Doppelhofer, G., Miller, R. I., & Sala-i-Martin, X. (2000). Determinants of Long-Term Growth: A Bayesian Averaging of Classical Eestimates (BACE) Approach. NBER Working Paper Series.
Gylfason, T., & Zoega, G. (2001). Natural Resources and Economic Growth: The Role of Investment. CEPR Discussion Paper No. 2743.
Lederman, D., & Maloney, W. F. (2007). Natural Resources: Neither Curse nor Destiny. Washington, DC: World Bank and Stanford University Press.
Ossowski, R., Villafuerte, M., Medas, P., & Tomas, T. (2008). Managing the Oil Revenue Boom: The Role of Fiscal Institutions. IMF Occasional Paper 260.
Sunley, E. M., Baunsgaard, T., & Simard, D. (2002). Revenue from the Oil and Gas Sector: Issues and Country Experience. Fiscal Policy Formulation and Implementation in Oil-Producing Countries.
The dataset was adjusted in a number of ways. The following are treated as data errors and excluded from the estimation sample: negative values for nominal GDP, GDP at constant prices, government consumption, exports, imports, population, employment, exchange rate, and terms of trade; and absolute values greater than 100 percent for dependency ratio, population growth, and for fiscal balance, government spending, current account balance and trade balance as a percent of GDP. Trade weight data are replaced by the data reported by country authorities whenever there is a large discrepancy.
When equations (1) and (2) are solved for six different levels of Y*, from $500 to $3,000, with increments of 500, both the static and the dynamic models’ error variances indicated that the optimal model is for a Y* value around $2,000. It should also be noted that the error variance across these models is quite small, particularly for values of Y* of $1,500 and above. Due to space limitations only the estimation results of Y* = 2000 is reported in this table.
However, one should note the caveats of using averaged data rather than annual data. First, the start date and duration of business cycles across countries may not overlap with the start dates of the 9-year averages. Second, 9-year averaging excludes many countries that have few time series data and eliminates variations in the dataset which would decrease the efficiency of the econometric estimates.
In the estimations, fiscal balance is lagged in order to breakdown the impact of growth on the cyclical component of the fiscal balance. Nevertheless, there may be some persistence over time. However, on should note that the coefficient on fiscal balance can be asymmetric, i.e. fiscal austerity improves growth but fiscal deficits may not be as detrimental, especially if they are financed by aid.
As shown in Appendix, Correlation across Variables of Interest, institutional indicators are highly and significantly correlated with income at the annual data frequency. In order to reduce the multicollinearity issues, 9-year averages are used for these estimations.
Reduced form of model (5) is obtained through the specification that minimizes the error variance of the regression.