Information about Middle East Oriente Medio

5. Fiscal Policy for Macroeconomic Stability

Raphael Espinoza, Ghada Fayad, and Ananthakrishnan Prasad
Published Date:
November 2013
  • ShareShare
Information about Middle East Oriente Medio
Show Summary Details

5.1 Introduction

Macroeconomic stability is important for long-term growth and the literature has discussed several channels through which volatility can reduce growth.1 Firms are likely to make important mistakes when planning is done in an uncertain environment (Ramey and Ramey 1991). As a result, when investment decisions are irreversible or when it is costly to switch production, investors may decide to wait before investing in order to gather more information about the future (McDonald and Siegel 1986; Bertola 1994). Firms also hit liquidity constraints more frequently in volatile environments. When financial institutions are underdeveloped, liquidity constraints limit investment (Aghion et al. 2010).

The empirical literature has found indeed that a one percentage point increase in volatility tends to reduce annual growth by around 0.3 percent, and the negative link is stronger for countries with poor institutions, intermediate levels of financial development, and for countries that do not implement countercyclical fiscal policies—all characteristics that apply to some extent to the GCC (Hnatkovska and Loayza 2005; Kose et al. 2006; van der Ploeg and Poelhekke 2009).

In the GCC as in many other oil-exporting countries, growth has been very volatile over the last thirty years, and this is why in Chapter 2 we attributed part of the disappointing growth in TFP to this factor. In the period 1976–90, the standard deviation of GDP growth exceeded 7 percent in all countries but Qatar, above what is typical for either advanced or developing countries (Table 5.1). The recent period has been more favorable. Nevertheless, the standard deviation of growth per capita has been around or above 4 percent in all the GCC countries but Bahrain, whereas before the Great Recession volatility had decreased to less than 2 percent in advanced economies.

Table 5.1.Standard deviation of GDP growth per capita in percent, 1976–2007
CountryGDP per capitaGDP per capitaNon-oil GDP per capita
Saudi Arabia7.75.53.3
OECD (median country)
Oil exporters (median country)
Other developing countries (median country)
Source: PWT 7, IMF, and authors’ calculations
Source: PWT 7, IMF, and authors’ calculations

Policymakers can use both fiscal and monetary policy to stabilize economic cycles. However, since the GCC countries have long pegged their currencies to the US dollar,2 domestic interest rates in the region have closely tracked US rates. As a result, conventional monetary policy loses power to stabilize the economy. Chapter 6 analyzes this issue further, whereas in this chapter we show the importance of fiscal policy as a tool for macroeconomic stabilization.

The GCC economies faced the Great Recession—as the 2008–9 world recession became known—in a strong position, supported by several years of high oil prices and fiscal surpluses that alimented the build-up of government savings and foreign reserves. Despite a major drop in government revenues in 2009, fiscal space was available for countercyclical measures, and overall GCC real government spending remained high, supporting non-oil growth. Also in 2009, Saudi Arabia in conjunction with other G20 countries assigned several percent of GDP of spending to a fiscal stimulus package motivated by the need to kick-start the world economy. We assess in this chapter the effectiveness of fiscal policy and the use that the GCC countries have made of this tool in the last thirty years.

There is little recent research on the effect of fiscal policy in the GCC3 and therefore having an estimate of the impact of spending shocks is important. Our focus is on spending as opposed to revenues because most of the revenue comes from hydrocarbon exports and little is derived from non-oil taxes. Hence, fiscal policy is tantamount to expenditure policy in the region. We show, using panel and simple Vector Auto-Regression (VAR) models, that fiscal expenditure is a major driver of growth cycles in the non-oil economy. We find that the fiscal multiplier (i.e. the volume of domestic economic activity that is generated by a dollar of government spending) is positive and statistically significant, and varies between 0.2 and 0.5 the first year spending is increased. Expenditure on public investment (capital expenditure) is found to have stronger effects than current expenditure (public wages, spending on goods and services, etc.).

The VAR models also show that 20 to 60 percent of the variance of unexpected non-oil GDP growth can be explained by fiscal shocks. To a large extent these shocks reflected oil revenues. Indeed, our results suggest that fiscal consolidations, forced by low oil prices, were an important cause of GDP contraction in 1986–7 and in 1998–2000, whereas an expansionary fiscal stance would have pushed activity after the First Gulf War and in the five years between 2003 and 2008. In addition, several GCC governments ran expansionary policies during the Great Recession with the objective of mitigating the negative impact of external factors.

We first provide some background on government spending in the GCC (section 5.2) and on the recent literature on fiscal multipliers (section 5.3). The estimates of fiscal multipliers, using both panel and VAR techniques, show that government spending has a strong effect on non-oil growth (section 5.4). In the last sections of the chapter, we discuss whether fiscal policy has been pro-cyclical or countercyclical, and assess the contribution of fiscal policy to economic cycles in the region, using variance and historical decompositions.

5.2 Background

In the GCC, governments exert a strong influence on the economy and government spending accounts for between 35 and 80 percent of non-oil GDP (see Table 5.2). Current expenditure forms the bulk of government spending, but capital expenditure represents nonetheless more than 12 percent of spending in all GCC countries, and reaches 30 percent of GDP in Oman and Qatar. Government revenues come from hydrocarbon exports and investment income and as a result, non-oil fiscal revenues and in particular income taxes, corporate taxes, and VAT are small (less than a third; see again Table 5.2). Government spending is therefore the main instrument of fiscal policy.

Table 5.2.Characteristics of GCC economies and government expenditure
Average 2000–9BahrainKuwaitOmanQatarSaudi ArabiaUAE
Government expenditure/non-oil GDP0.360.790.660.640.660.36
Current expenditure/total expenditure0.760.880.720.700.800.83
Capital expenditure/total expenditure0.
Hydrocarbon-related revenues/total revenues0.770.730.840.660.860.73
Investment income/total revenuesb0.
Taxes and others/total revenues0.
Government expenditure and oil prices R-square (1991-2007)a0.100.290.300.550.790.20
Imports of goods and services/nominal GDP0.710.350.390.350.340.68
Share of non-national in total population (2005)0.410.620.240.720.270.79

Regressions of the growth rate of nominal spending on contemporaneous and lagged oil price inflation.

Investment income for Bahrain includes other capital revenues.

Source: IMF and authors’ calculations

Regressions of the growth rate of nominal spending on contemporaneous and lagged oil price inflation.

Investment income for Bahrain includes other capital revenues.

Source: IMF and authors’ calculations

Because government revenues coming from the non-oil economy are small, government spending is eventually constrained by oil revenues and the historical pattern has indeed been that expenditure followed oil prices. The R-square of the regression of government expenditure on oil prices is highest in Qatar and Saudi Arabia, where it reaches 80 percent (Table 5.2). Indeed, Fasano and Wang (2002) show using a VAR model of revenues and expenditure that around 80 percent of the variance of expenditure shocks in Saudi Arabia can be explained by shocks in revenues. The proportion would be around 65 percent for Qatar. In the last crisis, as oil prices collapsed, the announcements from several countries that the government would maintain or even increase expenditure despite low oil prices was therefore a significant change from past practice.

This policy shift raises the question of the effect of government spending on activity. Theory suggests several determinants of the size of fiscal multipliers, depending on the model chosen. In a neoclassical model of a closed economy, Woodford (2011) shows that the size of the fiscal multiplier is lower the harder it is to employ new resources and the easier it is for the private sector to reduce consumption and investment to leave space for government spending (in that case government spending is “crowding out” private spending). In addition, if monetary policy does not react to fiscal expansion (if the interest rate is constant), the multiplier reaches 1 because private spending is fully determined by intertemporal optimization, and optimal private-sector decisions are unchanged if interest rates are fixed (Woodford 2011). In a Keynesian model, multipliers are lower in more open economies because a larger faction of the increase in government spending is spent on imports, which do not contribute to domestic production. In a Mundell-Fleming model, multipliers are larger under fixed exchange rate regimes because the interest rate is not affected by the fiscal expansion. Ilzetzki, Mendoza, and Végh (2010) have indeed shown using a panel of forty-four countries that multipliers are larger in closed economies and in economies with fixed exchange rates.

How does the GCC fare with respect to the determinants of the fiscal multiplier identified in the literature? Imports are large in the GCC—between 35 and 70 percent (Table 5.2)—and it is therefore possible that “leakages” from government spending to imports reduce the effect of government spending on growth. In other words, government money spent on wages or on capital expenditure could finance imports and therefore not contribute to GDP. In particular, stimulus spending would take in many cases the form of capital spending that would require importing machinery and material. On the other hand, the GCC currencies are fixed to the US dollar, and therefore movements in interest rates would not dampen the effect of fiscal policy.

Finally, it is relatively easy in the GCC to mobilize factors of production because a large portion of the new labor force comes from abroad and the governments are wealthy and do not need to tap into private funds to spend. With a large foreign workforce, even in the public sector, increases in government spending could leak out as outward remittances and not be spent in the domestic economy. The multipliers could therefore be small, as in many open emerging economies, but the issue has to be settled empirically.

5.3 The Empirical Literature on Fiscal Multipliers

The estimation of many macroeconomic relationships raises the concern of potential endogeneity of the explanatory variables. In the particular case at hand, the relationship of interest is that between an indicator of economic growth and an indicator of fiscal spending.

The problem is that the relationship might be bidirectional (i.e., fiscal spending influences economic growth and vice versa). Failing to control for the endogeneity of spending may lead to the well-known simultaneity bias. Economists have identified two sources of endogeneity when estimating fiscal multipliers:

  • (a) In good times, spending is reduced (lower employment benefits) and taxes are higher (automatic fiscal stabilizers) and therefore the correlation between spending and activity tends to be negative, and as a result the multipliers are underestimated by OLS. We can safely rule out this channel in the GCC since, as we have already shown, fiscal policy boils down to government spending and there is very little “automatic” spending in the region (unemployment benefits are insignificant).
  • (b) Systematic countercyclical policies may give the impression that fiscal expansions have limited effects since they will be implemented during bad times (endogenous fiscal policy). This source of endogeneity is likely to be of a lesser importance in the GCC than in other countries because, historically, spending has been limited by oil revenues and driven by diversification objectives.

The literature has tackled the issue of endogeneity from two angles (see also the surveys in Spilimbergo et al. 2008 and Spilimbergo et al. 2009).

  • (a) Case studies have looked at specific experiments. For instance, Romer and Romer (2008) analyzed the effect of tax policy changes using detailed information on policy decisions to identify the size and timing of policy changes in the US. This allows them to identify structural shocks (changes in taxes exogenous to the business cycle) and estimate the effects on GDP in an OLS quarterly model (a VAR is also analyzed for robustness).
  • (b) VAR identification procedures have been used in an attempt to take into account the endogenous fiscal response. Blanchard and Perotti (2002) estimated a quarterly VAR on the US with output, tax revenues, and total spending, and the key to the identification procedure was to recognize that the use of quarterly data virtually eliminates the endogeneity bias: fiscal policy cannot react fast enough to be endogenous to current quarter activity. The second element in their identification procedure was to estimate directly the behavior of the automatic fiscal stabilizers (applying the OECD method; see Giorno et al. 1995) to constrain the structural coefficient and thus deduce some reduced-form coefficients in the VAR. More sophisticated identification procedures exist and do not need to assume contemporaneous exogeneity. For instance, Mountford and Uhlig (2008) use sign restrictions in a standard VAR model to identify fiscal shocks as positive shocks to fiscal spending that are uncorrelated with monetary and business cycle shocks.4

The literature that covers a cross-section of countries has taken the VAR route to reduce the importance of the endogeneity bias, since finding detailed information on specific spending programs for many countries is cumbersome. For instance, IMF (2008) estimates impulse responses for the G7 countries using quarterly VARs with Choleski ordering of the output gap, the GDP deflator, the structural fiscal balance, and the cyclical fiscal balance. Perotti (2005, 2006) uses a different identification procedure but the methodology is also based on a quarterly structural VAR of fiscal variables, GDP, the GDP deflator, and the ten-year nominal interest rate.

Panel methods have also been common. Ilzetzki and Végh (2008) used Instrumental Variables models and system GMM on quarterly data from forty-nine countries to investigate both the size of multipliers and the importance of pro-cyclicality, and found that although IV models were inconclusive, the GMM models suggested that pro-cyclicality existed but that fiscal multipliers had also been underestimated, especially for developing economies, for which the elasticity was estimated to reach 0.73, implying a multiplier of around 1 for the GCC.

By and large, much of the fiscal multiplier literature has considered advanced economies. A recent survey is provided by Romer (2011). The literature on emerging markets is not as vast, and has not considered the GCC separately. The next section provides our estimates of fiscal multipliers using panel and VAR models for this group of countries.

5.4 Econometric Estimates of Fiscal Multipliers

We follow the literature in estimating panel models for the region and VAR models country by country. The description of the data used is presented in Table 5.3. Panel models allow us to increase statistical power by pooling data from the six GCC countries. As a result, it is easy to distinguish in these models the impact of current expenditure from that of capital expenditure, and to control for many external factors. However, panel models assume homogeneity in the estimated elasticities, and the identification of fiscal policy shocks is easier with VAR models. In addition, the variance and historical decomposition of shocks in time series models, discussed in section 5.5, is also useful to provide a narrative of the drivers of growth cycles.

Table 5.3.Data sources
Non-oil real GDP growth (GCC excl. Saudi Arabia)IMF (2010b) and IMF Article IV country reports (1980-90)
Saudi Arabia non-oil real GDP growthSaudi Arabia Monetary AuthorityNational Accounts, constant prices (base 100 in 1999)
Government expenditure (GCC excl. Saudi Arabia)IMF (2010b) and IMF Article IV country reports (1980-90)Central government, deflated by the CPI
Saudi Arabia government expenditureSaudi Arabia Monetary AuthorityCentral government, deflated by the CPI
Inflation (excl. Saudi Arabia)IMF (2010b) and IMF Article IV country reports (1980-90)CPI inflation
Saudi Arabia inflationSaudi Arabia Monetary Authority
Oil priceIMF (2010a)WTI (West Texas Intermediate)
World interest rateIMF (2010a)Fed Funds Rate

5.4.1 Panel Model

The data for the six members of the GCC are presented in Figure 5.1. Some co-movement between non-oil growth and spending appears, except in the UAE. In the UAE, fiscal data have gaps because there is no consolidated budget. We eventually dropped the data for the UAE as they drastically reduced the significance of the panel results.

Figure 5.1.Non-oil real GDP growth and real growth in total government expenditure

Source: IMF and IMF country reports

We first estimate a simple panel model of non-oil real GDP as a function of growth in public expenditure using Pooled OLS (Table 5.4, POLS, columns 1, 5, and 9), random effects (RE, columns 2, 6, and 10), and fixed effects (FE, columns 3, 7, and 11). We also estimated a model excluding contemporaneous spending, so as to check the robustness of the results with regard to endogeneity of government spending (lag FE model, columns 4, 8, and 12).5

Table 5.4.GCC panel fiscal multipliers—dependent variable: non-oil real GDP growth
Total ExpenditureCapital ExpenditureCurrent Expenditure
Δ Expenditure0.233***0.233***0.233***
Δ Expenditure (t-1)0.160***0.161***0.161***0.201**
Δ Expenditure (t-2)0.0891***0.0892**0.0897*0.0908**
Δ Capital0.0513***0.0513***0.0511**
Δ Capital0.0459**0.0459***0.0457***0.0336*
Expenditure (t-1)[2.430][6.794][6.692][2.512]
Δ Capital0.0453***0.0453***0.0450***0.0286**
Expenditure (t-2)[3.510][8.971][8.764][4.056]
Δ Capital0.0337***0.0337***0.0336***0.0316**
Expenditure (t-3)[2.687][4.784][4.758][4.146]
Δ Current0.235***0.235***0.241***
Δ Current0.103**0.103***0.108**0.103*
Expenditure (t-1)[2.427][4.030][3.674][2.190]
Δ Current0.04360.04360.04720.0437
Expenditure (t-2)[1.543][1.041][1.045][1.300]
*** p< 0.01, **p < 0.05, * p< 0.1, t-statistics in brackets
S-T multiplier0.330.330.330.290.330.330.330.220.440.440.450.19
L-T multiplier0.690.690.690.421.
Source: IMF, country authorities, and authors’ estimates
Source: IMF, country authorities, and authors’ estimates

For all regressions including the lag FE models, the coefficients for contemporaneous, lagged, and twice-lagged fiscal expenditure are positive and significant. For the models that include capital expenditure only, the third lags are also significant (all t-statistics presented are computed using standard errors robust to heteroskedasticity). Overall, our estimates suggest that the short-term fiscal multiplier6 is around 0.3 for total government expenditure (columns 1 to 4), 0.2–0.3 for capital expenditure (columns 5 to 8), and 0.2–0.4 for current expenditure (columns 9 to 12). A possible explanation for why the multiplier on capital expenditure is weaker than that for current expenditure is the relatively long gestation lags for capital formation. It can take several years of investment before productive capacity is operational.

Indeed, the long-run multiplier estimates suggest that the effect of capital expenditure on non-oil GDP is significantly higher than that for current expenditure: the long-run multiplier on capital expenditure is 0.6–1.1 versus 0.3–0.7 and 0.4–0.7 for current and total spending, respectively. Endogeneity did not seem to be a major issue since removing contemporaneous spending did not change statistical significance, although the multipliers were found to be smaller.

We tested the robustness of these results to the inclusion of several control variables: inflation, oil prices, interest rates, and world growth (Table 5.5). Oil prices can affect non-oil growth either via the petrochemical industry (the demand for petrochemicals is highly correlated to oil prices), via increased liquidity and business confidence, or via fiscal expenditure—but this latter channel should ideally be attributed to government spending. Contemporaneous oil prices were not significant but lagged oil price changes were found to affect non-oil growth. The contemporaneous correlation between fiscal spending and growth is therefore entirely attributed to the effect of fiscal policy when the contemporaneous oil price is dropped (a more appropriate identification of shocks is done thanks to the VAR historical decomposition; see next section).

Table 5.5.GCC panel fiscal multipliers, controlling for inflation and oil prices
Total Expenditure (FE)Capital Expenditure (FE)Current Expenditure (FE)
Δ Expenditure0.233***0.206**0.218***0.192***
Δ Expenditure (t-1)[4.934] 0.161***[4.453] 0.150***[5.680] 0.121**[5.179] 0.111**
Δ Expenditure (t-2)0.0897*0.09410.04200.0451
Δ Capital0.0511**0.0422**0.03460.0266
Δ Capital0.0457***0.041 8***0.0298**0.0269*
Expenditure (t-1)[6.692][7.122][2.892][2.218]
Δ Capital0.0450***0.0377***0.0297**0.0230*
Expenditure (t-2)[8.764][7.395][3.403][2.295]
Δ Capital0.0336***0.0327***0.02000.0198
Expenditure (t-3)[4.758][4.760][1.689][1.619]
Δ Current0.241***0.208***0.171*0.1 31
Δ Current0.108**0.0924**0.04790.0285
Expenditure (t-1)[3.674][3.467][0.771][0.432]
Δ Current0.04720.05890.005890.0157
Expenditure (t-2)[1.045][1.076][0.161][0.376]
Δ Oil Price (t-1)0.0438**0.03980.0778**0.0683**0.0696**0.0686**
Observations1561511551501 391 341381 331441 39143138
*** p < 0.01, ** p < 0.05, * p < 0.1, t-statistics in brackets
S-T multiplier0.330.290.310.160.330.
L-T multiplier0.690.640.540.501.130.990.730.620.740.670.420.33
Source: Country authorities and authors’ estimates
Source: Country authorities and authors’ estimates

The Fed Funds rate and its lags, as well as world GDP growth were not significant. Finally, current inflation was strongly correlated with growth and was also included. Overall, the short-run multipliers would remain around 0.2–0.3 in most models, and the long-run multiplier would be 0.6 for capital spending and 0.3 for current spending.

5.4.2 Fiscal Policy and Economic Cycles

We now investigate the importance of fiscal shocks on the dynamics of GDP using Vector Auto-Regressions. We estimate VAR models for each country linking real world GDP, real government expenditure (we follow Ilzetzki and Végh 2008 in deflating all fiscal variables by the CPI), and non-oil real GDP.7 In a VAR, it is important to keep the number of variables to its minimum. Adding endogenous variables in particular (such as inflation) is very costly in degrees of freedom. We therefore did not include as control variables the Fed Funds Rate, which was not significant in the panel, nor inflation, because it was not essential to a discussion of fiscal multipliers. We did however include world growth in order to disentangle external from domestic components when interpreting growth cycles. In addition, we checked the robustness of our results to including oil prices as an exogenous variable.

The model was estimated on 1980–2008 data since data was unavailable for most countries before that period. Data was available for Saudi Arabia, but the coefficients were unstable when using data between 1968 and 1975. The variables were converted into logarithms8 so that the ratio of the impulse responses can be thought of as elasticities. The short-term multiplier (one year) is then obtained by dividing the elasticity by the ratio of spending to GDP. The two-year multiplier is obtained by assuming the expenditure shock is maintained over two years.

For each country in the GCC, the following VAR is estimated by OLS for a vector yt of world GDP, real government expenditure, and non-oil real GDP. The VAR has three lags,9 and additional exogenous variables Xt are added as control variables:

The identification procedure is based on a Choleski orthogonalization of the covariance matrix of errors Σ, with fiscal expenditure ordered before non-oil growth, i.e., it is assumed there is no immediate effect of non-oil growth on fiscal expenditure. This identification procedure assumes that because budgets are voted much ahead of their implementation, the causality runs from spending to non-oil growth and not the other way around. The assumption is common in the literature and easily justifiable when using quarterly VARs, since fiscal policy cannot be adjusted within a quarter. But we only have annual data in the GCC, and therefore the VARs are estimated on annual data. The assumption is therefore a stronger one. World growth is ordered first in the VAR, to capture the effect of the global environment.

In addition, we checked how robust the results are to including oil prices as an exogenous variable. It is not always easy to disentangle shocks to oil prices from shocks to government spending because GCC spending has been strongly correlated with oil revenues. Table 5.6 summarizes the elasticities of government spending to oil prices found in the VAR when oil prices were included. In Saudi Arabia, as also noted in Table 5.1, the link between spending and oil prices is very strong. On the other hand, oil price was not significant in the UAE regression. Our preferred specification, presented in Figures 5.2 to 5.10, is however the one without oil prices because we think the main channel via which oil revenues affect the non-oil is via government spending.

Table 5.6.Elasticity of spending to a permanent increase in oil price in the VARs
BahrainKuwaitOmanQatarSaudi ArabiaUAE
1 year0.27−−0.20
2 year0.−0.10

Figure 5.2.Fiscal multiplier (impact of total government spending on non-oil GDP)

Figure 5.3.Impact of world growth on non-oil growth in the GCC

Figure 5.4.Pro-cyclicality of fiscal policy: Response of government spending to non-oil GDP shocks, in percent deviation from baseline

Note: Impulse Response Function for a permanent 1 percent shock to non-oil real GDP; 16th and 84th percentile error bands

Figure 5.5.Saudi Arabia (non-oil GDP growth on LHS scale, contributions on the RHS)

Figure 5.6.UAE (non-oil GDP growth on LHS scale, contributions on the RHS)

Figure 5.7.Kuwait (data post-1996 is from IMF; data pre-1996 is from UN). Dummies for estimation in 1991 and 1992

Figure 5.8.Qatar (non-oil GDP growth on LHS scale, contributions on the RHS)

Figure 5.9.Oman (non-oil GDP growth on LHS scale, contributions on the RHS)

Figure 5.10.Bahrain (non-oil GDP growth on LHS scale, contributions on the RHS)

We show in Figure 5.2 the effect of a shock to government spending (normalized to a permanent one dollar increase) on non-oil GDP. The multiplier is estimated to be around 0.3–0.5 in the year of the shock in Bahrain, Kuwait, Oman, and Qatar, and to increase to around 0.7 after two years in Kuwait, Qatar, and the UAE. The multipliers are statistically different from 0 in all countries at the 90th percent confidence level. When adding oil prices as a control variable, the multipliers are similar in the first year but tend to be higher in the second year (1 for Kuwait and Qatar, and 2 for Bahrain and the UAE). These results are roughly in line with the results of the panel model: dropping the UAE (which was dropped in the panel given the poor quality of data), the average multiplier is 0.3 in the first year and 0.5 after two years (1.0 when oil prices are added as a control variable). There is some heterogeneity in the data, which could not be captured by the panel model. In particular, the multiplier is smaller in Saudi Arabia (0.1 after one year and 0.2 after two years, irrespective of whether oil prices are included) and larger in Bahrain and the UAE.

Positive shocks in the world economy spill over to the GCC (Figure 5.3). The elasticity of individual countries to GDP is higher than 1 in the small and very open economies of Qatar, UAE, and Oman. The elasticity is however lower in Saudi Arabia, Kuwait, and Bahrain. The main difference, when adding oil prices, is that the elasticity of non-oil growth to world growth is much lower for Qatar and is positive in the first year after the shock for Kuwait.

5.5 Contribution of Fiscal Policy to Economic Cycles

We use the same VARs to assess whether fiscal policy has been pro-cyclical or countercyclical in the region. Our measure of cyclicality is the orthogonalized impulse response of government spending to shocks in non-oil GDP. One should remember however that since our identification strategy assumed that fiscal policy cannot react within a year to news on non-oil GDP growth, the VARs may underestimate the reactivity of fiscal policy. Figure 5.4 presents the impulse response functions, normalized for a permanent 1 percent shock to non-oil GDP.

Fiscal policy appears to have been historically countercyclical in Saudi Arabia and in Oman. In Saudi Arabia, a 1 percent (permanent) negative shock in non-oil GDP would have triggered a strong expansion in government spending in the next year, although the increase would have been quickly offset the subsequent year (surprisingly, this result is robust to including oil prices as a control variable). In the other four countries of the GCC, fiscal policy would have been pro-cyclical, and the elasticity of spending to non-oil growth would have been high in Qatar and in Bahrain. When controlling for oil prices, the elasticity of spending to growth is not significantly different from 0 in Kuwait and Qatar, which suggests that our finding of pro-cyclicality in these two countries is really due to the importance of oil revenues in driving both government spending and non-oil growth.

We use the VARs to assess to what extent government expenditure drives non-oil GDP cycles. We first look at the Forecast Error Variance Decomposition (FEVD), which summarizes the contribution of the orthogonalized shocks10 to the variance of the stochastic component of non-oil GDP growth (Table 5.7). In Qatar and Kuwait, government spending would have contributed more than 60 percent to this variance (50 percent when controlling for oil prices). In the other GCC countries, government spending would have contributed to between 17 and 38 percent of the variance (17–28 percent when controlling for oil prices).

Table 5.7.Non-oil GDP growth: Forecast Error Variance Decomposition two years ahead
BahrainKuwaitOmanQatarSaudi ArabiaUAE
World growth0.060.000.410.170.260.18
Govt expenditure0.380.650.170.620.170.28
Non-oil GDP0.560.350.420.210.570.53
Note: In Kuwait, the regression of non-oil growth on oil prices also includes two dummy variables for the years 1991 and 1992.
Note: In Kuwait, the regression of non-oil growth on oil prices also includes two dummy variables for the years 1991 and 1992.

World growth contributed less to volatility than might be expected given the high elasticities estimated earlier, because world growth is actually much more stable than growth in the GCC. The FEVD attributes 18 to 26 percent of the unexpected variance in GDP to shocks in world growth (except for Oman where the share goes to 41 percent). The Kuwait and Bahrain economies would not have been affected by world growth shock. These results are again mostly unaffected by adding oil price as a control variable.11

We now compute the historical decomposition of non-oil GDP growth. A historical decomposition relies on the moving average representation of a VAR:

where y^t+j is the deterministic part of yt+jyt+jF|t, is the forecast of yt+j based on information available at time t, and Vt+1,t+j is the component in yt+j that is due to shocks that occurred between t and t+j. Ut are the innovations that have been obtained thanks to the Choleski decomposition of Σ and Ψs is the matrix containing the orthogonalized impulse responses of the VAR.

The historical decomposition is useful because it allows us to explain growth and cycles from past shocks of world growth, fiscal spending, and oil prices using the coefficients Ψs (the impulse response functions, summarized in Figures 5.2 and 5.3), the structural shocks ut, and the deterministic part y^t+j, which is a function of oil prices in particular.

According to the historical decomposition, restrictive fiscal policy was the cause of the GDP contractions in Saudi Arabia in 1986 and in the mid-1990s (Figure 5.5). Expansionary policies would have pushed activity in the late 1980s and in the years 2005–10, despite negative contributions from the world business cycle. These cycles in spending were strongly related to oil prices, although since 2000 the oil price boom has been large enough for the Saudi government to be able to save. As a result, spending has been more stable and the 2009 announcement that fiscal policy would explicitly aim at stabilizing growth goes one step further. The VAR does indeed show that Saudi government policy was countercyclical in 2009–10.

In the UAE and in Kuwait, growth has been very volatile (Figure 5.6 and Figure 5.7) and fiscal policy as well as external shocks mattered. The boom in the UAE in 1993 has been attributed to a spike in government spending, but much of the growth in 2003–7 is unexplained, although some of it can be attributed to a favorable external environment. The UAE is a very open economy, and its service sector includes ports and airlines that are highly dependent on world trade. The recent fall in growth is clearly associated with the global crisis. In Kuwait, the First Gulf War affected government operations but public spending and growth rebounded in 1993. Oil prices and government spending also stimulated growth in 2003–5 but fiscal policy has been contractionary since 2010–11.

In Qatar, external and fiscal shocks have contributed to the high volatility of the economy (Figure 5.8). Non-oil growth was low in the first half of the 1990s as oil prices fell and the world economy slowed down. Government spending was restrained in the late 1990s and this drove down GDP growth from 1997 to 2003, until the persistent increase in oil prices allowed expansionary fiscal policies. The boost in public investment, aimed at scaling up gas production, contributed significantly to pushing non-oil GDP growth to 40 percent in 2006. The global downturn and the moderation in oil prices since 2008 have brought back growth to single digits.

Oman has suffered two recessions in the last thirty years (see Figure 5.9). The first one, in 1986–7, can be attributed to the fall in oil prices, which also constrained government spending. The second recession, in 1999, was again brought about by tight fiscal policy (oil prices had been low for two years) although domestic factors (a fall in private consumption and the completion of the large LNG project and of the Salalah port) also contributed.

Finally, for Bahrain, the VAR is only useful to explain the sources of growth after 2000 (before that, the historical decomposition attributes the shocks to domestic sources, as can be seen from Figure 5.10). The strong growth recorded between 2001 and 2009 was to a large extent due to government spending, high oil prices, and a positive external environment. The global crisis stopped this positive cycle, despite the government maintaining its spending, and domestic factors, including bank deleveraging and social unrest, have contributed to pulling the economy down since 2009.

5.6 Conclusion

In the GCC, governments wield considerable control over the economy and the main instrument of macroeconomic management remains government spending, in countries with fixed exchange rates and a small tax base (and therefore negligible automatic stabilizers). It is therefore important to assess the effectiveness of fiscal policy. The existing literature does not provide readily available estimates for the size of fiscal multipliers, i.e, the increase in GDP that can be obtained by decreasing the government balance by one dollar. Depending on country characteristics (open versus closed economy, fixed versus flexible exchange rate regimes) and on the instrument of fiscal policy (taxes versus government expenditure), and depending on the theoretical framework (Keynesian versus neoclassical) and on the econometric model used (VAR, case studies, etc.), multipliers have been found to be anywhere between 0 (e.g., IMF 2008) and 1 (e.g., GMM model in Ilzetski and Végh 2008).

We estimated several models for the GCC and found that the short-term multipliers are between 0.2 and 0.4 in most specifications. The long-term capital spending multipliers would be larger, and they were estimated between 0.6 and 1, whereas current spending multipliers in the long run would be between 0.2 and 0.4 according to the majority of our models.

The effect of government spending in the GCC is likely to depend on the distribution between capital and current spending and the import content of the programs implemented. For Saudi Arabia, where spending is scaling up, government expenditure will be distributed between capital spending in the oil infrastructure and utility sectors (with a fairly high import content) and current spending focused on second-generation reforms (education, health, the judiciary), areas that are likely to have a lower import content and therefore stronger multipliers. Econometric estimates provide merely baseline numbers and it is therefore important to apply judgment, based on the detailed distribution and timing of expenditure, when assessing the effect of different spending programs.

We also discussed the contribution of fiscal policy to economic stabilization. The VARs suggest that fiscal policy would have been countercyclical in Saudi Arabia and Oman. However, for the other countries of the region, policy would have been pro-cyclical. Enhancing fiscal management is important for all oil producers that seek to limit a volatility-driven resource curse. Fiscal frameworks can help (see IMF 2012) and there is scope for improvement in the region.


    AghionP.AngeletosG. M.BanerjeeA. and ManovaK.(2010). “Volatility and growth: Credit constraints and the composition of investmentJournal of Monetary Economics57 (3): 24665.

    • Search Google Scholar
    • Export Citation

    BlanchardO. and PerottiR.(2002). “An empirical characterization of the dynamic effects of changes in government spending and taxes on outputQuarterly Journal of Economics117: 132968.

    • Search Google Scholar
    • Export Citation

    BertolaG.(1994). “Flexibility, investment and growthJournal of Monetary Economics34: 21538.

    EspinozaR. and SenhadjiA.(2011). “How strong are fiscal multipliers in the GCC? An empirical investigation.” IMF Working Paper 11/61. Washington DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    FasanoU. and WangQ.(2002). “Testing the relationship between government spending and revenue: Evidence from the GCC countries.” IMF Working Paper 02/201. Washington DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    GiornoC.RichardsonP.RoseveareD. and NoordP.van den (1995). “Estimating potential output, output gaps and structural budget balances.” OECD Economics Department Working Papers No. 152. Paris: OECD Publishing.

    • Search Google Scholar
    • Export Citation

    HnatkovskaV. and LoayzaN.(2005). “Volatility and growth” in J.Aizenman and B.Pinto (eds) Managing Economic Volatility and Crises: A Practitioner’s Guide. Cambridge: Cambridge University Press65100.

    • Search Google Scholar
    • Export Citation

    IlzetzkiE.MendozaE. and VéghC.(2010). “How big (small?) are fiscal multipliers?NBER Working Paper No. 16479. Cambridge, Mass.: National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation

    IlzetzkiE. and VéghC.(2008). “Procyclical fiscal policy in developing countries: Truth or fiction?Mimeo. University of Maryland.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF) (2008). “Fiscal policy as a countercyclical tool” chapter 5 in World Economic OutlookOctober.Washington DC: International Monetary Fund15996.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF) (2010a). World Economic Outlook: Rebalancing Growth April 2010. Washington DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF) (2010b). Regional Economic Outlook: Middle East and Central Asia October 2010. Washington DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF) (2012). “Towards a GCC monetary union: The role of fiscal policy and institutions.” Washington DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    KoseM. A.PrasadE. S.TerronesM. E. (2006). “How do trade and financial integration affect the relationship between growth and volatility?Journal of International Economics69: 176202.

    • Search Google Scholar
    • Export Citation

    McDonaldR. and SiegelD.(1986). “The value of waiting to investQuarterly Journal of Economics101 (4): 70727.

    MountfordA. and UhligH.(2008). “What are the effects of fiscal policy shocks?NBER Working Paper No. 14551. Cambridge, Mass.: National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation

    NakibullahA. and IslamF.(2007). “Effect of government spending on non-oil GDP of BahrainJournal of Asian Economics18 (5): 76074.

    • Search Google Scholar
    • Export Citation

    PerottiR.(2005). “Estimating the effects of fiscal policy in OECD countries.” CEPR Discussion Paper No. 4842. London: Centre for Economic Policy Research.

    • Search Google Scholar
    • Export Citation

    PerottiR.(2006). “Public investment and the golden rule: Another (different) look.” IGIER Working Paper No. 277. Milan: Bocconi University Innocenzo Gasparini Institute for Economic Research.

    • Search Google Scholar
    • Export Citation

    PloegF.vander and PoelhekkeS.(2009). “Volatility and the natural resource curseOxford Economic Papers61: 72760.

    RameyG. and RameyV.(1991). “Technology commitment and the cost of economic fluctuations.” NBER Working Paper No. 3755. Cambridge, Mass.: National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation

    RomerC.(2011). “What do we know about the effects of fiscal policy? Separating evidence from ideology.” Speech at Hamilton CollegeNovember72011.

    • Search Google Scholar
    • Export Citation

    RomerC. and RomerD.(2008). “The macroeconomic effects of tax changes: Estimates based on a new measure of fiscal shocks.” Unpublished ms. University of CaliforniaBerkeley.

    • Search Google Scholar
    • Export Citation

    SpilimbergoA.SymanskyS.BlanchardO. and CottarelliC.(2008). “Fiscal policy for the crisis.” IMF Staff Position NoteDecember292008 SPN/08/01.

    • Search Google Scholar
    • Export Citation

    SpilimbergoA.SymanskyS. and SchindlerM.(2009). “Fiscal multipliers.” IMF Staff Position NoteMay202009 SPN/09/11.

    WoodfordM.(2011). “Simple analytics of the government expenditure multiplierAmerican Economic Journal: Macroeconomics3: 135.

    • Search Google Scholar
    • Export Citation
1This chapter is an extension of a paper written with Abdelhak Senhadji (Espinoza and Senhadji 2011). The authors are grateful to A. Senhadji for allowing us to use parts of his work in the present chapter.
2Kuwait is an exception, as its currency has been pegged to a basket of currencies since 2007.
3An exception is Nakibullah and Islam (2007) who study the effect of fiscal policy in Bahrain. They find that temporary shocks have little effect whereas permanent shocks are important. They take into account the global environment of Bahrain by controlling for US fiscal shocks and find that a US expansionary shock has negative spillovers on Bahrain.
4See, for instance, the discussion on VAR identification in Perotti (2005) for more details.
5We also used different IV models (not reported). We tried oil prices as an instrument, but because high oil prices also benefit the petrochemical industry (the production of which is included in the non-oil GDP data), improve confidence, and ease financial conditions, exogeneity of the instrument was unlikely. Lagged spending proved also to be a weak instrument. As a result, we found the lag FE model, in which endogeneity is removed by construction, to be a safer alternative.
6The fiscal multiplier is computed as the elasticity divided by the ratio of capital (or current or total) expenditure to GDP. The elasticity is α = (ΔY/Y)/(ΔG/G) and therefore the multiplier is ΔY/ΔG = α Y/G. The ratio Y/G has evolved over time, which is why we use the historical average to calibrate it (see Table 5.2).
7We use total government expenditure, despite the criticism formulated in Ilzetzki and Végh (2008) against the inclusion of interest payments and transfers in the data, because we do not have such detailed description of expenditure before 1990. In any case, both transfers and interest payments represent a very small portion of spending in the GCC (interest payments averaged 4.8 percent of total spending and rarely exceeded 10 percent of spending).
8The variables are cointegrated, which is why estimation in log level remains appropriate. The VAR results were very similar when using growth rates instead of log levels.
9For Saudi Arabia, the AIC criteria and the LR tests suggest using three lags while the BIC (Schwartz) criteria suggested using two lags only. There were no major differences on the impulse response functions using two or three lags. The three-lag structure was kept identical for all counties.
10i.e., after identifying the shocks by Choleski ordering, with world growth order first, followed by government expenditure and non-oil GDP shocks.
11Note however that the variance that is being decomposed is different when oil prices are added as control variables, since the FEVD is a decomposition of the unexpected/stochastic component of a variable. Also, the FEVD cannot identify oil price shocks since the oil price was included as an exogenous variable (the historical decomposition is more interesting for this purpose—see Figures 5.5 to 5.10 and the related discussion). It is nonetheless clear from simple correlations that oil prices matter directly and indirectly via government spending. The “reduced-form” relationship between oil prices and non-oil growth can be measured by the R2 in the univariate regression of oil prices on non-oil growth, which has been above 40 percent for all countries.

    Other Resources Citing This Publication