This Selected Issues paper analyzes the income dispersion and comovement in the Eastern Caribbean Currency Union region. It finds that incomes are diverging, with the Leeward Islands converging to a higher income level than the Windward Islands. The paper examines the macroeconomic impact of trade preference erosion on the Windward Islands and demonstrates the substantial impact from preference erosion on growth, trade balances, and fiscal positions. The paper also analyzes the size of the informal economy in the Caribbean.

Abstract

This Selected Issues paper analyzes the income dispersion and comovement in the Eastern Caribbean Currency Union region. It finds that incomes are diverging, with the Leeward Islands converging to a higher income level than the Windward Islands. The paper examines the macroeconomic impact of trade preference erosion on the Windward Islands and demonstrates the substantial impact from preference erosion on growth, trade balances, and fiscal positions. The paper also analyzes the size of the informal economy in the Caribbean.

V. Domestic Investment and the Cost of Capital66

A. Introduction

115. Despite relatively high levels of investment, economic growth has been disappointingly low in many Caribbean countries (Figure V.1).67 There are many possible explanations, including that the linkages between foreign direct investment (including in the tourism sector) and the rest of the economy are weak, or that the return on public investment is less than on private investment (Khan and Kumar, 1997; Bouton and Sumlinski, 2000). Alternatively, there may be crowding out of domestic private investors.

Figure V.1:
Figure V.1:

GDP Growth and Gross Investment-to-GDP, 1995–2004 1/

Citation: IMF Staff Country Reports 2007, 097; 10.5089/9781451811704.002.A005

Source: IMF, International Financial Statistics.1/ Unweighted arithmetic average of constant price GDP annual growth rates.

116. While the role of foreign direct investment has been much debated in the region, private domestic investment could prove to be one important factor to unlock growth. If private domestic investment (PDI) were to increase, then the returns to capital spending by public agencies may also increase.68 PDI could also improve the linkages between the domestic economy and foreign-owned capital. If PDI could potentially play an important role in driving growth, what factors might cause it to increase?

117. Against this background, this chapter assesses the factors that have affected private domestic investment in the Caribbean. The main focus is on the cost of capital, which includes the cost of debt, equity, and the impact of taxes. The analysis suggests that private domestic investment is sensitive to the cost of capital in the region, yet is unresponsive to public investment or foreign direct investment (FDI). The results suggest that public policy should focus on the cost of capital (CoC) as a means of raising PDI rather than tax concessions that can be opaque, highly discretionary, and penalize domestic investors (Chai and Goyal, 2006; Sosa, 2006).69

118. The structure of this chapter is as follows. Section B presents the investment model and data. Estimation issues and empirical results are then discussed in the following sections, and Section E contains some concluding comments.

B. The Model and Data

119. The basic model derives from the Hall-Jorgenson approach in which firms invest in order to maximize the present value of all future cash flows. In this simple model, there are no capital stock adjustment costs, and the optimal capital stock (k, in logs) is defined as a simple function of the firm’s output (y, in logs), the post-tax cost of capital (ρ) and the elasticity of capital to labor as σ (also defined as the user cost elasticity of capital):

kt=a+ytσρt.(1)
A05lev2sec2

Box V.1 provides a brief overview of the data construction, with the key challenge being the creation of capital stock and cost of capital series. To measure the cost of capital, the OECD STAN database was used to construct investment deflators and depreciation rates for the capital used by sectors across countries, which were then weighted by the sector’s real value added in each country to obtain country-specific aggregates. The capital stock was estimated using the standard accumulation identity and the method first developed by Nehru and Dhareshwar (1993), taking into account the effect of natural disasters.70

120. Table V.1 illustrates that investment-to-GDP ratios across the ECCU region have been relatively high, particularly in the smaller ECCU countries. This reflects the strength of foreign direct investment, due in large part to global growth in the tourism sector, which has offset stagnant or declining PDI as a share of GDP. Figure V.2 also illustrates that the private investment rate in the region has been exceptionally volatile, partly reflecting the effect of natural disasters, but also that the rate has been stagnant or declining in many countries over the past decade.

Table V.1.

CARICOM: Gross Investment, 1985–2003 1/2/

(Percent of GDP)

article image
Sources: National authorities; IMF; World Bank; and author’s calculations.

Unweighted averages of the current price investment ratios for the countries in each group.

Private domestic investment (PDI) also includes investment by public enterprises.

Data end at 2003 as this is the last year in which data was available for all countries.

Figure V.2.
Figure V.2.

Private Investment Rate (I/K) and Economic Depreciation (Dep.)

(Local currency 2000 prices, percent per annum) 1/

Citation: IMF Staff Country Reports 2007, 097; 10.5089/9781451811704.002.A005

Sources: National authorities; Organization for Economic Cooperation and Development; and author’s calculations.1/ I and K deflated by the local currency capital goods deflator. Excludes effects of natural disasters.

Data Used in the Regression Analysis

The regression analysis estimated the impact of various factors on the level of the real private domestic capital stock. Theory suggests that the optimal level of capital should be fully determined by just two variables: real output and the real cost of capital. Other variables were included as controls. See Roache (2006) for more details.

Private capital stock: estimated using techniques described in Nehru and Dhareshwar (1993). This is based on the perpetual inventory method, with capital stock in any year equal to the amount of undepreciated capital in the previous year, plus new investment. In almost all cases, only nominal investment is provided by the authorities. The OECD STAN database was used to estimate an investment goods deflator for each country. This was also used, together with the EM-DAT natural disasters database, to estimate real depreciation.

Cost of capital: estimated using the cost of debt, equity, and the burden of taxation. The cost of debt was based on commercial bank lending rates as published in International Financial Statistics. The cost of equity was estimated from a capital asset pricing model, largely based on the required risk premium for the tourism sector. The weights of debt and equity were estimated by calculating the proportion of the capital stock that was financed by outstanding credit to the domestic private sector. The tax factor represents the overall corporate tax burden on the marginal unit of investment, including the present value of future allowance for depreciation. Tax data was sourced from international tax surveys conducted by global accounting firms, and from the International Bureau of Fiscal Documentation. In all cases, the published statutory rates of taxation were used in the calculation of effective marginal rates.

Other variables: Output, public investment, foreign direct investment (FDI), the real effective exchange rate, external debt-to-GDP ratio, and credit to the private sector: sourced from the authorities and IMF staff reports, where appropriate. The public investment and FDI deflators were estimated using the OECD STAN database.

Correlation statistics are presented in Table V.2 below. All variables are given in log first differences multiplied by 100 (as used in the regression) and, with the exception of the real effective exchange rate and the external debt-to-GDP ratio, are deflated.

121. The financing cost of capital appears to have been rising in many countries, reflecting debt, equity, depreciation, and taxation (Figure V.3). In the absence of a deep and liquid corporate debt and stock market, estimates for the cost of debt and equity from standard approaches, such as the capital asset pricing model (CAPM), are difficult to obtain. There is also no consensus on what assets should be used when estimating the cost of equity. To overcome the lack of financial markets, weighted average country CAPM betas were constructed using global asset prices (see Roache, 2006, for details).71

Figure V.3.
Figure V.3.

Financing Cost of Capital, 1985–2004 1/2/

(Percent per annum)

Citation: IMF Staff Country Reports 2007, 097; 10.5089/9781451811704.002.A005

Sources: National authorities; IMF; Organization for Economic Co-operation and Development; Datastream; Price Waterhouse; and author’s calculations1/ Includes the cost of debt, equity, and depreciation. Trinidad and Tobago is presented in place of Jamaica as extreme country-specific volatility in Jamaica makes it difficult to compare to regional peers.2/ The tax factor represents the overall corporate tax burden on the marginal unit of investment including the present value of future allowances for depreciation.

122. The cost of capital has been rising across much of the region in recent years. The increase in the depreciation rate (largely related to an increasing weight for technology equipment) has not been offset by a reduction in interest rates or the cost of equity. The average financing cost of capital for the sample countries in 2004 was 15.2 percent. Using the same methods, the approximate cost of capital for U.S. corporations in 2004 was 8.2 percent.72

123. To construct CAPM betas for each country, the share in GDP of what might be regarded as global industries, such as tourism, mining, and energy, were used to weight the betas from the global sector equity index on the world equity market. The remainder was accounted for by the beta of the global emerging markets index. This downplays the role of country-specific factors in determining the risk premium. This approach reflects the opportunity cost of equity capital when investors face an open capital account and are considering projects in certain industries.73 Investment opportunities in these small countries are often related to expectations regarding a key sector (e.g., tourism or energy), and this is likely to outweigh country-specific issues in many cases.

124. One interesting development in the regional beta has been the increasing weight of tourism and the effect this has had on risk. After the crises during 1997–99 in several emerging markets, the systemic risk has subsequently declined. However, this has not fully fed through to all areas of the Caribbean. One reason is that the beta of the global tourism sector rose sharply after the terrorist attacks of September 11, 2001. Equity assets in this sector are now performing as highly leveraged plays on the overall stock market cycle. 74

125. Compared with industrial country standards, the estimated level of equity funding in the Caribbean is high.75 Figure V.4 shows the private sector business debt-to-capital ratio for each country. The highest is in St. Kitts and Nevis (likely influenced by the sugar company, a public enterprise) and the lowest is Jamaica, which has suffered from long periods of intense real interest rate volatility. The average across the sample countries is 38 percent. This compares with an 80 percent ratio among the companies in the U.S. S&P 500.

Figure V.4.
Figure V.4.

Debt-to-Capital Ratios

(In percent)

Citation: IMF Staff Country Reports 2007, 097; 10.5089/9781451811704.002.A005

Sources: National authorities; ECCB; OECD; and author’s calculations.

126. Figure V.5 shows how the overall CoC using this method compares across countries. The figures for each country have been weighted by the average source of finance. For example, the weighted cost of debt in Jamaica is very low because almost all PDI is estimated to be financed from equity (e.g., retained earnings).

Figure V.5.
Figure V.5.

Components of the Weighted Average Real Cost of Capital

(Average percent cost for 2000–04) 1/

Citation: IMF Staff Country Reports 2007, 097; 10.5089/9781451811704.002.A005

Sources: National authorities and Fund staff estimates.1/ Contributions weighted by the proportion of capital financed from each source, with the percentage financed by debt being: Antigua and Barbuda 25%; St Kitts and Nevis 76%; St. Lucia 56%; St. Vincent and the Grenadines 42%; Barbados 21%; Jamaica 11%; and Trinidad and Tobago 25%. The three—year average cost of debt for Jamaica is used due to significant volatility. Depreciation and capital prices are excluded.

C. Estimation Issues

127. An unbalanced panel of seven countries using annual data with a maximum span of 1980 to 2004 was used for the estimation of equation (1). For all specifications, two samples were used: (i) the full sample of seven Caribbean countries; and (ii) the sample excluding Trinidad and Tobago, because of the very heavy weight of the energy sector in Trinidad’s economy.76 77

128. The strategy is to estimate equation (1) as a structural model, relating the long-run level of capital to output and the cost of capital. Two drawbacks of this approach are that it does not take into account shorter-run adjustments, and the regressors (output and the cost of capital) may be endogenous. To address the latter issue, an instrumental variables approach is taken.

129. Equation (1) was estimated in first differences using the generalized method of moments (GMM). Lagged values of the endogenous variables—such as the capital stock, output, and the cost of capital—were used as instruments.78 The GMM weighting matrix was chosen to account for cross-equation correlation.

D. Empirical Results

130. Equation (1) was initially estimated in first differences including only output and the cost of capital as regressors (Table V.3, columns 1 and 2). While consistent with the theory, the null that the instruments were valid was rejected for most sets of instruments chosen. When the set of regressors was expanded to include the change in the level of external public debt to GDP and the real exchange rate (both considered as endogenous and instrumented for using lagged values), the model’s performance improved significantly (Table V.3, columns 3–10).

Table V.2.

Correlation Matrix

(First Difference)

article image
Source: Author’s calculations.Note: See Box V.1 for variable source and definitions.
Table V.3.

GMM Regression Results: Dependent Variable is Private Domestic Capital Stock 1/

article image
Source: Author’s calculations.

* and ** denote significance at the 10 percent and 5 percent levels respectively, using standard errors that are robust to serial correlation. The Generalized Method Movements (GMM) weighting matrix accounted for cross-sectional heteroscedasticity. Estimates include dummy variables for the year of, and the year following, a natural disaster (taken from the EM-DAT database of CRED, 2006).

The Sargan test p-value shows the probability that the instruments are valid; i.e. uncorrelated with the residuals.

Full denotes all sample countries, ex-TTO denotes excluding Trinidad and Tobago.

131. The estimates showed that the user cost elasticity (UCE), which represents the sensitivity of private capital to the cost of capital, was negative and significant. The GMM estimated UCE for the full sample is shown in Table V.3 (columns 3–10). The UCE was estimated to be around -0.12, compared to the most recent estimates in the literature, which range from -0.18 to -0.03, and was stable across alternative specifications of the equations. Moreover, the model using the cost of capital also outperformed those using real interest rates (Roache, 2006).

132. The effects of output, public investment and foreign direct investment were insignificant. This suggests there is little evidence that a rise in aggregate demand brings forth additional private sector capital. It also points to weak linkages between the domestic private sector and the public and foreign-owned capital stock (Table V.3, columns 5–8). Multicollinearity may, to some extent, be dampening coefficient estimates for the latter two factors, particularly for public investment. As Table V.2. shows, public investment is correlated with output and the external debt ratio.

133. The coefficient on the change in the external public debt-to-GDP ratio was negative but statistically insignificant (Table V.3, columns 3–10). This provides some weak evidence of a public debt overhang, with large changes in the external debt ratio depressing the growth in the domestic capital stock. Caution is warranted in interpreting these results, particularly given the role of natural disasters. For example, when an island is hit by a hurricane it is likely that the capital stock will decline through damage. Meanwhile, the debt ratio will increase as the authorities attempt to smooth consumption. This is partly controlled for by dummy variables, the removal of extreme outliers, and the instrumenting of the debt ratio change using lagged changes.

134. There is little evidence of credit rationing. Following Lederman and others (2003), real credit to the private sector was included in a number of specifications; under credit rationing, the coefficient on this variable should be positive and significant, while that on the cost of capital should remain negative. The coefficients were actually negative and insignificant and there was no meaningful impact on the estimates of the user cost elasticity (Table V.3, columns 9–10).

E. Conclusions

135. This chapter has three main conclusions:

  • Private domestic investment is sensitive to the user cost of capital. Local real interest rates, the corporate tax burden, and global financial market conditions all have a major influence on the private domestic capital stock.

  • Real output does not affect private domestic investment (PDI). In addition, the multiplier effects of public investment and FDI for private domestic investment are weak.

  • Public policy designed to raise private domestic investment levels should focus on creating the conditions for a lower cost of capital. Policy should focus less on relying on public investment or FDI to stimulate domestic private sector growth.

136. These results suggest that policies to reduce the cost of capital could help raise private domestic investment across CARICOM countries. Currently, policy in some parts of the region may be based too heavily on the presumption that only public investment and the attraction of more FDI, through costly tax concessions, can jumpstart economic growth. The evidence suggests that public investment and FDI have had little impact on stimulating private domestic investment.

137. Public policy is likely to be more effective in reducing the cost of capital by creating conditions that allow the pre-tax financing cost to decline. Reducing the role for the public sector might ease crowding out effects, particularly in the domestic banking system. Structural financial sector reform measures might also be considered, such as easing foreclosure laws. This would reduce risks for financial institutions and allow banks to provide credit at lower rates to domestic businesses.

Appendix I: Model of Aggregate Investment

Despite a large literature, there is no consensus on the appropriate form of an empirical investment equation, particularly for developing economies. Much of the focus in this chapter is on the cost of capital and the Jorgenson approach allows explicit derivation of a cost term. We start with the representative firm’s problem, which is to maximize a value function:

Vt=Et{Σs=t(1+r)ts[(1u)psF(Ksp,KsG,Ns)wsNs(1Γ)(1τ)pskIsp]}(A1)
A05app01

This is subject to the usual capital accumulation constraint. It is assumed that private investment (I P) only becomes useful private capital (KP) with a one-period lag:

Ks+1p=Isp+(1δ)Ksp(A2)
A05app01

It is assumed that the firm’s production function includes both public (KG) and foreign (KF) capital as arguments. This model is being applied here to small, open economies. Consequently, the effect of personal taxation of dividend flows and capital gains is ignored since for a small open economy, the savings and investment decisions may be assumed to be separable, implying that personal taxation affects the incentive to save, but has no relevance for the investment decision. Across most of the region, the tax authorities apply the capital allowance to a capital base that is eroded exponentially. Then the present value of tax benefits may be written as:

Γ=η+uα˜+(1θηωα˜)Σv=t+1(1+r+α)tvuα(A3)
A05app01

There are some examples, such as Barbados, where the tax authorities apply the allowance to a capital base that is subject to straight-line erosion. Present value of deductions may then be written as:

Γ=η+uα˜+(1θηωα˜)Σv=t+1T(1+r)tvuT1(A4)
A05app01

where in the above equations: r ≡ nominal weighted average financing cost of capital (debt and equity); u ≡ corporate tax rate; p ≡ output price; pK ≡ capital goods price; w ≡ nominal wage; N ≡ labor; Γ ≡ present value of tax benefits for investment; η ≡ investment tax credit rate, in the form of a rebate/direct grant; θ ≡ proportion of tax credit that reduces the tax-depreciable capital base; α ≡ immediate capital allowance rate; ω ≡ proportion of the immediate capital allowance that reduces tax-depreciable capital; τ ≡ tariff on imported capital equipment; δ ≡ economic (real) depreciation rate; a ≡ tax depreciation/ capital allowance rate; and T ≡ assumed useful life if capital equipment for tax deduction purposes.

It is assumed that the firm’s production function is linearly homogeneous and that it is using the optimal labor input and faces competitive output markets. Firms will choose capital to maximize this value. Rearranging the first-order condition and denoting the real percent change in the price of capital as q and the inflation effect p, this can be written as:

Et{Pt+1}.Fk=(1Γ)(1+ι)(1u)[PtK(r+δ(1+Etq)(1+Etπ))(1+Etπ)PtK(1+Etq)+Ptk].
A05app01

It is common in the literature to let the expected real change in the capital goods price be equal to zero. This implies that the first-order condition can be expressed as the marginal product of capital equal to the post-tax user cost of capital, denoted by P:

Fk=ptkpt(1+Etπ)(1Γ)(1+τ)1u(rEtπ+δ(1+Etπ))=pt(A5)
A05app01

The final stage is to assume a CES form for the production function so that we may solve for K. This produces the basic equation used as the foundation of the empirical work that follows. Denote the log of the post-tax cost of capital as ρ and the user cost elasticity of capital (UCE) as σ. Then the first-order condition can be expressed as:

kt=a+ytσρt.
A05app01

References

  • Bouton, L., and M. Sumlinski, 2000, “Trends in Private Investment in Developing Countries, 1970–98,” IFC Discussion Paper, Number 41 (Washington DC: International Finance Corporation).

    • Search Google Scholar
    • Export Citation
  • CRED, 2006, “EM-DAT: The OFA/CRED International Disaster Database 1990–2005,” http://www.cred.be/emdat, Centre for Research on the Epidemiology of Disasters.

    • Search Google Scholar
    • Export Citation
  • Chai, J., and R. Goyal, 2006, “Tax Concessions and Foreign Direct Investment in the Eastern Caribbean Currency Union,” in R. Sahay, D.O. Robinson and P. Cashin (eds), The Caribbean: From Vulnerability to Sustained Growth (Washington DC: International Monetary Fund), pp. 25884.

    • Search Google Scholar
    • Export Citation
  • Everhart, S., and M. Sumlinski, 2001, “Trends in Private Investment in Developing Countries 1970–2000,” IFC Annual Report, Number 44 (Washington DC: International Finance Corporation).

    • Search Google Scholar
    • Export Citation
  • Khan, M., and M. Kumar, 1997, “Public and Private Investment and The Growth Process in Developing Countries”, Oxford Bulletin of Economics and Statistics, Vol. 59, pp. 6988.

    • Search Google Scholar
    • Export Citation
  • Lederman, D., A. Menendez, G. Perry, and J. Stiglitz, 2003, “Mexican Investment after the Tequila Crisis: Basic Economics, ‘Confidence’ Effects, or Market Imperfections?,” Journal of International Money and Finance, Vol. 22, pp. 13151.

    • Search Google Scholar
    • Export Citation
  • Nehru, V., and A. Dhareshwar, 1993, “A New Database on Physical Capital Stock: Sources, Methodology and Results,” Revista de Analisis Economico, Vol. 8, pp. 3759.

    • Search Google Scholar
    • Export Citation
  • Rasmussen, T., 2006, “Natural Disasters and Their Macroeconomic Implications”, in R. Sahay, D.O. Robinson and P. Cashin (eds), The Caribbean: From Vulnerability to Sustained Growth (Washington DC: International Monetary Fund), pp. 181205.

    • Search Google Scholar
    • Export Citation
  • Roache, S., 2006, “Domestic Investment and the Cost of Capital in the Caribbean,” IMF Working Paper No. 06/152, (Washington DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Sosa, S., 2006, “Tax Incentives and Investment in the Eastern Caribbean,” IMF Working Paper No. 06/23, (Washington DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
66

Prepared by Shaun Roache.

67

CARICOM (Caribbean Community and Common Market) comprises 15 member states and five associate members. The subjects of this study are seven CARICOM member states: Antigua and Barbuda, Barbados, Jamaica, St. Lucia, St. Kitts and Nevis, St. Vincent and the Grenadines, and Trinidad and Tobago.

68

The interaction between private and public investment in developing countries is complex. The empirical literature has provided ambiguous results. For example, a survey by Everhart and Sumlinski (2001) shows that the literature is almost evenly split on whether public investment crowds in or crowds out the private sector.

69

The ECCU in this paper refers to the six independent countries that comprise the Eastern Caribbean Currency Union: Antigua and Barbuda, Dominica, Grenada, St. Kitts and Nevis, St. Lucia, and St. Vincent and the Grenadines.

70

The effect of natural disasters, an important factor for the region, was included using estimates from the EM-DAT database of CRED (2006). These are clearly subject to measurement error. However, account has to be taken of disasters as they are a factor that, in the ECCU region, have caused damage equivalent to around 2 percent of GDP annually in recent years (Rasmussen, 2006). Dummy variables for the year of the disaster, plus the following year, were also included, although for the most extreme outliers, the observations were removed.

71

The CAPM beta represents the undiversifiable systemic risk of the investing in the country. Given that the sample countries run open capital accounts, it is assumed that all idiosyncratic risk may be diversified away.

72

This used the S&P 500 debt-to-assets ratio, an average BAA corporate bond yield of 6.4 percent, the statutory 35 percent corporate tax rate, an approximate tax factor of 1.14, and an equity risk premium over the U.S. ten-year treasury note of 2.8 percent.

73

For example, Caribbean investors could replicate a global sector investment by purchasing an exchange-traded fund that is indexed to a global sector benchmark.

74

To assess whether the September 11 effects influenced the results, estimations included a common post-September 11 dummy variable.

75

In the case of debt financing, with most business sector borrowing in CARICOM taking the form of bank loans rather than tradable bonds, the nominal book value of these loans is taken as the level of debt funding. The share of equity is then the residual.

76

Over the sample period, it was estimated from national authority and IMF data that the value added of the energy sector averaged 28 percent.

77

Standard panel unit root and cointegration tests suggest that most sample variables are non-stationary and not cointegrated—the results are available from the author. The absence of cointegration could reflect the possibility that factors outside the neoclassical model also determine the long-run capital stock.

78

The cost of capital was treated as endogenous, since the large real interest rate component is likely to be contemporaneously correlated with output shocks. Although, for a small open economy, the interest rate is typically regarded as exogenous, interest rates in the region have shown a lack of sensitivity to the nominal anchor, the U.S. dollar, in recent years. Other instruments used included FDI, central government investment, and credit to the private sector.

Eastern Caribbean Currency Union: Selected Issues
Author: International Monetary Fund