Coal Mining and the Resource Curse in the Eastern United States

Abstract

We measure the effect of resource-sector dependence on long-run income growth using the natural experiment of coal mining in 409 Appalachian counties selected for homogeneity. Using a panel data set , we find a one standard deviation increase in resource dependence is associated with 0.5-1 percentage point long-run and a 0.2 percentage point short-run decline in the annual growth rate of per capita personal income. We also measure the extent to which the resource curse operates through disincentives to education, and find significant effects, but this "education channel" explains less than 15 percent of the apparent curse.*The authors thank Seth Wiggins for providing essential GIS expertise in obtaining and analyzing coal quality and surface-mining data.

An investigation of the resource curse in Indonesia

“…To the extent that even some within country studies do find evidence of a resource curse (e.g. Papyrakis and Gerlagh (2007), James and Aadland (2011) and Douglas and Walker (2016) Notes: FE = Fixed Effect; FD = First Difference; Instrument 1 is continous abundance level instruments plus changes of physical production; Instrument 2 is binary abundance level instruments plus changes of physical production;…”

“…Second, most previous empirical studies have tended to use dependence measures. We begin with a measure of mining dependence (which includes coal, oil, gas, and all minerals, including quarrying) similar to what has been employed by within-country papers such as Douglas and Walker (2016) and Papyrakis and Gerlagh (2007) We ultimately elected to exclude 2003 and 2004 because of evidence of some unevenness in the quality of the district level data in these years of political transition and revisions to the planned revenue sharing Law 33/2004 that were implemented in 2005. In short, only by 2005 were both elections and revenue sharing effectively implemented by all districts, and missing observations cease.…”

“…Aside from being widely used in the resource curse literature, FD also provides a "bridging" model for attempts to address potential endogeneity of resource dependence measures using cross sectional instruments. We take a 10 year difference, and retain 2005 as a baseline year to control for initial district GRDP as commonly implemented in growth and resource curse studies (Sachs and Warner, 1995;Douglas and Walker, 2016;Edwards, 2016b). The inclusion of initial GRDP tests for convergence in income between districts as suggested in traditional growth theory (Barro (1991) and Temple (1999)).…”

“…James & Aadland (2011) similarly find negative effects using United States counties. Douglas and Walker (2016) focus on coal dependence among Appalachian counties in the United States using a panel data set over four decades, and also find negative effects on growth. Guo, Zheng, and Song (2016) similarly find negative, albeit weak linkages between resource dependence and output using panel data at the provinvial level in China.…”

“…Edwards employs mining and quarrying's share of total GRDP for each district in Indonesia using cross-sectional data, in 2009. We use this as our first resource dependence measure in Indonesia, for the years 2005 to 2015.However, some within-country studies have instead tried to capture resource dependence as the reliance of governments on resource rents or royalties, particularly in studies of large decentralised systems such as Brazil where central governments transfer revenues from extraction back to local districts based on "producer origin" (Casselli and Michaels (2013),Bjorvatn, et al (2012), Rusli (2014, 2016) andDouglas and Walker (2016).7 We therefore try three additional measures, namely district government dependence on mining revenues (MINREV) from oil, gas and coal, or just from oil and gas (OILGASREV), or just from coal (COALREV). These data are obtained from the Indonesian Ministry of Finance and the Audit Investigation Board (BPK) from 2005 to 2015.…”