Fiscal sustainability and dollarization: the case of Ecuador

ABSTRACT

This article tries to disentangle the dynamic relationships between fiscal variables and economic activity in a small emerging economy characterized by full dollarization, namely, Ecuador. We find that fiscal policy in Ecuador seems to be sustainable, explained by its policy of debt payment through oil revenues, rather than by a fiscal discipline that dollarization is supposed to encourage. The non-oil tax revenues variable is a purely adjusting variable. This result suggests that in a dollarized country that cannot benefit from the ‘seignorage’ revenues, the reliance on volatile oil revenues and on smoothing tax revenues leaves the economy’s fiscal sustainability vulnerable.

KEYWORDS:

• Latin American countries
• Ecuador
• fiscal policy sustainability
• dollarization
• debt
• cointegrated VAR

JEL CLASSIFICATION:

• C32
• E62
• H60

Acknowledgements

The authors would like to thank Katarina Juselius, Søren Johansen and Helena Chuliá for helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ In the early 1990s, Ecuador introduced various structural reforms that provided a certain degree of macroeconomic stability at least until the middle of that decade. However, a number of endogenous shocks – including, an inefficient fiscal policy and increasing financial dollarization – and exogenous shocks – including the impact of the climate oscillation, el Niño and international oil prices – immersed the country in a period of economic stagnation that saw macroeconomic imbalances increase (Jacome 2004). At the end of the twentieth century, Ecuador experienced one of the most serious crises in the history of the Republic with inflation rates being recorded at 30% per month. The government intervened in the banks and many public deposits were frozen. Internationally, Ecuador’s standing was not good; it was in arrears with its private creditors and bondholders, while the International Monetary Fund, the World Bank and the Inter-American Development Bank withheld important loans that might have supported the Ecuadorian balance of payments. The country was in urgent need of radical measures that would stabilize expectations, avoid acute currency depreciation and hyperinflation, and restore economic and financial activity. At the same time, the government was in urgent need of radical measures that would allow it to escape being overthrown. At its head, President Mahuad faced the challenges of severe social and economic crisis – real GDP fell 7.3%, unemployment rose from 11% to 15% and an active indigenous movement called for political and economic reform. In an attempt to switch the focus from political issues to economic matters, he concluded that the radical solution was dollarization.

² See Alesina and Barro (2001), Berg and Borensztein (2000) and De Nicoló, Honohan, and Ize (2005), and among other authors.

³ See http://www.bce.fin.ec/documentos/PublicacionesNotas/Notas/Dolarizacion/dolarizar.html

⁴ There is a very abundant literature examining the advantages and disadvantages of each exchange rate regime and its optimal choice (see, for example, Calvo and Reinhart (2002), Levy-Yeyati and Sturzenegger (2003), or Reinhart and Rogoff (2004), to name a few).

⁵ In addition, some empirical studies do not find evidence of fiscal discipline in countries with fixed exchange rate regimes (see Goldfajn and Olivares (2000) or Duttagupta and Tolosa (2006)), either.

⁶ This is a very odd result in Latin American countries, that might be understood in the context of a very active management of their public debt by Ecuadorian authorities, since the literature on emerging economies has provided evidence of a lack of responsible credit policies of both governments and financial institutions in these countries, which has fostered important debt crisis in those countries (see Aggarwal (1996), Edwards (1989, 2008) and FDCI (1997) and among others).

⁷ At the peak of a devastating economic crisis, Ecuador was forced to default on its Brady bonds ($6.6 billion of the total debt) in the summer of 1999. The restructuring process, officially implemented in August 2000, resulted in a reduction of close to 40% in the face value of the tendered bonds. After this, Ecuador focussed its fiscal policy on debt reduction. The 2002 Fiscal Responsibility, Stabilization and Transparency Act, created the Stabilization Fund for Social and Productive Investment and Debt Reduction (FEIREP), a special trust fund managed by the Central Bank. The FEIREP funds earmarked 70% for debt-buyback operations; 20% to stabilize oil revenues and for emergency spending, and 10% for education and health spending. The Fund was replaced in 2005 by the Special Account for the Productive and Social Reactivation, Development of Science and Technology and the Fiscal Stabilization (CEREPS). The 70% earmarking to debt reduction was reduced to 35%. The debt-to-GDP ratio fell from 86% by end-2000 to about 34% by end-2006. However, the government’s targeted debt reduction policy caused the revalorization of its international bonds, making the debt buyback even more onerous and sparking President Correa’s debt repudiation rhetoric. In December 2008 the debt-to-GDP ratio fell to around 23%. The public external debt was at its lowest level for over three decades. Nevertheless, Ecuador decided to default again, emphasizing that it was ‘unwilling’ rather than ‘unable’ to pay.

⁸ Ray and Kozameh (2012) and The World Bank (2004) offer more details about these fiscal expansive programs.

⁹ Currently, in the aftermath of the world financial and economic crisis that followed Lehman Brothers collapse in 2008, Ecuador has already reversed (from 2012) the abovementioned downward trend.

¹⁰ Literature is not so scarce on the effects of fiscal policy in other forms of extreme rate regimes such as a Currency Boards, though. See, for instance, Grandes and Reisen (2003) in the case of Argentina, and Vladimirov and Neycheva (2009) in the case of Bulgaria.

¹¹ The World Bank (2004) also emphasize that dollarization has left fiscal policy as the unique device to counteract shocks, and propose a reduction in expenditure entitlements and the consolidation of a reliable oil stabilization fund in order to achieve more flexibility in fiscal policy.

¹² The advantages of this methodology over others to analyse the inter-linkages and feedback effects among variables are clearly presented in Hoover, Johansen, and Juselius (2008).

¹³ El Anshasy and Bradley (2012) find that, in the long run, the higher the oil prices, the larger government spending, while in the short run government expenditure rises less than proportionately to the increase of oil revenues.

¹⁴ Actually, an analogy between oil revenues and aid inflows can be made, since both variables are affected by external shocks; the former depends on price volatility, and the latter on donors’ goodwill.

¹⁵ The initial debt equals the expected present value of future primary surpluses, if and only if discounted future debt converges to zero (Bohn 2005).

¹⁶ Even though Equation (4) does not conceive the possibility of default, it can be applied to our empirical study since our sample does not include the Ecuadorian 1999’s default and, the default of 2008 was due to the ‘unwilling to pay’ instead of the ‘incapacity to pay’ of the Ecuadorian government. Equation (4) would include this situation in the IBC as follows:

$E\left(\sum _{j=0}^{\alpha }\frac{s_{t+j}}{{\left(1+r-g\right)}^j}\right)>0$

In this sense, it is important to note that from the seminal paper by Eaton and Gersovitz (1981), models that include the probability of default also restrict its occurrence to the inability to service debt due to the scarcity of fiscal revenues (see, for instance, Jahjah and Montiel 2003).

¹⁷ Collignon (2012) adopting the fiscal reaction function for European countries ${\mathrm{\Delta }s}_t=\alpha ({def}_t-z_1)+\beta ({debt}_t-z_2)$ relates the deficit and debt ratios with the primary surplus. Z₁ and Z₂ are the target reference values for the deficit and debt ratios, respectively, under the Stability and Growth Pact; α and β are the adjustment speed coefficients by which governments respond to the deviation from the deficit and debt ratio, respectively.

¹⁸ Alvarado, Izquierdo, and Panniza (2004) point out that increasing resource exploitation to pay the debt does not affect sustainability since it is assumed that oil reserves have the same return as the government’s other financial assets and liabilities.

¹⁹ This concept of debt sustainability is presented by Hamilton and Flavin (1986) and Trehan and Walsh (1991). However, Bohn (1998) shows the existence of cointegration between debt, primary balance and other variables non debt determinants of the primary surplus, such as the level of temporary government spending (GVAR) and a business cycle indicator (YVAR).

²⁰ Aggregate real income (Y) grows at the constant rate ρ and government expenditure (G) at the constant rate ϒ. Clearly, ρ = ϒ is the only specification within the constant growth rate setup that allows for a positive and finite steady-state value of G/Y (see Barro 1979).

²¹ This means that all economic agents expect energy revenues not to change over the remaining life of the oil reserves:, ${EN}_t=E_{\hspace{0.17em}}\left[\hspace{0.17em}\right)\underset{0}{\overset{m}{\int }}{EN}_t{e}^{-rt}{dt}_{\hspace{0.17em}}\left|\left(I_t\right]\right)$ where t = m when the country’s energy resources are exhausted, and $I_t$ is the information available at time t, including the state of the economy (see Kia 2008).

²² This assumption, however, is unsustainable based on OPEP’s Annual Statistical Bulletin which states that Ecuador has about 8.24 billion barrels of proven reserves and an exportable trend of 334,000 barrels per day in 2011, i.e., 70% of its production.

²³ In growth models with optimizing households, others authors (see Ramsey (1928) among them) use a similar kind of assumption when considering that the representative, infinite-lived household seeks to maximize their overall utility.

²⁴ Bohn (1998) shows that the feedback obtained from the debt to tax and government spending ratios is statistically significant and economically relevant; Romer and Romer (2010) claim that the effect of a US tax shock on output depends on whether the change in taxes is motivated by the government’s desire to stabilize the debt or not; and Favero and Gavazzi (Favero and Giavazzi 2007) also find that interest rates depend on future monetary policy and the risk premium, both variables being affected by the debt dynamics. Hence, the absence of an effect of fiscal shocks on the long-term interest rates, a frequent outcome in VAR-based research that omits debt level, may be due to a misspecification.

²⁵ Perotti (2002) describes the four approaches to identify fiscal shocks that have been used in the literature: (1) the ‘narrative approach’; (2) the Cholesky ordering; (3) the sign restrictions on the impulse responses rather than the linear restrictions on the contemporaneous relations between reduced form innovations and structural shocks and finally (4) the structural VARs.

²⁶ We exclude data corresponding to the first year after dollarization since the different economic variables were still adjusting to the new exchange rate regime. Detailed charts regarding the evolution of the variables for the pre-dollarization versus the post-dollarization period can be found in ‘Estadísticas Macroeconómicas. Presentación estructural’. Ecuador Central Bank (2013).

²⁷ The oil sector accounts for about 50% of Ecuador’s export earnings and about one-third of all tax revenues. Besides, the inclusion of the oil revenues as an endogenous variable prevents using other variables such as international oil prices, since its high correlation would have entailed multicolinearity problems.

²⁸ See Lütkepohl (2004).

²⁹ The Economic Activity Index, which is calculated on a monthly frequency, involves production variables that show the trend of the economic activity. It is calculated as a quantum production indicator and has the mathematic form of Laspeyres index. It takes into account industries that cover more than 60% of GDP such as banana, coffee, cacao, fishing, oil refiner, manufacture, electricity, construction, transport or financial services. See: http://www.bce.fin.ec/index.php/component/k2/item/313-indice-de-actividad-econ%C3%B3mica-coyuntural-ideac for more details on its calculation.

³⁰ Marí Del Cristo and Gómez-Puig (2013) by means of a VECM analysis suggest that rising rates of imports from trade partners other than the United States and subsequent real effective exchange rate depreciations are causing the pass-through to move away from zero in Ecuador since it is importing inflation from its main trading partners (most of them emerging countries with appreciated currencies).

³¹ The CVAR framework does not require all series to be I(1). All that is required is that they are at most I(1). An I(0) variable in a CVAR model means a cointegrating vector on its own. Using the trace test, we reject non-stationarity for Ecuadorian inflation (which is expected due to the fact that its value is close to zero throughout all the sample period), but we do not reject it for the rest of variables included in our models.

³² These variables are not cyclically adjusted since fiscal variables have been collected on a monthly frequency basis, while GDP is only quarterly or annually reported by the Central Bank of Ecuador. If the intention was to isolate the cyclical effect from fiscal variables, we would had to create monthly GDP by interpolation techniques, with the risk of underestimating or overestimating the business cycle effect. Neither creating quarterly or annually fiscal variables to adjust them is a suitable option because, first, we need the maximum of observations to ensure enough degrees of freedom to carry on diverse hypothesis on the CVAR, and second, we would lose valuable information intrinsic in monthly frequency.

³³ Simulation studies have shown that valid statistical inference is sensitive to the violation of some of the assumptions, including parameter non-constancy, autocorrelated residuals and skewed residuals, while quite robust to others, such as excess kurtosis and residual heteroscedasticity (see Rahbek, Hansen, and Dennis (2002) and Cheung and Lai (1993)).

³⁴ We use the likelihood ratio test procedure $-2lnQ\left(H_k/H_{k+1}\right)=T\left(ln\left|{\hat{\mathrm{\Omega }}}_k\right|-ln\left|{\hat{\mathrm{\Omega }}}_{k+1}\right|\right)$ where $H_k$ is the null hypothesis that the model needs k lags and $H_{k+1}$ is the alternative hypothesis that the VAR needs k + 1 lags. This test is distributed as χ² with p² degrees of freedom. However, we also observe the Akaike, the Schwartz and the Hannan-Quinn information criteria to finally decide the optimal number of lags, as well as the Lagrange Multiplier test to check for left-over residual autocorrelation in each VAR(k).

³⁵ The total external debt ratio was reduced from 106% GDP at the end of 1999 to around 98% in 2000 (see Quispe-Agnoli and Whisler 2006). In June 2009, the Correa government defaulted on $3.2 billion of foreign public debt and then completed a buyback of 91% of the defaulted bonds (see Sandoval and Weisbrot 2009).

³⁶ After dollarization, in spite of the elimination of money creation, during 2000 and 2001 both inflation and active interest rates still remained above international levels, just after these years Ecuador experienced a reduction in interest rates as Figure A4 in Appendix I shows.

³⁷ Two contributing factors to the positive fiscal performance occurred in 2006. First, in April 2006 Congress approved a reform to the Hydrocarbons Act stipulating that when crude prices exceed the level agreed in the contract with each private company, the State will be entitled to 50% of the revenues from those oil exports. Second, in May 2006 the Minister for Energy and mines announced the termination of the contract between State and Occidental Petroleum with important oilfields (Block 15 of the Amazon region and the unified fields of Eden Yuturi and Limoncocha) now being managed by the state-owned enterprise PETROECUADOR (see ECLAC 2006).

³⁸ Our non-normal residuals (from AIR_t, LDEBT_GDP_t) present positive and negative skewness less than 0.14, which is inside the range suggesting a normal population (see Doane and Seward 2011).

³⁹ In order to identify the long run structure we need to impose at least r(r−1) restrictions on $\beta$ vectors.

⁴⁰ This is to be expected since the total debt-to-GDP ratio is unrelated to the path of its fiscal surpluses but to the 1999’s default.

⁴¹ The model’s specification has changed to require only three lags and the transitory dummy dum0609_t. The inclusion of interests in government expenditure (LTGOV_t) does not change the main results.

⁴² These two hypotheses have been used in the previous study by Martins (2010).

⁴³ We also include the estimations of imposing H₁ and H₄ in Table A8. Both hypotheses have been used in Section “Robustness of results” to examine the robustness of results.

⁴⁴ Since we have five variables, the B matrix adds 25 new coefficients. The assumption that u~IN (0, I) implies ((p*(p + 1)/2) = 15) 15 restrictions on B (5 unit coefficients on the diagonal elements and 10 zero restrictions on the off-diagonal elements). Six additional restrictions ((p−r)*r = 6) are necessary to separate transitory from permanent shocks, and four more restrictions are required to achieve a just-identified structural MA model. These four extra restrictions are essential because there are two possible sequences of the transitory shocks and two possible sequences of the permanent shocks. A single specification can be obtained by imposing three exclusion restrictions on the common trends and one exclusion restriction on the transitory impulse responses.

⁴⁵ The first two rows give the combinations which make up the transitory shocks. Note that the second transitory shock is primarily given by shocks to government expenditure. Hence, government expenditure can be regarded as a source of transitory and permanent shock. It adjusts to long run relations as a transitory shock (See Table A8) but also it can be considered a pushing force that put the system out of equilibrium generating shocks with long run impact on the rest of the variables.

⁴⁶ Recall that the first set of impulse response corresponds to the structure of H_3.

⁴⁷ Norway has opted for the real income approach or ‘bird in hand’. The income from the sale of the oil revenues become financial assets and only the interest which emanates from these assets is spent. This spending rule not only is de-linked to oil price volatility but also ensures the use of oil income by future generations.

Additional information

Funding

This work was supported by the by the Government of Spain under grant number ECO2013-4836.