The oil cycle and the tax-spend hypothesis: the case of Angola

Abstract

The intertemporal relationship between oil revenues, real government spending and real output over the oil cycle is investigated for the case of Angola, the second largest oil producer in Sub-Saharan Africa. The results of a trivariate VAR, impulse response functions and variance decomposition analysis provide empirical support for the tax-spend hypothesis. In addition, the Angolan economy is found victim of the resource curse, as real output does not respond to aggregate demand shocks such as changes in real government spending. The policy implications for Angola are 2-fold. First, to alleviate the constraints associated with the resource curse, Angola needs to find ways to diversify its economy. Second, given the direction of causality found between government expenditures and oil revenues, in order to avoid fiscal disequilibria, efficient management of oil revenues is as important as controlling public spending.

Acknowledgements

I would like to thank, without implicating, comments and suggestions offered by Jorge T. Araújo, Manoel Neto da Costa, Steve Kyle, Peter Nicholas and Jon Shields on an earlier draft. The findings, interpretations and conclusions expressed in this article are entirely mine and they do not necessarily represent the views of the World Bank Group, its Executive Directors, or the countries they represent and should not be attributed to them.

Notes

¹ Diamonds come at a distant second place. The corresponding numbers for diamonds are: 5.5% of GDP; 8% of total exports and 1.6% of total government revenues (the latter as of 2001).

² Experimentation with other information methods such as BIC and AIC, or a general to specific method (GTS) did not change the conclusions about unit roots.

³ In their original article, Hodrick and Prescott (1997) considered quarterly data as being the standard frequency of time series data and suggested that, for such a frequency, the value of λ should be 1600. It is worth noticing here that a λ of zero, corresponds to an extreme situation in which the time series would be detrended. Conversely, as λ tends to infinity, the HP trend tends toward a deterministic time trend.

⁴ There are certain preconditions that must be observed for a successful implementation of a petroleum revenue fund and an assessment as to whether these are present or not in Angola is beyond the scope of this article. For a thorough discussion of this subject, please refer to Davis et al. (2003) and Katz et al. (2004).

⁵ Additional Granger-causality tests confirmed the existence of unidirectional causality running from oil revenues to government spending, corroborating the support to the tax-spend hypothesis for the case of Angola. The null of nonGranger-causality from oil revenues to real expenditures attracted a p-value of 0.003, while the null of nonGranger-causality from real expenditures to oil revenues could only be rejected with a very high level of significance (p-value of 0.211).

⁶ A result which has also been found for the case of Mexico (Tijerina-Guajardo and Pagán, 2003).

⁷ For an assessment of different ways in which the Dutch disease can operate in the economy, see Hausmann and Rigobon (2003).

⁸ It is reasonable to assume that if government demand is channelled to the domestic economy by means of increased investments in infrastructure and human resources, then the nonoil domestic sector will in the medium until the long run eventually respond, leading to increases in GDP.