The resource curse exorcised: Evidence from a panel of countries

TL;DR: The authors evaluated the impact of major natural resource discoveries since 1950 on GDP per capita using panel fixed-effects estimation and resource discoveries in countries that were not previously resource-rich as a plausibly exogenous source of variation.

About: This article is published in Journal of Development Economics.The article was published on 2015-09-01 and is currently open access. It has received 212 citations till now. The article focuses on the topics: Per capita & Resource curse.

Summary

1 Introduction

• Since the seminal work of Sachs & Warner (1995), a near-consensus has formed supporting the existence of a “resource curse”, the counter-intuitive finding that countries rich in natural resources tend to experience slower growth.
• Another form treats institutions as exogenous to resource wealth, and the interaction between resources and institutions explains the divergent outcomes of resourcerich countries (Robinson et al 2006, Mehlum et al 2006, Bulte et al 2011).
• This method uses a data-driven algorithm to find a weighted combination of control countries that best replicates the pre-treatment behavior of a single treatment country.
• The results on GDP per capita, education and infant mortality are consistent with the average positive effects found with the fixed effects model.
• Michaels & Lei (2011) examine whether giant oil field discoveries (defined as containing 500 million barrels of recoverable reserves) leads to armed conflict.

2 Background of Oil Discovery

• This section provides a brief history of the oil industry, with emphasis on how and when production spread geographically, and what factors drove further exploration.
• One clear consequence of this dynamic was the push into Africa in the 1950s.
• Up to that point drilling at the depths involved in North Sea drilling had never even been attempted, but this discovery fortuitously coincided with a new generation of offshore technology that made it viable.
• To summarize, major oil discoveries in previously non-producing nations have been driven to a great extent by global factors exogenous to any one country, particularly technology advance and enormous growth in global demand (along with, of course, geographic luck of the draw).
• Hence there may be some caveats to the design of this paper, but there does not appear to be an obvious mechanism that would systematically bias results, especially given the country and region-year fixed effects used in all regressions.

3 Empirical design

• Regions are assigned according to World Bank country groups where applicable.
• This is done so that differences in the treatment effect over time are not driven by different compositions of treatment countries identifying each event-time coefficient, an especially important consideration given the small number of treatment countries.
• There are six countries that matched the above definition in terms of hydrocarbon production but are not included in the treatment group.
• One exception to this rule is Nigeria, which saw production drop to nearly zero shortly after the event year as defined above (1965), so in this case I assign the second such year (1969), after which production proceeds to surge upwards.
• This may differ from the event year if a country initially produces a very small amount of oil, but is a good indicator when at least some drilling infrastructure was in place.

4.1 GDP Growth

• Column 1 of Table 3 shows the regression results for the specification in equation (1).
• In Column 2 I run the same specification using Penn World Table 7.0 GDP data.
• Because of less extensive GDP data coverage in Penn World Tables, there are five treatment countries that do not have pre-extraction data (Algeria, Gabon, Libya, Oman, and Yemen), and thus cannot contribute to identifying a treatment effect and are dropped from the sample.
• Due to the diminished treatment group, the event study specification yields especially noisy results, and are suggestive of a downward trend possibly causing spurious differencein-difference estimates.
• 18 While again noting the data limitations of these regressions, it appears that more investment is indeed taking place in treatment countries due to discovery, but the results on non-hydrocarbon GDP suggest that most or all of this 18in the difference-in-difference specification, not shown, the effect is positive but not significant.

4.3 Education and Infant Mortality

• The positive results for growth still leave questions on distributional effects.
• Educational outcomes can also give a sense as to whether governments have reinvested resource revenues into public goods.
• When the authors trim the treatment group to include only non-OECD countries, which mostly come from regions where educational metrics are typically exceptionally low at the beginning of the sample period, there is a statistically significant increase of 0.89 average years of schooling.
• Since infant mortality data does not begin until 1955, I must exclude two countries (Gabon and Algeria) to obtain a balanced panel going back to ten years before exploitation, leaving ten treatment countries.
• The results, shown in Figure 17,21 are mixed, with three countries experiencing positive effects, two experiencing little to no effect, and six experiencing negative effects, with the negative impacts being generally larger in magnitude than the positive ones.

4.4 Democracy

• The last outcome I examine is whether the degree of democracy is affected by resource discovery and extraction.
• Further, three treatment countries (Algeria, Gabon, and Yemen) do not have democracy data prior to the 21Yemen is again excluded as its pre-extraction infant mortality rate is higher than all regional controls.
• The coefficient suggests that a treatment country is 19% less likely to be a democracy post-exploitation than if no discovery had occurred, however the estimate is not statistically significant.
• In this case the estimates are similar but more precise and both estimates are significant at a 5% level.
• This figure shows the unconditional average democracy scores for treatment and control countries through real time, with both treatment and control countries limited to Africa and the Middle East.

5 Conclusion

• This paper takes a novel approach to estimating the impact of natural resources, using modern discoveries, longitudinal data, and more sophisticated empirical methods to provide a more rigorous test of the existence of the resource curse than has been heretofore performed.
• Contrary to the past literature, I find positive growth effects for modern resource discoveries in lesser developed countries, and no effect for highly developed countries.
• As a whole, treated countries failed to invest the windfall into diversification of the economy, and the nonhydrocarbon sector suffered.
• Equatorial Guinea, with one of the highest levels of GDP per capita in the region and yet one of the highest poverty rates, is a stark demonstration of the perils of using GDP per capita as an overall measure of welfare.
• As- 22Botswana was a democracy in 1970, and Yemen was not measured.

Figures

Table 1: Treatment Countries

Figure 3: Oil Price and Event Years

Figure 4: Event Study, GDP per capita

Figure 5: Event Study, GDP per capita, Long Panel

Table 5: Event Study: GDP/capita

Figure 16: Event Study, Infant Mortality rate, Non-OECD Treatments

Figure 14: Synthetic Control Results, Average Years Schooling, non-OECD

Figure 15: Synthetic Control Results, Average Years Schooling, OECD

Table 6: Heterogeneous Treatment effects: Non-OECD Treatments

Figure 11: Event Study, Investment/GDP, Discovery Event Year

Figure 12: Event Study, Average Years Schooling, Non-OECD Treatments

Figure 13: Unconditional Average Years Schooling, Non-OECD countries

Figure 8: Synthetic Control Results, Non-hydrocarbon GDP per capita, non-OECD

Figure 10: Event Study, Investment/GDP

Figure 9: Synthetic Control Results, Non-hydrocarbon GDP per capita, OECD

Table 7: Difference in Difference: Non-hydrocarbon GDP/capita

Table 8: Difference in Difference: Investment Ratio

Table 9: Difference in Difference: Average Years Schooling

Table 2: Initial Characteristics as Predictors of Discovery

Figure 6: Event Study, GDP per capita, Discovery Event Year

Figure 7: Synthetic Control Results, Selected Countries

Frequently asked questions

Q1. What have the authors contributed in "The resource curse exorcised: evidence from a panel of countries" ?

This paper evaluates the impact of major natural resource discoveries since 1950 on GDP per capita and other economic and social indicators. I further test these outcomes with synthetic control analysis, yielding results consistent the fixed-effects model. 

Q2. What is the pressing research question going forward?

Perhaps the most pressing research question going forward is how equitably resource-driven growth is distributed within countries. 

Q3. What is the common form of the institutions-driven resource curse?

One form of the institutions-driven resource curse is that resource discovery subsequently weakens institutions and thus growth (Ross 2001, Leite & Weidmann 2002, Sarr et al 2011). 

Q4. What is the common culprit of the Dutch resource curse?

One commonly cited culprit is the socalled “Dutch disease” (a term coined in 1977 after the natural gas boom in the Netherlands), whereby resource exports increase exchange rates, reducing the competitiveness of exporters in the manufacturing sector (Sachs & Warner 1995, Gylfason, et al 1999, Sala-i-Martin & Subramanian 2003). 

Q5. What is the average effect of excluding Botswana?

As Botswana was one of the fastest-growing countries in the world over the period studied thanks to its diamond industry, the average effect when excluding it is smaller but still economically significant, and statistically significant at a 10% level. 

Q6. Why are there five countries that do not have pre-extraction data?

Because of less extensive GDP data coverage in Penn World Tables, there are five treatment countries that do not have pre-extraction data (Algeria, Gabon, Libya, Oman, and Yemen), and thus cannot contribute to identifying a treatment effect and are dropped from the sample. 

Q7. What was the reason why Africa remained unexplored?

Africa up to that point remained largely unexplored, partly due to remoteness and lack of infrastructure, but also because prospects were thought to be sparse (in another sign that oil prospecting was still a highly imperfect science, shortly prior to the Algerian discovery a prominent professor of geology at Sorbonne announced that he was “so sure that there was no oil in the Sahara that he would happily drink any drops of oil that happened to be found there”). 

Q8. What is the reason why Africa was under-explored?

Africa wasunder-explored entering the post-war period at least partially due to lack of infrastructure, but to the extent that this was a region-wide phenomenon, the region-year fixed effects in this paper’s regressions control for it.