have common cycles. There is no need to test for common cycles based on calculating the square canonical correlations. In this case, the existence of 3 cointegration vectors (i.e., 4 common trends) in a first-order system suggests the existence of 4 cofeature vectors, i.e., 3 common cycles. Figures 1 and 2 plot the 4 common trends and 3 common cycles estimated for the seven countries in the sample. In essence, Caribbean economies are found to be exposed to these common trends and common cycles, although individual exposure to a particular common trend or cycle could be very different, as will be revealed below.

Figure 1.
Figure 1.

Three Common Cycles

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

Source: Authors’ calculations.
Figure 2.
Figure 2.

Four Common Trends

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

Source: Authors’ calculations.

Decomposition into cycles and trends

Since the number of common trends (four) and the number of common cycles (three) add up to 7, the number of variables, we can decompose real GDP series into trend and cycle components for each of the seven countries in the sample, which are plotted in Figures 3 and 4 together with trend/cycle components derived from the standard HP filter. The common trends and common cycles approach has a tendency to derive the trend component with a relatively high volatility, as seen in the original application to the U.S. consumption by Vahid and Engle (1993).

Figure 3.
Figure 3.

Caribbean Countries: Cyclical Components of Real GDP

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

Source: Authors’ calculations.
Figure 4.
Figure 4.

Caribbean Countries: Trend Components of Real GDP

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

Source: Authors’ calculations.

Growth elasticities

It would be reasonable to assume that the U.S. has been driving the common trends and common cycles these small open Caribbean economies are exposed to. Following Roache (2008), we use ordinary least squares (OLS) to estimate how much growth in the Caribbean would respond to cyclical and trend shocks in the U.S. The two basic equations to estimate cyclical and trend growth elasticities, respectively, are as follows5:

Δyicycle=ci+αiΔyUScycle+βΔyUStrend+εi(10)
Δyitrend=ci+αiΔyUScycle+βΔyUStrend+εi(11)

Main estimation results are summarized as follows (Table 4), with diagnostics of these models shown in Table 5:

Table 4.

Growth Elasticities in the Caribbean 1/

article image
Source: Authors’ calculations.

Elasticity of the cyclical and trend component of growth in each Caribbean economy to the cycle and trend in the U.S., with ***,**, and * indicating significance at the 1, 5, and 10 percent levels, respectively.

Table 5.

Diagnostics of Growth Elasticity Models

article image
Source: Authors’ calculations.

P-value of the test statistic if the null hypothesis of no autocorrelation were true.

  • Growth trend and cycle in the ECCU are found to synchronize closely with those of the U.S., with estimated cyclical and trend growth elasticities close to 1. Indeed, the simple growth correlation between the U.S. and ECCU (0.4) is statistically significant and the highest among all Caribbean economies in the sample.

  • Barbados and Trinidad and Tobago are also found to be significantly affected by both the trend and cycle in the U.S., although directions of linkages vary. Consistent with the result of simple correlation analysis, Trinidad and Tobago’s trend and cycle are negatively related to those of the U.S. Barbados’ cycle is positively affected by that of the U.S. while its trend is negatvely associated with the U.S. trend. This helps explain the small GDP growth correlation between Barbados and the U.S. (0.21).

  • The elasticity of the cycle for Barbados and Guyana to the U.S. is greater than 1, which would imply an exaggerated effect of the U.S. fluctuations on these economies. On the contrary, Jamaica’s economic activity is fairly inelastic to fluctuations in the US economy, which is consistent with the findings of Murray (2007).

  • In sum, Caribbean countries in the sample appear to differ in terms of the impact of the U.S. business cycle and trend on them. This is in constrast with Roache’s (2008) study on Central America, which suggested that Central America has its own regional trend while the cycle is driven by the U.S. There are Caribbean economies such as ECCU that have been heavily affected by the U.S. short-run and long-run economic movements. Guyana and Jamaica, however, appear to have their own growth trends, although their cycles are influenced by that of the U.S. The decoupling of the Guyana and Jamaica trends cound be a result of the long periods of economic crisis experienced by these two countries, and the associated contractionary policies pursued, which could have blunted the response to changes in the U.S. In the case of Belize, we find that its growth trend, not cycle, is heavily affected by that of the U.S. Together, these results suggest that Caribbean economies may not be as homogenous as generally perceived.

B. Spillovers from the U.S. to the ECCU

This section uses VAR analysis to quantify spillovers and transmission channels from the U.S. to ECCU economies.

  • First, we use standard VAR analysis to estimate the impact of a 1 percent U.S. growth shock on the growth of each ECCU economy. Since quarterly GDP growth data are unavailable for the ECCU countries, the VAR analysis uses annual data with a lag of one. To see the evolution of spillovers over time, we also conduct estimation for three sample periods: 1963–2007, 1976–2007, and 1989–2007. We divide the sample period this way because the ECCU countries pegged their common currency (the Eastern Caribbean dollar) to the U.S. dollar starting from 1976, and 1989 is the year from which data on stayover tourist arrivals from the U.S. to the ECCU are available.

  • Second, we look at channels through which U.S. shocks could affect the ECCU. Three possible channels—trade, financial, and commodity prices—are considered. As noted earlier, for the ECCU countries the trade channel is through tourist arrivals from the U.S., rather than traditional trade in goods. The unavailability of reliable data on remittances does not allow us to explore remittances as a transmission channel.

We find that spillovers from the U.S. are large drivers of business cycles in the ECCU countries, which is consistent with what we have found above using the common trends and common cycles approach. Moreover, the magnitude of spillovers has strengthened over time.

  • For the entire sample period of 1963–2007, the ECCU economies as a whole are found to respond to a 1 percent U.S. growth shock by 0.4 percentage point in the first year. The estimated response increased to 0.7 and 0.8 percentage point for the subsample periods of 1976–2007 and 1989–2007, respectively, suggesting that the impact of the U.S. business cycles on the ECCU economies has strengthened over time (Figure 5).

Figure 5.
Figure 5.

ECCU: Responses to One Percent U.S. Growth Shock

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

  • Similar results are obtained for individual ECCU economies (Figure 6). For the whole sample period 1963–2007, individual countries’ responses to a one percent U.S. growth shock range from 0.3 to 0.7 percentage point in the first year. Moreover, the spillover effects are found to have strengthened over time. The estimated responses a one percent U.S. growth shock, using a sub-sample period of 1976–2007, have increased markedly for every single ECCU economy, reaching 0.4–1.4 percentage point. With only the exception of St. Vincent and the Grenadines, the estimated responses are higher for the period of 1989–2007.

Figure 6.
Figure 6.

ECCU: Country Responses to One Percent U.S. Growth Shock

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

To identify spillover channels, we use annual growth of stayover arrivals from the U.S. to capture the trade channel and U.S. interest rates (three-month T-bill rate and the yield on ten-year government bonds) to capture the financial channel. The world commodity prices used are annual percent changes of the WEO fuel and nonfuel commodity indexes.

  • For all ECCU countries, adding financial indicators or world commodity prices to the basic VAR equation does not alter much the estimation results, i.e., ri and ri,j are almost the same, suggesting that financial channel and commodity prices do not account for much of the spillovers from the U.S. to the ECCU. This result is not surprising, considering the very limited degree of financial integration of the ECCU with the U.S., despite the peg to the U.S. dollar.

  • There is only limited evidence that trade (i.e., tourism) might be the spillover channel. In the case of Antigua and Barbuda, the largest ECCU economy, annual growth of tourist arrivals from the U.S. helped explain about a half of the response of Antigua and Barbuda to a 1 percent growth shock in the U.S. in the first year (Figure 7). However, this result does not hold for the other five ECCU Fund member countries—adding annual growth rates of tourist arrivals from the U.S. does not change much the impulse responses estimated under the basic VAR equations.

Figure 7.
Figure 7.

Antigua and Barbuda: Responses to One Percent U.S. Growth Shock

Citation: IMF Working Papers 2009, 071; 10.5089/9781451872187.001.A001

V. Conclusions and Policy Implications

Using two different approaches, this paper finds that ECCU economies are very sensitive to both temporary and permanent movements in the U.S. economy, and that such linkages have strengthened over time. Based on these results, the ECCU economies cannot be expected to escape from the impact of the current economic downturn in the U.S. There is, however, less clear-cut evidence on the transmission channels. United States monetary policy does not appear to be an important channel of influence, reflecting the relative stability of ECCU interest rates and the low elasticity of spending to interest rates changes. In addition, tourism is important for only one ECCU country (Antigua and Barbuda). More research in this direction is warranted.

The strong sensitivity of economic activity in the ECCU to U.S. fluctuations would require implementation of offsetting policies. Monetary policy is muted because the ECCU maintains a hard peg against the U.S., reducing monetary policy independence. Accordingly, fiscal policy would need to take the brunt of any adjustment. However, given high debt levels and extremely tight fiscal position in most countries, further exacerbated by recent food and fuel shocks, there might not be sufficient fiscal room to maneuver. It is therefore important to continue efforts at fiscal consolidation to reduce the high public debt levels and create room for countercyclical fiscal policy in the future.

Structural reforms would also be particularly important to increase the flexibility of the economies to respond to external shocks. These reforms should be aimed at reducing the rigidities in the economy that limit the supply responses and create asymmetries over the business cycle. Inflexible labor markets, product market imperfections and inefficiencies in the investment climate have been identified as some of the major sources of rigidities.

The diversity of responses of Caribbean economies to shocks in the U.S. would suggest that there is still some way to go to achieve the convergence necessary for regional monetary integration. In particular, a proposed monetary union that includes Trinidad and Tobago and ECCU members could prove to be challenging. Monetary policy under a proposed common currency for the region will need to take account of the disparate responses of the economies to external shocks and may need to be complemented by fiscal action.

References

  • Cashin, P., 2006, “Key Features of Caribbean Business Cycles,” in R. Sahay, D. Robinson and P. Cashin (eds.), The Caribbean: From Vulnerability to Sustainability, Washington DC: International Monetary Fund, pp. 83121.

    • Search Google Scholar
    • Export Citation
  • Craigwell, R. and A. Maurin, 2002, “Production and Unemployment Cycles in the Caribbean: The Case of Barbados and Trinidad and Tobago,” (unpublished, Research Department, Central Bank of Barbados).

    • Search Google Scholar
    • Export Citation
  • Grenade, K. and W. Moore, 2008. “Co-Movements Between Foreign and Domestic U.S. Interest Rates In a Fixed Exchange Rate Regime: The Case Of The ECCU and U.S.,” Applied Econometrics and International Development, Euro-American Association of Economic Development, Vol. 8(1), pp. 11930.

    • Search Google Scholar
    • Export Citation
  • Hernandez, J., 2004, “Business Cycles in Mexico and the United States: Do They Share Common MovementsJournal of Applied Economics, Vol. VII, No. 2, pp. 30322.

    • Search Google Scholar
    • Export Citation
  • Kandil, M., 2009, “Does Demand Volatility Lower Growth and Raise Inflation? Evidence from the Caribbean,” forthcoming IMF Working Paper.

    • Search Google Scholar
    • Export Citation
  • Kluyev, V., 2008, “Real Implications of Financial Linkage between Canada and the United States,” IMF Working Paper, WP/08/23.

  • Mamingi, N., 1999, “Testing for Convergence and Common Features in International Output: The Case of the Eastern Caribbean Countries,” Journal of Eastern Caribbean Studies, Vol. 24, pp. 1540.

    • Search Google Scholar
    • Export Citation
  • Murray, A., 2007, “Modeling the Jamaican Business Cycle: A Structural Vector Autoregression Approach.” (Unpublished paper presented at the 38th Conference of the Caribbean Centre for Money and Finance).

    • Search Google Scholar
    • Export Citation
  • Mishra, P., 2006Emigration and Brain-Drain: Evidence from the Caribbean,” IMF Working Paper, WP/06/25.

  • Rappoport, H., and F. Docquier, 2004, “The Economics of Migrant Remittances,” Stanford Center for International Development Working Paper, No. 236.

    • Search Google Scholar
    • Export Citation
  • Roache, S., 2008, “Central America’s Regional Trends and U.S. CyclesIMF Working Paper, WP/08/50.

  • Swinston, A., and T. Bayoumi, 2008, “Spillovers Across NAFTA,” IMF Working Paper, WP/08/3.

  • Vahid, F. and R. F. Engle, 1993, “Common Trends and Common Cycles”, Journal of Applied Econometrics, Vol. 8, No. 4, pp. 34160.

1

We would like to thank Shaun Roache for generously sharing his program code and insightful discussions. We are also grateful to Paul Cashin, Sam Ouliaris, and seminar participants at the 2008 Eastern Caribbean Central Bank/IMF Research Workshop and the Caribbean Centre for Money and Finance 2008 Annual Conference for comments and suggestions. All errors are authors’ responsibilities.

2

See for example Mamingi (1999) and Craigwell and Maurirn (2000). Cashin (2006) notes that such transformations yield distorted estimates of the growth cycle.

3

Grenade and Moore (2008) show that there is long-run convergence between interest rates in the U.S. and the ECCU, which is consistent with interest rate parity with the U.S. In the short-run, changes in the Fed Funds rate have an almost immediate effect on interest rates in the ECCU.

4

This approach implicitly assumes that additional variables are independent to those endogenous variables in the VAR.

5

In a few cases where serial autocorrelation is detected using standard tests, lagged dependent variables are added to the basic estimation equations.

ECCU Business Cycles: Impact of the U.S.
Author: Ms. Yan M Sun and Mr. Wendell A. Samuel