The recent global financial crisis has drawn renewed attention to the effectiveness of fiscal policy, as many countries implemented fiscal stimulus measures to boost economic activity. The effectiveness of fiscal policy is often assessed by the size of fiscal multipliers, which measure a change in output caused by an exogenous change in government spending or tax revenue. This chapter estimates fiscal multipliers for the Caribbean using quarterly data for 14 Caribbean countries,1 and investigates key determinants of the size of the multipliers. The results show that fiscal multipliers in the sample countries are modest, and that the high levels of trade openness and public debt account for their modest size.
Fiscal Multipliers
Different multipliers are used in the literature, depending on the time frame considered. The most frequently used measure is the impact multiplier, which is defined as ΔYt/ΔGt, where ΔYt is a change in output and ΔGt is a change in government expenditure in period t. The impact multiplier measures the increase in output generated by an additional dollar in government spending in period t. Another frequently used notion is the cumulative multiplier, which is defined as
Since fiscal stimulus packages can only be implemented over time and there may be lags in the economy’s response, the cumulative multiplier may be more accurate in capturing the impact of fiscal policy. We can define tax-revenue multipliers in the same way.
Interpreting fiscal multipliers requires caution, because they are not deep structural parameters. Instead, they consist of policy reactions and structural parameters. That is, fiscal multipliers depend on various factors that can differ from case to case, such as the fiscal policy instrument, its duration, its associated fiscal adjustments, the stance of monetary policy, and country-specific circumstances. Multipliers are therefore best interpreted as empirical summaries of average output reactions following exogenous changes in government spending or tax revenue.
Theories of Fiscal Multipliers
Different theories provide different mechanisms for fiscal multipliers, but they all highlight the importance of hours worked in explaining the short-run multipliers.2 Since the capital stock cannot be adjusted instantaneously, only hours worked can increase total output in the short term. Therefore, the short-run multipliers are essentially accounted for by the response of equilibrium hours worked to a fiscal policy and the extent to which those hours translate to output.
The Keynesian tradition explains fiscal multipliers through demand-side effects. Assuming that an economy is constrained by demand, not by supply, Keynesian theory says that government spending increases demand, which raises incomes, which in turn leads to private sector spending. In the extreme case where wages and prices do not respond, the multiplier is given by 1/(1 − mpc) for spending and —mpc/(1 − mpc) for taxes, where mpc is the marginal propensity to consume. The multipliers become smaller according to the extent of price adjustment, and zero in another extreme case where wages and prices are infinitely responsive. The multipliers also depend on the monetary policy; they become larger when the central bank keeps the nominal interest rate constant. In a new Keynesian model calibrated to the U.S. economy, Christiano, Eichenbaum, and Rebelo (2011) show that when government spending rises for 12 quarters while the nominal interest rate is kept constant, the multiplier is about 1.6 on impact and attains a peak at around 2.3.
The neoclassical tradition provides a mechanism for fiscal multipliers through supply-side effects. In neoclassical models, economic agents react to a fiscal policy if it changes the present value of income or intertemporal trade-offs. For example, a tax cut with a future tax increase in a nondistortionary way does not affect the present value of income, so its multiplier is zero (Ricardian Equivalence). On the other hand, a fiscal policy that involves distortionary adjustments can generate either positive or negative effects on output. For example, when government spending temporarily rises but a current distortionary labor tax also increases to keep the budget balanced, the impact multiplier can be negative (Baxter and King, 1993). If an increase in government spending is financed with a deficit, the impact multiplier is larger because of intertemporal substitution effects: economic agents work more today because they expect a higher tax in the future.
Empirical Studies on Fiscal Multipliers
The size of the fiscal multiplier varies with countries, time periods, and circumstances. Previous studies have suggested a number of factors that influence the size, including trade openness, public debt level, exchange rate regime, and state of the economy.3
-
Trade openness: Fiscal multipliers are smaller for countries with a larger propensity to import (IMF, 2008; Ilzetzki, Mendoza, and Végh, 2009; Barrel, Holland, and Hurst, 2012), because additional demand created by the fiscal policy will “leak” through imports.
-
Public debt level: Fiscal multipliers are smaller for countries with higher levels of public debt (Ilzetzki, Mendoza, and Végh, 2009; and Kirchner, Cimadomo, and Hauptmeier, 2010), because expansionary fiscal policies in countries with large public debt imply a risk to fiscal sustainability and the need for fiscal tightening in the near future.
-
Exchange rate regime: Fiscal multipliers are larger for countries with a fixed exchange rate regime (Ilzetzki, Mendoza, and Végh, 2009). This is because the “leakage” through the currency appreciation caused by a fiscal expansion would be minimal in a country with a fixed exchange rate regime.
-
State of the economy: Fiscal multipliers are larger during economic recessions, when there is substantial slack in the economy (Auerbach and Gorod-nichenko, 2012a, 2012b; Baum, Poplawski-Ribeiro, and Weber, 2012; Blanchard and Leigh, 2013; IMF, 2012). Larger multipliers reflect larger shares of liquidity-constrained (hand-to-mouth) households and firms (Gali, López-Salido, and Vallés, 2007; Parker, 2011).
The range of estimated fiscal multipliers varies considerably. Baunsgaard and others (forthcoming) review a large number of empirical studies and report that the spending multipliers in the first year range between 0.3 and 2.1 for the United States and between 0.3 and 1.8 for the European countries. They also report a range for the tax multiplier, which is −0.7 to −1.4 for the United States and -0.5 to 0.7 for Europe. Kraay (2012) finds that the first-year spending multiplier for developing countries is around 0.4. Ilzetzki, Mendoza, and Végh (2009) find that economies with relatively low degrees of openness to trade (measured as exports plus imports as a proportion of GDP) have a cumulative multiplier of around 1.6 after 24 quarters, but relatively open economies have an almost zero multiplier. They further find that the cumulative multiplier is about 1.5 after 24 quarters for economies under fixed exchange-rate regimes, but it is essentially zero in economies under flexible exchange regimes.
Fiscal multipliers in the Caribbean are low, according to the existing empirical studies. Guy and Belgrave (2012) find that the cumulative multipliers are less than 0.3 after 24 quarters in their sample countries over the period 1980–2008. They find that the first-year spending multiplier is 0.14 for Barbados, 0.11 for Jamaica, 0.18 for Trinidad and Tobago, and negative for Guyana. The cumulative spending multiplier after 24 quarters is at 0.2 for Barbados, 0.3 for Jamaica, and negative for Guyana and Trinidad and Tobago. Guy and Belgrave (2012) argue that although public expenditures are initially increased in a downturn to stimulate productive sectors, they are typically not sustained due to the constraints of declining revenues and high debt ratios.
Gonzalez-Garcia, Lemus, and Mrkaic (2013) estimate fiscal multipliers in the Eastern Caribbean Currency Union (ECCU) and find that the multipliers of consumption and investment spending range from 0.1 to 0.3 on impact and from 0.2 to 0.62 after 24 quarters; and that only the long-run investment multiplier is statistically significantly different from zero. These modest Keynesian effects in the Caribbean countries could be due to a combination of factors discussed in the literature, such as the degree of openness4 and high debt levels.
Methodology and Data
We estimate fiscal multipliers using structural vector autoregression (SVAR) models and quarterly data for Caribbean countries.5 Specifically, we consider multipliers of spending and tax revenue of central governments, as well as those of spending components: consumption spending and capital spending. The impact and cumulative multipliers are constructed using the dynamic response of GDP to shocks in fiscal variables, which are the estimated SVAR impulse response functions. Appendix 8.1 provides data sources and definitions of variables, such as “government spending,” and Appendix 8.2 describes our empirical model.
Our analysis focuses on fiscal multipliers for a group of countries instead of those for individual countries. Although fiscal multipliers in individual countries could be heterogeneous and informative, obtaining reliable estimates for them is difficult for many Caribbean countries given the limited number of observations available. In addition, Ilzetzki, Mendoza, and Végh (2009) and Gonzalez-Garcia, Lemus, and Mrkaic (2013) illustrate that panel VAR techniques are useful in providing characteristics of “average” multipliers for a group of economies and for analyzing the determinants of the size of fiscal multipliers by comparing those in different groups.
Estimated Fiscal Multipliers
Effects of an Expenditure Shock
Fiscal stimulus policies boost real GDP, but their effects tend to be short-lived in the Caribbean. Figures 8.1 and 8.2 show the estimated impulse response function (solid line) and its 95 percent confidence interval (dotted lines). The results, based on the 14 Caribbean countries studied, indicate that real GDP increases for about a year after a positive shock in government spending. Specifically, a one-standard-deviation shock in government spending increases real GDP by 20 basis points per quarter for about a year, but the effect tapers off afterwards.
Response of GDP to a Spending Shock, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the impulse response functions (solid line).Response of Spending to a Spending Shock, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the impulse response functions (solid line).The spending multiplier in the Caribbean is small on impact and modest in the long run.6 The impact multiplier of government spending is 0.13, which means an additional dollar of government consumption generates only 13 cents of additional output in the quarter when it is implemented. The cumulative multiplier grows modestly to its peak at 0.6 in the seventh quarter and then converges to its long-run value of 0.53, which can be considered a moderate size7 (see Figure 8.3). The cumulative multiplier converges to the long-run value in two years because the output response to a spending shock becomes almost zero after two years, as we can see in the “response” charts (Figures 8.1 and 8.2).
Cumulative Spending Multiplier on GDP, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the cumulative multiplier (solid line).“Leakages” through imports are important for understanding the modest level of the spending multiplier in the region. The trade openness there, defined as exports and imports in percent of GDP, exceeds 100 percent on average during 1990–2012. In fact, the Caribbean average of trade openness is the second highest among all regions during this period (Figure 8.4). As discussed in Endegnanew, Amo-Yartey, and Turner-Jones (2012), the proportion of imports in domestic consumption is high in small states, including most Caribbean and Pacific island countries, due to their small domestic market and the tendency toward a high degree of specialization in production. The effects of a fiscal policy would be limited if the policy led to importing additional goods and services from abroad instead of stimulating domestic consumption and investment. Ilzetzki, Mendoza, and Végh (2009) empirically show that trade openness is one of the critical determinants of the size of fiscal multipliers.
Openness to Tradeby Global Region, 1990–2012
(Average, percent of GDP)
Sources: World Economic Outlook database; and author’s calculations.Note: Openness to trade is defined as exports plus imports in percent of GDP. Each bar indicates the average in each region during 1990–2012.By contrast, the spending multiplier on imports in the Caribbean is actually sizable and significant. It is defined as the cumulative change in imports over the cumulative change in government spending at some horizon, and in our sample it is estimated for countries where quarterly import data are also available.8 The spending multiplier on imports is 0.4 on impact, increases at its peak to 0.97 in the seventh quarter, and converges to its long-run value of 0.83 (Figure 8.5). That is, the “leakage” of an additional one dollar in fiscal spending is 4 cents in the first quarter and 83 cents in the long run. This result supports the previous studies that show lower multipliers in economies that are open to trade.
Cumulative Spending Multiplier on Imports, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the cumulative multiplier (solid line).The public debt level is another critical factor affecting the size of multipliers suggested in the literature and relevant for the Caribbean region. We divide the sample countries into two groups—a high-debt group and a low-debt group—and estimate and compare the multipliers across groups. The high (low) debt group includes countries whose average central government debt-to-GDP ratio during the sample period is above (below) 65 percent. Figure 8.6 identifies these two groups.
Central Government Debt-to-GDP Ratio by Country
(Percent)
Sources: National authorities; and author’s calculations.Note: Figure shows the average of the debt-to-GDP ratio during the sample period used for the estimation: Antigua and Barbuda, 2003–11; The Bahamas, 1997–2011; Barbados, 2003–06; Dominica, 2000–11; Grenada, 2003–11; St. Lucia, 2000–11; St. Vincent and the Grenadines, 2000–11; Suriname, 2002–08; Trinidad and Tobago, 2000–11. See also Appendix Table 8.1.We also confirm that the level of public debt is another critical factor explaining the relatively small spending multiplier in the region.9 Results show that the cumulative multipliers of government spending are lower for high-debt countries, which is consistent with the evidence in the literature (Figures 8.7 and 8.8). The cumulative multiplier for high-debt countries becomes essentially zero after two years, while the one for low-debt countries converges to 0.77 and is statistically significant. This result suggests that fiscal expansions may have weakened fiscal sustainability and decreased the confidence of economic agents when the public debt level is high. Consumers and investors might act in a precautionary fashion against a fiscal stimulus policy if they are concerned about debt sustainability.
Cumulative Spending Multiplier (High Debt), Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the cumulative multiplier (solid line).Cumulative Spending Multiplier (Low Debt), Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the cumulative multiplier (solid line).Effects of a Tax Cut
In estimating the tax multiplier, we need the estimates of the within-quarter elasticity of taxes with respect to output for all countries10 (see Appendix 8.2 for the technical explanation). In this analysis, as a bold reference, we take the tax elasticity estimate of 2.08 for the U.S. economy from Blanchard and Perotti (2002). That is, in our analysis here, we assume that tax revenue responds to unexpected output movements within the quarter at an elasticity of 2.08 (Figure 8.9). Considering that several countries in the Caribbean do not have broad-based consumption taxes and/or income taxes (dos Santos and Bain, 2004), one can argue that the within-quarter tax elasticity in the Caribbean could be much lower than that in the United States. Therefore, one can interpret our estimate of the tax multiplier as an upper bound for the actual tax multiplier in the Caribbean.
Response of GDP to a Negative Tax Shock, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the impulse response functions (solid line). The underlying elasticity of tax with respect to output is set at 2.08.A tax cut boosts real GDP immediately, but its effect is not persistent. Figure 8.10 shows the estimated impulse response functions to a negative shock in tax revenue (solid line) along with 95 percent confidence intervals (dotted lines). Real GDP rises by 80 basis points after a one-standard-deviation negative shock in tax revenue, but the effect dies out after two years.
Response of Tax Revenue to a Negative Tax Shock, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the impulse response functions (solid line). The underlying elasticity of tax with respect to output is set at 2.08.The estimated tax multiplier is also modest in the region (Figure 8.11). The impact tax multiplier of −0.51 indicates that a tax reduction of one dollar would deliver an additional output of 51 cents. This is slightly larger than the impact multiplier of government spending and statistically significant, but it is still less than one. In addition, as discussed above, if the true elasticity of taxes with respect to output in the Caribbean is smaller than the U.S. elasticity of 2.08, the implied impact multiplier of tax revenue is less than 0.51. The cumulative tax multiplier quickly converges to its long-run value at 0.62 as the output response to a tax reduction tapers off after the first quarter. The point estimate of the long-run tax multiplier is larger than the long-run spending multiplier of 0.53, but it is not significantly different from zero.
Cumulative Tax Multiplier on GDP, Caribbean Region Average
Source: Author’s calculations.Note: Dotted lines indicate the 95 percent confidence interval of the cumulative multiplier (solid line). The underlying elasticity of tax with respect to output is set at 2.08.Summary and Conclusion
This chapter addressed one of the important macroeconomic policy questions: To what extent do governments’ spending expansions or tax cuts stimulate economic activities in the Caribbean? We examined this question using panel structural vector autoregression (SVAR) methods and quarterly data on 14 Caribbean countries.
The estimated fiscal multipliers in the Caribbean are modest, consistent with evidence from existing empirical studies (Guy and Belgrave, 2012; and Gonzalez-Garcia, Lemus, and Mrkaic, 2013). The impact multiplier of government spending is 0.13 and the cumulative multiplier is 0.53 after 24 quarters. The tax multiplier is 0.51 on impact and 0.62 after 24 quarters.
The modest size of fiscal multipliers can be accounted for by the region’s high levels of trade openness and public debt. We find that the “leakages” through imports are actually positive and statistically significant. In addition, we find that the cumulative spending multiplier is essentially zero in the subgroup of high-public-debt countries, while it is 0.77 and statistically significant in the subgroup of low-public-debt countries.
One important caveat is that the estimated multipliers are not deep structural parameters, so their interpretation needs caution. As discussed in the opening section of this chapter, the empirical literature emphasizes that fiscal multipliers can differ with countries, time periods, and circumstances because they are made of response functions as well as deep parameters. The estimated multiplier is the average output response to an exogenous change in a fiscal policy during the sample period, and it is not necessarily the predicted effectiveness of a fiscal policy under consideration.
Appendix 8.1. Data Sources and Processing
Data are sourced from various publications from central banks, their web sites, and the IMF’s International Financial Statistics database. We construct a panel data set, which includes quarterly data of fiscal variables and GDP for 14 Caribbean countries. Appendix Table 8.1 reports the countries included in our analysis and the sample periods of main variables.
Fiscal variables used in the analysis are defined as follows. Tax revenue (Tax) is defined as the central government’s tax revenue net of taxes on international trade. We exclude taxes on international trade because our focus is on the effects of tax cuts on domestic activities. Government spending (Spen) is defined as the sum of consumption spending (Cons) and capital spending (Capx) by the central government. Cons is defined as current expenditure net of interest payments and transfers.
For countries without quarterly GDP data, it is imputed using the interpolation method of Chow and Lin (1971). The interpolation method uses variation of quarterly series, which are relevant to the economy, in imputing the quarterly variation of real GDP. Among available quarterly series, we select a set of variables used for the imputation for each country, considering its main economic activities and the performance of the regression model. The selected variables, which are reported in Appendix Table 8.2, typically include bank credit to the private sector, the number of stay-over tourists (for tourism countries), oil prices, and some indicators of U.S. economic activities. Nominal data, such as private credit and compensation to employees, are deflated by the consumer price index (CPI) of each economy.
Series used for the SVAR estimation are deflated, seasonally adjusted, and detrended. Nominal data are deflated using the CPI index. We take the natural logarithm of all variables, use the SEATS algorithm for seasonal adjustment, and detrend the series using the Hodrick-Prescott filter.
Summary of Quarterly Data Series Used for Analysis
Summary of Quarterly Data Series Used for Analysis
| The Bahamas | Barbados | Belize | Jamaica | Trinidad and Tobogo | |
|---|---|---|---|---|---|
| Real GDP | … | 1990:Q1–2011:Q4 | 1994:Q1–2011:Q4 | 1996:Q1–2011:Q4 | 2000:Q3–2011:Q3 |
| Tax revenue | 1998:Q1–2011:Q4 | 2003:Q2–2007:Q1 | 2005:Q2–2011:Q4 | 2003:Q2–2011:Q4 | 1991:Q1–2011:Q4 |
| Spending | 1999:Q1–2011:Q4 | 2003:Q2–2007:Q1 | 2005:Q2–2011:Q4 | 2003:Q2–2011:Q4 | 1991:Q1–2011:Q4 |
| Imports of goods and services | 2007:Q1–2011:Q4 | … | … | … | … |
| Policy interest rate | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| CPI | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| Real GDP | Suriname | Anguilla | Antigua and Barbuda | Dominica | Grenada |
| Tax revenue | 2002:Q1–2008:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 |
| Spending | … | 2008:Q1–2011:Q4 | 2004:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 |
| Imports of goods and services | … | … | 2005:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2006:Q1–2011:Q4 |
| Policy interest rate | 1990:Q4–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| CPI | 1990:Q1–2011:Q4 | 1998:Q1–2011:Q4 | 1998:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| Montserrat | St. Kitts and Nevis | St. Lucia | St. Vincent and the Grenadines | ||
| Real GDP | … | … | … | … | … |
| Tax revenue | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 | |
| Spending | 2010:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | |
| Imports of goods and services | … | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2001:Q1–2011:Q4 | |
| Policy interest rate | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | |
| CPI | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
Summary of Quarterly Data Series Used for Analysis
| The Bahamas | Barbados | Belize | Jamaica | Trinidad and Tobogo | |
|---|---|---|---|---|---|
| Real GDP | … | 1990:Q1–2011:Q4 | 1994:Q1–2011:Q4 | 1996:Q1–2011:Q4 | 2000:Q3–2011:Q3 |
| Tax revenue | 1998:Q1–2011:Q4 | 2003:Q2–2007:Q1 | 2005:Q2–2011:Q4 | 2003:Q2–2011:Q4 | 1991:Q1–2011:Q4 |
| Spending | 1999:Q1–2011:Q4 | 2003:Q2–2007:Q1 | 2005:Q2–2011:Q4 | 2003:Q2–2011:Q4 | 1991:Q1–2011:Q4 |
| Imports of goods and services | 2007:Q1–2011:Q4 | … | … | … | … |
| Policy interest rate | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| CPI | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| Real GDP | Suriname | Anguilla | Antigua and Barbuda | Dominica | Grenada |
| Tax revenue | 2002:Q1–2008:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 |
| Spending | … | 2008:Q1–2011:Q4 | 2004:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 |
| Imports of goods and services | … | … | 2005:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2006:Q1–2011:Q4 |
| Policy interest rate | 1990:Q4–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| CPI | 1990:Q1–2011:Q4 | 1998:Q1–2011:Q4 | 1998:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
| Montserrat | St. Kitts and Nevis | St. Lucia | St. Vincent and the Grenadines | ||
| Real GDP | … | … | … | … | … |
| Tax revenue | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2003:Q1–2011:Q4 | |
| Spending | 2010:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | |
| Imports of goods and services | … | 2000:Q1–2011:Q4 | 2000:Q1–2011:Q4 | 2001:Q1–2011:Q4 | |
| Policy interest rate | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | |
| REER | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | |
| CPI | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 | 1990:Q1–2011:Q4 |
Variables Used in the Interpolation of Quarterly Real GDP
Variables Used in the Interpolation of Quarterly Real GDP
| Variables | ||
|---|---|---|
| The Bahamas | Private credit; stay-over tourists; oil price; U.S. real GDP. | |
| Anguilla | Private credit; stay-over tourists; oil price; U.S. index of industrial production. | |
| Antigua and Barbuda | Private credit; stay-over tourists; oil price; U.S. index of industrial production (-1). | |
| Dominica | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| Grenada | Private credit; stay-over tourists; oil price; U.S. index of industrial production. | |
| Montserrat | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| St. Kitts and Nevis | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| St. Lucia | Private credit; stay-over tourists; Domestic exports; U.S. real compensation to employees (-1). | |
| Suriname | Private credit; oil price; gold price; U.S. real GDP. | |
| St. Vincent and the Grenadines | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
Variables Used in the Interpolation of Quarterly Real GDP
| Variables | ||
|---|---|---|
| The Bahamas | Private credit; stay-over tourists; oil price; U.S. real GDP. | |
| Anguilla | Private credit; stay-over tourists; oil price; U.S. index of industrial production. | |
| Antigua and Barbuda | Private credit; stay-over tourists; oil price; U.S. index of industrial production (-1). | |
| Dominica | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| Grenada | Private credit; stay-over tourists; oil price; U.S. index of industrial production. | |
| Montserrat | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| St. Kitts and Nevis | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
| St. Lucia | Private credit; stay-over tourists; Domestic exports; U.S. real compensation to employees (-1). | |
| Suriname | Private credit; oil price; gold price; U.S. real GDP. | |
| St. Vincent and the Grenadines | Private credit; stay-over tourists; U.S. real compensation to employees (-1); U.S. real GDP (-1). | |
Appendix 8.2. Structural Vector Autoregression Models and Estimation Methods
Our baseline SVAR model is the following system of equations, which is similar to the specification in Gonzalez-Garcia, Lemus, and Mrkaic (2013):
where Yit is a two-variable vector including a fiscal variable of interest (spending, tax revenue, consumption spending, or capital spending) and GDP of country i in quarter t; Zi,t is an N-dimensional vector of control variables, which are exogenous to the system; and εi,t is a vector of two exogenous shocks.11 In order to control for the effects from external shocks and monetary policies, Zi,t includes the real effective exchange rate, the U.S. GDP, and the interest rate of monetary policy. The matrix A allows for the possible simultaneous effects across the endogenous variables Yi,t. The matrices Ct-p and Dt-p capture the effects from the pth lag of the endogenous variables Yt-p and the exogenous variables Zt-p on the current endogenous variables Yt. The matrix B is diagonal, which contains standard deviations of exogenous shocks. We estimate the system (A1) by panel OLS regression with fixed effects.
The number of lags is set at four for both endogenous and exogenous variables in all estimations. We computed five criteria in determining how many lags to include in the equation: Akaike information criterion, Schwarz information criterion, modified sequential likelihood ratio criterion, final prediction error criterion, and Hannan-Quinn criterion. However, they were not very useful because the suggested number of lags differs across methods and often takes a maximum number of lags considered. Therefore, we use four lags in all analyses because the results do not differ substantially from those using other suggested numbers of lags12 and because it is often the choice of other related studies, including Ilzetzki, Mendoza, and Végh (2011) and Gonzalez-Garcia, Lemus, and Mrkaic (2013).
The SVAR model (1) needs additional assumptions in order to identify the effect of fiscal policies. The identification issue exists because fiscal policies and output could affect each other. We use the short-term restriction proposed by Blanchard and Perotti (2002): it takes more than one quarter for policymakers and the legislatures to learn about and to respond to a GDP shock. This implies that the contemporaneous effect of a GDP shock on government spending and tax revenue is only through automatic feedback, which is zero for government spending. For tax revenue, Blanchard and Perotti (2002) estimate within-quarter elasticity of taxes with respect to output using detailed data on taxes and tax bases from 1947:Q1 to 1997:Q4 in the United States. For the analysis in this chapter, we use their estimated elasticity in generating the baseline tax results. Therefore, the estimated tax multiplier can be overestimated because the true tax elasticity in the Caribbean may be smaller due to narrower tax bases, as is discussed in this chapter.
References
Auerbach, Alan J., and Yuriy Gorodnichenko, 2012a, “Fiscal Multipliers in Recession and Expansion,” Paper prepared for the NBER conference, “Fiscal Policy after the Financial Crisis,” Milan, December 2011.
Auerbach, Alan J., and Yuriy Gorodnichenko, 2012b, “Measuring the Output Responses to Fiscal Policy,” American Economic Journal, Vol. 4, No. 2, May.
Barrel, Ray, Dawn Holland, and Ian Hurst, 2012, “Fiscal Consolidation: Part 2. Fiscal Multipliers and Fiscal Consolidations,” OECD Economics Department Working Paper No. 933 (Paris: Organization for Economic Co-operation and Development, February).
Baum, Anja, Marcos Poplawski-Ribeiro, and Anke Weber, 2012, “Fiscal Multipliers and the State of the Economy,” IMF Working Paper 12/286 (Washington: International Monetary Fund).
Baunsgaard, T., A. Mineshima, M. Poplawski-Ribeiro, and A. Weber, forthcoming, “Fiscal Multipliers,” in Post-Crisis Fiscal Policy, ed. by C. Cottarelli, P. Gerson, and A. Senhadji (Boston: MIT Press).
Baxter, Marianne, and Robert G. King, 1993, “Fiscal Policy in General Equilibrium,” American Economic Review, Vol. 83, pp. 315–34.
Blanchard, Olivier, and Daniel Leigh, 2013, “Growth Forecast Errors and Fiscal Multipliers,” IMF Working Paper 13/1 (Washington: International Monetary Fund).
Blanchard, Olivier, and Daniel Leigh, and Roberto Perotti, 2002, “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output,” Quarterly Journal of Economics, Vol. 117, pp. 1329–68.
Chinn, Menzie David, 2013, “Fiscal Multipliers,” Working Paper No. 2015–02, La Follette School of Public Affairs.
Chow, Gregory, and An-Loh Lin, 1971, “Best Linear Unbiased Interpolation, Distribution, and Extrapolation of Time Series by Related Series,” Review of Economics and Statistics, Vol. 53, pp. 372–75.
Christiano, Lawrence J., Martin Eichenbaum, and Sergio Rebelo, 2011, “When Is the Government Spending Multiplier Large?” Journal of Political Economy, Vol. 119, pp. 78–121.
dos Santos, Paulo, and Laurel Bain, 2004, “Survey on the Caribbean Tax Systems,” Paper prepared for the 18th General Assembly and Technical Conference of the Caribbean Organization of Tax Administrators (COTA), Nassau, The Bahamas, July 2004.
Endegnanew, Yehenew, Charles Amo-Yartey, and Therese Turner-Jones, 2012, “Fiscal Policy and the Current Account: Are Microstates Different?” IMF Working Paper 12/51 (Washington: International Monetary Fund).
Galí, Jordi, J. David López-Salido, and Javier Vallés, 2007, “Understanding the Effects of Government Spending on Consumption,” Journal of the European Economic Association, Vol. 5, No. 1, pp. 227–70.
Gonzalez-Garcia, Jesus, Antonio Lemus, and Mico Mrkaic, 2013, “Fiscal Multipliers,” Chapter 8 in The Eastern Caribbean Economic and Currency Union: Macroeconomics and Financial Systems, ed. by A. Schipke, A. Cebotari, and N. Thacker (Washington: International Monetary Fund).
Guy, Kester, and Anton Belgrave, 2012, “Fiscal Multipliers in Microstates: Evidence from the Caribbean,” International Advances in Economic Research, Vol. 18, No. 1, pp. 74–86.
Ilzetzki, Ethan, Enrique G. Mendoza, and Carlos A. Végh, 2009, “How Big Are Fiscal Multipliers?” CEPR Policy Insight, No. 39.
Ilzetzki, Ethan, Enrique G. Mendoza, and Carlos A. Végh, 2011, “How Big (Small?) Are Fiscal Multipliers?” IMF Working Paper 11/52 (Washington: International Monetary Fund).
International Monetary Fund, 2008, “Fiscal Policy as a Countercyclical Tool,” Chapter 5 in World Economic Outlook, October (Washington: International Monetary Fund).
International Monetary Fund, 2012, “Are We Understanding Short-Term Fiscal Multipliers?,” Box 1.1., Chapter 1 in World Economic Outlook, October (Washington: International Monetary Fund).
Kirchner, Markus, Jacopo Cimadomo, and Sebastian Hauptmeier, 2010, “Transmission of Government Spending Shocks in the Euro Area: Time Variation and Driving Forces,” ECB Working Paper No. 1219 (Frankfurt am Main: European Central Bank).
Kraay, Aart, 2012 “Government Spending Multipliers in Developing Countries: Evidence from Lending by Official Creditors,” World Bank Policy Research Working Paper No. 6099 (Washington: World Bank).
Parker, Jonathan A., 2011, “On Measuring the Effects of Fiscal Policy in Recessions,” NBER Working Paper 17240 (Cambridge, Massachusetts: National Bureau of Economic Research).
Perotti, Roberto, 2005, “Estimating the Effects of Fiscal Policy in OECD Countries,” CEPR Discussion Paper No. 4842, January (London: Centre for Economic Policy Research).
Ramey, Valerie A., 2011, “Can Government Purchases Stimulate the Economy?” Journal of Economic Literature, Vol. 49, No. 3, pp. 673–85.
Spilimbergo, Antonio, Steve Symansky, and Martin Schindler, 2009, “Fiscal Multipliers,” IMF Staff Position Note No. 09/11 (Washington: International Monetary Fund).
Appendix 8.1 shows the sample countries and periods included in the analysis.
Ramey (2011) and Chinn (2013) provide comprehensive surveys of theoretical work.
Baunsgaard and others (forthcoming) provides a comprehensive literature review on key determinants of the size of the fiscal multipliers.
Endegnanew, Amo-Yartey, and Turner-Jones (2012) show that fiscal policies affect the current account in microstates; a strengthening of the fiscal balance improves the current account.
The effects of fiscal policies are identified in a SVAR model by assuming that fiscal authorities require at least one quarter to respond to output shocks (Blanchard and Perotti, 2002).
The 95 confidence intervals of the cumulative multipliers are computed by Monte Carlo simulations based on 500 replications.
The cumulative multiplier of 0.53 after 24 quarters can be considered as moderate. For example, Perotti (2005) finds that the cumulative spending multiplier ranges from 0.3 to 1.4 after 20 quarters for five Organization for Economic Co-operation and Development member countries.
These countries are Antigua and Barbuda, The Bahamas, Dominica, Grenada, St. Kitts and Nevis, St. Lucia, and St. Vincent and the Grenadines.
Using different thresholds, such as 50 or 60 percent of debt to GDP ratio, does not change our results.
The tax elasticity changes the magnitude of the initial responses.
We use imports instead of GDP when we estimate the multipliers on imports.
Although the suggested number of lags varies, four lags are sometimes actually suggested.
