ABSTRACT
This paper analyses if and how oil and gas developments foster in-migration of workers into boomtowns. In particular, we focus on the workers’ human capital, as a way to help local growth. Using a zero-inflated negative binomial model, we find that oil and gas shocks, on average, take three years to significantly impact migration flows into boomtowns. The migration response is heterogeneous with a disproportionately higher positive effect for medium-high human capital workers. The types of human capital gained by rural and sparsely populated boomtowns can have important policy implications for their long-run growth and economic resilience.
ACKNOWLEDGEMENTS
This paper is based on a chapter of Isha Rajbhandari’s doctoral dissertation work titled ‘The Impacts of Oil and Gas Developments on Local Economies in the United States' at the Ohio State University. We thank the editor and reviewers for their constructive feedback.
DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author(s).
Notes
1. For a review of the effects of internal migration, see Faggian et al. (Citation2017).
2. For a review of the literature on determinants of interregional migration trends in the US, see Rajbhandari and Partridge (Citation2018).
3. Although endogeneity in the decision process for the location of oil and gas production, it is much less concerning in our empirical design as discussed below.
4. Appendix: Figure A1 in the online supplemental data includes a map depicting the geographical locations of shale oil and gas plays in the lower 48 states.
5. Vachon (Citation2016) and Wilson (Citation2022) do not account for the impacts of origin and destination characteristics on migration trends, as well as, the effect of shale gas developments on migration of individuals based on their education.
6. Although, the ACS data from 2012 to 2019 is publicly available, the post-2011 dataset uses the 2010 Census Geographic definition to define PUMA and migration PUMAs, which differ from the 2000 definition used in this study. Given that there is no obvious overlap and consistency between the two definitions, we are unable to update the dataset.
7. Although we follow Tsvetkova and Partridge (Citation2016), their study does not account for the impacts on interregional migration patterns.
8. Partridge and Rickman (Citation2003) specifically focus on Colorado, Louisiana, Montana, Oklahoma, Texas, West Virginia and Wyoming based on the energy production of the late 1970s and early 1980s.
9. Based on the definition from IPUMS USA. https://usa.ipums.org/usa-action/variables/PUMA#description_section.
10. PUMAs and migration PUMAs follow different geographical boundary definitions. We use the variable PUMARES2MIG, provided by the IPUMS-USA, to convert the PUMAs into migration PUMAs before calculating the gross migration flows.
11. Using change in housing location to define migration only includes permanent movers and excludes commuters and transient workers who do not change their housing location.
12. Appendix: Figure A2 presents examples of how migration PUMAs are geographically related to counties in different states.
13. County level data was aggregated at the migration PUMA level with appropriate weights. Refer to Appendix: Section A.1 and Table A1.
14. The Compulsory School Attendance laws are state-mandated and require children of certain age to attend school. The maximum age limits under these laws range from 16–18 years old. Based on the information obtained from Home School Legal Defense Association (HSLDA). http://www.hslda.org/docs/nche/Issues/S/State_Compulsory_Attendance.asp
15. One advantage of our approach is that it is consistent with the random utility maximisation approach developed by McFadden (Citation1974), giving it a microeconomic theoretical interpretation. The random utility model leads to the application of the conditional logit model for various destination choices, which can be shown to produce results asymptotically equal to the Poisson approach (see Guimarães et al., Citation2000).
16. The descriptive statistic is displayed in Appendix: Table A2. We also ran a formal over dispersion test, which demonstrate that the data is over dispersed. Likewise, the likelihood ratio test of alpha = 0 (dispersion parameter of count data) values included in corroborate the finding.
17. The descriptive statistics of yearly interregional migration inflows to shale boom and non-boom migration PUMAs, disaggregated by educational groups, are discussed in Appendix: Section A.2 and Table A3.
18. The marginal effects of key variables are displayed in Appendix: Table A4.
19. The result suggests that if the industry mix growth rate increases by one standard deviation in the destination region, the difference in the log of expected number of in-migrants is expected to significantly increase by 0.09.
20. Robustness checks that account for the spatial spillover effects of shale oil and gas development and the impact of unemployment rate on migration flows are discussed in Appendix: Section A.3 and Tables A5, A6, A7 and A8.
21. Based on direct oil and gas employment. It is calculated as: .
22. The direct oil and gas employment used in this study is the sum of employment in the oil and gas extraction (NAICS code: 2111) and support activities for mining (NAICS code: 2131) as used by Weinstein and Partridge (Citation2011), Weinstein (Citation2014) and Tsvetkova and Partridge (Citation2016).
23. Following Bartik (Citation1991), the industry mix growth rate variable is calculated as: Where refers to the employment share of industry S (where S does not include NAICS 2111 or NAICS 2131) in migration PUMA in the beginning of the period and refers to the national employment growth rate in industry S during the period.