Are Energy Endowed Countries Responsible for Conditional Convergence?

• Author: Oliver, Matthew E.

1. INTRODUCTION

   Economic growth and its underlying causes rank among the most exhaustively researched topics in economics over the past century. It is difficult to overstate the importance of growth, especially for developing countries. A prominent vein within the economic growth literature has been the study of convergence--the idea that poorer countries tend to grow faster than richer ones, implying GDP per capita should converge across countries over time (Sala-i-Martin, 1996; Barro and Sala-i-Martin, 2004).

   This paper's objective is to study economic convergence and fossil fuel (FF) endowments. To this end, we present a simple yet compelling contrast in economic convergence (and non-convergence) patterns within two distinct samples of countries: those with and without significant FF resource endowments. Standard tests indicate ft-and cr-convergence among FF-endowed countries and, by contrast, non-convergence among non-FF-endowed countries. To our knowledge, ours is the first study to explicitly conduct and compare convergence tests for energy resource endowed and non-endowed countries, thus comprising a novel contribution to the literature.

   The striking difference in convergence patterns across these two sets of countries suggests several important implications for the study of economic growth and convergence, particularly with regard to the effects of FF development on growth. There has been considerable debate in the literature as to whether FF resources aid or hinder growth. We show that not only did FF-endowed countries grow faster on average over our sample period than non-endowed countries, but they also appear to have formed a 'convergence club' in which the lower-income members were able to grow considerably faster than--and thus converge with--developed economies. Developing countries without FF endowments are found to have grown slowly over our sample period, failing to converge on average with richer nations or enjoy the economic benefits thereof. Despite this observed pattern, we are careful to point out that our results, while suggestive, cannot be interpreted as a direct causal link between FF endowments and convergence. Moreover, not all FF-endowed countries may have converged; we still observe significant heterogeneity in growth rates among low-income FF-endowed countries. While we do not rule out the possibility that a 'resource curse' may explain some of the observed heterogeneity in growth rates among FF-endowed developing countries, our main convergence results are robust to controlling for resource curse effects. Our findings suggest that, on average, these potentially negative effects on growth do not outweigh the economic benefits of FF endowments.

   A broader implication of our analysis is that as calls grow to rein in fossil fuel use to combat climate change, such calls are likely to be met with resistance by developing countries seeking to exploit valuable FF resources as part of a long-term growth strategy. Countries with FF reserves tend to use them, and have been shown to invest less in the development of non-hydro renewable energy sources (Johnsson et al, 2019). If, as our results suggest, FF endowments--and, by extension, their extraction and utilization--are positively related to growth, leaders of developing economies might understandably be reluctant to forego economic growth in the name of decarbonization.

   The remainder of the paper is structured as follows. After a concise review of the relevant literature on economic convergence and resource curse effects in Section 2, Section 3 describes our country-level data on economic growth and several other key variables, and our criteria for identifying countries as having significant FF endowments. Section 4 lays out our empirical design. Results are presented and discussed in Section 5. Section 6 explores potential mechanisms underlying our main finding and discusses the limitations of our research design. Section 7 concludes by offering a few broad policy implications. An online appendix contains supplementary tables and figures.

2. RELATED LITERATURE

   Johnson and Papageorgiou (2020) provide a comprehensive survey of the convergence literature, in light of which we briefly discuss the most seminal or recent papers here.

   The convergence hypothesis emerges from a standard neoclassical production function with diminishing marginal returns to capital. Initial GDP per capita is predicted to be inversely related to subsequent GDP per capita growth (Solow, 1956). Seminal empirical studies such as Barro and Sala-i-Martin (1992); Mankiw et al. (1992); Sala-i-Martin (1996); and Sala-i-Martin (2006) found evidence of convergence across countries or U.S. states. However, empirical evidence of this predicted relationship between GDP per capita and subsequent growth rates may be elusive when a broad sample of countries is considered (Baumol, 1986; Barro, 1991; Dowrick, 1992). This inconsistency prompted the observation that the neoclassical model does not in fact predict absolute convergence, but instead predicts convergence only after controlling for the determinants of the steady state. After including human capital into a country's production function, Mankiw et al. (1992) found that this conditional convergence occurs at approximately the rate the model predicts. A common criticism of the Mankiw et al. (1992) analysis, however, is that it tests for convergence based on differences between countries' GDP per capita levels and U.S. GDP per capita. More reliable estimates of conditional convergence at various levels (e.g., countries, U.S. states) are available in Barro and Sala-i-Martin (2004), whose basic analytical framework we employ here. (1) According to the 'iron rule of convergence', countries eliminate gaps in real per capita GDP at a rate of around 2% per year, implying that it takes 35 years for half of an initial gap to vanish and 115 years for 90% to disappear (Barro, 2015).

   In linking growth in developing economies to FF endowments, our work also contributes to the literature on the 'resource curse' hypothesis (van der Ploeg, 2011; Venables, 2016), particularly with respect to FF resource dependency. However, despite short-run stimulus from resource booms, (2) long-run dependence on natural resource development may lead to slower economic growth--a phenomenon known as the 'resource curse'. Based on early analyses of cross-country data, this result was considered "a reasonably solid fact" (Sachs and Warner, 2001). Subsequent studies, however, produced mixed empirical evidence, some generally corroborating the resource curse hypothesis (Auty and Mikesell, 1998; Gylfason et al., 1999; Lederman and Maloney, 2007; Papyrakis and Gerlagh, 2007), and others finding mixed or limited evidence (Brunnschweiler and Bulte, 2009; Manzano and Rigobon, 2001). (3) James (2015), using the value of crude oil and gas production relative to GDP as a proxy for resource dependence across a wide sample of countries, suggests the resource curse is instead a "statistical mirage" that reflects only the slow growth of the FF extraction sector. Others have found that oil endowments have had significant positive impacts on GDP (or income) per capita levels, in direct contradiction to the resource curse hypothesis (Alexeev and Conrad, 2009; Michaels, 2010; Smith, 2015). (4) Exploiting exogenous spatial variations in sedimentary basins associated with petroleum deposits, Cassidy (2019) finds that oil production raises GDP per capita without significantly harming non-resource sectors, but that it is associated with reduced levels of democracy, increased corruption, and lower tax revenues as a share of GDP.

3. DATA

   Our sample contains annual data for 127 countries over the interval 1970-2014. Countries were selected based on whether GDP per capita and consumption per capita data were available for the full sample period. Summary statistics are presented in Table 1. Appendix Table 1 provides source information and a brief description of each variable.

   Panel A of Table 1 shows country characteristics. In total, these 127 countries constitute an overwhelming share of the global population. To see this, we take simple five-year averages of each country's population during two periods: 1970-74 and 2010-14, respectively the beginning and end of our 45-year sample. (5) During the 1970-74 period, our sample encompasses over 88% of the global population, increasing to over 90% by 2010-14. (6)

   Countries were identified as FF-endowed based on absolute levels of annual (i) oil and gas reserves, and (ii) coal production. These data were obtained from the 2018 BP Statistical Review of World Energy (BPStat) (BP, 2018). Oil reserves are measured in billions of barrels (Bbl) of proved reserves. Gas reserves are measured in trillion cubic meters (Tcm) of proved reserves. Coal production is measured in millions of tons of oil equivalent (Mtoe). Note, however, that for our main specifications we are not concerned with the size of countries' FF endowments--either in absolute terms or relative to population or to each other--but instead with identifying a given country as having an endowment or not.

   Of the 56 countries in our sample identified as having a FF endowment, 37 have oil, 40 have natural gas, and 26 have coal, where several countries have endowments of two or all three FFs. (7) A list of all FF-endowed countries in our sample is provided in Appendix Table 2. As expected, the geographic dispersion of oil and gas endowments is highly correlated (although not perfectly correlated), whereas the correlations between oil and coal endowments and between gas and coal endowments are lower (Appendix Table 3).

   Next, Panel B of Table 1 shows summary statistics on resource endowments, again using two five-year averages: 1985-89 and 2013-17. Sample countries identified as FF-endowed accounted for roughly 90% of global oil reserves and around 67% of world gas reserves during...