Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Resources Policy 66 (2020) 101599

Contents lists available at ScienceDirect

Resources Policy
journal homepage: http://www.elsevier.com/locate/resourpol

The economic impact of oil and gas development in the Permian Basin:
Local and spillover effects
Haoying Wang
Department of Business and Technology Management, New Mexico Tech, 801 Leroy Pl, Socorro, NM, 87801, USA

A R T I C L E I N F O A B S T R A C T :

Keywords: The Permian Basin is one of the most prolific oil and natural gas geologic basins in the US. With soaring pro­
Oil production duction in the last decade, the economic impact of oil and gas development in the basin has become a pressing
Natural gas policy question to answer. This paper presents the first basin-wide study that examines both the employment
Spatial spillover
effect and the income effect of the Permian Basin development. We consider not only the local impact but also
Industrial spillover
Permian basin
the spatial spillover effect and the industry-level spillover effect. To correct for the estimation bias due to the
Economic development potential simultaneity between drilling decisions and economic activities, an instrumental variables (IV)
regression model is proposed. We find that both the employment effect and the income effect of shale devel­
opment in the basin are highly significant. We also show that there are significant spatial spillover effects and
spillover effects onto the indirect industries. Relevant policy implications for the long-run economic prosperity in
the Permian Basin are discussed.

1. Introduction investment. According to a recent report by the US Energy Information


Administration (EIA) (2016), the total onshore capital costs per well
The revolution in drilling and hydraulic fracturing has led to large- range from $4.9 million to $8.3 million, which typically covers land
scale development of several major shale formations across the US acquisition, capitalized drilling, completion and facilities costs, lease
since the 2000s. The unprecedented fossil energy development has operating expenses, and gathering, processing and transport costs.
reshaped not only the energy market dynamics but also the regional Normally, investors expect a reasonable return from the investment. In
economies involved. Among these regions, the Permian Basin is leading the case of shale development, due to the high demand for capital and
the energy development race, especially when it comes to oil produc­ labor in a short timeframe, these required inputs often move from
tion. It is one of the most prolific oil and natural gas geologic basins in another region to the region that is going to be developed. When the
the US. The New Mexico portion of the basin also contains the nation’s resources start depleting, those inputs and the profit withdraw back to
largest known concentration of potash reserves (Austin, 1980). Since the the original region or move to another region. A critical policy question
late 2000s, the production in the Permian Basin has soared both in is what has been left after the resources being developed and depleted.
production levels and as percentages of total US production (see Fig. 1). To answer the question, it is important to conduct an economic impact
The drilling activities in the basin have maintained at a high level as analysis. The information can help local governments and organizations
indicated by the rig count in Fig. 1. This rapid development has brought craft policies and sustain the development. The literature suggests that
considerable improvement to the revenue of local governments (Iglesias, the economic impact estimates can show great heterogeneities varying
2018). However, it is unclear how big and how widespread the eco­ by shale formations (Barth, 2013). Therefore, a basin-specific or
nomic impact is in terms of employment and income in the entire region. region-wide economic impact analysis is often necessary.
This question is important partly because the semi-arid landscape makes The economic impact of shale development has been long concerned
the traditional economic activities (e.g., agriculture) in the region in the literature. The first batch of studies that focus on major shale
different (potentially weaker) than other regions. formations in the U.S. came out as industrial reports. The peer-review
The prevailing notion of ‘resource curse’ often adds a question mark literature in energy economics and policy quickly found that most of
to the economic impact of shale oil and gas development. A shale the industrial reports have overestimated the positive economic im­
development process is usually associated with intensive capital pacts. For example, Kinnaman (2011) argues that the economic impact

E-mail address: haoying.wang@nmt.edu.

https://doi.org/10.1016/j.resourpol.2020.101599
Received 4 September 2019; Received in revised form 22 December 2019; Accepted 18 January 2020
Available online 25 January 2020
0301-4207/© 2020 Elsevier Ltd. All rights reserved.
H. Wang Resources Policy 66 (2020) 101599

estimates in these reports are often overstated due to questionable as­ portion and the New Mexico portion of the basin. Previous studies (e.g.
sumptions in methodology. Weber (2012) also suggests that the number Weber, 2012; Lee, 2015) often ignore the New Mexico portion of the
of jobs created by developing the Fayetteville and Marcellus shale gas basin (a four-county area, see Fig. 1). Among the four New Mexico
formations likely has been overestimated in the industrial reports. One counties of active production, Lea and Eddy counties are the most
reason is that the quantitative models (e.g., IMPLAN) that these reports intensively drilled/fracked area in the entire basin. The four-county area
deploy work best when considering modest or marginal changes in produces over 20% of the total basin-wide oil output. IV regression is
economic activities (Kinnaman, 2011). What happens, in reality, is that used to correct for the potential bias due to the simultaneity between
billions of direct investment are injected into the regional economy in as drilling decisions and economic activities. We construct the instru­
quick as a few years, which is almost surely going to change the structure mental variables by taking advantage of the spatial variation of
of the economy. Another possible reason is the lack of consideration of pre-existing oil/gas formations and the temporal variation of techno­
the spillover effects. Increased energy-related employment and income logical change. We find that both the employment effect and the income
can spill over to surrounding regions and other indirect industries effect of shale development are highly significant in the Permian Basin.
through more purchases of intermediate goods and induced local We also show that there are significant spatial spillover effects and
spending (Munasib and Rickman, 2015). The spatial spillovers are also spillover effects onto the indirect industries. Relevant policy implica­
found important in energy transitions (Noseleit, 2018). Often time, little tions for the long-run economic prosperity in the Permian Basin region
is known about these spillover effects. are discussed.
This study contributes to the literature by conducting a basin-wide
economic impact analysis with considerations for both the spatial 2. Literature review
spillover effects and the industry-level spillover effects. We compile a
panel data of 67 counties (from 1998 to 2016) in the Permian Basin and The economic impact of shale development has been well studied in
its periphery to study both the local effect and the spillover effects of the recent literature, mostly on employment and income effects. Wein­
shale development. The shale revolution did not start until around 2010 stein (2014) presented a review of the early literature of shale devel­
in the Permian Basin (Fig. 1). Our longer study period helps to capture opment and its economic impact in the US. Marchand and Weber (2018)
the boom of shale development. The study area covers both the Texas provided a more comprehensive review pooling economic impact

Fig. 1. Rig count, oil and natural gas production in the Permian Basin (2007–2018).
Data source: US EIA. https://www.eia.gov/petroleum/drilling/, accessed Jan 15, 2019.

2
H. Wang Resources Policy 66 (2020) 101599

studies worldwide with a focus on shale gas and oil development. The heterogeneities in the economic impact of shale development, even
economic impact estimates, however, show a great deal of heterogene­ within the same country like the US. For example, there is a major dif­
ity. There are three major reasons for the differences in estimates. First, ference in terms of land ownership between the Permian Basin and the
the economic impact often depends on the geologic conditions of the Marcellus Shale. The Permian Basin has quite some public lands, espe­
given formation. A simple explanation is that geologic conditions affect cially in the New Mexico portion of the basin where the governments are
development costs and hence the profitability of related industries. This open to shale development. In the Marcellus shale, most of the drillings
variation in geologic conditions can be exploited to tackle endogeneity happen on private land. A few studies in the literature have revealed the
issues in empirical regression analysis (Weber, 2012; Fetzer, 2014; regional differences in the economic impact of shale development,
Feyrer et al., 2017). Second, many studies only considered either oil directly or indirectly. For instance, Ashmoore et al. (2016) show that
production or natural gas production, which ignores the interaction there is a significant regional variation in how newspapers cover the
between the extraction and transport processes of two resources. When impacts of shale gas development including the economic, environ­
both resources have significant production, it is important to consider mental, and social impacts. In Weinstein (2014) and Brown (2014),
the impact of oil and gas development together. Third, there is consid­ including state-fixed effects significantly changes estimation results
erable variation in the analytical methodology employed. Two meth­ suggesting regional differences in economic impacts indirectly. McCol­
odological (in)considerations can affect the economic impact estimates lum and Upton Jr. (2018) find significant regional differences in the
substantially: the spillover effects and the simultaneity between pro­ positive impact on local economic conditions using mortgage payments
duction decisions and economic activities. data from shale development regions.
Still, several major findings emerged from the literature. First, the
economic impact tends to be strong in the short run but fades out in the 3. Method and data
long run. Haggerty et al. (2014) is one of the early studies that document
such a pattern. The study shows that the short-run positive economic 3.1. Conceptual framework
impact from oil and gas specialization decreases as a county remains
focused on oil and gas production. The pattern has also been confirmed A common challenge in quantifying the economic impact of oil and
by a few other studies (e.g., Weinstein, 2014; Betz et al., 2015; Komarek, gas development is the misalignment between production activities and
2016; Tsvetkova and Partridge, 2016). Venables (2016) gives two ex­ population distribution. For example, Eddy County in New Mexico is one
planations for the disappearing economic impact in the long run: tech­ of the most productive areas in the Permian Basin, but with only a
nical difficulty in handling resource revenues and lack of governance in population of slightly over 50,000 according to the 2010 Census.1 This
resisting the short-run spending pressures and committing to the suggests that many workers in the oil and gas fields are commuters. A
long-run investment. piece of direct evidence is the increase in demand for hotel rooms and
Second, the employment effect is more pronounced compared to the hence the increase in hotel room rates (Christopherson and Rightor,
income effect. Paredes et al. (2015) show that the employment effect is 2012). One solution is to capture the spatial effects with a spatial
more pronounced and robust compared to the income effect although autoregressive model (e.g. Lee, 2015). However, the economic inter­
both types of effects are small in the Marcellus Shale region. Other pretation of the estimated spatial autoregressive parameter often can be
studies also suggest that in general the employment effect is more pro­ difficult in practical applications (Wang, 2018). Another solution is to
nounced (e.g., Munasib and Rickman, 2015). One possible explanation model the relationship between economic outcome and production ac­
is that the increased income can be distributed through different chan­ tivities at a larger spatial scale (Feyrer et al., 2017). Ideally, the spatial
nels, which makes it difficult to track (Feyrer et al., 2017). Another scale should be large enough to include the commuting zone. At the
reason is that income is more volatile given that it can be easily trans­ same time, such a spatial aggregation process smooths some
ferred to other regions compared to jobs. cross-sectional variations out of the data. Therefore, the size of the
Third, there is a crowding out effect in the local area (e.g., county) as spatial aggregation needs to be carefully chosen. In this study, we follow
the direct industries (e.g., mining and transportation) expand. The the second solution.
crowding out effect happens when previously productive sectors are Another important aspect to consider when designing the model
“crowded out” by industries with higher costs and even reduced framework is the relationship between direct and indirect industries. In
competitiveness (Karl, 2004). For example, Fleming and Measham general, the mining industry and the transportation industry are
(2015) show that a coal seam gas development area in Australia expe­ considered being the direct industries of oil and gas development. Other
rienced a decline in local agricultural jobs. Cosgrove et al. (2015) find a industries such as professional services and real estate are considered
similar crowding out effect by studying the border counties between being indirect industries. Intuitively, we expect employment and salary
New York and Pennsylvania in the Marcellus Shale region. In the income in the direct industries to go up. Employment and salary income
meantime, there is also a positive spillover effect from the direct in­ in the indirect industries may go up (complimentary) or go down
dustries to certain indirect industries. For example, Allcott and Keniston (competitive). For example, Fleming and Measham (2015) showed that
(2018) find evidence against a natural resource curse using a agricultural jobs have decreased following a natural gas development
long-period sample from the U.S. – a county with one standard deviation boom while construction and professional services experienced positive
additional oil and gas endowment enjoyed a one-percent higher real job spillovers. Most of the studies in the literature essentially estimate a
wage between 1969 and 2014. net economic impact on all industries. A few studies distinguish between
Lastly, government policy plays an important role in fully material­ the direct industries and the indirect industries (e.g., Fleming and
izing the potential economic benefits of oil and gas development. De Measham, 2015; Feyrer et al., 2017).
Silva et al. (2016) show that government policy and support are very In this study, we start with a two-way fixed effects model to control
important in developing shale gas formation, especially from the for any county-specific effects and region-wide trends in oil and gas
perspective of lowing the development costs. Effective government development. The benchmark model is set up as the following:
policy can also help to reduce market inefficiency and market uncer­
yit ¼ βXit þ ηPOPit þ δi þ γt þ εit (1)
tainty that may affect the development of energy resources (Brown et al.,
2016). Evans et al. (2019) discussed the Community Benefit Provisions
Program in Wales (UK) suggesting the need for policy intervention when
trying to embed locally the economic benefits from unconventional
shale development. 1
https://www.census.gov/quickfacts/eddycountynewmexico, accessed 20
Differences in policy environments can also help to explain January 2019.

3
H. Wang Resources Policy 66 (2020) 101599

where yit denotes the economic outcome and Xit denotes oil or oil boom did not start until around 2010 in the Permian Basin. The study
equivalent production volume in the given county i and year t. POPit is area includes 67 counties (with seven in New Mexico, as shown in
the total population in the given county i and year t. δi represents the Fig. 2).
time-invariant county fixed effects. γ t represents the time-varying tem­ In the production data, crude oil and natural gas are reported sepa­
poral effects capturing anything unique to a particular year (e.g., an rately. Note that oil production volume includes crude oil and conden­
exogenous shock to the whole regional economy). εit represents the sate; natural gas production volume includes natural gas and casing
unobserved random effects on the economic outcome. Different from head gas. In this study, we consider two specifications of the production
many existing studies where oil and gas production is measured in its volume: oil only and oil þ gas (oil equivalent). In the Permian Basin, oil
market value, we use physical units to measure production output production on average is about twice as much as (oil equivalent) natural
directly in this study. The main reason is that the estimated model can be gas production. Given the potential interactions between oil and natural
readily used for forecasting purposes. In general, shale production vol­ gas at both the production level and the market level, it is reasonable to
ume is much easier to estimate than oil and gas prices hence the value of convert natural gas volume into oil-equivalent barrels (BBLs). This study
the production. The prices are highly volatile even in the short run. uses the standard unit conversion factor recommended by the US EIA:
one MCF natural gas � 0.18 BBL oil.2 The income variables that are
3.2. Empirical strategy measured in dollar amount have been converted to constant dollars (in
2015 USD) using the standard GDP deflator (from the US Bureau of
This study follows the IV regression approach adopted by Fetzer Economic Analysis) throughout the analysis.
(2014) and Feyrer et al. (2017). The estimate of β from the benchmark In the estimation, employment variables are the total employment as
model in (1) is likely biased due to the endogeneity of production vol­ of March of the year. To reflect the lag in economic response, the in­
ume Xit . As Feyrer et al. (2017) point out, the oil and gas production dependent variables ​ Xit and POPit are all one-year lagged. Income
level depends on both the abundance of the resource and the firms variables are the per job payroll income computed as the total payroll
willing to develop it. The industry’s willingness-to-develop is driven by amount divided by the total employment. To examine the spatial spill­
several factors. One of the factors is the cost of development such as the over effects in employment and income, we consider both the local
land leasing cost and severance tax (De Silva et al., 2016). Such factors county economic impacts and the economic impacts within a sur­
can affect local employment and income directly through, for example, rounding zone (excluding the county in the center). The range of the
policy channels, which cause simultaneity between drilling decisions surrounding zone varies from 100 miles to 200 miles. 100-mile is a
and economic outcomes. Therefore, an unbiased estimate of β requires reasonable choice of the commuting zone (Feyrer et al., 2017). We
instrument variables for production volume Xit . Following Fetzer (2014) expect the economic impacts to decline as we move away from the
and Feyrer et al. (2017), we first estimate: county in the center. The economic outcome in the surrounding zone is
computed based on a weighted average of all counties falling into the
lnðXit þ 1Þ ¼ αi þ θZit þ εit (2) zone.3 The weights are chosen based on the standardization of the in­
verse of distance squared (distance from the given county in the zone to
where αi is a dummy variable for each county and Zit consists of a set of
the county in the center). The weights have a non-linear relationship
interaction variables between year dummy variables and a variable with the distance. As mentioned in the conceptual framework subsection
indicating the percentage of county land area inside a major sub-basin.
3.1, such a spatial aggregation more or less smooths out the
This is similar to the dummy variables of shale play-year combination cross-sectional variation in the data. This restricts one from choosing a
used in Feyrer et al. (2017). Note that here we use sub-basins instead of
very large surrounding zone.
shale plays to construct the instrument variables. Fig. 2 plots the extent To see the difference between the direct industries and the indirect
of all major sub-basins in the Permian Basin. In the case of the Permian
industries, we separate the sample into two based on the two-digit
Basin, the extent of sub-basins represents the distribution of oil and gas NAICS (North American Industry Classification System) codes. Similar
resources better than the extent of major shale plays. Based on the es­
to Feyrer et al. (2017), the direct industries include mining and trans­
timates from the model in (2), the set of instrument variables can be portation (21 and 48). The indirect industries include all other in­
recovered through an exponential transformation:
dustries. Based on region-wide averages for the study period (Tables A1
α i þ bθZit Þ
b it ¼ expðb
X 1 (3) and A2), the top five indirect industries by employment growth are
warehousing (49), educational services (61), accommodation and food
The instrument variables constructed above approximate the tem­ services (72), health care and social assistance (62), professional, sci­
poral variations of productivity under a given geologic condition (e.g., entific, and technical services (54). The top five indirect industries by
within a sub-basin). Specifically, here we rely on two sources of varia­ salary income growth are real estate rental and leasing (53), wholesale
tions exogenous to the region: the change of industry productivity over trade (42), construction (23), administrative and support and waste
time and the spatial variation in pre-existing oil/gas formations. To management and remediation services (56), professional, scientific, and
account for the unobserved differences in policy environment (e.g., technical services (54). We also consider all industries pooled together.
taxes, environmental regulations), the standard errors are clustered at Note that what we estimate in this study is the net economic impact on
the state level by year. each category of the industries. We cannot tell the inter-flows between
industries. Still, the estimates can inform us on the spillover effects from
3.3. Data the direct industries to the indirect industries. Table 1 summarizes all of
the key variables in regression analysis. Tables A.1 and A.2 in the Ap­
We compile the data used in this study from four sources: (1) County- pendix show employment and income growth by two-digit NAICS in­
level employment and payroll income data from the US Census County dustries. Both the direct industries and most of the indirect industries
Business Pattern database; (2) Texas county-level oil and natural gas have experienced substantial growth in employment and annual income
production data from the Railroad Commission of Texas; (3) New during the study period. For instance, the employment and income
Mexico oil and natural gas production data provided by the Petroleum
Recovery Research Center at New Mexico Tech; (4) annual county-level
population estimates from the US Census. The associated GIS shapefiles 2
https://www.eia.gov/energyexplained/units-and-calculators/, accessed 12
are obtained from the US Census and the US EIA. The sample period December 2019.
covers 1998–2016, which updates previous studies by including the 3
A county is considered inside the given surrounding zone if its geographic
shale boom period in the Permian Basin. As shown in Fig. 1, the shale center falls into the zone.

4
H. Wang Resources Policy 66 (2020) 101599

Fig. 2. The study area and the extent of major sub-basins in the Permian Basin.
Data source: US EIA.

growth in the mining industries (21) from the 1998–2002 period to the
2012–2016 period were 112% and 38%, respectively.
Table 1
Summary statistics of variables. 4. Results
Variable # of Mean Std. Dev.
OBS Tables 2 and 3 show the benchmark fixed effects model regression
Annual oil production (lagged, million BBLs) 1273 5.84 9.75 results for employment effect and income effect, respectively. The esti­
Annual oil-equivalent production (lagged, 1273 10.76 17.48 mation results correspond to equation (1). Each cell in the table repre­
million BBLs)a sents a separate estimate regressing employment (or income) on the
Annual natural gas production (lagged, million 1273 27.84 52.67
volume of energy production (oil or oil equivalent). Panel A presents the
MCFs)
Employment (all industries) 1257 9588.05 29,410.40 estimates with oil production only, which is supposed to be the domi­
Annual income per job (in $1000, all 1223 33.81 11.73 nating driver of the economic impacts. Panel B presents the estimates
industries) with oil and natural gas (converted to oil equivalent) production
Employment (direct industries) 1168 953.18 2546.61 together. Columns (1)–(3) show the economic impacts on the direct
Annual income per job (in $1000, direct 800 41.33 20.75
industries)
industries, the indirect industries, and all industries from left to right.
Employment (indirect industries) 1254 8826.61 27,894.78 The first row in each panel represents the within county economic im­
Annual income per job (in $1000, indirect 1199 19.90 10.38 pacts (no spatial spillover effects). Each rest of the rows represents the
industries) spatial spillover economic impacts from the county in the center to the
County population estimate (lagged, in 1000) 1273 31.72 97.66
specified surrounding zone excluding the given county in the center.
Sample period (years) 1998–2016 Note that the income effects in Table 3 and throughout the paper are
Total number of counties 67 measured in $1000 and they are converted to constant values in 2015
a
Note: According to the US EIA and the SPE standard: one MCF natural gas � USD.
0.18 BBL oil. Due to the potential endogeneity associated with the production
volume variables, the estimates in these benchmark models are likely
biased. However, we can still make several observations on the general
pattern of economic impacts. First, the local (within-county) economic

5
H. Wang Resources Policy 66 (2020) 101599

Table 2
Employment effect – fixed effects model.
Panel A: Oil only Direct Industries (1) Indirect Industries (2) All Industries (3)

County 69.6465 (12.0600)*** 26.4787 (11.0362)*** 94.3752 (22.5508)***


Zone - 100 miles 33.9845 (9.2054)*** 11.6594 (7.2542) 45.6104 (16.2343)***
Zone - 125 miles 31.3622 (8.2080)*** 11.8711 (6.5293)* 43.2035 (14.5152)***
Zone - 150 miles 28.7552 (7.4613)*** 11.3465 (5.9478)* 40.0675 (13.2163)***
Zone - 175 miles 26.2196 (6.8362)*** 10.6479 (5.3531)** 36.8098 (12.0038)***
Zone - 200 miles 24.4881 (6.5004)*** 9.7116 (5.0018)* 34.1629 (11.3325)***

Panel B: Oil þ Gas

County 34.5171 (6.5616)*** 8.1075 (5.3885) 40.3949 (11.3310)***


Zone - 100 miles 20.0293 (5.1973)*** 5.7223 (4.0643) 25.6564 (9.1024)***
Zone - 125 miles 17.9325 (4.5481)*** 5.7567 (3.5659) 23.6593 (7.9454)***
Zone - 150 miles 16.5999 (4.1743)*** 5.7705 (3.2826)* 22.3377 (7.3279)***
Zone - 175 miles 15.3827 (3.8683)*** 5.5175 (2.9625)* 20.8689 (6.6994)***
Zone - 200 miles 14.5222 (3.7050)*** 5.1153 (2.7970)* 19.6102 (6.3889)***

Note: 1. Throughout the paper, asterisks (*, **, ***) indicate statistical significance at 10%, 5%, and 1% level, respectively, unless otherwise noted. 2. Each cell in the
table represents a separate regression corresponding to equation (1). Population and fixed effects estimates are suppressed to save space, available upon request.

Table 3
Income effect – fixed effects model.
Panel A: Oil only Direct Industries (1) Indirect Industries (2) All Industries (3)

County 0.0264 (0.0560) 0.1451 (0.0340)*** 0.2523 (0.0610)***


Zone - 100 miles 0.0742 (0.0372)** 0.0730 (0.0135)*** 0.2829 (0.0583)***
Zone - 125 miles 0.0873 (0.0335)*** 0.0670 (0.0135)*** 0.2362 (0.0467)***
Zone - 150 miles 0.0846 (0.0304)*** 0.0597 (0.0113)*** 0.2052 (0.0396)***
Zone - 175 miles 0.0731 (0.0270)*** 0.0532 (0.0103)*** 0.1858 (0.0360)***
Zone - 200 miles 0.0606 (0.0247)** 0.0469 (0.0093)*** 0.1687 (0.0330)***

Panel B: Oil þ Gas

County 0.1338 (0.0505)*** 0.0626 (0.0310)** 0.1155 (0.0320)***


Zone - 100 miles 0.0278 (0.0221) 0.0382 (0.0068)*** 0.1526 (0.0335)***
Zone - 125 miles 0.0377 (0.0194)* 0.0336 (0.0065)*** 0.1253 (0.0264)***
Zone - 150 miles 0.0364 (0.0170)** 0.0310 (0.0057)*** 0.1090 (0.0224)***
Zone - 175 miles 0.0329 (0.0151)** 0.0275 (0.0052)*** 0.0987 (0.0201)***
Zone - 200 miles 0.0263 (0.0138)* 0.0242 (0.0047)*** 0.0891 (0.0184)***

Note: 1. Each cell in the table represents a separate regression corresponding to equation (1). Population and fixed effects estimates are suppressed to save space,
available upon request. 2. The unit of estimate here is $1000/million BBLs.

impacts are stronger than the spatial spillover effects as expected. Sec­ insignificant in some specifications. The only cases where we see income
ond, the spatial spillover effects decrease as the distance increases (the effect being insignificant or even negative are the specifications related
size of the surrounding zone gets larger), which again fits the economic to the direct industries. As discussed before, these are likely biased es­
intuition. Third, oil production has a much larger economic impact than timates. These results may be also driven by the inclusion of natural gas
natural gas on a physical oil-equivalent base. The difference is clear by production. Though we cannot compare the magnitudes of the two ef­
comparing Panel A and Panel B in the tables.4 The result holds for both fects, it is reasonable to conclude that overall the oil and gas production
employment and income. This is consistent with the findings in the in the Permian Basin has significant economic impacts in the short run.
literature. As Feyrer et al. (2017) point out, the fact that almost all Note that this study has not much to say about the long-run economic
natural gas is transported by pipeline can explain why oil production has impact. The literature suggests that the economic impact of shale
a much stronger economic impact than natural gas production. Oil often development, in general, diminishes in the long run. However, recent
requires being transported through railway or trucking. Lastly, the studies based on historical data also find evidence against a natural
economic impact is stronger in the direct industries compared to the resource curse in the US suggesting that the indirect industries can
indirect industries. This is particularly true for the employment effect. It benefit from oil and gas development in the long run (e.g., Allcott and
is consistent with findings in the literature and it is expected in a short Keniston, 2018).
run economic analysis like the current study. The county population Tables 4 and 5 show the IV regression results. The estimates are
variable has a significant positive estimate consistently across all of the organized in the same way as in Tables 2 and 3 Note that all IV re­
specifications for the employment effect as expected. For the income gressions within each panel share the same first stage regression given
effect, the estimates of the population variable are mixed. The estimates that the key independent variable in question – the volume of energy
(available upon request) are suppressed to save space. production – stays the same. The IV results suggest that the benchmark
Different from some of the existing studies, we find that both the fixed effects model underestimates the economic impacts substantially.
employment effect and the income effect are highly significant in most This is consistent with the findings in Feyrer et al. (2017). We interpret
of the specifications. The indirect industries’ employment effect is the estimated economic impacts based on the IV regression results.
Tables A3 – A6 in the Appendix report estimation results without control
for the county population.
4
The results with natural gas production only are suppressed to save space, As for the employment effect, the oil only specification suggests a
available upon request. The economic impact of natural gas production has total local effect (within-county, all industries) of 203 jobs (2.12% of the
similar patterns but with smaller magnitudes as expected.

6
H. Wang Resources Policy 66 (2020) 101599

Table 4
Employment effect – instrumental variables regression.
Panel A: Oil only Direct Industries (1) Indirect Industries (2) All Industries (3)

County 131.3663 (10.7751)*** 69.8543 (12.2018)*** 202.6360 (19.7873)***


Zone - 100 miles 74.7552 (3.2862)*** 37.1879 (4.3067)*** 112.3450 (6.6094)***
Zone - 125 miles 66.7021 (2.9623)*** 33.4658 (3.7469)*** 100.5336 (5.8759)***
Zone - 150 miles 59.3495 (2.6501)*** 30.0961 (3.3396)*** 89.7898 (5.2391)***
Zone - 175 miles 53.7577 (2.4728)*** 26.6638 (3.0033)*** 80.7033 (4.8075)***
Zone - 200 miles 50.3698 (2.3520)*** 24.9062 (2.7371)*** 75.5780 (4.4859)***
First stage F-statistic 368.94

Panel B: Oil þ Gas

County 112.8289 (10.1907)*** 53.5092 (11.4570)*** 167.6490 (18.6686)***


Zone - 100 miles 60.7036 (3.2305)*** 26.1251 (4.0788)*** 87.2523 (6.4042)***
Zone - 125 miles 54.6794 (2.9007)*** 23.8239 (3.5499)*** 78.8908 (5.6861)***
Zone - 150 miles 48.9049 (2.5911)*** 22.3174 (3.1608)*** 71.5600 (5.0596)***
Zone - 175 miles 43.8165 (2.4197)*** 19.5754 (2.8432)*** 63.6686 (4.6455)***
Zone - 200 miles 40.5802 (2.3047)*** 18.0062 (2.5936)*** 58.8898 (4.3426)***
First stage F-statistic 414.28

Note: Each cell in the table represents a separate IV regression corresponding to equation (3). Population and fixed effects estimates are suppressed to save space,
available upon request.

basin-wide county average, hereinafter) for every one million BBLs of Among those, the impact on the direct industries ranges from $483/job
new oil production (17.1%). The total employment impact can be to $325/job. The impact on the indirect industries ranges from $180/job
decomposed into a direct industry impact of 131 jobs and an indirect to $117/job. All of the impacts are statistically significant. As expected,
industry impact of 70 jobs. As for the spatial spillover effects, the total the oil þ gas specification gives smaller income impact estimates. The
employment effect ranges from 112 jobs to 76 jobs for every one million total local effect is $384/job (1.14%) for every one million BBLs of new
BBLs of new oil production. Among those, the impact on the direct in­ oil-equivalent production (9.29%). It can be decomposed into a direct
dustries ranges from 75 jobs to 50 jobs. The impact on the indirect in­ industry impact of $462/job and an indirect industry impact of $284/
dustries ranges from 37 jobs to 25 jobs. As expected, the oil þ gas job. Both are highly significant. As for the spatial spillover effects, the
specification gives smaller employment impact estimates. The total local total income effect ranges from $426/job to $242/job for every one
effect is 168 jobs (1.75%) for every one million BBLs of new oil- million BBLs of new oil-equivalent production. Among those, the impact
equivalent production (9.29%). It can be decomposed into a direct in­ on the direct industries ranges from $337/job to $223/job. The impact
dustry impact of 113 jobs and an indirect industry impact of 54 jobs. As on the indirect industries ranges from $154/job to $100/job.
for the spatial spillover effects, the total employment effect ranges from One noticeable result here is the crowding out effect in the indirect
87 jobs to 59 jobs for every one million BBLs of new oil-equivalent industries. The significant differences in income impact estimates be­
production. Among those, the impact on the direct industries ranges tween the direct industries (column (1), Table 5) and the indirect in­
from 61 jobs to 41 jobs. The impact on the indirect industries ranges dustries (column (2), Table 5) suggest a possible talent loss in the
from 26 jobs to 18 jobs. indirect industries. Even though employment in the indirect industries
As for the income effect, the oil only specification suggests a total does increase due to the high demand for intermediate goods and ser­
local effect (within-county, all industries) of $455/job (1.35%) for every vices, workers with better skills and education are likely to switch from
one million BBLs of new oil production (17.1%). The total income the indirect industries to the relatively high-paying direct industries.
impact can be decomposed into a direct industry impact of $504/job and The remaining vacancies and new demand in the indirect industries are
an indirect industry impact of $331/job. Both are highly significant. As then filled in with workers less skilled and of less education. Such local
for the spatial spillover effects, the total income effect ranges from labor market dynamics cannot be observed through aggregate statistics
$448/job to $266/job for every one million BBLs of new oil production. such as employment or average salary income directly, which suggests

Table 5
Income effect – instrumental variables regression.
Panel A: Oil only Direct Industries (1) Indirect Industries (2) All Industries (3)

County 0.5040 (0.1478)*** 0.3309 (0.0380)*** 0.4552 (0.0733)***


Zone - 100 miles 0.4832 (0.0479)*** 0.1801 (0.0142)*** 0.4475 (0.0343)***
Zone - 125 miles 0.4394 (0.0438)*** 0.1641 (0.0126)*** 0.3797 (0.0311)***
Zone - 150 miles 0.3862 (0.0392)*** 0.1438 (0.0115)*** 0.3224 (0.0285)***
Zone - 175 miles 0.3543 (0.0366)*** 0.1302 (0.0105)*** 0.2925 (0.0267)***
Zone - 200 miles 0.3253 (0.0344)*** 0.1170 (0.0098)*** 0.2660 (0.0253)***
First stage F-statistic 368.94

Panel B: Oil þ Gas

County 0.4619 (0.1394)*** 0.2838 (0.0358)*** 0.3837 (0.0688)***


Zone - 100 miles 0.3368 (0.0456)*** 0.1536 (0.0134)*** 0.4260 (0.0320)***
Zone - 125 miles 0.3031 (0.0417)*** 0.1416 (0.0119)*** 0.3493 (0.0292)***
Zone - 150 miles 0.2721 (0.0373)*** 0.1240 (0.0108)*** 0.2943 (0.0267)***
Zone - 175 miles 0.2449 (0.0349)*** 0.1122 (0.0099)*** 0.2671 (0.0251)***
Zone - 200 miles 0.2228 (0.0327)*** 0.1004 (0.0093)*** 0.2420 (0.0237)***
First stage F-statistic 414.28

Note: 1. Each cell in the table represents a separate IV regression corresponding to equation (3). Population and fixed effects estimates are suppressed to save space,
available upon request. 2. The unit of estimate here is $1000/million BBLs.

7
H. Wang Resources Policy 66 (2020) 101599

the importance of examining the spillover effects at the industry level. allocated as stimulus investments for the development of geothermal
Different types of spillover effects from the direct industries onto the energy for agriculture and food production. In this case, the new in­
indirect industries have been identified in the literature (Komarek, dustry only depends on the oil and gas development at its start-up stage.
2016; Feyrer et al., 2017; Allcott and Keniston, 2018). By the time oil and gas resources deplete, the new industry should have
already been well established.
5. Policy discussion Lastly, we should emphasize the importance of choosing a proper
methodology for the economic impact analysis. An overestimated eco­
Similar to other studies, we have shown that the short-run economic nomic impact can lead to wrong choices in project cost-benefit analysis
impact of developing fossil energy in the Permian Basin is significant. and policymaking. The consequences can be many, such as granting too
We find that both the employment effect and the income effect of shale many permits or drilling in a high cost but low yield areas. The whole
development are highly significant in the basin. This study also con­ cost-benefit formula would still depend mainly on firms’ expectations
tributes to the literature by examining two types of spillover effects as about the future energy market. The policy implication is that given a
part of the economic impact: the spatial spillover and the industrial more accurate estimate on the economic impact state and local gov­
spillover. Both spillover effects are found significant. The existence of ernments can regulate the pace of the development properly. Another
the spillover effects makes the question on how to embed the potential related methodological issue is on choosing the right study area. Many of
economic benefits from shale development a much more important the existing studies choose their study areas based on state boundaries.
question. To residents and organizations in the region, this is highly The problem with such a choice is that the spatial distribution of oil and
policy-relevant. Most parts of the Permian Basin have low population gas resources usually does not align with the state boundaries. The
density and a sensitive semi-arid desert ecosystem. Historically, the Permian Basin is a good example of this. There are only 4–5 New Mexico
economic development in the region depends heavily on local natural counties falling into the extent of the basin. However, it happens that
resources and the stability of the ecosystem. It has been well studied that two of these counties (Lea County and Eddy County) are the most pro­
shale development can bring serious impacts on various natural re­ ductive ones in the entire basin. If an economic impact study chooses to
sources essential to the ecosystem (e.g. Vidic et al., 2013; Muehlenbachs focus on the oil and gas development in western Texas only, then the
et al., 2015). Therefore, a key policy implication is to invest revenue conclusion from the analysis is likely going to be biased. Therefore,
raised from shale development in protecting the local natural resources between basin-wide and statewide, we argue that a basin-wide study
and building environmental stewardship including community engage­ area is a more appropriate choice.
ment as proposed by Evans et al. (2019). Only by doing so can we have a
foundation for sustainable development when the fossil energy re­ 6. Concluding remarks
sources start depleting. This also suggests that local governments and
economic development organizations should focus more on place-based It is estimated that the production capacity in the Permian Basin
policies and programs. could reach over five million BBLs/day in 2020 (Tagle, 2018). The USGS
Exploring the difference between employment effect and income (2018) estimates that there are 46.3 billion BBLs oil, 281 billion MCFs
effect provides another angle for transforming short-run economic natural gas, and 20 billion BBLs natural gas liquids in the most pro­
benefits to long-run economic prosperity. As been shown in this study ductive sub-basin – the Delaware Basin (see Fig. 2). Accounting for
and others, the short-run impact on employment is significant across possible technology improvement in extraction and transportation,
almost all industries. However, the impact on per job salary income is these estimates suggest that the production in the Permian Basin could
not. Largely, an increase in the number of jobs represents an improve­ last for another 20–30 years. An important question then becomes how
ment at the extensive margin while an increase in income represents an the economic landscape would look like in the region after that – the
improvement at the intensive margin. A common economic develop­ long-run economic prosperity in the Permian Basin region. Admittedly,
ment challenge faced by regions like the Permian Basin is the slow the long-run economic prosperity is difficult to forecast because it does
growth of the population and the low stock of the labor force. Therefore, not simply depend on resource endowments. Other factors also matter,
both tax policy and economic development policy should focus more on such as human capital, technology advances, and institutions. In this
creating and maintaining jobs, especially in the indirect industries. The study, we seek to derive relevant policy implications for the long-run
jobs created in the direct industries will almost surely be gone after the economic development from a short-run economic impact analysis.
resources are depleted. However, the jobs in the indirect industries can Based on the findings from this study and the literature, we make
sustain and grow in the long run if under a supportive economic three major policy suggestions aiming at inspiring strategies for the
development environment. As Venables (2016) points out, a prevailing long-run economic prosperity in the Permian Basin region. First, at least
issue is that many state and local governments find it difficult to resist some portion of the revenue collected from shale development should be
short-run spending pressures. Thus, a large portion of the revenue from invested in building the environmental and natural resources steward­
oil and gas development may be allocated to short-run consumptive uses ship for the future. Second, governments and economic development
(e.g., health care) and less is left for long-run investment purposes (e.g., organizations should focus on creating jobs in indirect industries. Third,
infrastructure). local governments should use raised revenue to incubate and foster new
In light of findings from this study, the significant industrial spillover industries that are not directly tied to oil and gas development. When it
suggests that the state and local governments should play a strong role in comes to developing a new industry, state governments usually play a
supporting the periphery and indirect industries. As Table 4 shows, major role. It is also worth noting that shale development is not only
about one-third of the increase in jobs goes to the indirect industries. about tax revenue, jobs, and income, but also about other aspects of
Moreover, the ratio of jobs going to the indirect industries stays stable as society. In future research, other aspects such as the environmental
we move away from the central county to the surrounding zones. When impact (e.g., wildlife, ecosystem) and the social impact (e.g., education)
oil and gas development is active, there is no doubt that the indirect should also be concerned.
industries can do well. The question is can the prosperity of the indirect
industries sustain. A relevant policy implication here is that more sup­ Funding
port should be given to industries that are not directly tied to oil and gas
development but tied closely to other local or regional resources. For This research did not receive any specific grant from funding
example, geothermal energy is another important resource in New agencies in the public, commercial, or not-for-profit sectors.
Mexico and it is not directly related to oil and gas development. A
portion of the state revenue collected from oil and gas production can be

8
H. Wang Resources Policy 66 (2020) 101599

Declaration of competing interest Fetzer, T., 2014. Fracking Growth. London School of Economics and Political Science.
CEP Discussion Paper 1278.
Feyrer, J., Mansur, E.T., Sacerdote, B., 2017. Geographic dispersion of economic shocks:
None. evidence from the fracking revolution. Am. Econ. Rev. 107 (4), 1313–1334.
Fleming, D.A., Measham, T.G., 2015. Local economic impacts of an unconventional
energy boom: the coal seam gas industry in Australia. Aust. J. Agric. Resour. Econ.
CRediT authorship contribution statement 59 (1), 78–94.
Haggerty, J., Gude, P.H., Delorey, M., Rasker, R., 2014. Long-term effects of income
Haoying Wang: Conceptualization, Methodology, Data curation, specialization in oil and gas extraction: the US West, 1980–2011. Energy Econ. 45,
186–195.
Software, Formal analysis, Visualization, Writing - original draft, Iglesias, D., 2018. New Mexico Cost Burden on the Oil and Gas Extraction Industry. New
Writing - review & editing. Mexico Legislative Finance Committee Hearing Brief. June 4, 2018. https://www.
nmlegis.gov/handouts/ALFC%20060418%20Item%203%20LFC%20Hearing%20Br
ief_Effective%20Tax%20Rates%20on%20Oil%20and%20Gas%20Industry.pdf.
Acknowledgement: (Accessed 10 December 2019).
Karl, T.L., 2004. Oil-led development: social, political, and economic consequences.
The author would like to thank Ms. Martha Cather of the Petroleum Encycl. Energy 4, 661–672.
Kinnaman, T.C., 2011. The economic impact of shale gas extraction: a review of existing
Recovery Research Center at New Mexico Tech for her generous data studies. Ecol. Econ. 70 (7), 1243–1249.
support. Komarek, T.M., 2016. Labor market dynamics and the unconventional natural gas boom:
evidence from the Marcellus region. Resour. Energy Econ. 45, 1–17.
Lee, J., 2015. The regional economic impact of oil and gas extraction in Texas. Energy
Appendix A. Supplementary data Pol. 87, 60–71.
Marchand, J., Weber, J., 2018. Local labor markets and natural resources: a synthesis of
the literature. J. Econ. Surv. 32 (2), 469–490.
Supplementary data to this article can be found online at https://doi. McCollum, M., Upton Jr., G.B., 2018. Local labor market shocks and residential mortgage
org/10.1016/j.resourpol.2020.101599. payments: evidence from shale oil and gas booms. Resour. Energy Econ. 53,
162–197.
Muehlenbachs, L., Spiller, E., Timmins, C., 2015. The housing market impacts of shale
References gas development. Am. Econ. Rev. 105 (12), 3633–3659.
Munasib, A., Rickman, D.S., 2015. Regional economic impacts of the shale gas and tight
Allcott, H., Keniston, D., 2018. Dutch disease or agglomeration? The local economic oil boom: a synthetic control analysis. Reg. Sci. Urban Econ. 50, 1–17.
effects of natural resource booms in modern America. Rev. Econ. Stud. 85 (2), Noseleit, F., 2018. Renewable energy innovations and sustainability transition: how
695–731. relevant are spatial spillovers? J. Reg. Sci. 58 (1), 259–275.
Ashmoore, O., Evensen, D., Clarke, C., Krakower, J., Simon, J., 2016. Regional Paredes, D., Komarek, T., Loveridge, S., 2015. Income and employment effects of shale
newspaper coverage of shale gas development across Ohio, New York, and gas extraction windfalls: evidence from the Marcellus region. Energy Econ. 47,
Pennsylvania: similarities, differences, and lessons. Energy Res. Soc. Sci. 11, 112–120.
119–132. Tagle, R., 2018. Permian Oil and Gas Takeaway Capacity Improvements on Horizon. htt
Austin, G.S., 1980. Potash in New Mexico. N. M. Geol. 7–9, 1980 (February). ps://info.drillinginfo.com/permian-oil-and-gas-takeaway-capacity-improvement
Barth, J.M., 2013. The economic impact of shale gas development on state and local s-on-horizon/. (Accessed 10 January 2019).
economies: benefits, costs, and uncertainties. New Solut.: J. Environ. Occup. Health Tsvetkova, A., Partridge, M.D., 2016. Economics of modern energy boomtowns: do oil
Pol. 23 (1), 85–101. and gas shocks differ from shocks in the rest of the economy? Energy Econ. 59,
Betz, M.R., Partridge, M.D., Farren, M., Lobao, L., 2015. Coal mining, economic 81–95.
development, and the natural resources curse. Energy Econ. 50, 105–116. US EIA, 2016. Trends in U.S. Oil and Natural Gas Upstream Costs. Independent Statistics
Brown, J.P., 2014. Production of natural gas from shale in local economies: a resource & Analysis. US Energy Information Administration, Washington, DC. March 2016.
blessing or curse? Econ. Rev. Fed. Reserv. Bank Kans. City 2014 (1), 119–147. USGS, 2018. USGS Announces Largest Continuous Oil Assessment in Texas and New
Brown, J.P., Fitzgerald, T., Weber, J.G., 2016. Capturing rents from natural resource Mexico. https://www.usgs.gov/news/usgs-announces-largest-continuous-oil-assess
abundance: private royalties from US onshore oil & gas production. Resour. Energy ment-texas-and-new-mexico. (Accessed 10 January 2019).
Econ. 46, 23–38. Venables, A.J., 2016. Using natural resources for development: why has it proven so
Christopherson, S., Rightor, N., 2012. How shale gas extraction affects drilling localities: difficult? J. Econ. Perspect. 30 (1), 161–184.
lessons for regional and city policy makers. J. Town & City Manag. 2 (4), 1–20. Vidic, R.D., Brantley, S.L., Vandenbossche, J.M., Yoxtheimer, D., Abad, J.D., 2013.
Cosgrove, B.M., LaFave, D.R., Dissanayake, S.T., Donihue, M.R., 2015. The economic Impact of shale gas development on regional water quality. Science 340 (6134),
impact of shale gas development: a natural experiment along the New York/ 1235009.
Pennsylvania border. Agric. Resour. Econ. Rev. 44 (2), 20–39. Wang, H., 2018. The spatial structure of farmland values: a semiparametric approach.
De Silva, P.N.K., Simons, S.J.R., Stevens, P., 2016. Economic impact analysis of natural Agric. Resour. Econ. Rev. 47 (3), 568–591.
gas development and the policy implications. Energy Pol. 88, 639–651. Weber, J.G., 2012. The effects of a natural gas boom on employment and income in
Evans, N., Jones, C., Munday, M., Song, M., 2019. Economic effects in the UK periphery Colorado, Texas, and Wyoming. Energy Econ. 34 (5), 1580–1588.
from unconventional gas development: evidence from Wales. Energy 166, Weinstein, A., 2014. Local labor market restructuring in the shale boom. J. Reg. Anal.
1037–1046. Pol. 44 (1), 71–92.

You might also like