Fiscal Rules for Resource Windfall Allocation
The Case of Trinidad and Tobago
Author: Keyra Primus

Contributor Notes

Author’s E-Mail Address: KPrimus@imf.org

Managing resource revenues is a critical policy issue for small open resource-rich countries. This paper uses an open economy dynamic stochastic general equilibrium model to analyze the transmission of resource price shocks and a shock to resource production in the Trinidad and Tobago economy. It also applies alternative fiscal rules to determine the optimal allocation of resource windfalls between spending today and saving in a sovereign wealth fund. The results show that spending all the resource windfall on consumption and investment creates more volatility and amplifies Dutch disease effects, when compared to the case where all the excess revenues are saved. Also, neither a policy of full spending nor full saving of the surplus revenue inflows is optimal if the government is concerned about both household welfare and fiscal stability. In order to minimize deviations from both objectives, the optimal fiscal response suggests that a larger fraction of the resource windfalls should be saved.

Abstract

Managing resource revenues is a critical policy issue for small open resource-rich countries. This paper uses an open economy dynamic stochastic general equilibrium model to analyze the transmission of resource price shocks and a shock to resource production in the Trinidad and Tobago economy. It also applies alternative fiscal rules to determine the optimal allocation of resource windfalls between spending today and saving in a sovereign wealth fund. The results show that spending all the resource windfall on consumption and investment creates more volatility and amplifies Dutch disease effects, when compared to the case where all the excess revenues are saved. Also, neither a policy of full spending nor full saving of the surplus revenue inflows is optimal if the government is concerned about both household welfare and fiscal stability. In order to minimize deviations from both objectives, the optimal fiscal response suggests that a larger fraction of the resource windfalls should be saved.

I. Introduction

A key issue facing policymakers in resource-rich developing countries is the prudent management of natural resource wealth. Resource revenue is difficult to manage due to highly volatile commodity prices and production discoveries. This volatility leads to increased revenue fluctuations and overall macroeconomic instability as it creates boom-bust cycles in natural resource-rich countries (Asik (2013)). Revenue volatility is the key reason why fiscal policy has been pro-cyclical in some resource-abundant countries, such as Trinidad and Tobago (see Artana et al. (2007), International Monetary Fund (2012a) and Céspedes and Velasco (2014)).1 Also, the exhaustibility of non-renewable resources poses uncertainty about future income and complicates fiscal planning. This raises concern about how living standards are maintained once resources are depleted. The exhaustibility and volatility of natural resource revenue therefore pose great challenge to policymakers and raise concern about how much of the resource wealth to consume or save.

There are different views on the management of natural resource revenues. The Permanent Income Hypothesis (PIH) approach recommends that a resource-rich country should sustain a constant flow of consumption that is equal to the implicit return on the present value of future resource revenue (International Monetary Fund (2012a)). Another approach, the Bird-in-Hand policy, suggests that resource revenue should be used to accumulate financial assets in a sovereign wealth fund, and only the interest accrued from these assets should be spent. Also, it has been argued that because citizens own the resources, the resource rents should be transferred to them in the form of direct transfer programmes or conditional cash transfer schemes (Gelb and Grasmann (2010)).2 Furthermore, Takizawa et al. (2004) examined the Hand-to-Mouth rule, which posits that countries can be better off spending all their resource wealth upfront if the initial capital stock is low. Other studies have noted that resource revenue should be saved in the form of government financial assets, which can then be used to make domestic and international loans (Collier et al. (2010)).

Given the infrastructure gaps and capital scarcity in resource-rich developing countries, saving all the resource windfalls impose severe constraints for these economies. By contrast, spending all the resource windfalls can make these countries more susceptible to boom-bust cycles and create macroeconomic instability. These bring the issue of optimal fiscal management of resource windfalls to the fore. Several researchers provide formal discussions on the management of natural resource revenue. Contributions along these lines include Collier et al. (2010), Venables (2010), van der Ploeg (2011), van der Ploeg and Venables (2011, 2013), and van der Ploeg (2012). Some of these studies also address the issue of optimal allocation of resource windfalls, using arbitrary allocation rules to determine how much of the windfall should be saved. One limitation though is because of the nonstochastic nature of the models used in these studies, they are unable to determine the optimal allocation based on measures of volatility. At the same time, existing stochastic models that examine the transmission of resource price shocks focus on combining fiscal and monetary policy to mitigate Dutch disease effects, and the implications of using natural resource revenue for public investment. Therefore, these studies did not examine the critical issue of optimal resource windfall allocation (see Dagher et al. (2012), Berg et al. (2013), Richmond et al. (2013) and Samake et al. (2013)).

Agénor (2016) is the first paper to provide a methodological contribution to the literature on the issue of optimal allocation of resource windfalls in a Dynamic Stochastic General Equilibrium (DSGE) model using a social loss function defined in terms of consumption volatility and fiscal or macroeconomic stability. The Agénor framework incorporates a range of externalities associated with public infrastructure, which include a direct complementary effect with private investment and lower distribution costs, to capture the constraints faced by low-income countries. Additionally, public capital is subject to congestion and absorption constraints. The key insight of Agénor’s analysis is that the optimal allocation rule of resource windfalls involves internalizing a dynamic volatility trade-off: spending less today tends to reduce volatility today in the economy, but the greater the proportion of the windfall that is saved, the greater the proceeds from these assets that governments can spend later on, and the greater the volatility that is injected back in the economy over time. The slope of this trade-off depends in general on the structure of the model and the parameters that characterize the economy, including the accumulation rule for foreign assets. The optimal policy (that is, the optimal share of a resource windfall that must be accumulated today in a sovereign fund) minimizes a social loss function defined earlier.3 Because Agénor’s analysis is fundamentally a methodological contribution, developed with a new oil producer in mind, it is important to apply some of the features of this model to a mature resource producing country.

Although the Trinidad and Tobago economy has been producing oil for over 100 years, it only established an interim sovereign wealth fund in 2000, which was later formalized in 2007. Despite the fact that the sovereign wealth fund specifies rules regarding deposits into the fund, these guidelines were not based on any rigorous framework but rather on adhoc rules which may not have taken specific issues such as household welfare and fiscal stability into account. Thus, a key issue facing policymakers in Trinidad and Tobago is how to determine the optimal allocation of resource windfalls between spending today and saving in the sovereign wealth fund, so welfare can be improved and at the same time there can be a lasting impact on development (Velculescu and Rizavi (2005); Williams (2013)). The aim of this study is to examine the transmission of energy price and production shocks, and to determine the optimal allocation of resource windfalls between spending and saving. To do so this paper applies a modified version of the model developed in Agénor (2016) to the Trinidad and Tobago economy. The contribution of this research is that it is the first country application of the Agénor framework. This paper is also the first attempt to provide a rigorous assessment of how much of the resource windfall should be used for consumption and savings in a general equilibrium framework which takes some of the features of the Trinidad and Tobago economy into account.

This study departs from Agénor (2016) in the following ways: distribution costs are excluded because of the low cost of transport fuel in Trinidad and Tobago; there is no complementary effect with private investment; the model accounts for domestic consumption of natural resource products; the framework includes imperfect capital mobility; and the overall primary balance to output ratio (rather than the nonresource primary balance to output ratio) is the key fiscal indicator. Further, owing to the fact that Trinidad and Tobago is a country with absorptive capacity concern, public capital is subject to absorption constraints which affects investment efficiency. The results show that spending all the resource windfall on consumption and investment creates a lot of volatility, whereas saving all the windfall reduces volatility and mitigates Dutch disease effects initially, but increases volatility later, as interest income is spent. As noted earlier, this dynamic volatility trade-off is the key insight of the analysis in Agénor (2016). Moreover, if the government is equally concerned about household welfare and fiscal stability, the optimal rule suggests that the government should save about 80 percent of the excess resource revenues. In general, the greater the concern for fiscal stability, the larger the proportion of the surplus resource revenue that should be saved. These findings provide evidence that fiscal policy can help to reduce the effects of resource price and production shocks.

The rest of this paper is organized as follows. Section II provides some background information on the natural resource sector in Trinidad and Tobago. Section III presents the model and Section IV outlines the key steady-state and log-linearized equations of the model. Section V provides a discussion of the calibration for the Trinidad and Tobago economy. The dynamic transmission of resource price shocks under alternative fiscal rules is examined in Section VI, while Section VII presents the determination of the optimal allocation of resource windfalls between spending and saving. In Section VIII, sensitivity analysis is provided to test the robustness of the results obtained from the optimal allocation rule. Penultimately, Section IX examines the transmission and optimal allocation of windfalls emanating from a shock to resource production. The final section summarizes the key results and discusses their implications for fiscal policy in Trinidad and Tobago.

II. Background

Trinidad and Tobago is a high-income economy4, endowed with vast energy resources (oil and natural gas). The economy is classified as being “resource-rich”, given the significant share of export earnings and government revenue obtained from oil and natural gas.5 The heavy dependence on the fortunes of the energy sector makes the economy highly vulnerable to energy price shocks. Table 1 shows the economic contribution of the energy sector to the Trinidad and Tobago economy. Since 2000, the economy has become more fiscally dependent on the energy sector—which accounts for over 80 percent of merchandise export earnings. However, although the energy sector is a major source of wealth, it accounts for less than 4 percent of the labour force, because the capital intensive nature of oil and gas industries cannot provide substantial employment opportunities.

Table 1.

Economic Contribution of the Energy Sector (percent, 2000 — 2012)

article image
Note: Government revenues only include taxes and royalties paid by companies in the exploration and production business and the refinery business.Source: Central Bank of Trinidad and Tobago.

Over the last decade, Trinidad and Tobago has benefited from surpluses on its fiscal accounts, supported largely by buoyant energy prices. High energy prices have been accompanied by increased government expenditure, which are likely to be unsustainable if oil and gas prices decrease dramatically. Between 2000 and 2013 for example, government expenditure to GDP increased by over 10 percent. In addition, the higher level of government spending, coupled with lower non-energy revenue, has caused a deterioration in the non-energy fiscal deficit as a ratio of GDP, which deteriorated from 2.4 percent in 2000 to 10.2 percent in 2013. Although there have been some fluctuations in the overall primary balance as a percent of GDP, it recorded surpluses for most of the period, with the exception of 2009 and 2012.

The abundance of oil and gas in the Trinidad and Tobago economy has caused a decline in the non-energy traded goods sector. Figure 1 shows that since 1996, the non-energy tradable sector has been constantly shrinking as a share of GDP. At the same time, the energy sector has expanded, making the economy more resource-dependent and increasing the risks associated with commodity price shocks. Despite the decline in the relative size of the nontradable sector over the period, it still accounts for the largest share of total output. Overall, the characteristics of the production structure provide supporting evidence to Dutch disease effects, which appear to be a permanent feature of the economy given the historical preponderance of oil and natural gas in government revenue and export receipts. The inflow of capital from the oil boom has caused the real exchange rate to appreciate (see Figure 2), which in turn resulted in a loss of international competitiveness in nonresource tradable goods.

Figure 1.
Figure 1.

Production Structure in Trinidad and Tobago

Citation: IMF Working Papers 2016, 188; 10.5089/9781475536775.001.A001

Source: Central Statistical Office of Trinidad and Tobago.
Figure 2.
Figure 2.

Trinidad and Tobago’s Real Effective Exchange Rate (2005=100; increase in index represents an appreciation)

Citation: IMF Working Papers 2016, 188; 10.5089/9781475536775.001.A001

Source: International Monetary Fund Database.

In natural resource-rich developing countries and developing countries in general, public resources are often wasted. Trinidad and Tobago, like many resource-rich developing countries, has a low efficiency of public investment. Dabla-Norris et al. (2012) estimated the efficiency of public investment to be 0.275, which indicates that more than 70 percent of investment spending is unproductive. Moreover, despite the abundant resource wealth, infrastructure in the economy is inadequate and poor when compared to other high-income countries (Artana et al. (2007)). The poor infrastructure facilities in the economy are primarily due to underinvestment. Having realized this, the government has recently been increasing investment in infrastructure through the Public Sector Investment Programme (PSIP) in an attempt to improve the level of infrastructure. However, governance reforms are also critical, in light of the poor efficiency of investment spending alluded to earlier.

The government of Trinidad and Tobago formally established a sovereign wealth fund—The Heritage and Stabilisation Fund (HSF)—in March 2007.6 The purposes of the Fund include to sustain public expenditure capacity during periods of revenue downturn, caused by a fall in crude oil or natural gas prices and to provide savings for future generations. According to the deposit rules for the HSF, quarterly deposits are made to the fund when actual petroleum revenues in each quarter of the financial year exceed the estimated petroleum revenues for that quarter by more than 10 percent and when actual revenues exceed estimated revenues by less than 10 percent.7 Furthermore, a minimum of 60 percent of the excess total revenues shall be deposited to the Fund in any financial year. Withdrawals are permitted from the Fund in cases where the petroleum revenues collected in any financial year fall below the estimated petroleum revenues for that financial year by at least 10 percent.8 In 2013 total assets in the HSF were approximately 19 percent of GDP.

A key issue on the agenda of policymakers in Trinidad and Tobago is to determine an appropriate deposit rule for the HSF that is backed by a rigorous framework. This is particularly important given the decline in oil production and prices, as well as the fall in Trinidad and Tobago’s exports of LNG to the U.S.—arising from an increase in U.S. shale gas production. These developments have caused a decline in energy sector revenues. It is therefore important to determine the share of the excess revenue that should be deposited into the Fund, to provide a balance between immediate consumption and savings.9

III. The Model

The framework considered is an open economy general equilibrium model with three production sectors: a nonrenewable resource sector (which represents the oil and natural gas sector and is identified with superscript O), a nonresource tradable sector (identified with superscript T), and a nontradable sector (identified with superscript N). Resource output is a flow endowment that is owned by all citizens, where the government acts as the trustee or custodian for the resources. Some of the resource products are consumed domestically (by households), and the rest are exported. Tradable output and nontradable output are produced competitively. The tradable good can either be consumed or invested, whereas the nontradable good is a pure consumption good.

Households purchase and consume both tradables and nontradables, whereas the government buys the nonresource tradable and nontradable goods and consumes only nontradables. Private investment consists of tradables only, whereas public investment consists of both tradables and nontradables. As is common in developing countries, public capital is subject to absorption constraints, which affect the efficiency of public investment (See Agénor (2010, 2012)). The model also accounts for imperfect intersectoral capital mobility, and both households and the government have imperfect access to world capital markets.

In the model, prices are flexible and the resource price is exogenously determined outside the home country. The world price of a unit of the nonresource tradable good is unity and purchasing power parity (PPP) holds at the wholesale level and retail level for tradable goods.

A. Total Output

Total domestic output, Yt, measured in foreign currency, is given by,

Yt=YtT+zt1YtN+PtOYtO,(1)

where YtT,YtN,YtO denote nonresource tradable output, nontradable output and natural resource output, respectively. PtO is the world resource price and zt1 is the real exchange rate.

B. Tradable Production

Labour, LtT, capital, KtT, and public capital, KtG are used to produce tradable goods. The production function of tradables is given by,

YtT=(LtT)β(KtT)1β(KtG)ωT,(2)

where β ∈ (0,1) and ωT > 0. The first-order conditions for the economy-wide wage rate, wt, and rental rate of capital in the tradable sector, rtK,T, take the standard form,

wt=β(YtTLtT),(3)
rtK,T=(1β)(YtTKtT).(4)

C. Nontradable Production

Nontradable goods are produced using labour, LtN, private capital, KtN, and public capital. The production function is given by,

YtN=(LtN)η(KtN)1η(KtG)ωN,(5)

where η ∈ (0, 1), ωN > 0. The elasticity of output of nontradables with respect to public capital is assumed to be same in both production sectors, so that ωN = ωT. The first-order conditions are,

ztwt=η(YtNLtN),(6)
ztrtK,N=(1η)(YtNKtN),(7)

where rtK,N is the rental rate of capital in the nontradable sector.

D. Resource Production and Prices

In the model, natural resource output follows an exogenous stochastic process:

YtO=(Yt1O)ρYOexp(εtYO),(8)

where ρYO ∈ (0, 1) is the autoregressive coefficient, and ϵtYO a normally distributed random shock with zero mean and a constant variance.

The international resource price, PtO, follows an exogenous process given by

PtO=(Pt1O)ρPOexp(ϵtPO),(9)

where ρPO ∈ (0, 1) is the autoregressive coefficient, and ϵtPO a normally distributed random shock with zero mean and a constant variance.

E. Households

In the first stage, households determine the optimal level of total consumption, and in the second stage, the optimal level of consumption chosen is allocated between spending on tradable goods and nontradable goods. The objective of the representative household is to maximize the following utility function,

Ets=0Λs{(Ct+s)1ς11ς1ηL1+ψ(Lt+s)1+ψ},(10)

where Et is the expectations operator conditional on the information available in period t, and Λ ∈ (0, 1) denotes the discount factor. The term ς represents the intertemporal elasticity of substitution for consumption, whereas ψ is the inverse of Frisch elasticity of labour supply, and ηL > 0 is a preference parameter.

There is imperfect capital mobility across production sectors. The accumulation equation for the stock of private capital is given by,

KtP=(1δP)Kt1P+It1PΓ(KtP,Kt1P),(11)

where ItP is private investment, δP ∈ (0, 1) gives a constant rate of depreciation, and Γ() is a capital adjustment cost function specified as,

Γ(KtP,Kt1P)=0.5κ(KtPKt1P1)2Kt1P,(12)

where κ > 0 measures the magnitude of adjustment costs.

Households own both types of firms but do not earn any profit from them because of perfect competition. Their net income consists of after-tax nonresource income and after-tax resource income. The households’ end-of-period budget constraint is given by,

Dt+1P=(1+rtW)DtP(1τNO)(YtT+zt1YtN)(13)ψO(1τO)PtOYtO+Ct+ItP+TtL,

where DtP represents foreign-currency debt, rtW is the world interest rate, τNO ∈ (0, 1) denotes the nonresource tax rate, τO ∈ (0, 1) is the resource tax rate, and ψO ∈ (0, 1) is the share of the non-taxed resource windfall that domestic households (as opposed to nonresidents) receive.

Each household maximizes lifetime utility with respect to Ct,Lt,KtP, and DtP. Thus, maximizing (10) subject to (11) to (13) yields the following first-order conditions,

Ctς1=Λ(1+rtW)Et(Ct+1ς1),(14)
Lt=[(1τNO)wtηLCtς1]1ψ,(15)
Et{[κ(Kt+1PKtP1)+1]1[(1τNO)rt+1K+1δP+κ2(Δ(Kt+2P)2(Kt+1P)2)]}=1+rtw,(16)

where Δ(Kt+2P)2=(Kt+2P)2(Kt+1P)2. Equation (14) is the standard Euler equation, (15) defines labour supply, and (16) shows the expected return on capital is related to the world interest rate.

Private consumption is a bundle of tradable consumption, CtT, and nontradable consumption, CtN,

Ct=(CtN)θ(CtT)1θ,(17)

where θ ∈ (0, 1). The representative household maximizes (17) subject to the static budget constraint,

Ct=CtT+zt1CtN.(18)

The first-order conditions are given by,

CtN=θztCt,(19)
CtT=(1θ)Ct.(20)

Tradable consumption consists of a bundle of natural resource products, CtTO, and nonresource related goods CtTNO,

CtT=(CtTO)θT(CtTNO)1θT,(21)

where θT ∈ (0, 1), and the budget constraint for tradable goods is,

CtT=CtTNO+PtOCtTO.(22)

Maximizing (21) subject to (22) the solution is given by,

CtTO=θT(PtO)1CtT,(23)
CtTNO=(1θT)CtT.(24)

F. Government

The government collects resource revenue, TtO, nonresource revenue, TtNO, and lump-sum taxes, TtL. It also receives interest income on the stock of foreign-currency assets, Ft, held in a sovereign wealth fund. The interest rate accrued on the assets in the sovereign wealth fund is rtF. Thus, total government revenue, Tt, is given by,

Tt=TtO+TtNO+TtL+rtFFt.(25)

As noted earlier, resource output is taxed at the rate τO, and the tax rate on nonresource output is τNO. Thus, resource revenue and nonresource revenue collected each period are,

TtO=τOPtOYtO,(26)
TtNO=τNO(YtT+zt1YtN).(27)

Therefore, (25) can be written as,

Tt=τOPtOYtO+τNO(YtT+zt1YtN)+TtL+rtFFt.(28)

Government spending, Gt, is allocated in fixed fractions to investment, ItG, and consumption, CtG,

ItG=vGGt,(29)
CtG=(1vG)ztGt,(30)

where vG ∈ (0, 1). Government spending in foreign-currency terms is,

Gt=ItG+zt1CtG.(31)

In the log-linearized system, where variables are defined as deviations from the steady state, the definition of government spending will depend on the fiscal rule at hand, whereas in the steady state, government spending is calculated as a constant fraction, ψG ∈ (0, 1), of output.

Investment spending is allocated in fixed shares between spending on nontraded goods, ItG,N, and nonresource traded goods, ItG,T:

ItG,N=vG,NztItG,(32)
ItG,T=(1vG,N)ItG,(33)

where vG,N ∈ (0, 1). Thus total public investment, ItG, is given by,

ItG=ItG,T+zt1ItG,N.(34)

The public capital stock is given by,

KtG=(1δG)Kt1G+φt1It1G,(35)

where δG ∈ (0, 1) is the depreciation rate and φt is an indicator of efficiency of spending on infrastructure, as first proposed in Agénor (2010). The efficiency parameter—which captures absorption constraints—is negatively related to the ratio of public investment to public capital,

ϕt=ϕ0(ItGKtG)ϕ1,(36)

where ϕ1 > 0.

The government’s flow budget constraint is,

Dt+1G=(1+rtW)DtG+GtTt,(37)

where DtG is the government’s foreign-currency denominated debt.10

The overall primary balance, OPBt, is defined as,

OPBt=TtO+TtNO+TtLGt.(38)

G. World Interest Rate and Risk Premium

The market cost of foreign borrowing, rtW, depends on the world risk-free (constant) rate, rW,R, and a risk premium, PRt,

rtW=(1+rW,R)(1+PRt)1.(39)

In line with the literature on sovereign debt spreads for developing countries (see Agénor and Montiel (2015)), the premium is positively related to the government net debt to total output ratio,

PRt=(DtGYt)pr1,(40)

where pr1 > 0. Therefore, an increase in total output lowers the risk premium.

H. Market-Clearing Conditions

The market-clearing condition of the nontradable sector is,

YtN=CtN+CtG+ItG,N.(41)

The labour market equilibrium condition is,11

Lt=LtN+LtT.(42)

The CES aggregator for total private capital is given by,

Kt1P=[ζK(KtT)(ηK1)/ηK+(1ζK)(KtN)(ηK1)/ηK]ηK/(ηK1).(43)

The aggregate rental rate of capital is,

rtK=[(ζK)ηK(rtK,T)1ηK+(1ζK)ηK(rtK,N)1ηK]1/(1ηK).(44)

The asset accumulation rule is,

Ft+1=(1φF)Ft+χTtO,(45)

where ϕF ∈ (0, 1) represents a management fee levied on the stock of assets held in the sovereign wealth fund and χ ∈ (0, 1) is the fraction of the resource windfall saved in the sovereign wealth fund.

The current account balance is given by,

Dt+1Ft+1=(1+rtW)DtYtT+CtT+ItP+ItG,T(46)(1+rtF{1ν}φF)Ft[ψO+(1ψO)τO]PtOYtO,

where Dt=DtP+DtG denotes total debt and ν ∈ (0, 1) is the fraction of the management fee that goes to domestic agents.

As in Agénor (2016), the competitive equilibrium in this framework consists of sequences of allocations {CtN,CtT,ItP,Dt,Ft,LtN,LtT,KtP,KtN,KtT,Gt}t=0, final good and factor prices, {wt,rtK,rtK,T,rtK,N}t=0, such that, taking as given K1P,K1G,D1,F1, the exogenous processes {PtO,YtO}t=0, constant policy parameters χ, τO, τNO, vG, and vG,N, and constant public debt,

a) {Ct,CtN,CtT,Lt,ItP,DtP,KtP}t=0 solve households’ optimization problem;

b) {LtN,KtN} solve the nontradable good firm’s optimization problem;

c) {LtT,KtT} solve the nonresource tradable good firm’s optimization problem;

d) the government sets a sequence of total spending {Gt}t=0, its components {CtG,ItG}t=0; a sequence of lump-sum taxes {TtL}t=0; and a sequence of assets {Ft}t=0, held in the sovereign wealth fund so that its flow and lifetime budget constraints are satisfied; and

e) market-clearing conditions for nontradable goods, labour, private capital, and nonresource tradable goods are satisfied.

IV. Steady State and Log-Linearization

This section presents some of the key steady-state and log-linearized equations of the model.

Total consumption in the steady state is,

C=1(1θ)[YTrWDIPIG,T+rFF{1ν}φFF+[ψO+(1ψO)τO]POYO].

The steady-state world interest rate is given by the standard equation,

rW=1Λ1.

The real exchange rate is solved from the equilibrium condition between supply and demand of nontradables,

z=1θC[YNCGIG,N].

In the steady state, the risk premium is given by,

(1+PR)(1+rW,R)=1+rW.

The model is solved by log-linearizing each variable around the steady state. Variables with a hat represent percentage point deviations for interest rate variables from the steady state, and log-deviations around a non-stochastic steady state for the other variables.

Total output in log-linear form is given by,

Y^t=1Y[YTY^tT+z1YN(Y^tNz^t)+POYO(P^tO+Y^tO)].

The resource price and resource production are,

P^tO=ρPOP^t1O+εtPO,
Y^tO=ρYOY^t1O+εtYO.

Total consumption is given by the standard equation,

EtC^t+1=C^t+ςr^tW.

Log-linearizing private investment gives,

I^tP=1IP{DPD^t+1PFF^t+1(1+rW)DP(r^tW+D^tP)+YTY^tTCTC^tTIG,TI^tG,T+(1+rF)F(r^tF+F^t){1ν}φFFF^t+[ψO+(1ψO)τO]POYO(P^tO+Y^tO)}.

The overall primary balance represents total revenues less noninterest government spending,

OPB^t=1OPB[TOT^tO+TNOT^tNO+TLT^tLGG^t].

Public capital is,

K^tG=(1δG)K^t1G+ϕIGKG(ϕ^t1+I^t1G).

Efficiency of public investment depends positively on the public capital stock and is negatively related with public investment,

ϕ^t=ϕ1[K^tGI^tG].

V. Calibration

The model is calibrated using data for Trinidad and Tobago because of the importance of the resource sector to the economy, and the critical need to determine how resource windfalls should be managed—as highlighted in Section II. The main data sources are The Central Bank of Trinidad and Tobago, The Central Statistical Office of Trinidad and Tobago, The Ministry of Energy and Energy Affairs of Trinidad and Tobago, and The Ministry of Finance of Trinidad and Tobago. In cases where data and country-specific parameters are not available, estimates from other studies are used.

A summary of the benchmark set of parameters is provided in Table 2. Considering the parameters characterizing the household behaviour, the intertemporal discount factor, Λ, is set at 0.972 based on estimates of real interest rates.12 The intertemporal elasticity of substitution, ς, is 0.2 (Agénor and Montiel (2015)), and the preference parameter for labour, ηL, is set at a low value of 0.2. The Frisch elasticity of labour supply, ψ, is calibrated at 12, implying an inelastic labour supply. The share of nontradables in total private consumption, θ, is set at 0.55. This is the same value used in Pieschacón (2012), and it is in line with the share of nontradable goods reported in the Household Budget Survey (HBS) for Trinidad and Tobago.13 Using data from the HBS, the share of household spending on oil and gas products in total tradable consumption, θT, is calculated to be 0.06. The adjustment cost parameter for private investment, κ, is set at 30, whereas the depreciation rate for private capital, δP, is 0.045, in line with estimates in the literature. The share of capital in the nonresource tradable sector, ζK, is calibrated at 0.6, to reflect the fact that the nonresource tradable sector is more capital intensive. Furthermore, the elasticity of substitution between nonresource traded and nontraded goods, ηK, is set to 0.5. The share of the nontaxed resource windfall that domestic households receive, ϕO, is set as 88.4 percent, given that the profits repatriated by nonresidents for the period 2007-2010, (1 − ψO), is 11.6 percent.

Table 2.

Calibrated Parameter Values: Benchmark Case

article image

Given that the resource commodities are oil and natural gas, the degree of persistence in resource production is calculated similar to Agénor (2016). Therefore, assuming that proven oil and natural gas reserves may last about 15 years, ρYO is calibrated at 0.912.14,15 For energy prices, the degree of persistence (ρPO) is 0.93, in line with empirical estimates (see Maliszewski (2009)).16 Also, in the nonresource sector, as production in the nontradable sector is more labour intensive, the elasticity of production with respect to labour in that sector, η, is set at a value of 0.65; this is greater than the elasticity of production in the tradable sector, β, which is equal to 0.6. The elasticities with respect to public capital, ωT and ωN, both take a value of 0.17.

Moreover, for the government, revenue from the oil and gas sector represents 17.9 percent of GDP on average for the period 2009-2012. Given that the value of oil and gas production for the same period is 44 percent of GDP, then the tax rate on energy income, τO, is calculated as 41 percent.17 Using data for the period 2009-2012 on the tax revenue-to-GDP ratio, the tax rate on non-energy income, τNO, is calibrated at 14.7 percent. Furthermore, data for the same period reveal that government spending is 13.8 percent of total output; hence this value is used for ψG. Using information from the Ministry of Planning and Sustainable Development (2012), the actual amount the government spent on infrastructure investment during the fiscal year 2012 was used to estimate the initial share of infrastructure investment in government spending, vG, to be 0.151. The parameter that captures the allocation of investment in infrastructure to nontraded goods, vG,N, is set at 0.41 based on the estimate of the share of nontradables in total investment reported in Bems (2008) for Trinidad and Tobago. In addition, the efficiency parameter for public investment, ϕ, is 0.275 based on the value reported in Dabla-Norris (2012) for Trinidad and Tobago, whereas the absorption constraint elasticity for public investment, ϕ1 is set at a low value of 0.05. The rate of depreciation of public capital, δG, is equal to 0.035, in line with Dagher et al. (2012).

To calculate the average interest rate earned by the country’s sovereign fund, rF, data from the Heritage and Stabilisation Fund Quarterly Investment Report were used.18 The nominal interest rate for 2009-2012 was 5.1 percent on average. Given that in the same period, inflation in the U.S. was 1.6 percent on average, the real return on the sovereign wealth fund is set at 3.5 percent. The risk-free world interest rate, rW,R, is computed as 1.0 percent, based on the real yields on U.S. treasury bonds issued in 2014. To calculate the world interest rate, rW, I used the nominal yield on recent sovereign bonds issued by Trinidad and Tobago on the international financial market in December 2013. Given that the bonds had a yield of 4.375 percent, the real bond rate is 2.875—accounting for a 1.5 percent average U.S. inflation rate for 2013. Therefore, from (39), the risk premium (in foreign-currency terms) is calculated as 1.86 percent. The elasticity of the risk premium with respect to the debt to output ratio, pr1, is set at a low value of 0.25. Furthermore, to manage the assets in the Heritage and Stabilisation Fund, a fee—which is set as a fraction of the assets in the Fund—is paid to the external fund managers and to the Central Bank of Trinidad and Tobago, who is the manager of the Fund. In line with the recent data on sovereign wealth funds, the total fee paid for managing the Fund, ϕF, is set at 1.10 percent; whereas, the share of the management fee that goes to residents, ν, is 0.80 percent.

VI. Dynamics of Resource Price Shocks

This section examines the transmission of a positive temporary shock to commodity prices under two “extreme” fiscal rules: the first fiscal rule considers the case where the government spends all the excess revenue from the windfall; and in the second fiscal rule, all the resource windfall is saved in a sovereign wealth fund. The simulations show the percent deviation of the variables from their steady-state values, with the exception of the risk premium and the rental rate of capital, which are expressed in percentage points. In the first fiscal rule, the government spends all the excess revenue from the windfall. This is quite common in many resource-rich countries that have not established a sovereign wealth fund, or any other formal mechanism to manage the proceeds from natural resources. In the second fiscal rule, all the resource windfall is saved in a sovereign wealth fund. Under both rules, it is assumed that public debt is constant and lump-sum taxes adjust to clear the government budget.

A. Full Spending of Resource Windfall

The full spending experiment corresponds to the Hand-to-Mouth policy. This experiment is consistent with the view that governments in developing countries should use natural resource revenue to address their development needs. This is particularly important in capital scare economies that have infrastructure deficits, and poor education and health care services. Hence, under this rule, the government spends all the windfall on consumption and investment, so government spending rises by the amount of the windfall, and there is no asset accumulation in the sovereign wealth fund. Formally,

GG^t=TOT^tO,(47)
F^t=0,(48)

where (48) corresponds to χ = 0 in (45).

Lump-sum taxes are solved residually from the government budget constraint, (37), using (28),

T^tL=1TL[TNOT^tNOTOT^tO(1+rF)F(r^tF+F^t)+FF^t+GG^t+(1+rW)DGr^tW].

Using (47) and (48), and with r^tF=0, lump-sum taxes are given by,

T^tL=1TL[TNOT^tNO+(1+rW)DGr^tW].(49)

Figure 3 shows the general equilibrium effects of a 5 percent temporary increase in resource prices. On impact of the shock, there is a fiscal effect, which causes an immediate increase in government resource revenues, and in turn leads to higher government spending, as well as a rise in public investment. The rise in government spending dominates the increase in resource revenues, thereby reducing the overall primary balance and the nonresource primary balance. Also, on impact of the shock, there is a temporary wealth effect created by higher income to household. The wealth effect causes households to increase total private consumption. The higher level of current consumption increases the demand for leisure and lowers labour supply. Thus, employment falls in the tradable and the nontradable sectors. The expansion in aggregate demand for nontradable goods leads to a real appreciation and causes the product wage in that sector to increase. The nonresource tradable sector shrinks because of the resource movement effect, as well as a result of the real appreciation which reduces the competitiveness of the nonresource tradable goods. Overall, under the full spending experiment, Dutch disease effects are significant. The expansion in demand for nontradable goods increases production of nontradables, as well as nonresource revenues.

Figure 3.
Figure 3.

Full Spending of Resource Windfall

Citation: IMF Working Papers 2016, 188; 10.5089/9781475536775.001.A001

Absolute deviations from baseline, unless otherwise indicated.

Upon impact of the shock, total output increases which in turn reduces the risk premium and the world interest rate. The drop in the interest rate exerts downward pressure on the aggregate rental rate of capital and increases private investment and the total stock of physical capital. The lower interest rate also amplifies the increase in private consumption today, through the intertemporal effect. Initially, there is also a temporary reallocation of capital from the tradable sector to the nontradable sector. This can be attributed to the real appreciation which dampens the effect of the increase in the rental rate of capital in the nontradable sector, bringing about a higher stock of capital in that sector. However, over time the increase in capital in the nontraded goods sector quickly dissipates.

Due to absorption constraints, the higher level of public investment reduces the efficiency of public investment and leads to a marginal increase in the public capital stock.19 The slow rate of accumulation of both private and public capital causes the public-private capital ratio to remain unchanged for a while before falling overtime.

B. Full Saving of Resource Windfall in Sovereign Wealth Fund

The full saving rule corresponds to the Bird-in-Hand policy, which has been discussed in the literature. In this case all the resource revenue is accumulated in a sovereign wealth fund and only the interest income generated from the fund is used to finance government spending on consumption and investment, in proportion of initial spending allocations. A key point to note is that saving from natural resource rents can be used as a stabilization buffer to smooth fluctuations that can emanate from future resource revenue shocks. In this experiment, which corresponds to χ = 1 in (45), government spending is,

G^t=1G[(1+rF)F(r^tF+F^t)FF^t],(50)

and the accumulation rule for the stock of assets is given by,

F^t=1F[(1φF)FF^t1+χTOT^t1O].(51)

The equation for lump-sum taxes, which excludes resource revenues, can be written as,

T^tL=1TL[TNOT^tNO(1+rF)F(r^tF+F^t)(52)+FF^t+GG^t+(1+rW)DGr^tW],

where using (50), lump-sum taxes are also determined by (49).

Figure 4 shows the simulations of a 5 percent temporary shock to resource prices under the full saving rule—compared to the full spending rule. Notably, if all the resource windfall is saved, Dutch disease effects are eliminated, and volatility in the fiscal variables is reduced. In comparison to the full spending rule, under the full saving rule government spending rises slowly, causing public investment to increase at a slower pace and government consumption to fall. The marginal and gradual rise in public investment reduces the absorption pressures; hence the efficiency of public investment falls by substantially less compared to the full spending experiment. The overall primary balance records a surplus in this case, and as a fraction of output the drop in the balance is mitigated. The sovereign fund assets as a fraction of output increases to around 30 percent of output.

Figure 4.
Figure 4.

Full Saving Experiment versus Full Spending Experiment

Citation: IMF Working Papers 2016, 188; 10.5089/9781475536775.001.A001

Absolute deviations from baseline, unless otherwise indicated.

Also, with the full saving rule, the increase in total output—which is less than the full spending case—causes the risk premium to fall, thereby lowering the cost of borrowing abroad, which in turn raises consumption today but by less than the previous case. The positive effect on consumption raises the demand for leisure and lowers labour supply but by less than the full spending case. The overall impact on aggregate demand is mitigated so the appreciation of the exchange rate is less significant. In comparison to the full spending experiment, the rental rate of capital in the tradable sector falls by substantially less before increasing marginally. However, the drop in private capital in the nontradable sector is more substantial under the full saving experiment. The aggregate private capital stock rises because of the higher level of private investment. Overall, the fall in both employment and capital in the nontradable sector causes a contraction in the production of nontradables, which in turn lowers the product wage in the nontradable sector. Also, the drop in the nonresource tradable output is slightly less when all the windfall is saved. The contraction in the production of both nonresource tradables and nontradables lowers the increase in nonresource tax revenues, and mitigates the rise in total output. Given the lower increase in total output, volatility in the risk premium is lower, which in turn reduces fluctuations in the world interest rate and consumption.

VII. Optimal Allocation of Resource Windfalls

An important practical issue for Trinidad and Tobago is how to determine the optimal allocation of the resource windfall between spending on consumption and investment, and saving in a sovereign wealth fund. Because of the volatility of resource revenue flows, it is necessary for some of the windfall to be set aside as a precautionary liquidity buffer. To examine this issue, a partial spending rule is considered whereby a fraction of the resource windfall, χ, is saved—when there is a 5 percent temporary increase in resource prices. Under the partial spending approach, the asset accumulation rule is given by (51), and government spending and lump-sum taxes are adjusted to account for the share of the windfall that should be allocated to spending, 1 − χ. Thus,

G^t=1G[(1χ)TOT^tO+(1+rF)F(r^tF+F^t)FF^t],(53)
T^tL=1TL[TNOT^tNO(1χ)TOT^tO+(1+rW)DGr^tW].(54)

A. Social Loss Function

Using a similar approach to Agénor (2016), to determine the optimal level of resource windfalls that should be saved, χ, the partial spending rule is applied to minimize a social loss function defined as a weighted geometric average of the volatility of private consumption, σCχ, normalized to its steady-state value, CSS, and the volatility of the overall primary balance to output ratio, σOPBYχ, normalized to its steady-state value, OPBYSS.20 The criterion used therefore accounts for both household welfare, which is affected by volatility of private consumption, and fiscal stability. Owing to the fact that in Trinidad and Tobago consumption is highly volatile, an important concern to policymakers is to minimize welfare losses. The overall primary balance is used as the fiscal indicator because Trinidad and Tobago has a long reserve horizon; therefore, the aim is to manage revenue volatility.21,22 The social loss function is given by,

LtS(χ)=(σCχCSS)μ(σOPBYχOPBYSS)1μ,(55)

where μ ∈ (0, 1). The loss is calculated using the asymptotic variances, for μ and χ both varying between 0 and 1 with a grid of 0.1. If the government is mainly concerned about fiscal stability then μ = 0; whereas, if the government sets policy only on the basis of household welfare, μ = 1.

Table 3 presents the results of the social loss function, with the optimal values in red. The results reveal that if the government is mainly concerned about fiscal stability, then all the excess revenue should be saved (χ = 1). In the case where there are equal weights on consumption volatility and fiscal volatility (μ = 0.5), then χ = 0.8 which implies that 80 percent of the resource windfall should be saved. If the government is only concerned about consumption volatility (μ = 1), then χ = 0.6. Contrary to Agénor (2016), these findings indicate that if the government is concerned more about household welfare, a greater fraction of the excess revenue should be spent, as this can help to improve welfare. An analysis of the data for Trinidad and Tobago shows that as the share of revenues from the energy sector increases, the share of social expenditure rises. Similarly, Spatafora and Samake (2012) found that commodity price shocks are associated with a significant increase in social expenditure in commodity-exporting developing countries.

Table 3.

Optimal Allocation of Resource Windfalls under Social Loss Function

article image
Source: Author’s calculations.Notes: C denotes consumption; OPB/Y is the overall primary balance to output ratio; and Rel SD is the Relative Standard Deviation.

B. Alternative Specification of Loss Function

This section extends the social loss function given in (55) to obtain a generalized loss function similar to Agénor (2016). In addition to the volatility of private consumption, the generalized loss function includes a broader measure of macroeconomic volatility defined in terms of a weighted average of the volatility of the overall primary balance to output ratio and the volatility of the real exchange rate, σZχ, scaled to their respective steady-state values. The generalized loss function is given as,23

LtG(χ)=(σCχCSS)μ[(σOPBYχOPBYSS)0.8(σZχZSS)0.2]1μ,(56)

where ZSS represents the steady-state value for the real exchange rate.

Table 4 illustrates the results of the optimal value of χ, using weights of 0.8 and 0.2 on the fiscal indicator and the real exchange rate, respectively, when the generalized loss function is used. These weights were chosen to consider a government that—while being concerned with real exchange rate volatility—remains mainly focused on mitigating fiscal instability as a source of macroeconomic instability. The findings show that the optimal allocation parameter is lower in general. Therefore, if the government is concerned solely about macroeconomic stability, μ = 0, 80 percent of the windfall should be saved, whereas if the main focus is on consumer welfare, μ = 1, 60 percent of the excess revenue should be saved.

Table 4.

Optimal Allocation of Resource Windfalls under Generalized Loss Function

article image
Source: Author’s calculations.Notes: C denotes consumption; OPB/Y is the overall primary balance to output ratio; Z is the real exchange rate; Rel SD is the Relative Standard Deviation.

Figure 5 shows the volatility of consumption, the overall primary balance to output ratio and the real exchange rate, with χ varying between 0 and 1. In the case where χ = 0 (the full spending rule) the overall primary balance to output ratio is highly volatile, but as more of the windfall is saved (as χ tends to 1) fiscal volatility is consistently reduced. In a similar way, consumption is more volatile under the full spending rule, but as the proportion of resource revenue saved in the sovereign fund increases volatility is reduced. However, because the interest income from the assets held in the sovereign fund increases, more resources are injected back into the economy. This leads to higher spending on consumption and investment over time and increases volatility once again. As noted earlier, this is the main insight from Agénor’s (2016) contribution and it explains why consumption volatility takes a convex shape. With regard to the real exchange rate, there is a gradual increase in volatility initially, but as more of the windfall is saved volatility rises because the higher interest income from the assets in the fund increases spending and creates pressure on the exchange rate.

Figure 5.
Figure 5.

Volatility of Consumption, the Real Exchange Rate and the Overall Primary Balance to Output Ratio as a Fraction of the Resource Windfall Saved

Citation: IMF Working Papers 2016, 188; 10.5089/9781475536775.001.A001

Note: χ represents the share of the resource windfall saved in the Sovereign Wealth Fund.

VIII. Sensitivity Analysis

This section tests the robustness of the results for the optimal value calculated for χ using the social loss function in Section VII (A)—which is the benchmark case. To conduct this exercise, I consider changes in some parameter values to assess: a higher degree of capital mobility; less resources to domestic residents via a lower share of the management fee; and tighter absorption constraints. I also examine an alternative specification of the risk premium and an investment-only spending rule.

A. Higher Degree of Capital Mobility

Table 5 shows the optimal values for the social loss function when there is an increase in the elasticity of substitution between KtN and KtT,ηK, from 0.5 to 0.8. If private capital is more mobile across sectors then it is much easier to shift resources between the production of traded and nontraded goods. A higher degree of capital mobility will increase the volatility of a commodity price shock on output and consumption, and should therefore require a higher optimal χ. The results presented in Table 5 show that if μ = 0.5 the optimal value is 0.9 compared to 0.8 in the benchmark case (Table 3). If more emphasis is placed on fiscal stability (μ = 0 or μ = 0.1), the optimal value for χ is 1.0—which is the same value in Table 3. Hence, to better distinguish these results, a smaller grid of 0.01 was done for χ varying between 0.9 and 1.0. The results (which are not reported) show that if μ = 0.1 a higher optimal value of 1.0 is required compared to a value 0.97 under the benchmark case.

Table 5.

Social Loss Function: Higher Degree of Capital Mobility [ηK from 0.5 to 0.8]

article image
Source: Author’s calculations.Notes: C denotes consumption; OPB/Y is the overall primary balance to output ratio; Rel SD is the Relative Standard Deviation.

B. Lower Share of Management Fee to Residents

If residents receive a lower share of the management fee, it means that a larger proportion of the windfall will leave the country to nonresidents. This therefore reduces the wealth effect to household and lowers aggregate demand, thereby reducing on impact volatility in consumption, the real exchange rate and output, so less (given the form of the loss function and the nature of the government’s optimization problem) of the windfall should be saved. The results, which are reported in Table 6, show that as the share of the management fee that goes to residents, ν, is reduced from 0.8 percent to 0.5 percent, the optimal value for χ is lower. For example, if the sole concern is about fiscal stability, the optimal χ is 0.9 compared to 1.0 under the benchmark case (see Table 3). Also, if the government is concerned about consumption volatility and fiscal volatility equally (μ = 0.5), then the optimal value for χ is 0.7 as compared to 0.8. But when the focus shifts more towards household welfare, the results show that more of the windfall should be spent. Thus, if μ = 0.8, the optimal value is 0.4 as compared to 0.7 in Table 3. Intuitively, because households have less income the magnitude of the wealth effect is smaller, so consumption increases by less, which in turn reduces volatility. Because a higher weight is attached to consumption volatility (or household welfare), then the government can afford to spend more and save less.

Table 6.

Social Loss Function: Lower Share of Management Fee to Residents [ν from 0.8 to 0.5]

article image
Source: Author’s calculations.Notes: C denotes consumption; OPB/Y is the overall primary balance to output ratio; Rel SD is the Relative Standard Deviation.

C. Higher Incidence of Absorption Constraints

Due to absorption constraints in developing countries, an increase in public investment causes efficiency to fall. A lower efficiency of public investment should therefore reduce the volatility of the public capital stock, which in turn will reduce fluctuations in macroeconomic variables, and hence require a lower optimal χ. The results from this experiment (which are not reported here to save space) show that when φ1 increases from 0.05 to 0.06 the optimal values are the same as the benchmark case (Table 3). To examine this closer I calculate the optimal values using a finer grid of 0.01 for χ varying between 0.9 and 1.0. The results show that with μ = 0.2, the optimal value is 0.94 compared with 0.95 in the benchmark case, thereby implying a lower optimal value with a higher absorption constraint. However, this effect is not very strong in the present case.

D. Alternative Specification of the Risk Premium

Consider now a different specification of the risk premium, where government debt is scaled by total nonresource output, instead of total output. Thus, equation (40) is now specified as,

PRt=(DtGYtT+zt1YtN)pr1.(57)

Given the inherent volatility and uncertainty of resource revenues, they can be seen as a weakness and may not be considered by markets in determining the premium countries pay on international capital markets. Therefore, by using (57), the effect of the shock on the risk premium and the world interest rate will be mitigated. Thus, volatility in the interest rate will be reduced, which in turn, will lower volatility in total output and consumption—thereby implying a lower optimal χ.

Table 7 shows that under the new specification for the risk premium, the optimal value for χ is reduced. If the government is concerned about fiscal volatility, the optimal value is 0.9 compared to 1.0 in Table 3. Also, as more emphasis is placed on household welfare, the optimal value falls. For example, if μ = 0.9, then 50 percent of the windfall should be saved compared to 70 percent in Table 3.

Table 7.

Social Loss Function: Alternative Specification of t