SSRN Id2939603 PDF

Drilling and Debt
Journal of Finance, forthcoming

ERIK P. GILJE,
The Wharton School, University of Pennsylvania and NBER
ELENA LOUTSKINA,
University of Virginia, Darden School
and
DANIEL MURPHY
University of Virginia, Darden School
August 27, 2019
Abstract
This paper documents a previously unrecognized debt-related investment distortion. Using
detailed project-level data for 69 firms in the oil and gas industry, we find that highly levered firms
pull forward investment, completing projects early at the expense of long-run project returns and
project value. This behavior is particularly pronounced prior to debt renegotiations. We test several
channels that could explain this behavior and find evidence consistent with equity holders
sacrificing long-run project returns to enhance collateral values and, by extension, mitigate lending
frictions at debt renegotiations.
The authors would like to thank Steve Baker, David De Angelis, Michael Faulkender, Vincent Glode, Rustom Irani,
Don Keim, Doron Levit, Song Ma, Greg Nini, Krishna Ramaswamy, Julio Riutort, Michael Roberts, Phil Strahan,
Mathieu Taschereau-Dumouchel, Luke Taylor, Hongda Zhong, and seminar participants at Wharton; Dartmouth
(Tuck); University of Virginia (Darden); IU Bloomington; GWU; the 2016 MIT Junior Faculty Conference; 9th annual
FSU SunTrust Beach Conference; 2017 PNC University of Kentucky Finance Conference; 14th Annual Conference
in Financial Economic Research IDC; 2017 Frontiers of Finance Conference; 2017 Financial Intermediation Research
Society Conference; and Finance – UC 12th International Conference for helpful comments. The authors would also
like to thank Julia Tulloh for her valuable research assistance, and the Rodney L. White center for financial support.
Electronic copy available at: https://ssrn.com/abstract=2939603

Understanding how debt affects a firm’s investment decisions is one of the central questions in
finance. Incentive conflicts between debt and equity claimants have the potential to result in
inefficient and value destroying decisions (Jensen and Meckling, 1976). Existing theoretical and
empirical work has largely focused on the size, prevalence, and mitigation of investment
distortions linked with the traditional agency costs of debt such as underinvestment and risk-
shifting. 1
In this paper, we use new detailed project-level data to document a previously
unrecognized debt-related investment distortion. We show that high-leverage firms facing debt
renegotiations pull forward investment and complete projects early at the expense of project net
present value (NPV). These negative project-level investment distortions likely aim to enhance
collateral values, thereby mitigating lending frictions and increasing firms’ access to finance.
Overall, our results suggest that by increasing collateral values, high-leverage firms (dependent
on asset-based loans) mitigate financing frictions arising at debt renegotiations. Our findings,
thus, also highlight a previously unexplored hidden cost of collateral-based financing.
Identifying how debt affects the actions of firms is empirically challenging. First, it is
difficult to observe actions at the project level and identify how these actions affect firms’ cash
flows. Second, even if one can observe detailed actions by managers, assessing whether a
decision is value maximizing requires a clear, unambiguous counterfactual decision and its value
to be observable. Third, leverage and the composition of credit agreements are not randomly
assigned, and omitted endogenous variables could be related to both firm-level investment
decisions and leverage, making it problematic to infer a causal relationship. Finally, the
1
Theoretical work focused on these issues includes, e.g., Aghion and Bolton, (1992), Bolton and Scharfstein (1990),
Hart and Moore (1995, 1998), Jensen and Meckling (1976), and Myers (1977). Empirical work includes Andrade
and Kaplan (1998), Parrino and Weisbach (1999), Eisdorfer (2008), Almeida et al (2011), and Gilje (2016).

distortive effects of debt tend to manifest only in the presence of market frictions that intensify as
leverage increases (see, e.g., Stein, 2003). 2
We exploit an empirical setting that allows us to make significant progress on each of
these challenges. Specifically, we use detailed project-level completion decisions from North
American shale oil drilling projects to explore how oil and gas companies with different levels of
leverage react to the severe contango episode that began in December 2014. This setting has
several advantages.
First, we observe high-frequency project-level company decisions and can quantify the
effects of completing an individual oil well versus delaying completion of said oil well. Our
dataset contains detailed project-level data on 3,557 North American shale oil wells operated by
69 publicly traded oil and gas firms. We know the date of well spudding (project start), well
completion (first production and project cash flow), as well as the precise location of the well.
Second, the 2014-2015 contango episode offers a clear instance in which completion
should be delayed. During this period, oil spot prices were significantly lower than oil futures
prices. In February 2015, for example, six-month oil futures prices were 11% higher than spot
prices (Figure 1). The futures curve provides us with a clear counterfactual as it allows us to
estimate the value of delaying completion. We show that a decision to accelerate well
completion and to start oil production early is not value maximizing at the project level. 3
Third, the detailed project-level data allows us to make progress in controlling for the
endogeneity of potential differences in investment opportunities across high-leverage and low-
leverage firms (Lang, Ofek, and Stulz, 1996; Ahn, Denis, and Denis, 2006). Using precise well
2
In a frictionless world, investment is independent of firm leverage and depends only on investment opportunities
(Modigliani and Miller, 1958).
3
Production from shale oil projects is highest during the first month and declines in each subsequent month.
Consequently, pricing at the time of initial production is a key determinant of project-level returns.

location data, we control for differences in investment opportunities (well/project quality) via a
set of granular township (6-by-6-mile) geography fixed effects.
Finally, we exploit the fact that a subsample of oil and gas firms secure debt via asset-
based credit agreements (commonly referred to as asset-based lending, or ABL) that are
characterized by seasonal credit renegotiations that expose our sample firms to credit market
frictions. These pre-scheduled renegotiations are not a consequence of covenant violations,
distress, or defaults, but rather are a standard part of the debt contract lifecycle, defined in
advance. The timing of credit renegotiations and related exposure to lending frictions is thus
plausibly exogenous with respect to the oil price contango period. We exploit the staggered
timing of ABL renegotiations to assess the role that asset based lending and leverage may have
on project completion. 45 out of 69 firms in our sample are subject to such credit agreement
renegotiations, including all but one high-leverage firms (those in the top quintile of the leverage
distribution).
The renegotiations and the contango period serve as the foundation of two separate
difference-in-differences (DiD) analyses. While we do not offer an instrument for leverage, the
combination of two different DiD empirical designs and granular geographic fixed effects
imposes significant hurdles on non-debt-related interpretations of our results.
In comparing well completion decisions pre- and during contango, we find that high-
leverage firms start producing, on average, 1.0 months earlier during contango, resulting in about
a 4.8% loss in the project net present value (NPV), or $124,000 per project. 4 This effect is non-
linear and is concentrated among the high-leverage firms (consistent with Gilje, 2016; Chava and
Roberts, 2008; Bharath and Shumway, 2008 who document other debt related investment
effects). It cannot be explained by firms’ heterogeneity in profitability, size, and market-to-book.
More importantly, the results cannot be fully explained by the mere presence of ABL: the effect
4
The legend for Figure 2 offers a detailed discussion of assumption behind the well-level NPV calculations pre-
contango and during contango.

manifests within the sample of ABL firms, indicating that leverage plays an important role in the
documented investment distortion. For an average firm, the aggregate effect of accelerating
investments across the projects we focus on in this study is equivalent to a deadweight loss of
about 1.2% of equity value.
The results of the second DiD analysis, that focuses on ABL renegotiations, yields
similar conclusions: in the month prior to a pre-scheduled renegotiation or amendment, high-
leverage firms complete 145% more wells than they complete during the month following
renegotiation (relative to other firms in the sample). This number compares to 63% for low-
leverage firms. Furthermore, high-leverage firms exposed to debt renegotiations complete 36%
more wells prior to the renegotiation month than they complete during the month following
renegotiation when benchmarked to other high-leverage firms that are not exposed to
renegotiations and related market frictions. The latter result indicates that lending frictions (at
renegotiations) play an important role in linking high leverage and investment distortions.
While the notion that debt affects firms’ investment decisions has been prominently
featured in the finance literature, the evidence we document is inconsistent with traditional
distortions such as asset substitution (see, e.g., Gilje (2016) for an overview) and debt-overhang-
related underinvestment (see, e.g., Melzer (2017) for an overview). Our results point to a
previously unrecognized debt-related distortion as the documented deadweight loss stems from
acceleration (not delay) of investments.
The short-term transparent nature of the decision to accelerate production in relation to
debt renegotiations limits the number of potential economic forces capable of explaining the
results. At every renegotiation, firms in our sample are effectively refinancing existing asset-
based credit agreements, which exposes them to traditional lending frictions stemming from
asymmetric information and moral hazard (Aghion and Bolton, 1992; Stulz and Johnson, 1985;
Hart and Moore, 1998). Covenants and collateral are prominently featured in the prior literature
as a mechanism that mitigates lending frictions ex ante (at the point of debt origination), expands

access to, and lowers the cost of debt. 5 Yet both can intensify frictions ex post. Covenant
violations lead to significant reductions in firms’ investments (Chava and Roberts, 2008; Nini et
al., 2009) and curtail bank line-of-credit availability (Sufi, 2009). Declines in collateral values
can lead to smaller investments (Chaney, Sraer, and Thesmar, 2012). In the second half of the
paper, we evaluate whether our results can be driven by market frictions linked with (i) binding
covenants and related immediate cash flow needs; and (ii) collateral constraints.
First, we hypothesize that facing binding cash flow or covenant constraints, firms distort
their real decisions in an attempt to accelerate cash flows. After all, firms facing binding
covenant restrictions benefit from access to immediate cash (Faulkender and Wang, 2006) and
earnings management (Dichev and Skinner, 2002). We find that the project-level cash flow
patterns and the backward-looking nature of the covenant metrics, almost mechanically, rule out
binding-covenant-based explanations. 6 The acceleration of well completions should have a

negative effect on a firm’s immediate cash availability as it takes 20 months for new production-
related EBITDA to cover the $3.5 million capex associated with a well completion. Furthermore,
the detailed data on liquidity ratios and borrowing availability suggests that despite firms having
high leverage, neither current ratios, interest coverage ratios, nor credit facility availability
indicate impending liquidity challenges at the time of credit renegotiations.
5
For covenants-related arguments, see, e.g., Jensen and Meckling (1976); Smith and Warner (1979), Aghion and
Bolton (1992) and Dewatripont and Tirole (1994). For collateral-related arguments, see, e.g., Benmelech and
Bergman (2009) Berger and Udell (1990), John, Lynch, and Puri (2003), Jimenez, Salas, and Saurina (2006).
6
While acceleration of well completion by one month indeed accelerates the cash in-flows associated with
producing a well (average EBITDA = $0.35 million in the first month of production), early completion also
accelerates $3.5 million capex necessary to complete the well. Capex (funded by debt on the margin) necessary to
complete a well adversely affects a firm’s near-term financial covenant ratios, as debt immediately goes up. The
cash flow from a well, however, is not immediately recognized as the covenant metrics typically rely on trailing 12
months EBITDA numbers.

Second, we hypothesize that high-leverage firms forego higher-NPV projects to mitigate
lending frictions and expand their access to finance via expanding their collateral base. While
collateral mitigates lending frictions at the point of debt origination, collateral constraints
(similar to covenants) have the potential to expose borrowers to additional frictions at the point
of debt renegotiations. This effect should be more pronounced for high-leverage firms and firms
more dependent on asset-based borrowing (Berger and Udell, 1990). In our setting, well
completions directly affect the collateral values ascribed by lenders, and therefore affect a firm’s
ability to secure/expand access to finance.

To test this channel, we exploit two types of cross-sectional heterogeneity across
individual oil wells in terms of their ability to add to a firm’s collateral. First, we show that prior
to debt renegotiations, high-leverage firms complete wells that have 43% higher daily production
relative to those completed by the same firms after debt renegotiations. Second, we exploit the
fact that newly completed wells offer a verifiable signal about the value of neighboring
undeveloped oil reserves. The geographic proximity of undeveloped reserves to producing
well(s) reduces asymmetric information and allows lenders to assign higher collateral values to
even undrilled non-producing wells. We find that prior to ABL renegotiations, high-leverage
firms are more likely to complete wells on leases with neighboring undeveloped oil reserves
rather than wells on previously developed oil leases. 7 These results are consistent with firms
distorting their investments to enhance their collateral values, thereby mitigating lending
frictions at renegotiation and expanding access to finance.
We conduct a set of tests to assess the stability of our results. First, we recognize that our
analysis is based on a small sample of firms (69) and conduct an array of tests that evaluate
whether the results are driven by a few high-leverage firms in our sample. Second, we test
whether the results are robust to alternative definitions of leverage. Finally, we implement
7
Appendix A1 provides a detailed discussion and a numerical example that quantifies the effect of well completion
on a firm’s collateral value and borrowing base.

placebo tests for alternative periods to assess the importance of the effect of contango on wells’
NPVs.
Overall, our findings highlight a novel debt-related investment distortion that arises at the
intersection of high leverage, renegotiations (and related exposure to lending frictions), and the
presence of collateral constraints for firms that rely heavily on ABL. This distortion is
inconsistent with the traditional risk-shifting and debt-overhang channels. It speaks to
overinvestment and manifests in the absence of imminent technical or financial default. Prior
literature argues that covenants and renegotiations lead to Pareto-improving reallocation of

control rights and better investment outcomes (e.g., Chava and Roberts, 2008; Roberts and Sufi,
2009; Benmelech and Bergman, 2009) and advocates that debt restrictions are value enhancing
(Nini, Smith, and Sufi, 2009; Denis and Wang, 2014) ex ante. We add to this literature by
documenting that collateral constraints and debt renegotiations can lead to negative investment
distortions ex post: high-leverage firms in our sample distort investment even before the
renegotiations and in the absence of covenant violations.
We also add to the expansive literature exploring the role collateral plays in alleviating
the financial frictions and thus enhancing firms’ access to capital. Benmelech and Bergman
(2009) show that collateral allows firms to reduce the costs of external debt financing and
expand debt capacity. Our study highlights previously unexplored hidden costs of collateral-
based financing, linked to the real investment activity of firms.
Finally, our findings relate to the studies exploring the mechanisms contributing to oil
production being inelastic in the short-run (Anderson, Kellogg, and Salant, 2016), which has
important implications for understanding the macroeconomic effects of oil price shocks (e.g.,
Kilian and Murphy, 2012). Existing studies attribute the inelastic supply to the traditional oil
production technology characterized by high-cost production adjustments (Anderson, Kellogg,
and Salant, 2016). However, even with the onset of new production technology (fracking) and
associated lower costs of production adjustment, firms’ oil production remained inelastic (Lehn

and Zhu, 2016). Our results suggest that the contractual nature of debt financing creates frictions
preventing downward adjustment in oil production when oil prices fall. We thus offer further
support to the argument that mounting debt of the oil and gas industry firms is becoming an
important factor in the global oil price dynamic (Domanski et al, 2015).
I. Institutional Background and Data
The shale oil industry offers an attractive setting in which to examine the relationship
between leverage and investment decisions. The oil price futures curve provides firms (and
econometricians) with the expected benefit associated with completing a well at any given date.
The detailed high-frequency data on the precise timing of well completion allow us to evaluate
both the investment outlays and expected project benefits at the firm-project-level. Granular
information on well locations and the ability to observe timing of all oil extraction stages allows
us to control for a variety of alternative factors linked to investment opportunities. The ability to
match the project-level data with firm characteristics allows us to link firms’ leverage and
investment responses to changes in expected project value.
In this section, we provide more details on each individual component of our empirical
setting. We first discuss the contango episode we exploit. We then offer a detailed discussion of
fracking projects in the oil and gas industry and describe the project-level data we utilize.
Finally, we discuss how the contango episode affected the NPVs of individual projects in our
data.
A. Institutional Setting and Oil Price Contango
In this study we exploit the unique evolution of spot and futures prices in the oil market
between November 2014 and June 2015. In late 2014, abrupt changes in the oil price futures
curve, due to the decision by OPEC not to support oil prices, dramatically affected the expected
profits from completing new oil wells early versus waiting to complete the wells. The decline
started in late 2014 and was rapid with spot prices falling below $50 per barrel in January of

2015. At that point the oil market changed from traditional backwardation to severe contango. In
January the 2-year futures price exceeded the spot price by nearly 30% (Panel A of Figure 1) and
the 6-month futures prices exceeded the spot price by over 10%. It was a drastic change from the
backwardation in oil markets just two months prior. The 2015 futures curve exhibited dramatic
deviations from spot prices across all maturities. Panel B of Figure 1 compares the futures curve
as of February 2015 to one as of September 2014.
[Insert Figure 1]
The 2014-2015 contango episode was unprecedented as futures prices exceeded the spot
price by more than three standard deviations relative to historical data. It was not only severe in
terms of futures price deviation from the spot prices, but it was also abnormally long, extending
through the rest of 2015. In April 2015, spot prices experienced a slight increase reaching $60,
consistent with the upward price trajectory projected by futures prices as of the beginning of
2015. By mid-year, however, spot prices declined again and the market remained in strong
contango through 2015 and into early 2016.
The severe contango episode of 2014-2015 played an important role in companies’
assessment of future expected cash flows from oil production operations. Futures prices are
traditionally used by both academics and practitioners alike as a good estimate of expected future
oil prices. Kellogg (2014) prominently features this assumption under risk neutral traders and
efficient aggregation of market information conditions. Consistent with the information content
of oil futures in the post-recession era, most oil producers use futures prices in forming their
expectations (Society of Petroleum Evaluation Engineers, 1995).
Consider the well completion incentives under backwardation and contango. Under
normal market backwardation, oil producers have incentives to accelerate oil production in the
current period because spot prices exceed that of expected future prices. In contrast, during the
severe contango episode, oil producers have a disincentive to initiate new production and bring
new (fracking) wells online as the value of oil produced six months in the future was expected to
10

be at least 11% higher than the value of oil produced in early 2015. More importantly, the
pricing uncertainty associated with the decision to delay production could be fully eliminated.
The oil producers can enter into a futures contract (at significantly higher prices due to contango)
and store the oil underground while waiting to complete a well (at no or minimal marginal
cost). 8,9
Consistent with these arguments, we exploit the sharp shift to contango in
December 2014/January 2015 as an exogenous shock to U.S. producers’ incentives to complete
oil wells. Specifically, we expect the onset of contango to create a disincentive for firms to start
production in favor of opening the available reserves later.
B. Shale Oil Drilling Overview
In 2003, a surprise technological breakthrough combined horizontal drilling with
hydraulic fracturing (“fracking”), enabling development of natural gas shale. In 2009 the
innovation extended into oil development and dramatically reshaped the U.S. oil industry. While
prior to 2009 shale oil production contributed minimally to global oil supply, the technological
change resulted in an increase of U.S. oil production from 5.4 million barrels per day in 2009 to
9.4 million barrels per day by the end of 2014. This increase represented 52% of the overall
increase in oil production globally. Shale oil development has been one of the largest economic
8
One can argue that oil and gas firms often hedge their production 12 to 24 months out and, thus, need to produce
early to deliver on the ex-ante hedges. However, these hedges do not affect current production decisions in our
setting. If a firm sold forward oil in 2014 for delivery in early 2015, a lower cost alternative to initiating production
early on a well would be to close out the position by buying physical barrels for delivery at current low spot prices.
In this sense, the obligation to deliver based on a prior hedge that is in the money is independent of the decision to
complete new wells.
9
During the 2014-2015 contango period physical storage above ground was constrained. Oil in storage hit record
levels and many tank farms were full, making above ground storage options prohibitively costly.
11

transformations that the U.S. economy has experienced, resulting in an aggregate increase of
$2.5 trillion in U.S. equity market capitalization (Gilje, Ready, Roussanov, 2016).
By early 2014, 69 public oil and gas firms and more than 120 private firms were using
fracking technology. In fact, by mid-2014 these companies were producing oil from over
200,000 oil wells across five states with shale oil fracking operations: Texas, Oklahoma, North
Dakota, Colorado, and New Mexico. 10 In our analysis, we absorb the heterogeneity in
investment opportunities via township geographic fixed effects.
Shale oil resides in geologic formations up to two miles below the earth’s surface. To
extract these reserves, firms first have to secure mineral rights from respective property owners.
These oil leases allow energy companies to drill and then hydraulically frack shale formations to
free oil from shale rock. Shale oil well drilling is somewhat unique relative to other types of oil
extraction in that it requires two distinct project steps. First, the well must be spud/drilled.
Second, the well must be completed, which involves hydraulic fracturing. To spud a well, the
majority of the firms in the industry rent a drilling rig from specialized service providers. It takes
these service providers from 3 days to 3 weeks to drill onshore wells at an average cost of $2.97
million dollars. The spudded well can sit idle until the company decides to complete or “frack”
it. 11 The hydraulic fracturing process is separate from the drilling process and can occur any time
after initial drilling has been completed. It also involves using a contractor that specializes in
fracking (completing) wells. Firms that specialize in fracking include Halliburton, Baker Hughes,
and Schlumberger. It typically takes two or three days to complete the fracturing process and
costs about $3.5 million for an average well.
10
Figure A1 provides a snapshot of horizontal drilling activity within a township (6 miles by 6 miles square).
11
Typically, oil lease contracts with mineral owners allow an oil producer to maintain their lease so long as drilling
has started (i.e., wells are spud). Spudding involves drilling the well into the shale and inserting steel well casing and
cement down the hole. Once a well is spud, a shut-in royalty of as little as $10 per month is paid to the mineral
owner.
12

Production typically starts immediately after the well is completed for a number of
reasons. First, the completion specialist must test the well and report the initial production
volume and oil quality measurements to a respective regulatory body (e.g., Texas Railroad
Commission for the state of Texas). Once the well is fracked, disrupting the pressure in the well
is prohibitively costly: the proppant used to frack the well begins to disintegrate immediately
after a well is completed which reduces the amount of oil that can be recovered from a well.
Once a well is completed, production typically declines quickly. High initial pressure leads to
high initial production. As pressure is released from a well, the production quickly declines
month over month. Every day spent not pumping oil from a completed well is, in essence, a day
of lost production and cash flows. Consequently, oil prices at the beginning of a well’s
productive life are critical in determining the economic return of a well. Panel A of Figure 2
illustrates the declining after-tax cash flows based on the average production profile of 2,484
wells completed in North Dakota and the average January 2015 NYMEX futures curve.
[Insert Figure 2 here]
The declining production schedule plays an important role during the period of super-
contango. Panel B of Figure 2 illustrates the NPV of a decision to delay production by one to six
months using the above production volume for a representative North Dakota well. This NPV
calculation assumes a 10% discount rate, which is the rate that banks use to evaluate a firm’s
reserves, as well as the discount rate firms use to report reserves to the SEC. It also relies on the
average January 2015 NYMEX futures curve. 12 Figure 2 illustrates that delaying production
enhances individual project NPV. For example, a one month delay in production in January 2015
12
Given that all firms in our sample are publicly traded,operate in the same industry, and pursue similar project
types it is reasonable to assume that they face similar project level risk and hence discount rates. We have
implemented a comparison of NPV over a wide range of plausible discount rates (5% to 20%) and find that delaying
production by 1 month improves well-level NPV by between 3.3% and 11.1%.
13

increases a representative well NPV from $2,435,718 to $2,559,735. The $124,017 increase
constitutes a 4.8% increase in project NPV relative to immediate well completion.
The two-stage oil extraction process − spudding/drilling and completing/fracking −
provides a number of empirical design benefits. First, we can isolate the decision to start
production (complete a well) from that of exploratory drilling. Second, the ability to cap the well
for an indefinite period of time after spudding, but before the completion, clearly isolates two
operational decisions. The costs of drilling become a sunk cost at the time the
completion/production decision is made (Kellogg, 2014). Finally, the inability of firms to delay
the production after the well was completed and the need to report the completion date to the
regulatory body provides us with a fairly precise date of the completion decision. We exploit all
these features in building the sample of oil well projects for our analysis.
C. Well-level Data.
Our sample consists of 3,557 wells developed by 69 publicly traded oil and gas firms.
The well selection process is based on the two stages of the oil development process. We start
with the most comprehensive well drilling data set available, which is provided by RigData, Inc.
RigData relies on public filings and relationships with a wide set of drilling contractors to
precisely track the start of every shale well drilling operation in the United States. Our study
focuses on shale oil drilling in Texas, North Dakota, Oklahoma, New Mexico, and Colorado,
states in which 98% of U.S. shale oil drilling has taken place.
While RigData, Inc. provides the date of the first stage of well development
(drilling/spudding), it does not provide the well completion (fracking) and start of production
date. We augment the RigData by hand-collecting the completion date from regulatory filings on
well completion collected by the oil industry regulatory bodies in major shale oil producing
states. These include the Texas Railroad Commission (form W-2), the Oklahoma Corporation
Commission (form 1002A), the North Dakota Industrial Commission (form 6), New Mexico
Department of Energy and Minerals (html web form), and the Colorado oil and gas Conservation
14

Commission (html web form). We also cross check dates with completion information from
fracfocus.org.
Our empirical strategy is to evaluate how contango affects the completion decisions made
by individual firms. Consequently, we isolate the completion decision from the drilling decision
by focusing on oil wells that were spudded before the onset of contango – September 2014
through November 2014. We then evaluate the completion decisions for these wells during the
severe contango period of December 2014 through March 2015. To build the control group, we
augment this sample and consider wells spudded and potentially completed during the period of
(more typical) backwardation. Specifically, we include wells spudded in September 2013
through November 2013 and evaluate their completion decisions during December 2013 through
March 2014 to benchmark firms’ completion decisions. We maintain the same calendar months
to eliminate any explanations stemming from seasonal variation in oil production.
Our main variable of interest is the time between the spud date of a well and the
completion date. Panel A of Table I reports that this variable exhibits significant heterogeneity in
our sample. Some wells in our sample are completed as quickly as one month after the spud date.
Others sit idle for in excess of two years before production commences. The median time
between drilling and completion dates in our sample is 4 months.
To address investment-opportunity-based alternative explanations of our results we
exploit detailed information about the geographic location of the wells in our sample (land
survey section township range or latitude and longitude) provided by the respective state
regulatory bodies. This data allows us to control for well quality and investment opportunities as
shale geologic qualities are similar over the 6 mile by 6 mile areas (townships) over which we
conduct our comparisons. Finally, we supplement the well project drilling and completion data
with additional well characteristics such as well completion costs and production volumes. We
obtain this data from regulatory reports where available. For example, the North Dakota
15

Industrial Commission provides the production schedules for a subset of wells, but such detailed
data is not readily available from regulatory bodies of other states in our sample.
D. Firm-level Data.
We hand-match individual project-level data on well drilling and completion to a set of
public oil and gas companies in Compustat following Gilje and Taillard (2016). Panel A of
Table I reports core financial characteristics of the oil and gas companies in our sample including
the core variable of interest in this study – market leverage. Panel B splits the sample into two
sub-samples: high- and low-leverage firms. We define all firms in the top 20% of the market
leverage distribution as of the end of September 2014 as high-leverage firms. The remainder of
the sample we classify as low-leverage firms. Panel B of Table I illustrates that high-leverage
firms are smaller and are characterized by lower Tobin’s Q, however, they are comparable to the
low-leverage firms on other observable dimensions: profitability; capex, dividends, cash holding,
tangible assets, SGA and sales growth. Notably, high- and low-leverage firms within the ABL
sub-sample are of much more comparable size, yet exhibit some differences in capex. In later
empirical tests we ensure that differences in size, capex, and/or growth opportunities across high-
and low-leverage firms are unlikely to explain our results.
[Insert Table I here]
Finally, we hand collect detailed information about public firms’ credit lines and debt
contracts from SEC filings offered by EDGAR. We focus on the size of the credit line offered by
banks to firms in our sample, pre-set renegotiation dates, covenants and collateral constraints
imposed by the debt agreement. This information is collected for the years leading up to and
including the contango episode (2013 through 2016 financial reports).
II. Empirical Analysis and Results
In this section, we evaluate whether debt affects firms’ investment decisions by analyzing
well completion behavior during contango. In Section II.D we summarize tests assessing a
16

variety of concerns and alternative explanations for our results. Finally, in Section II.E, we
outline and empirically evaluate two potential channels through which firms may benefit from
accelerating investment. First, we test whether accelerating investment alleviates binding
financial covenants. Second, we evaluate whether accelerated investment may improve
collateral, thereby mitigating information asymmetry or moral hazard, and improving a firm’s
access to finance.
A. Super-contango and Well Completion Decisions
The analysis presented in this section compares the timing of well completion by high-
and low-leverage firms (first difference) during the periods of contango and backwardation
(second difference). This analysis, first, allows us to validate the core underlying premise of our
empirical design: contango creates a disincentive for (all) oil and gas firms to complete wells and
start production. Second, it allows us to evaluate whether, during the contango episode, high-
leverage firms complete wells quicker than low-leverage firms.
We focus on wells spudded in September –November 2014 and benchmark contango
completions against the completion decisions of wells spudded during September –November
2013 when markets were in backwardation. The 2013 sample serves as our baseline time period,
and allows us to difference out operator-specific differences in project completion time.
Panel A of Table II presents the results of a univariate comparison of the time to
completion decision across high-and low-leverage firms during periods of contango and
backwardation. We find that low-leverage oil producers, those in bottom four quintiles of the
leverage distribution, behave as expected. They delay well completion during super-contango by
about one month compared to well completion in 2013. The one month extension is statistically
and economically significant as it constitutes about a third of the average completion time in
2014/2015. In contrast, high-leverage (top quintile) producers do not delay production. The lack
of adjustment to completion time for high-leverage firms is consistent with the notion that debt
distorts firms’ real investments.

17

[Insert Table II here]
One can argue that the differences in completion times can be driven by other unobserved
heterogeneity rather than financial leverage such as differences in investment projects or growth
opportunities. High-leverage firms may have access to or choose to complete wells that require
less capital or are more/less productive. Table I shows that high- and low-leverage firms differ in
terms of their Tobin’s Q. Existing literature documents that investment opportunities could be
linked to leverage (Lang, Ofek, and Stulz, 1996; Ahn, Denis, and Denis, 2006): firms with bad
investment opportunities could also have high leverage. If this is the case, the incentives of high-
leverage firms to delay production might be weaker than what is captured by our analysis.
Panel A of Table III confirms that high- and low-leverage firms are indeed characterized
by different investment opportunities at the project (well) level. It shows that high-leverage
firms’ wells are characterized by lower production volumes and lower completion capex. To
address this issue we exploit the fact that well quality (production capacity and costs of
completion) is driven by characteristics of shale reservoirs, which we can control for via a set of
very granular 6×6 mile (township) geographic fixed effects.
[Insert Table III here]
Panel B of Table III illustrates the efficacy of 6×6 mile geographic fixed effects. It
documents that wells located at most within 8.5 miles (62+62)1/2 of each other (within one
township) do not exhibit dramatic differences in completion costs and production volume. We
find that while the sub-samples of high-and low-leverage firms exhibit economically significant
differences in well costs and production, these differences are not statistically or economically
different in a regression setting that incorporates fine geographic fixed effects. The combination
of very granular geography and firm-level fixed effects capture the heterogeneity in investment
opportunities across high- and low-leverage firms.
18

With the controls for investment opportunities, we implement a regression analysis that
examines the effect of contango on well completion times of high-versus-low-leverage firms
using the following regression equation:
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑡𝑡𝑡𝑡 𝐶𝐶𝐶𝐶𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛽𝛽1 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑜𝑜𝑡𝑡 + 𝛽𝛽2 𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝐿𝐿𝐿𝐿𝑣𝑣𝑖𝑖 × 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑜𝑜𝑡𝑡 +
+ 𝐹𝐹𝑖𝑖𝑖𝑖𝑖𝑖 𝐹𝐹𝐹𝐹𝑖𝑖 + 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺ℎ𝑦𝑦 𝐹𝐹𝐹𝐹𝑗𝑗 + 𝜖𝜖𝑖𝑖𝑖𝑖𝑖𝑖
where 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑡𝑡𝑡𝑡 𝐶𝐶𝐶𝐶𝐶𝐶𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖𝑖𝑖 is the number of months between the start of well 𝑗𝑗 spud date and
its first production date. The unit of observation is well j of firm i at time t. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑜𝑜𝑡𝑡 is an
indicator variable that equals one for wells spudded in September – November of 2014, the
period just prior to the emergence of contango in the oil market. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑜𝑜𝑡𝑡 is equal to zero for
wells spudded in September – November of 2013, our control backwardation year. 𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝐿𝐿𝐿𝐿𝑣𝑣𝑖𝑖 is
an indicator variable for whether the firm is in the top quintile of the leverage distribution.
𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝑖𝑖 and 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺ℎ𝑦𝑦 𝐹𝐹𝐹𝐹𝑗𝑗 are firm and geography (township level) fixed effects. The key
coefficient of interest is the coefficient 𝛽𝛽2 on the interaction term 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑜𝑜𝑡𝑡 ×𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝐿𝐿𝐿𝐿𝑣𝑣𝑖𝑖 which
indicates whether high-leverage firms initiate production on their wells sooner than low-leverage
firms during the super-contango period relative to the baseline period. The standard errors are
clustered at the firm level. 13
13
We acknowledge that a natural empirical experiment to evaluate an economic mechanism on the intersection of
leverage and credit agreements is DiDiD analysis focused on the triple interaction (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 × 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝑄𝑄5𝑖𝑖 ×
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛𝑛𝑛 𝐷𝐷𝑖𝑖 ). Yet collateral, covenants and mandatory renegotiations are not independent as they all
mitigate lending frictions that intensify with increasing leverage. In our sample, only one high-leverage firm does
not have a credit agreement. The outlier−American Eagle (AE)−had a $173.4 million bond offering (101% of the
equity value) in August 2014, just 6 month prior to the period covered in our study. During the contango episode,
AE extended its average completion period by 4.34 months, more than any other firm in our sample, even more than
low-leverage firms without ABL. The DiDiD estimate would be entirely driven by a single high-leverage firm (AE).
19

Panel A of Table IV reports the results where the dependent variable is the number of
months to completion. We find that in contango, firms in all but the top quintile of the leverage
distribution delay well completion by approximately one month. Column (1) demonstrates this
nonlinear dependence of wait times during contango on leverage. Firms in the top leverage
quintile pull forward completion relative to their less constrained peers. The difference in
completion times is even more pronounced when examining firms with high leverage and asset-
based lending (column 3). The results are statistically significant irrespective of the statistical
approach underlying the standard errors. Furthermore, the coefficient on

(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑖𝑖 × 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 ) in column (4) is negative, but not statistically
significant. The latter can be attributed to a non-linear effect of leverage, consistent with other
leverage distortions (i.e., risk shifting and debt overhang).
[Insert Table IV here]
Since all but one of the 13 firms in the top leverage quintile have ABL and are exposed to
pre-scheduled renegotiations, yet only about half of the firms in other leverage quintiles have
ABL, one can argue that the effect we document might be driven by asset-based borrowing
rather than leverage. To address this issue in Panel B of both Table II and Table IV we repeat our
empirical experiment after restricting our sample to firms with ABL. 45 of 69 firms in our
sample have such debt agreements. The results show that compared to low-leverage firms with
ABL, high-leverage firms with ABL accelerate well completion by 1 month during contango.
20

This evidence suggests that the documented investment distortion cannot be attributed to ABL
alone. Leverage plays an important role in acceleration of well completions. 14
B. Super-contango and Well Completion Decisions: Alternative Explanations
To assess explanations based on observed heterogeneity distinct from leverage (recall
Table I), we augment the analysis presented in Table IV and add interaction terms between the
contango dummy and other firm-level control variables measured prior to contango (as of
September 2014). Table V shows that heterogeneity in firms’ size, profitability, capex and
Tobin’s Q is unlikely to account for our results. Our core inferences do not change after we
control for respective interaction terms (Panel A). Even if we allow these firm characteristics to
affect well completion decisions in a non-linear fashion (via utilizing quintile dummies in
Panel B of Table V), we still find that high-leverage firms accelerate well completion by one
month relative to their low-leverage peers in the industry. Once again, we find that leverage has
both economically and statistically significant effects on well completion decisions in contango
irrespective of the statistical approach underlying the standard errors.
[Insert Table V here]

Overall, the lack of a completion time response for high-leverage firms is consistent with
the notion that debt distorts firms’ investment decisions in the presence of credit market frictions.
14
In implementing this analysis, we recognize that our results are based on a relatively small number of firms (69).
While ours is not the only study that utilizes small samples in this line of inquiry (Benmelech and Bergman (2009)
use a sample of 246 airplane certificates issued by 12 airlines), in small-sample settings standard asymptotic-based
statistical techniques may over-reject the null and bootstrapped standard errors are recommended to make proper
statistical inferences (Bertrand, Duflo, and Mullainathan, 2004). Appendix B reports all the regression analysis
reported in this paper under the cluster-bootstrapping procedure of Cameron, Gelbach, and Miller (2008) that allows
adjusting standard errors for small samples in settings similar to ours: a small number of subjects (clusters), each
with a large number of projects.
21

Based on the representative well-level cash flows presented in Figure 2, we estimate that by not
expanding time to completion by one month, high-leverage firms are foregoing 4.8% of project
NPV or $124,017 per project. The number is economically significant. Considering the number
of projects available for completions to each high-leverage firm during contango, acceleration of
project completion by one month has the potential to create a deadweight loss of 1.2% of equity
value for high-leverage firms.15
C. Debt Renegotiations
In this section we evaluate completion decisions of high- and low-leverage firms around
ABL renegotiations that by definition expose firms to credit frictions. Alongside further
validating our conclusions above, this analysis allows us to evaluate whether lending frictions
play an important role in firms’ decisions to accelerate oil production. Since the ABL
renegotiation dates are pre-scheduled about every 6 months, these dates can be considered
exogenous to the onset of contango and related changes in firms’ financial conditions. We
exploit the Spring 2015 debt renegotiations staggered through February, March, April, or May of
2015 and compare well completion decisions of high vs low-leverage firms (first difference)
around debt renegotiation dates (second difference).
In this setting, evaluating the time to completion is uninformative as for wells drilled in
September through November of 2014 the time to completion increases purely mechanically
from before renegotiation to after renegotiation. Consequently, we adapt our dependent variable
to capture the start of production decisions in an on/off fashion. Specifically, our dependent
variable 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 takes the value of 1 if a well j of firm i is completed
15
To quantify the average aggregate effect of one month completion acceleration we multiply the well-level
deadweight loss of $124,017 by the number of wells spudded by each high-leverage firm in our sample and then
divide the resulting estimate by the market value of equity of the respective high-leverage firms. 1.2% represents an
average deadweight loss normalized by equity across all high-leverage firms in our sample.
22

(turned on from idle to producing) in a given month t, and it is 0 in all other months (both before
the completion month and following the completion month). The unit of observation is well j,
firm i, month t. Effectively, if one averages the dependent variable within each firm i, within
each month t, one would estimate the share of firm i ex-ante spudded wells that started producing
in a given month t. If debt renegotiations intensify the distortive effects of leverage on
investment decisions we should observe high-leverage firms complete a higher percentage of
wells prior to debt renegotiations as compared to completion decisions of low-leverage firms.
Table VI reports the univariate analysis of completion decisions around renegotiation

dates for high- and low-leverage firms. We observe evidence consistent with distortive effects of
high leverage. The estimate of 0.22 at time (t-1) for high-leverage firms suggests that one month
prior to renegotiations high-leverage firms complete 22% of wells they had spudded in the fall of
2014. This number drops to 9% in the month after renegotiations, a change that is both
economically and statistically significant. To put this into perspective, high-leverage firms
complete 145% more wells in the month prior to renegotiation than the month of or one after the
renegotiation (0.22/0.09−1). In contrast, low-leverage firms exhibit a much smaller change in
well completion percentages around the debt renegotiations dates. We observe a drop from a
12% rate of completion prior to renegotiation to a 7% rate of completion after renegotiation.
[Insert Table VI here]
We confirm this pattern using regression analysis of the completion decisions relative to
debt renegotiations in event time, where we control for geography, firm and month fixed effects:
3
𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = � 𝛼𝛼m 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀ℎ𝑗𝑗,𝑖𝑖,𝑚𝑚,𝑡𝑡 +

𝑚𝑚=−3
+ � 𝛽𝛽m 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀ℎ𝑗𝑗,𝑖𝑖,𝑚𝑚 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝐿𝐿𝐿𝐿𝑣𝑣𝑖𝑖 + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝑖𝑖 + 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝐹𝐹𝐹𝐹𝑘𝑘 + 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀ℎ𝑡𝑡 + 𝜖𝜖𝑖𝑖𝑖𝑖𝑖𝑖
𝑚𝑚=−3
where 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is a dummy variable equal to 1 if well j of firm i starts
production in month t and this month is m months removed from the month of credit agreement
23

renegotiation, and 0 otherwise. For example, if firm i has a credit renegotiation in March of 2015
and well j starts is completed in April 2015 then 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 = 1 and
𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀ℎ = 0.
Firm-level fixed effects control for heterogeneity in firms’ characteristics. Since we only
consider oil and gas firms’ behavior over a six-month span, these fixed effects control for a wide
range of unobservable firm characteristics, including firms’ size, profitability, Tobin’s Q, and
capex. Geographic fixed effects control for heterogeneity in project quality and production
potential. Month fixed effects control for changes in economic conditions including changes in
oil prices and severity of contango. The control group for the baseline results reported in
specifications (1)-(3) of Table VII contains all low-leverage firms as well as high-leverage firms
that do not have pre-scheduled renegotiation during Spring of 2015. 𝛽𝛽m are the key coefficients
of interest. If debt renegotiations intensify the distortive effects of leverage on investment
decisions, we expect firms to complete more wells just prior to debt renegotiations (m=-1 or m=-
2) than after the renegotiation decisions (m=0 or m=1).
[Insert Table VII here]
The regression analysis results reported in Table VII offer conclusions similar to those
implied by the univariate analysis reported in Table VII. For example, column (2) of Table VII
suggests that between the month prior to and the month of renegotiations, high-leverage firms
exhibit an 11.3% drop in the rate of well completion (the coefficient drops from -0.022 to -0.135)
as compared to all non-ABL firms and other high-leverage firms not going through renegotiation
in a given month. This is similar in magnitude to the 0.13 change in completion rate documented
in Table VI. In contrast, the number of new wells being brought on line by the low-leverage
firms (column 3 of Table VII) does not vary across debt renegotiation dates in a statistically or
economically significant way (the coefficient changes from -0.007 in m=-1 to -0.006 in m=0).
In column (4) of Table VII we implement the analysis based solely on the sample of
high-leverage firms’ wells. In this setting high-leverage firms not going through renegotiations
24

provide the counterfactual for high-leverage firms facing renegotiations (the control group
consists of other high-leverage firms that have not (yet) been treated). Column (5) reports
identical analysis for low-leverage ABL firms. The results suggest that even in the presence of
lending frictions low-leverage firms do not distort investments. Yet lending frictions play an
important role in inducing high-leverage firms to accelerate well completion. This evidence is
consistent with high leverage leading to distortion only in the presence of lending frictions (in
our setting, induced by renegotiations) (Stein, 2003). The latter is important to account for in
identifying potential economic mechanisms driving our results.
D. Robustness and Alternative Interpretations
In this section we discuss a wide array of different tests we implemented to further assess
the sensitivity of our results to different specification assumptions. Results in this section are
reported both in the Online Appendix and the main paper.
(i) We start by testing whether our results are sensitive to alternative definitions of
leverage: (i) book leverage calculated as book value of debt normalized by total book value of
debt plus book value of equity; and (ii) (reverse) interest coverage ratio calculated as ratio of
interest expense divided by EBITDA. Both are measured as of September 2014. The interest
coverage ratio is reversed to ensure comparability of the results (highest quintile being
comparable to high leverage quintile firms). Appendix Tables A.I through A.IV offer results
consistent with those reported in Tables II, IV and VII: high-leverage firms accelerate well
completion by about one month relative to low-leverage firms, the investment distortion is more
pronounced in the month prior to ABL renegotiation.
(ii) Another typical concern that arises with any study that evaluates the effect of leverage
on firm financial decisions is that leverage is not randomly distributed. One can argue that our
results are due to an omitted variable affecting both debt and completion decisions (e.g., firm
size or investment opportunities) or due to firms choosing debt based on their business strategy
(reverse causality), leading to alternative interpretations of our tests. While we do not offer a
25

viable instrument for firms’ leverage, there are several pieces of evidence that support a debt-
based interpretation and undermine the argument that our results are due to omitted variable bias
or reverse causality.
First, our identification relies on the contango episode being unanticipated by firms and
the market. The 2014-2015 contango episode was exceedingly rare in that the difference between
the futures price and the spot price exceeded its average by over three standard deviations. To
test whether our results are indeed due to contango and are not a manifestation of long-term
trends where high-leverage firms accelerate well completion over time, we implement Placebo
tests. Appendix Table A.V reports the results of the analysis analogous to the one reported in
Table IV where instead of comparing firm completion behavior around the end of 2014 contango
episode, we look at placebo years. Panel A (B) compares the completion time of wells spudded
in September through November of 2013 (2012) and wells spudded in September through
November of 2012 (2011). The results show no significant discernable differences in well
completions across the time dimension (equivalent of contango dummy) or across high- and low-
leverage firms. These results provide further support for our empirical design. Furthermore, the
results make it more challenging to argue that firms could have anticipated the severe contango
of 2014 and chosen their leverage accordingly. This result does not support reverse-causality-
based explanations.
Second, the evidence of well completion activity around debt renegotiations (where we
control for firm fixed effect) also allows us to further mitigate concerns about unobserved
heterogeneity. It is difficult to argue that changes in completion behavior around debt
renegotiation dates could be attributed to economic channels that exclude debt-related forces. It
is also hard to offer an economic explanation based on, e.g., differential managerial behavior,
wholly unrelated to leverage, for the differential completion activity of high- and low-leverage
firms pre- and post-renegotiation after we control for firm-level fixed effects. Table VI shows
26

that high-leverage firms complete more wells than low-leverage firms pre-renegotiation (22%
versus 12%), but have similar completion rates post-renegotiation (9% versus 7%).
(iii) One might be concerned that the results we document are due to the
contemporaneous rapid drop in oil prices rather than firm leverage or contango itself. For
example, a drop in spot oil prices might lead to significant changes in asset values that vary
across high- and low-leverage firms (see Gilje and Taillard, 2016). There are several pieces of
evidence to suggest that a direct oil price change does not fundamentally alter the interpretation
of our results. First, our main tests are based on comparisons of investment decisions within the
same month in the assets of similar quality located in the same township, but owned by firms
with different leverage. Second, the key advantage of contango is that it creates an incentive to
wait to complete projects even when the oil prices are low.
(iv) While our empirical setting offers us a unique opportunity to evaluate firm
investment decisions at the project level, our analysis is based on a small sample of firms – only
69 public oil and gas firms are accounted for in our analysis. As such one can argue that the
results we document might be driven by a few (out of 13) high-leverage firms accelerating
completion for reasons unrelated to debt.
To assess this possibility we have conducted two types of analyses. First, we evaluate
whether the results presented in Table IV are driven by a single high-leverage firm in our
sample. Table VIII provides relevant statistics on each high-leverage firm as well as the
interaction coefficients corresponding to the Table IV column (2) specification after a respective
high-leverage firm is dropped from our sample. The resulting coefficients of interest fall in the
range of [-1.18,-0.86], and remain statistically significant.
[Insert Table VIII here]
Second, we evaluate whether the exclusion of up to 3 highest-leverage firms and
subsequent redefinition of the leverage quintiles affects the Table IV results. Table IX reports
this analysis. While we observe relatively stable magnitudes of the core coefficients of interest
27

within the sample of ABL firms (those exposed to credit renegotiations and related lending
frictions), we do see some erosion in magnitude and statistical significance of the coefficients
when we consider the full sample. The robustness tests for Panel A of Table IV (reported in
Panel A of Table IX) show the most significant compression in the core coefficient magnitude as
it reaches -0.6 at its lowest point versus -1.02 reported in Table IV. Similar analysis in Panel B of
Table IX, one that focuses on ABL firms, shows more stable coefficient magnitudes and
statistical significance: the coefficient vary between -0.962 and -1.007. This once again points to
the importance of renegotiations as a part of the economic mechanism underlying the

documented overinvestment. 16
[Insert Table IX here]
E. Leverage Effect on Investment Decisions: Economic Channels
The results presented in Tables II through VII are consistent with high leverage distorting
firms’ investment decisions, particularly so around renegotiations. The question remains as to the
exact economic channels underlying the effect of debt and renegotiations on sample firms’
investment decisions.
The rationale for delaying completion during contango is transparent and unambiguous,
and any finance professional should see the value left on the table as a result of accelerating
production. Barring lending frictions, firms should be able to renegotiate their credit agreements
in a way that allows them to delay productions and capture the additional value (through rent
sharing) under contango. This suggests that our results might be driven by lending frictions
stemming from asymmetric information and moral hazard (Aghion and Bolton, 1992; Stulz and
Johnson, 1985; Hart and Moore, 1995, 1998). On the moral hazard side, delaying well
completion might be perceived by lenders as managerial effort to divert funds to bad projects,
16
Table AVI extends these robustness tests to exclude up to 5 highest-leverage firms. Tables AVII and AVIII report
similar robustness tests for Tables VI and VII of the paper.
28

siphon them off to pay shareholders/management, invest in higher risk projects, or otherwise
misappropriate funds (Berger and Udell, 1990; John, Lynch, and Puri, 2003; Jiménez, Salas, and
Saurina, 2006). On the asymmetric information side, delaying well completion might be
interpreted as an attempt to misrepresent the value of non-producing reserves to secure continued
access to (asset-based) finance.
Covenants and collateral are prominently featured in the prior literature as contract
features that mitigate such lending frictions ex ante (at the point of debt renegotiations). Both
increase firms’ access to debt finance and lower the cost of capital (Chava and Roberts, 2008;
Benmelech and Bergman, 2009). Yet both intensify frictions ex post, particularly so for high-
leverage firms. Covenant violations lead to decline in firms’ investments (Chava and Roberts,
2008; Nini et al., 2009) and curtail bank line-of-credit availability (Sufi, 2009). Decline in
collateral values lead to smaller investments (Chaney, Sraer, and Thesmar, 2012).
In this section, we evaluate whether frictions linked with two aspects of oil and gas firms’
access to debt finance contributed to our results: (i) binding covenants and related cash flow
needs; and (ii) collateral constraints.
E.1 Cash Flow and Financial Covenants
First, we assess whether the decision to complete wells early is driven by binding
covenants and related immediate cash-flow needs. To avoid post covenant-violation investment
distortions (Chava and Roberts, 2008) firms accelerate cash flows and manage earning (e.g.,
Dichev and Skinner, 2002). One can hypothesize that facing binding covenants, firms can go
beyond merely adjusting their accounting approach and distort their real decisions in an attempt
to accelerate cash flows. Revenue shortage, associated with low oil prices in late 2014, might be
felt more acutely by high-leverage firms characterized by higher interest payments and
potentially binding covenant restrictions. This pressure might induce the high-leverage firms to
accelerate production and secure access to funds necessary to avoid real or technical default.
29

To evaluate this hypothesis, we analyze covenant and cash flow characteristics of firms in
our sample. We find that high-leverage firms in our sample do not face binding covenant
constraints. As of March 2015, these firms are characterized by an average interest coverage
ratio of 3.18 and an average current ratio (current assets/current liabilities) of 2.464. These
figures suggest that during our sample period high-leverage firms generate sufficient cash flows
to cover interest payments. Additionally, we collect detailed data on credit limits, and find that
high-leverage firms are characterized by spare debt capacity with them utilizing, on average,
only 37.3% of their credit lines’ credit facility. All these numbers indicate that while high-
leverage firms may not have the strongest balance sheets, they are neither vulnerable to
imminent default, nor do they face an impending liquidity crunch at the time of credit
renegotiation.
A careful analysis of cash flows associated with a representative well further undermines
a liquidity/cash-flow based explanation of our results. Acceleration of well completion by an
average of one month indeed accelerates the cash in-flows associated with a producing well
(average EBITDA = $0.35 million in the first month of production). However, early well
completion also accelerates $3.5 million necessary capex. Net, the early well completion should
have negative effects on immediate cash availability. In the medium term, acceleration of well
completion would also have a negative effect on a firm’s cash flows as the early-completed well
would generate lower revenue stemming from lower oil prices at the peak of the well operational
capacity yet require the same completion capex. Combined, early well completion is highly
unlikely to yield a better cash position (e.g., necessary for interest payments) for high-leverage
firms. Furthermore, the pattern of project-level cash flow, almost mechanically, rules out
covenant-based explanations. Given that financial covenant metrics are typically backward-
looking (e.g., current debt to trailing 12 month EBITDA), expending capital (funded by debt on
the margin) to complete a well would adversely affect a firm’s near-term financial covenant
ratios.
30

Overall, the short term nature of the decision we evaluate, the cash flow profile of a
representative well, and the collected covenant data allow us to reject the binding financial
covenants as driving our results.
E.2 Collateral Channel
The second economic mechanism stems from an observation that even in the absence of
binding (backward-looking) covenants, high-leverage firms have a more vested interest in
ensuring continued access to finance for future investments. In fact, one-year-ahead investment
plans reported by firms in their 2014 10Ks constitute, on average, 106% of end of the year cash
flows and, thus, require expanded access to credit. For ABL firms that drive our core results,
well completions alongside affecting the cash flows affect the value of pledgeable assets that can
be used to secure/expand firms’ access to finance. Both asymmetric information and moral
hazard lending frictions (at the point of renegotiation) create incentives for collateral generation
by firms at the expense of lower project-level NPVs. Ultimately, equity holders forgo project-
level returns in exchange for greater credit availability. This effect is more pronounced for riskier
firms more dependent on asset-based borrowing (Berger and Udell, 1990).
To evaluate this hypothesis we exploit a specific approach banks pursue in determining
the size of credit agreements in the oil and gas industry. For ABL firms in our sample, the
expected value of collateral determines the “borrowing base” that sets the upper limit on banks’
lines of credit established to finance oil explorations and production operations. Given the
asymmetric information about the quality of oil reserves between firms and banks, lenders apply
different haircuts to expected cash-flow values of developed and undeveloped wells, as well as to
non-producing reserves located in close vicinity of producing reserves versus those located in
geographic isolation.
Figure 3 offers Office of the Comptroller of the Currency (OCC) reserve classification
for oil and gas exploration and production lending as well as suggested “risk adjustment factors”
that the OCC recommends banks to apply to the NPVs of different types of reserves in
31

determining the borrowing base. 17 The OCC guidance suggests that the borrowing base is
adjusted up when (i) oil prices increase, (ii) new wells are completed and moved from proved
undeveloped to proved producing reserves, (iii) higher production volume wells are completed,
or (iv) newly completed wells mitigate asymmetric information about the value of other
(typically neighboring) non-producing reserves. Apart from directly affecting the collateral value
of completed wells, production decisions have the capacity to increase collateral values assigned
to the nonproducing and undeveloped resources.
[Insert Figure 3 here]
This suggests that spudded wells exhibit significant heterogeneity in their ability to affect
the borrowing base once completed, stemming from production capacity and ability to mitigate
the asymmetric information between firms and banks regarding the value of neighboring
nonproducing reserves. For example, if a firm completes a well on a lease with existing
producing wells (multi-well lease) it receives collateral credit for the newly added production
volume, but well completion does not offer additional information about neighboring (already
producing) reserves. In contrast, if the same firm decides to complete the first well on an oil
lease (single-well lease), it not only receives collateral credit for the newly added production
volume, but it also recognizes the benefits of reduced asymmetric information about the quality
of neighboring non-producing reserves. The wells on single-well leases, ceteris paribus, have
higher collateral impact. Appendix A1 offers a numerical example that compares the collateral
effect of completing a representative well on single- and multi-well leases.
[Insert Table X here]
We explore these two margins of collateral adjustment in our empirical tests. First, we
compare production volume of wells completed right before a debt renegotiation versus those
17
Comptroller’s Handbook for oil and gas Exploration and Production Lending, March 2016, Office of the
Comptroller of the Currency, https://www.occ.treas.gov/publications/publications-by-type/comptrollers-
handbook/pub-ch-og.pdf
32

completed right after. Univariate analysis presented in Table X shows that wells completed by
high-leverage firms just prior to a credit renegotiation have higher initial production volume than
wells completed after renegotiation. Specifically, prior to debt renegotiations firms initiate
production on wells that produce 417 barrels per day, versus 292 barrels per day right after debt
renegotiations.
[Insert Table XI here]
Second, we exploit the heterogeneous effect on collateral values of single- and multi-well
lease completions. We split our sample of wells based on these well definitions and compare
companies’ completion behavior within each well category. Table XI reports the results. We find
that high-leverage ABL firms accelerate completion on high-collateral-impact single-well leases
prior to renegotiations. This effect is economically and statistically absent among completions on
low-collateral-impact multi-well leases. Table XII further confirms these results in a regression
setting similar to that reported in Table VII. Column (1) shows that high-leverage ABL firms’
rate of high-collateral-impact well completion declines by 21% after the debt renegotiation
month (0) and indicates acceleration of completion prior to renegotiations. Column (4) indicates
a uniform low-collateral-impact well completion behavior around renegotiations date. We
observe similar evidence even if we implement the renegotiation-focused DiD analysis
separately for high- and low-leverage ABL firms (columns 2-3 and 5-6).
[Insert Table XII here]
As a final note, we address a concern that single-well lease and multi-well leases could
differ on other dimensions in addition to their impact on collateral values. For example, if single
well leases are, on average, higher-risk higher-reward type projects, then our evidence is
indicative of a risk-shifting rather than a collateral-based mechanism. Using production data on a
broad set of wells across public and private companies, we verify that single-well leases exhibit,
on average, both lower initial production and lower standard deviation of production realizations.
33

This evidence is not consistent with risk-shifting and confirms arguments presented by Gilje
(2016). 18
Taken together, the evidence indicates that ability of collateral to mitigate credit frictions
offers a potential mechanism underlying our core results. It suggests that in making completion
decisions during contango, high-leverage ABL firms are driven by a desire to expand their
access to finance.
III. Discussion and Conclusion
In this paper we document a new debt-related investment distortion and link this
distortion with firms’ incentives to create collateral. Our results suggest that by increasing
collateral values, high-leverage firms (dependent on asset-based loans) mitigate financing
frictions arising at debt renegotiations. We show that during the 2014/15 contango episode, high-
leverage firms complete wells and start production earlier than low-leverage firms. This
acceleration of production adversely affects the NPV of oil projects, which, according to our
estimates, is equivalent to a deadweight loss of about 1.2% of high-leverage firms’ equity value.
This adverse investment decision is particularly pronounced prior to (pre-scheduled) debt
renegotiations. High-leverage firms complete 145% more wells in the month leading to
renegotiations than in the month of and after renegotiations (when their rate of well completion
is equal to that of low-leverage firms). High-leverage firms, being more subject to lending
frictions, resort to accumulating collateral to secure and/or expand their access to finance.
The literature has long argued that riskier firms depend more on secured financing
(Berger and Udell, 1990). Collateral plays a much more important role in high-leverage firms’
18
Note that the evidence regarding the lower-risk and lower-return nature of single well leases is not inconsistent
with evidence indicating higher production volume of wells opened by high-leverage firms prior to debt
renegotiations. In view of the firm-level fixed effects, both results suggest that high-leverage firms tend to open
more productive, single wells prior to renegotiation than after renegotiation.
34

cost of debt financing and access to debt financing as compared to low-leverage firms
(Benmelech and Bergman, 2009). Low-leverage firms can potentially substitute away from
asset-based financing to other forms of financing. We find that high-leverage firms forego
project-level NPV to boost collateral values prior to renegotiations. In essence, our evidence
suggests that high-leverage firms pay (in a form of lower project NPV) to ensure continued
access to cheaper and more abundant credit.
Our study highlights previously undocumented hidden costs of collateral-based financing.
More than 70% of commercial and industrial loans are issued on a secured basis (Berger and
Udel, 1990). Even publicly traded firms rely on bank debt and exploit secured credit to a
significant extent (Rauh and Sufi, 2010). Secured or not, debt contracts are frequently
renegotiated even in the absence of distress or default (Denis and Wang, 2014; Roberts and Sufi,
2009). These renegotiations result in large changes in amounts, maturity, and pricing of the
contracts (Roberts and Sufi (2009)). Roberts and Sufi (2009) estimate that close to 90% of the
debt contracts are renegotiated before stated maturity. Our results, hence, provide important
implications for the wide set of firms that have asset-based debt and are subject to renegotiation
and refinancing-related lending frictions.
Finally, we would like to note that our findings also shed light on a puzzle of oil
production being inelastic in the short-run (Anderson, Kellogg, and Salant, 2016) which has
important implications for understanding the macroeconomic effects of oil price shocks (e.g.,
Kilian and Murphy, 2012). These studies attribute inelastic supply to the traditional oil
production technology characterized by high-cost production adjustments. However, even with
the onset of new production technology (fracking) and associated low cost of production
adjustment, firms’ oil production remained inelastic. Our results suggest that the contractual
nature of debt financing creates frictions preventing downward adjustment in oil production
when oil prices fall. We thus offer further support to the argument that mounting debt of the oil
and gas industry firms is becoming an important factor in the global oil price dynamic.
35

References
Ahn, Seoungpil, David J. Denis, and Diane K. Denis, 2006, Leverage and investment in
diversified firms, Journal of Financial Economics 79, 317-337.
Almeida, Heitor, Murillo Campello, and Michael S. Weisbach, 2011, Corporate financial and
investment policies when future financing is not frictionless, Journal of Corporate Finance
17, 675-693.
Anderson, Soren T., Ryan Kellogg, and Stephen W. Salant, 2016, Hotelling under pressure,
forthcoming, Journal of Political Economy.

Andrade, Gregor and Steven N. Kaplan, 1998, How costly is financial (not economic) distress?
Evidence from highly leveraged transactions that became distressed, Journal of Finance 53,
1443-1493.
Aghion, Philippe, and Patrick Bolton, 1992, An incomplete contracts approach to financial
contracting. Review of Economic Studies 59, 473-494.
Benmelech, Efraim, and Nittai K. Bergman, 2009, Collateral pricing, Journal of Financial
Economics 91, 339-360.
Berger, Allen N., and Gregory F Udell, 1990, Collateral, loan quality, and bank risk, Journal of
Monetary Economics 25, 21-42.
Bertrand, Marianne, Esther Duflo, Sendhil Mullainathan, 2004, How much should we trust
differences-in-differences estimates?, Quarterly Journal of Economics, February, 249-275.
Bharath, Sreedhar T. and Tyler Shumway, 2008, Forecasting default with the Merton Distance to
Default Model, Review of Financial Studies, 21(3): 1339-1369.
Bolton, Patrick and David S. Scharfstein, 1990, A theory of predation based on agency problems
in financial contracting, American Economic Review 80, 93-106.
Bolton, Patrick and David S. Scharfstein, 1996, Optimal debt structure and the number of
creditors, Journal of Political Economy 104, 1-25.
36

Cameron, A. Colin, Jonah B. Gelbach, and Douglas L. Miller, 2008, Bootstrap-Based
Improvements for Inference with Clustered Errors, Review of Economics and Statistics 90
(3): 414–27.
Chaney, T., D. Sraer, and D. Thesmar, 2012, The collateral channel: How real estate shocks
affect corporate investment, American Economic Review 102, 2381-2409.
Chava, Sudheer, and Michael R. Roberts, 2008, How does financing impact investment? The
role of debt covenants, Journal of Finance 63, 2085–2121.
Denis, David J., and Jing Wang, 2014, Debt covenant renegotiations and creditor control rights,
Journal of Financial Economics 113, 348-367.
Dichev, Ilia D. and Douglas J. Skinner, 2002, Large sample evidence on the debt covenant
hypothesis, Journal of Accounting Research 40, 1091–1123.
Domanski, Dietrich, et al., 2015, Oil and debt, BIS Quarterly Review March, 55-65.
Eisdorfer, Assaf, 2008, Empirical evidence of risk-shifting in financially distressed firms,
Journal of Finance 63, 609-637.
Eisfeldt, Andrea L. and Adriano A. Rampini, 2009, Leasing, ability to repossess, and debt
capacity, Journal of Financial Economics 22, 1621-1657.
Faulkender, Michael, and Rong Wang, 2006, Corporate financial policy and the value of cash,
The Journal of Finance 61, 1957-1990.
Gilje, Erik P., 2016, Do firms engage in risk-shifting? Empirical evidence, Review of Financial
Studies 29, 2925–2954.
Gilje, Erik P., Robert Ready, and Nikolai Roussanov, 2016, Fracking, drilling, and asset pricing:
Estimating the economic benefits of the shale revolution, NBER Working Paper No. 22914.
Gilje, Erik P. and Jerome P. Taillard, 2016, Do private firms invest differently than public firms?
Taking cues from the natural gas industry, Journal of Finance 71, 1733-1778.
Hart, Oliver, and John Moore, 1995, Debt and seniority: An analysis of the role of hard claims in
constraining management, The American Economic Review 85(3), 567-585.
37

Hart, Oliver, and John Moore, 1998, Default and renegotiation: a dynamic model of debt,
Quarterly Journal of Economics 113, 1-42.
Jensen, Michael C and William H. Meckling, 1976, Theory of the firm: Managerial behavior,
agency costs, and capital structure, Journal of Financial Economics 3, 305-360.
Jiménez, Gabriel, Vicente Salas, and Jesús Saurina, 2006, Determinants of collateral, Journal of
Financial Economics 81, 255-281.
John, Kose, Anthony W. Lynch, and Manju Puri, 2003, Credit ratings, collateral, and loan
characteristics: Implications for yield, Journal of Business 76, 371-409.

Kellogg, Ryan, 2014, The effect of uncertainty on investment: evidence from Texas oil drilling,
American Economic Review 104, 1698-1734.
Kilian, Lutz and Daniel Murphy, 2012, Why agnostic sign restrictions are not enough:
Understanding the dynamics of oil market VAR models, Journal of the European Economic
Association 10, 1166-1188.
Lang, Larry, Eli Ofek, and René M. Stulz, 1996, Leverage, investment, and firm growth, Journal
of Financial Economics 40, 3-29.
Lehn, Kenneth, and Pengcheng Zhu, 2016, Debt, investment, and production in the U.S. oil
industry: An Analysis of the 2014 oil price shock.” mimeo.
Melzer, Brian T., 2017, Mortgage debt overhang: Reduced investment by homeowners at risk of
default, Journal of Finance 72, 575-612.
Modigliani, Franco and Merton Miller, 1958, “The cost of capital, corporation finance, and the
theory of investment.” American Economic Review 48: 261-297.
Myers, Stewart C. 1977, Determinants of corporate borrowing, Journal of Financial Economics
5, 147-175.
Nini, Greg, David C. Smith, and Amir Sufi, 2009, Creditor control rights and firm investment
policy, Journal of Financial Economics 92, 400-420.
38

Parrino, Robert, and Michael S. Weisbach, 1999, Measuring investment distortions arising from
stockholder-bondholder conflicts, Journal of Financial Economics 53, 3-42.
Rauh, Joshua D., and Amir Sufi , 2010, Capital structure and debt structure, The Review of
Financial Studies 23(12), 4242–4280,
Roberts, Michael R., and Amir Sufi, 2009, Renegotiation of financial contracts: Evidence from
private credit agreements, Journal of Financial Economics 93, 159-184.
Society of Petroleum Evaluation Engineers, 1995, Survey of Economic Parameters Used in
Property Evaluations.
Stulz, René M. and Herb Johnson, 1985, An analysis of secured debt, Journal of Financial
Economics 14, 501-521.
39

Panel A: Crude Oil Futures Prices over Time
Panel B: February 2015 and September 2014 Oil Price Futures Curves
Futures curves before contango and during contango
1.3
Futures curve as of September 2014

Futures curve as of February 2015
1.25
futures price relative to current spot price
1.2
1.15
1.1
1.05
0.95
0 5 10 15 20
Maturity of futures contract (Months)
Figure 1: Crude Oil Futures Price Dynamic

Panel A plots the differences between the 6-month (2-year) futures and the spot price normalized by the spot price
between 2012 and 2016. It is based on crude oil futures prices reported by Bloomberg. The shaded area represents
the contango period that we focus on in our study. Panel B plots the crude oil futures curve normalized by the spot
price at two different points in time: September 2014, prior to the contango; and February 2015, during the contango
episode.
40

Figure 2: Financial Implications of Delayed Production
Panel A presents an individual well’s cash flows under two mutually exclusive investment decisions: immediate
well completion and delaying of well completion by 1 month. See Appendix Figure A2 for the details of the
production cash flow calculations. Panel B plots the estimated gain in project-level NPV as a result of a decision to
delay production by a given number of month as a percentage of NPV of an immediately completed well. The well-
level after tax cash flows are based on the January 2015 NYMEX futures curve and on the average production
profile of 2,484 wells completed in North Dakota in 2014. The calculations assume a discount rate of 10%.
41

oil and gas
Reserves
Proved Unproven
Undeveloped Probable Possible

Developed
(25%-50%) (0%) (0%)
Producing Nonproducing
(90%-95%) (65%-75%)
Behind the
Shut-in*
pipe*
Figure 3: OCC oil and gas Reserves Classification

This figure presents OCC reserve classifications and suggested risk adjustment factors (in round brackets) applied in
determination of a borrowing base for oil and gas exploration and production lending. The OCC recommends that
the risk adjustment factors are to be applied to the NPV of the reserve value estimates. Alternatively, lenders can
make volumetric adjustments by applying the risk adjustment factors to oil production volumes rather than NPV.
Additionally, the OCC recommends accounting for the uncertainty of the production volumes for non-producing and
undeveloped reserves. * indicates classes that apply only to traditional production technology and is not applicable
to fracking wells. The engineering complexity of completion as well as deteriorating pressure in fracking wells
effectively eliminates ability to “shut-in” production to wait for higher prices. Source: Comptroller’s Handbook for
oil and gas Exploration and Production Lending, March 2016, Office of the Comptroller of the Currency.
42

Table I: Summary Statistics
This table contains summary statistics for firm-level financial variables, well completion time, and
the number of wells per firm. The sample covers 69 oil producers that spud a total of 3,557 wells
during the relevant time period. Market leverage is defined as total book debt divided by equity
market cap plus debt. Profitability is defined as the quarterly earnings before interest, taxes,
depreciation, and amortization scaled by lagged assets. Assets is total assets, and Tobin's Q is the
market value of equity plus debt divided by book assets. The financials are as of the last quarter prior
to onset of contango (September 2014). Panel B compares characteristics of low-leverage (bottom
four quintiles of market leverage ratio) firms to characteristics of high-leverage (top quintile) firms in
full sample as well as for the sub-sample of ABL firms. Additional variable definitions for Panel B
are as follows: Capex is capital expenditures divided by lagged assets; Dividends are dividends paid
in the quarter divided by lagged assets; Cash is cash and short-term investments divided by assets;
Tangibility is net property, plant, and equipment divided by assets; SGA is selling, general, and
administrative expense divided by sales; and Sales Growth is year-over-year percentage increase in
sales. * indicates significance at the 10% level, ** at the 5% level, and *** at the 1% level.
Panel A: Full Sample Summary Statistics

N Mean Std. Dev. p25 p50 p75
Months from Spud to Completion 3557 4.58 2.60 3.00 4.00 5.00
Market Leverage (%) 69 30.65 18.88 17.40 25.61 41.31
Profitability (%) 69 4.24 1.67 3.13 3.83 5.09
Total Assets ($millions) 69 30,600 70,890 2,171 5,833 17,846
Tobin's Q 69 1.17 0.44 0.84 1.08 1.43
Number of Wells Pre-Contango 69 20.32 26.96 3.00 10.00 26.00
Number of Wells Contango 69 31.23 35.35 6.00 19.00 40.00
Panel B: High Leverage and Low Leverage Firm Characteristics (Averages)

Full Sample ABL Firms
Low High Low High
(as of September, 2014) Leverage Leverage Difference Leverage Leverage Difference
Market Leverage (%) 23.5 61.7 -38.2*** 26.3 61.9 -35.60***
Total Assets ($millions) 36,780 3,978 32,802 3,954 4,279 -325***
Tobin's Q 1.24 0.86 0.38*** 1.34 0.87 0.47***
Profitability (%) 4.31 3.92 0.39 4.64 3.95 0.69
Capex (% of Total Assets) 8.58 7.74 0.84 11.85 7.38 4.47**
Dividends (% of Total Assets) 0.29 0.24 0.05 0.15 0.27 -0.12
Cash (% of Total Assets) 4.26 4.58 0.32 2.77 3.87 -1.1
Share of Tangible Assets (%) 83.03 85.57 2.54 89.02 86.07 2.95
SGA (% of Sales) 8.80 8.55 0.25 7.75 8.23 -0.48
Sales Growth (%) 43.13 46.82 -3.69 60.04 46.82 13.22

Table II: Leverage and Production Decisions, Univariate Test
This table reports the average number of months firms wait to complete wells after they are spudded
(started) in each period by quintiles of firms' leverage. Specifically, the columns report the average
number of months wells sit idle prior to completion across different leverage thresholds. The "Super
Contango" sample contains wells spudded during September, October, and November of 2014, and
completed during the "Super Contango" period in December 2014 and early 2015. The "Pre-super
Contango" sample contains wells spudded during September, October, and November of 2013 and
completed in December 2013 and early 2014 (this period forms the pre-event control period). Panel
A includes all fims in our sample, and Panel B is restricted to ABL firms. * indicates significance at
the 10% level, ** at the 5% level, and *** at the 1% level.
Panel A: Full Sample Panel B: ABL Firms

Super Pre-super Super Pre-super
Market Leverage Contango Contango Difference Contango Contango Difference
Quintile 5 (Highest) 3.75 3.57 -0.18 3.67 3.53 -0.14
Quintile 4 5.31 3.64 -1.67*** 5.45 3.64 -1.81***
Quintile 3 5.13 4.05 -1.08*** 4.25 3.63 -0.61***
Quintile 2 4.92 4.20 -0.72*** 3.95 3.25 -0.70***
Quintile 1 (Lowest) 5.06 4.05 -1.01*** 4.89 3.06 -1.84***

Table III: Differences in Investment Opportunities
This table provides project-level comparisons of well costs and initial production volumes across both
high-leverage and low-leverage firms. High-leverage firms are defined as those in the top 20% of the
market leverage distribution. The remainder of the firm population is defined as low-leverage firms.
Panel A reports univariate analysis of well characteristics, while Panel B reports the results of the
regression analysis controlling for geography and time fixed effects. The sample is composed of wells
for which initial production and costs are available from regulatory disclosures to the Oklahoma
Corporation Commission in 2013, 2014, and 2015. The Oklahoma Corporation Commission offers the
broadest available data set that captures both well costs and initial production. These data overlap with
the sample used in the main tests of this study, but only covers wells located in Oklahoma. The unit of
observation is well j in year t . The regression specification includes both time and township-level
geographic fixed effects (no well is farther than 8.48 miles from another well in a given township). In
Panel B, the standard errors are clustered by firm and are reported in square brackets below the
coefficient estimates. * indicates significance at the 10% level, ** at the 5% level, and *** at the 1%
level.
Panel A: Well Univariate Analysis
High Low
Leverage Leverage Difference
Cost $3,885,710 $7,306,493 $3,420,783*
N 228 706
Initial Production (Barrels of Oil Per Day) 241 364 -123***
N 738 1276
Panel B: Controling for Firm and Township Fixed Effects

Initial Ln(Initial
Dependent Variable Cost Ln(Cost)
Production Production)
(1) (2) (3) (4)
High Leverage Dummy -130,709 -0.055 24.989 0.021
[273,682] [0.036] [28.912] [0.110]
6 Sq Mile Geog j Fixed Effects Yes Yes Yes Yes
Time t Fixed Effects Yes Yes Yes Yes
N 934 934 1322 1319
R2 0.395 0.419 0.395 0.419

Table IV: Leverage and Production Decisions
This table reports the results of a difference-in-difference analysis that evaluates how project
completion decisions vary across firms based on market leverage. The unit of observation is at the
well j , firm i , year t level. The dependent variable is the number of months between the date a well
is spudded (started) and the date when it is completed and production begins. The pre-contango
period (Contangot = 0) is composed of wells started in September, October, and November of 2013,
the year prior to the super-contango period. The contango period (Contangot = 1) is composed of
wells started in September, October, and November of 2014, just before the oil market entered
contango in December of 2014. Leverage quintiles are based on a firm's market leverage (total debt
divided by debt plus equity market cap) measured as of September 30, 2014. Panel A reports the
results for the full sample. Panel B reports the results for the sub-sample of ABL firms where the
township fixed effects are estimated using all available projects via SUR setting. Standard errors
clustered at the firm level are reported in parentheses. ***, **, and * represent statistical
significance at 1%, 5%, and 10% levels respectively.
Dependent Variable =
Panel A: Full Sample Panel B: ABL Firms
Months to Production
Contangot 1.090*** 1.077*** 1.262*** 1.202* 0.967*** 1.264**
(0.331) (0.210) (0.347) (0.675) (0.343) (0.508)
Contangot × Leverage Q2i -0.276 -0.588
(0.631) (0.714)
Contangot × Leverage Q3i 0.147 -0.506

(0.553) (0.992)
Contangot × Leverage Q4i 0.184 0.251

(0.428) (0.812)
Contangot × Leverage Q5i -1.014** -1.002** -1.303* -1.071**
(0.442) (0.383) (0.722) (0.423)
Contangot × Continuous Leveragei -1.088 -1.676*
(0.848) (0.955)
Firm i Fixed Effects Yes Yes Yes Yes Yes Yes
6 Sq. Mile Geo j Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table V: The Role of Other Firm Characteristics
This table reports results of the difference-in-differences analysis reported in Table IV conditional on other
firm characteristics and their interaction with the contango dummy. Panel A utilizes continuous control
variables (profitability, size and Tobin's Q), while Panel B exploits dummies for the lowest quintile of
respective firm characteristics. Columns (1) through (4) report the results for the full sample. Columns (5)
through (8) report the results for the sub-sample of ABL firms where the township fixed effects are estimated
using all available projects via SUR setting. Appendix Table A.X presents specifications with all interaction
terms in the same regression. Standard errors clustered at the firm level are reported in parentheses. ***, **,
and * represent statistical significance at 1%, 5%, and 10% level respectively.
Dependent Variable = Months to Production

Panel A: Controlling for Firm Characteristic: Continuous Approach
Contangot 0.589 2.288* 0.786 0.803** 1.133 3.372 2.415* 0.553
(0.541) (1.192) (0.588) (0.305) (0.933) (2.113) (1.242) (0.503)
Contangot×Q5 High-levi -0.921** -1.225*** -0.907** -0.992** -1.103** -1.142*** -1.624** -0.881*
(0.399) (0.391) (0.429) (0.381) (0.431) (0.409) (0.649) (0.465)
Contangot×Profitabilityi 10.827 -3.620
(10.130) (20.067)
Contangot×Log Assetsi -0.121 -0.283
(0.121) (0.244)
Contangot×Tobin's Qi 0.234 -1.067
(0.423) (0.857)
Contangot×Capexi 3.980* 3.426
(2.291) (2.208)
Firm i FE Yes Yes Yes Yes Yes Yes Yes Yes
6 Sq. Mile Geo j FE Yes Yes Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 3557 1244 1244 1244 1244
2
R 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54
Panel B: Controlling for Firm Characteristic: Quintile Dummy Approach
Contangot 1.062*** 1.069*** 1.061*** 1.317*** 0.941*** 0.997** 0.984*** 0.970***
(0.224) (0.217) (0.241) (0.208) (0.345) (0.383) (0.347) (0.342)
Contangot×Q5 High-levi -1.052** -1.005***-1.022***-1.247*** -1.204** -1.095** -0.811* -1.077**
(0.416) (0.373) (0.377) (0.379) (0.488) (0.450) (0.421) (0.423)
Contangot×Q1 Profitabilityi 0.144 0.340
(0.372) (0.512)
Contangot×Q1 Log Assetsi 0.204 -0.230
(0.658) (0.516)
Contangot×Q1 Tobin's Qi 0.068 -0.528
(0.333) (0.418)
Contangot×Q1 Capexi -0.806* -1.970***
(0.440) (0.342)
Firm i FE Yes Yes Yes Yes Yes Yes Yes Yes
6 Sq. Mile Geo j FE Yes Yes Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 3557 1244 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54

Table VI: Production Decisions and Debt Renegotiations, Univariate Tests
The table reports the results of univariate analysis that compares firms' well completion rates around
debt renegotiation dates. The sample contains wells of firms with asset-based lending and with
borrowing base collateral redeterminations scheduled during February, March, April, and May of
2015. The unit of observation is well j for firm i in month t . The variable of interest equals one if
production was initiated in a given month and zero otherwise for months before and after completion.
Therefore the statistics reported can be interpreted as the share of spudded wells that were completed
in a given month. High-leverage firms are firms in the top quintile of the market leverage distribution
as of September 2014, the last quarter prior to the beginning of the super-contango period. * indicates
significance at the 10% level, ** at the 5% level, and *** at the 1% level.
Probability of Well Starting Production Difference

Well Startst=-1 -
Time 0 = month of debt renegotiation Well Startst=0
-3 -2 -1 0 1 2 3+
High Leverage 0.25 0.18 0.22 0.09 0.09 0.04 0.04 0.13***
N 129 238 238 238 238 238 238
Low Leverage 0.15 0.18 0.12 0.07 0.03 0.03 0.11 0.05***
N 626 640 640 640 640 640 640
DifferenceHigh - DifferenceLow 0.08**

p-value 0.013

Table VII: Production Decisions and Debt Renegotiations
The table reports the results of the difference-in-differences analysis that compares firms' well completion
rates around debt renegotiation dates. The unit of observation is well j for firm i in month t . The dependent
variable is a dummy equal one if production was initiated in a given month and zero otherwise. For firms
with borrowing base collateral redeterminations scheduled during February, March, April, and May of 2015
we create a set of firm-level dummies capturing different months relative to the firm-specific credit
agreement renegotiation date (month 0). These dummies are equal to zero for wells of firms that are not
subject to credit renegotiations (columns 1 through 3). In column 2 (3) the sample consists of high-leverage
(low-leverage) firms and all non-borrowing-base firms. In column 4 (5) the sample consists of all high-
leverage (low-leverage) firms and we draw our identification from heterogeneity in renegotiation dates. The
direct effect of the high-leverage dummy is absorbed by the firm fixed effects. Standard errors clustered at
the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at 1%, 5%,
and 10% levels respectively.
Dependent Variable = Well Start (1 if well High Lev + Low Lev + High Lev Low Lev
starts producing in month, 0 otherwise) All
Non-ABL Non-ABL Only Only
(1) (2) (3) (4) (5)
Month=-2 to Renegotiation D t 0.020 -0.062 0.017 -0.064 0.017
(0.038) (0.061) (0.039) (0.065) (0.045)
Month=-1 to Renegotiation Dt -0.003 -0.022 -0.007 0.001 -0.008
(0.030) (0.049) (0.031) (0.065) (0.028)
Month=0 to Renegotiation Dt -0.005 -0.135** -0.006 -0.117* -0.007
(0.036) (0.050) (0.037) (0.061) (0.039)
Month=1 to Renegotiation D t -0.034 -0.107** -0.034 -0.086 -0.034
(0.029) (0.046) (0.030) (0.063) (0.029)
Month=2 to Renegotiation D t -0.014 -0.111*** -0.015 -0.132** -0.015
(0.031) (0.037) (0.031) (0.065) (0.032)
Month≥3+ to Renegotiation Dt 0.074 -0.092* 0.074 -0.146* 0.074*
(0.052) (0.046) (0.052) (0.083) (0.042)
High Levi × Month=-2 to Renegotiation D t -0.078
(0.068)
High Levi × Month=-1 to Renegotiation D t -0.010
(0.054)
High Levi × Month=0 to Renegotiation D t -0.122**
(0.057)
High Levi × Month=1 to Renegotiation D t -0.060
(0.047)
High Levi × Month=2 to Renegotiation D t -0.090**
(0.042)
High Levi × Month≥3+ to Renegotiation Dt -0.162**
(0.065)
Firm i Fixed Effects Yes Yes Yes Yes Yes
Month t Fixed Effects Yes Yes Yes Yes Yes
6 Sq Mile Geog j Fixed Effects Yes Yes Yes Yes Yes
N 20,297 15,051 18,755 1,569 18,728
R2 0.052 0.056 0.049 0.080 0.049

Table VIII: High-Leverage Firm Characteristics
The table summarizes characteristics of oil and gas companies in the high-leverage sub-sample. Alongside market
leverage and assets in $millions it reports (i) whether the firm has an asset base lending facility (column 3) ; (ii) the
number of wells each firm has in our sample (column 4); (iii) average completion rate of this firm before Contango
(column 6) and during Contango (column 7). Column (8) evaluates whether an individual firm drives our core result. It
reports the interaction coefficient between Contango and the high-leverage-firm indicator variable from the analysis
corresponding to specification (2) of Table IV, after the respective high-leverage firm is dropped from the treatment
group. * indicates significance at the 10% level, ** at the 5% level, and *** at the 1% level.
Months from Spud

ABL % of Wells
to Completion
Revolving in High Table 4 (2):
Market Assets Credit # of Leverage Before
Contango Excluding
Company Name Leverage $millions Facility Wells Quintile Contango Firm
(1) (2) (3) (4) (5) (6) (7) (8)
Triangle Petroleum Corp 0.518 1,550 YES 15 4.6% 4.75 6.14 -1.02**
Comstock Resources Inc 0.529 2,322 YES 11 3.3% 4.00 3.40 -0.97**
Sanchez Energy Corp 0.532 3,104 YES 21 6.4% 2.22 2.83 -1.03**
Petroquest Energy Inc 0.539 729 YES 1 0.3% 4.00 NM -1.00**
Penn Virginia Corp 0.543 2,722 YES 38 11.6% 4.38 3.40 -0.86**
Linn Energy LLC 0.552 18,927 YES 5 1.5% 2.00 3.33 -1.00**
American Eagle Energy Corp 0.583 373 NO 9 2.7% 4.33 8.67 -1.18***
Sandridge Energy Inc 0.603 6,978 YES 117 35.6% 3.05 3.27 -0.89*
Exco Resources Inc 0.629 2,402 YES 25 7.6% 4.50 5.08 -1.09***
Halcon Resources Corp 0.679 5,934 YES 29 8.8% 4.67 5.07 -0.89**
Swift Energy Co 0.719 2,675 YES 3 0.9% NM 6.33 -1.00**
Jones Energy Inc 0.766 1,721 YES 7 2.1% NM 3.43 -1.00**
Midstates Petroleum Co Inc 0.825 2,280 YES 48 14.6% 2.96 3.17 -1.02**

Table IX: Leverage and Production Decisions: The Role of Outliers
This table reports the results of a difference-in-difference analysis that evaluates how the time to project
completion varies across firms based on leverage. Each panel replicates the DiD regression analysis
reported in Table IV of the paper after respective number of highest market-leverage firms (top of the
leverage distribution) is droppped from the sample. In each panel the remaining firms are reassigned to new
leverage quintiles that reflect the remaning set of firms. Standard errors clustered at the firm level are
reported in parentheses. ***, **, and * represent statistical significance at 1%, 5%, and 10% level
respectively. All regressions include firm-level fixed effects and 6-square-mile geography fixed effects.
Dependent variable=Month to
Production
Panel A: Top 1 High-leverage Firm Dropped from the Sample
Contangot 1.091*** 1.079*** 1.238*** 1.205* 0.973*** 1.329**
(0.330) (0.213) (0.375) (0.675) (0.360) (0.556)
Contangot×Leverage Q5 -0.932* -0.923** -1.232 -1.007**
(0.477) (0.438) (0.748) (0.474)
Contangot×Continuous Leverage -0.974 -1.923

(1.056) (1.235)
Contango×Q1-Q4 Lev. Quintile Controls Yes No No Yes No No
N 3509 3509 3509 1196 1196 1196
2
R 0.54 0.54 0.54 0.54 0.54 0.54
Panel B: Top 2 High-leverage Firms Dropped from the Sample
Contangot 1.087*** 1.079*** 1.238*** 1.196* 0.973*** 1.329**
(0.331) (0.213) (0.375) (0.671) (0.360) (0.556)
Contangot×Leverage Q5 -0.936* -0.923** -1.240 -1.007**
(0.476) (0.438) (0.745) (0.474)
(1.056) (1.235)
N 3502 3502 3502 1189 1189 1189
R2 0.54 0.54 0.54 0.54 0.54 0.54
Panel C: Top 3 High-leverage Firms Dropped from the Sample
Contangot 1.086*** 1.088*** 1.238*** 1.204* 0.972*** 1.329**
(0.331) (0.227) (0.375) (0.673) (0.360) (0.556)
Contangot×Leverage Q5 -0.616 -0.612 -1.246* -1.005**
(0.428) (0.406) (0.746) (0.474)
(1.056) (1.236)
N 3499 3499 3499 1186 1186 1186
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table X: Well Initial Production Before vs. After Debt Renegotiation
This table reports the univariate analysis of well-level initial production across high-leverage
firms before and after debt renegotiation. Data on initial production from wells was collected from
a subsample of firms that have operations in Texas (available through the Texas Railroad
Commission) and in Oklahoma (available through the Oklahoma Corporation Commission). *
indicates significance at the 10% level, ** at the 5% level, and *** at the 1% level.
Initial Production = Barrels of Oil per Day

Before Renegotiation After Renegotiation Difference
High Leverage Firms 417.34 291.71 125.64*
N 151 41
Initial Production = Log (Barrels of Oil per Day)
Before Renegotiation After Renegotiation Difference
High Leverage Firms 5.57 5.23 0.34*
N 151 41

Table XI: High Collateral Impact vs. Low Collateral Impact, Univariate Tests
This table reports a univariate analysis similar to that reported in Table VI. Panel A reports the
results for high-collateral-impact single-well-lease completion rates. The sample of high-
collateral-impact wells contains wells on leases that have no other completed and producing
wells as of September 2014. In contrast, Panel B reports the results for low-collateral-impact
multi-well leases. In this analysis the samples contain spudded wells located on leases with other
completed and producing wells as of September 2014. * indicates significance at the 10% level,
** at the 5% level, and *** at the 1% level.
Well Starting Production Dummy Difference

Well Startst=-1 - Well
Time 0 = month of debt renegotiation Startst=0
-3 -2 -1 0 1 2 3+
Panel A: Single Well Lease (High Collateral Impact)
High Leverage 0.29 0.18 0.26 0.09 0.09 0.05 0.03 0.18***
Low Leverage 0.14 0.16 0.12 0.09 0.04 0.05 0.13 0.03
DifferenceHigh - DifferenceLow 0.15***
p-value 0.002
Well Starting Production Dummy Difference

Time 0 = month of debt renegotiation Startst=0
-3 -2 -1 0 1 2 3+
Panel B: Multi Well Lease (Low Collateral Impact)
High Leverage 0.17 0.18 0.15 0.10 0.07 0.02 0.07 0.05**
Low Leverage 0.17 0.19 0.12 0.06 0.03 0.01 0.09 0.06**
DifferenceHigh - DifferenceLow -0.01

p-value 0.891

Table XII: High Collateral Impact vs. Low Collateral Impact
This table reports the results of a regression form of difference-in-differences for completion activity of wells
around credit agreement renegotiations for high-collateral-impact (single lease wells) and low-collateral-impact
(multi-lease wells). The sample contains wells of firms with asset-based lending that have borrowing base
collateral redeterminations scheduled during February, March, April, and May of 2015. The unit of observation
is well j for firm i in month t . The dependent variable equals one if production was initiated in that month and
zero otherwise, including months before and after completion. A negative coefficient can be interpreted as a
relative decline in completion activity. Month = 0 is the month of the credit agreement renegotiation. All
regressions include firms that are not subject to credit renegotiations as part of the control group. This control
group is designed to control for general time trends in completion activity over this period. The direct effect of
the high leverage dummy is absorbed by the firm fixed effects. Standard errors clustered at the firm level are
reported in round parenthesis. ***, **, and * represent statistical significance at 1%, 5%, and 10% levels
respectively.
Dependent Variable = Well Start (1 if well High Collateral Impact Low Collateral Impact
starts producing in month, 0 otherwise) High Low High Low
All All
Leverage Leverage Leverage Leverage
(1) (2) (3) (4) (5) (6)
Month=-2 to Renegotiation Dt 0.023 -0.124 0.019 0.020 0.029 0.020
(0.044) (0.079) (0.045) (0.054) (0.083) (0.055)
Month=-1 to Renegotiation Dt -0.004 -0.046 -0.010 0.012 0.004 0.011
(0.041) (0.096) (0.041) (0.043) (0.062) (0.043)
Month=0 to Renegotiation Dt 0.011 -0.213*** 0.006 -0.003 -0.020 0.000
(0.054) (0.071) (0.054) (0.041) (0.061) (0.042)
Month=1 to Renegotiation Dt -0.023 -0.182*** -0.026 -0.030 -0.002 -0.026
(0.041) (0.061) (0.042) (0.033) (0.053) (0.033)
Month=2 to Renegotiation Dt 0.016 -0.179*** 0.014 -0.027 -0.017 -0.027
(0.049) (0.054) (0.049) (0.029) (0.048) (0.029)
Month≥3+ to Renegotiation Dt 0.106 -0.187*** 0.106 0.059 0.048 0.059
(0.064) (0.061) (0.065) (0.055) (0.057) (0.055)
High Levi × Month=-2 to Renegotiation Dt -0.144 0.013
(0.090) (0.097)
High Levi × Month=-1 to Renegotiation Dt -0.035 -0.001
(0.104) (0.072)
High Levi × Month=0 to Renegotiation Dt -0.211** -0.013
(0.084) (0.068)
High Levi × Month=1 to Renegotiation Dt -0.146** 0.041
(0.067) (0.053)
High Levi × Month=2 to Renegotiation Dt -0.188*** 0.018
(0.067) (0.046)
High Levi × Month≥3+ to Renegotiation Dt -0.288*** -0.009
(0.083) (0.074)
Month t Fixed Effects Yes Yes Yes Yes Yes Yes
6 Sq Mile Geog j Fixed Effects Yes Yes Yes Yes Yes Yes
N 11160 8868 10194 9137 6183 8561
R2 0.046 0.052 0.040 0.071 0.075 0.071

Online Appendix
Appendix A1
Effect of Well Completion on Collateral Values: Numerical Example
This Appendix offers a numerical example in an attempt to quantify the tradeoffs between
completing a well early and retaining liquidity for later well completion. We identify dimensions
that affect the tradeoffs and then discuss why these tradeoffs might differ across high-and low -
leverage firms.
First, completion of a new well affects the collateral value of this specific well and expands a
company’s access to credit. An average well in our sample costs about $3.5 million to complete
and generates about $8.907 million in PV of production-related incremental cash in-flows during
contango (note: the drilling and spudding capex represent sunk cost as we only consider spudded
wells). Multi well lease completion allows the lender to move a well from the Proven
Undeveloped resource category (see OCC guidance in Figure 5 of the paper) to the Proven
Developed Producing resource category. This results in an increase of the collateral value from,
on average, 37.5% of the PV to 92.5% of the PV, or $8.907 million × (92.5% - 37.5%) =
$4.899 million. Since a typical bank lends up to 50% of the collateral value, this leads to an
increase in the borrowing base of $2.449 million, which is below the completion investment of
$3.5 million.
If a well is completed on a single well lease, this well is moved from the Unproven resource
category (0% on OCC guidance) to the Proven Developed resource category. This results in an
increase in the collateral value of $8.239 million (=$8.907 million × 92.5%) and an increase in
the borrowing base of $4.119 million, which exceeds the completion cost of $3.5 million.
Second, completion of a well on a single-well lease also enhances the collateral value of other
resources on this lease. The remaining potential wells on the lease move from the Unproven
resource category to the Proven Undeveloped category (37.5% value discount). An average lease
in our sample has 5.437 wells. With the completion of one well, the remaining 4.437 wells are
now generating incremental collateral value (after factoring in the drilling and completion costs
of those wells) of $4.053 million (=4.437 × ($8.907 million - $3.500 million (completion cost) -
43

$2.971 (drilling cost)) × 37.5%) and command an increase in the borrowing base of
$2.026 million (50% of collateral value). Combined, completion of a single-well lease has the
potential to increase the borrowing base by $6.146 million (= $4.119 million + $2.026 million).
As such, a decision to invest $3.5 million on a single-well lease allows a firm to access
additional finance equal to 175% (=$6.146 million / $3.5 million) of the invested capital.19
From one perspective, these numerical examples illustrate that a completion of a (single-well
lease) well is beneficial for any company as it expands access to credit irrespective of its timing.
The advantage of our setting is that contango imposes additional costs on immediate well
completion as it reduces the PV of production related incremental cash flows. During contango,
firms have to trade-off the immediate expansion of their access to credit against a 5% reduction
in NPV of their current investments (see, Figure 5). While quantifying the value of continued
access to finance in this setting is challenging, it is clear that low-leverage firms with well-
underutilized borrowing bases would benefit significantly less from expanding their debt
capacity than high-leverage firms that might be cut-off from access to credit.
Finally, the increase in the borrowing base stemming from well completion has the potential to
affect the cost of new and existing debt even before renegotiations. A typical credit agreement in
our sample is characterized by a staggered interest rate schedule closely tied to the credit line
utilization rate. Firms typically experience a 25 bps increases in the cost of debt every time the
utilization rate moves from [0;0.25[ to [0.25; 0.5[ to [0.5; 0.75[ to [0.75; 0.9[ to [0.9; 1].20
Consequently, an increase in the borrowing base might allow firms to curtail their cost of
existing debt alongside securing access to additional debt capacity. While the interest rate
increases are gradual and affect both low- and high-leverage firms similarly, the marginal benefit
19
Notably, spending $3.5 million to repay debt does not compare favorably to investing into well completion. First,
it generates only $3.5 million of incremental debt capacity as it does not change the value of collateral and
borrowing base. Second, it deprives the company of the profits associated with a positive NPV project (well
completion).
20
See, e.g., https://www.sec.gov/Archives/edgar/data/1533924/000119312512268479/d365613dex101.htm or
https://www.sec.gov/Archives/edgar/data/1349436/000119312514380467/d808277dex101.htm.
44

of lowering the utilization rate via a $3.5 million investment is higher for high-leverage firms
that, in the absence of a borrowing base increase, might experience a 25 bps hike in the cost of
capital for significantly larger debt amounts. Such pricing of debt creates additional incentives
for high-leverage firms to accelerate production.
45

Figure A1: Density of Oil Wells in Woodford Shale, Oklahoma
This figure illustrates oil and gas development operations in the Woodford Shale, Oklahoma. Each square
represents a 6 mile by 6 mile area (township), which is composed of 36 individual drilling tracts (leases). Each dot
represents the wellhead, and each individual line illustrates the direction of horizontal wellbore for each individual
well. Each township is composed of 36 separate one mile by one mile drilling tracts, each of which could have a
single well or multiple wells. The wells are color coded by operator.
46

$350,000
$300,000
$250,000
$200,000
$150,000
$100,000
$50,000
$-
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Months
Figure A2: Production Cash Flows from a Representative Well

This figure plots the well-level after tax cash flows implied by January 2015 NYMEX futures curve and based on
the average production profile of 2,484 wells completed in North Dakota in 2014. The average well in North Dakota
has an initial monthly production of 12,961 BBLs of oil, which declines to 4,833 BBLs per month by the well’s 12th
month of production. We use actual average production data for the first 12 month. Beyond first 12 month, we
estimate production based on an Arps decline curve methodology which is consistent with Covert and Kellogg
(2017) and the guidelines of the Society of Petroleum Engineers. We adjust the revenue for state severance tax, lease
royalties, depreciation, monthly lease operating costs, and income taxes. The assumptions are as follows: (i)
severance tax of 3% is estimated sample average; (ii) DrillingInfo data provides an estimate of median royalty rate
in the area of 18.75%; (iii) the depreciation estimates are based on $6.47 million capex requirement per well which
includes drilling and completion capex; (iv) lease operating costs are based on estimates from 10-Ks. Finally, (v)
marginal income tax rate is assumed to be 38%.
47

Table A.I: Leverage and Production Decisions, Univariate Test: Alternative Measures of Leverage
This table reports the average number of months firms wait to complete wells after they are spudded (started) in each period by
quintiles of firms' leverage. Panel A reports the results using the book leverage measure, while Panel B utilizes firms' interest coverage
ratio to capture leverage. To aid comparability, firms sorted by interest coverage are sorted such that the lowest interest coverage firms
are the highest leverage. The empirical design is similar to that reported in Table II of the paper. Specifically, the columns report the
average number of months wells sit idle prior to completion across different leverage thresholds. The "Super Contango" sample
contains wells spudded during September, October, and November of 2014, and completed during the "Super Contango" period in
December 2014 and early 2015. The "Pre-super Contango" sample contains wells spudded during September, October, and November
of 2013 and completed in December 2013 and early 2014. This period forms the pre-event control period. * indicates significance at
the 10% level, ** at the 5% level, and *** at the 1% level.

Super Pre-super Super Pre-super
Difference Difference
Contango Contango Contango Contango
Panel A: Book Leverage
Leverage Quintile 5 (Highest Leverage) 3.71 3.45 -0.26 3.62 3.41 -0.21
Leverage Quintile 4 5.36 4.00 -1.36*** 5.21 3.83 -1.38***
Leverage Quintile 3 4.63 3.88 -0.75*** 3.79 3.52 -0.27*
Leverage Quintile 1 (Lowest Leverage) 5.22 4.12 -1.10*** 5.18 2.80 -2.38***
Panel B: Interest Coverage Ratio

Leverage Quintile 5 (Highest Leverage) 3.67 3.41 -0.26 3.58 3.37 -0.21
Leverage Quintile 1 (Lowest Leverage) 5.67 4.30 -1.37*** 5.81 3.14 -2.67***

Table A.II: Leverage and Production Decisions: Alternative Measures of Leverage
This table reports the results of a difference-in-difference analysis that evaluates how the time to project completion varies across firms based on
leverage. Panel A reports the results using book leverage measure, while Panel B utilizes firms' coverage ratio to capture leverage. To aid comparability,
firms sorted by interest coverage are sorted such that the lowest interest coverage firms are the highest leverage. The empirical design is similar to that
employed in Table IV of the paper. The unit of observation is at the well j , firm i , year t level. The dependent variable is the number of months
between the date a well is spudded (started) and the date when it is completed and production begins. The pre-contango period (Contangot = 0) is
composed of wells started in September, October, and November of 2013, the year prior to the super-contango period. The contango period (Contangot =
1) is composed of wells started in September, October, and November of 2014, just before the oil market entered contango in December of 2014.
Columns (1) through (3) reports the results for the full sample. Columns (4) through (6) reports the results for the sub-sample of ABL firms. Standard
errors clustered at the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels
respectively.

Contangot 1.204*** 1.039*** 0.175 2.016*** 0.840** 1.589*
(0.270) (0.214) (1.176) (0.523) (0.336) (0.836)
Contangot × Leverage Q2 -0.521 -0.855

(0.554) (0.821)
Contangot × Leverage Q3 -0.299 -1.955***

(0.469) (0.596)
Contangot × Leverage Q4 0.290 -0.446

(0.469) (0.608)
Contangot × Leverage Q5 -0.881* -0.710* -0.712*

(0.455) (0.423) (0.403)
Contangot × Continuous Leverage 1.505 -1.589

(2.140) (1.469)
Firm and 6 Sq. Mile Geog Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.180*** 1.080*** 1.011*** -0.075 0.978*** 0.632
(0.247) (0.210) (0.246) (0.460) (0.347) (0.387)
Contangot × Leverage Q2 -0.430 2.136***

(0.485) (0.638)
Contangot × Leverage Q3 0.125 0.737

(0.630) (0.667)
Contangot × Leverage Q4 0.065 1.278**

(0.454) (0.619)
Contangot × Leverage Q5 -1.122*** -1.021*** -1.097**

(0.388) (0.378) (0.428)
Contangot × Continuous Leverage -0.001 0.004

(0.003) (0.023)
Firm and 6 Sq Mile Geog Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 3557 3557 3557
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Table A.III: Production Decisions and Debt Renegotiations, Univariate Tests: Alternative Measures of Leverage
This table reports the results of univariate analysis that compares firms' well completion rates around debt renegotiation dates similar to the one reported in Table VI of
the paper. The sample contains wells of firms with asset-based lending and with borrowing-base collateral redeterminations scheduled during February, March, April,
and May of 2015. The unit of observation is well j for firm i in month t . The variable of interest equals one if production was initiated in a given month and zero
otherwise for months before and after completion. Therefore the statistics reported can be interpreted as the share of spudded wells that were completed in a given
month. Panel A reports the results using the book leverage measure, while Panel B utilizes firms' interest coverage ratio to capture leverage. High leverage firms are
firms in the top quintile of the respective leverage measures as of September 2014, the last quarter prior to the beginning of the super-contango period. * indicates
Probability of Well Starting Production Difference

Time 0 = month of debt renegotiation
Startst=0
-3 -2 -1 0 1 2 3+
High Leverage 0.27 0.21 0.25 0.09 0.10 0.05 0.05 0.16***
N 70 193 193 193 193 193 193
Low Leverage 0.16 0.17 0.11 0.07 0.03 0.03 0.10 0.04**
N 670 670 670 670 670 670 670
p-value 0.001
-3 -2 -1 0 1 2 3+
High Leverage 0.22 0.18 0.25 0.08 0.08 0.05 0.04 0.17***
N 97 208 208 208 208 208 208
Low Leverage 0.16 0.18 0.11 0.07 0.03 0.02 0.11 0.03**
N 643 655 655 655 655 655 655
p-value 0.000

Table A.IV: Production Decisions and Debt Renegotiations: Alternative Measures of Leverage
This table reports the results of an analysis similar to the one reported in Table VII of the paper. The difference-in-differences analysis evaluates firms'
well completion rates around debt renegotiation dates. The unit of observation is well j for firm i in month t . The dependent variable is a dummy equal
to one if production was initiated in a given month and zero otherwise. For firms with borrowing base collateral redeterminations scheduled during
February, March, April, and May of 2015 we create a set of firm-level dummies capturing different months relative to firm-specific credit agreement
renegotiation date (month 0). These dummies are always equal to zero for wells of control firms that are not subject to credit renegotiations. Panel A
reports the results using book leverage measure, while Panel B utilized firm coverage ratio to capture leverage. Standard errors clustered at the firm
level are reported in round brackets. ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels respectively.
Panel A: Book Leverage Panel B: Interest Coverage Ratio

Dependent Variable = Well Start (1 if well starts
producing in month, 0 otherwise)
High Leverage Low Leverage All High Leverage Low Leverage All
Month=-2 to Renegotiation Dummy -0.046 0.004 0.005 -0.055 0.015 0.016

(0.067) (0.039) (0.038) (0.067) (0.038) (0.038)
Month=-1 to Renegotiation Dummy -0.020 -0.012 -0.011 0.019 -0.018 -0.016
(0.062) (0.030) (0.029) (0.071) (0.029) (0.028)
Month=0 to Renegotiation Dummy -0.165*** -0.008 -0.008 -0.142** -0.006 -0.004
(0.060) (0.036) (0.036) (0.054) (0.036) (0.036)
Month=1 to Renegotiation Dummy -0.120* -0.038 -0.039 -0.112** -0.032 -0.033
(0.067) (0.029) (0.028) (0.044) (0.030) (0.030)
Month=2 to Renegotiation Dummy -0.126** -0.019 -0.019 -0.100** -0.018 -0.018
(0.054) (0.030) (0.030) (0.047) (0.030) (0.030)
Month≥3+ to Renegotiation Dummy -0.105 0.066 0.066 -0.091* 0.072 0.072
(0.065) (0.050) (0.050) (0.049) (0.051) (0.051)
High Leveragei × Month=-2 to Renegotiation -0.050 -0.067
(0.074) (0.073)
High Leveragei × Month=-1 to Renegotiation -0.001 0.041
(0.065) (0.076)
High Leveragei × Month=0 to Renegotiation -0.149** -0.131**
(0.063) (0.061)
High Leveragei × Month=1 to Renegotiation -0.064 -0.065
(0.065) (0.046)
High Leveragei × Month=2 to Renegotiation -0.097* -0.075
(0.056) (0.051)
High Leveragei × Month≥3+ to Renegotiation -0.167** -0.160**
(0.077) (0.066)
N 14,765 19,041 20,297 14,939 18,867 20,297
2
R 0.056 0.050 0.052 0.057 0.050 0.052

Table A.V: Leverage and Production Decisions: Placebo Tests, Prior Time Periods
This table reports the results of a difference-in-difference analysis that evaluates how the time to project completion varies across firms based
on leverage in earlier placebo time periods. Each panel replicates the DiD regression analysis similar to Table IV except conducted on the
2012-2013 (Panel A) and 2011-12 (Panel B) time periods, when there was no super contango. Firms are sorted into leverage quintile bins
based on their leverage as of Q3 in the placebo year. The regression specifications include firm and township (geography) fixed effects.
Standard errors clustered at the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at the 1%, 5%,
and 10% level respectively.
Dependent Variable = Dependent Variable =
Months to Production Ln(Months to Production)
Panel A: 2012-2013 Placebo Tests
Dummy2013 0.047 -0.184 0.104 0.014 -0.028 0.026
(0.512) (0.186) (0.348) (0.101) (0.034) (0.034)
Dummy2013 × Leverage Q2 -0.055 0.007
(0.569) (0.107)
Dummy2013 × Leverage Q3 -0.430 -0.109
(0.569) (0.111)
Dummy2013 × Leverage Q4 -0.353 -0.059
(0.596) (0.118)
Dummy2013 × Leverage Q5 0.245 0.473 0.068 0.108
(0.638) (0.432) (0.146) (0.113)
Dummy2013 × Continuous Leverage -0.862 -0.158
(1.085) (0.228)
6 Sq. Mile Geog j Fixed Effects Yes Yes Yes Yes Yes Yes
N 2803 2803 2803 2803 2803 2803
2
R 0.58 0.58 0.57 0.59 0.59 0.57
Panel B: 2011-2012 Placebo Tests
Dummy2012 0.222 0.116 -0.274 0.013 0.015 -0.040
(0.621) (0.294) (0.485) (0.130) (0.060) 0.098
Dummy2012× Leverage Q2 0.248 0.080
(0.759) (0.158)
Dummy2012 × Leverage Q3 -0.767 -0.119
(0.745) (0.158)
Dummy2012 × Leverage Q4 0.500 0.111
(0.748) (0.147)
Dummy2012 × Leverage Q5 0.525 0.631 0.091 0.088
(0.687) (0.407) (0.146) (0.087)
Dummy2012 × Continuous Leverage 1.611 0.227
(1.126) (0.229)
N 1928 1928 1928 1928 1928 1928
2
R 0.55 0.55 0.53 0.56 0.55 0.54

Table A.VI: Leverage and Production Decisions: The Role of Outliers
This table reports the results of a difference-in-difference analysis that evaluates how the time to project completion varies across firms based on leverage.
Each panel replicates the DiD regression analysis reported in Table IV of the paper after respective number of highest market-leverage firms (top of the
leverage distribution) is droppped from the sample. In each panel the remaining firms are reassigned to new leverage quintiles that reflect the remaning set
of firms. The unit of observation is at the well j , firm i , year t level. The dependent variable is the number of months between the date a well is spudded
(started) and the date when it is completed and production begins. The pre-contango period (Contangot = 0) is composed of wells started in September,
October, and November of 2013, the year prior to the super-contango period. The contango period (Contangot = 1) is composed of wells started in
September, October, and November of 2014, just before the oil market entered contango in December of 2014. Leverage quintiles are based on a firm's
market leverage (total debt divided by debt plus equity market cap) measured as of September 30, 2014. Columns (1) thorugh (3) report the results for the
full sample. Columns (4) thorugh (6) report the results for the sub-sample of ABL firms where the township fixed effects are estimated using all available
projects via SUR setting. Standard errors clustered at the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at the
1%, 5%, and 10% level respectively.
Dependent variable=Month to Production Full Sample ABL Firms

Contangot 1.091*** 1.079*** 1.238*** 1.205* 0.973*** 1.329**
(0.330) (0.213) (0.375) (0.675) (0.360) (0.556)

(0.454) (0.838)
(0.553) (0.993)
(0.630) (0.712)
Contangot × Leverage Q5 -0.932* -0.923** -1.232 -1.007**
(0.477) (0.438) (0.748) (0.474)
Contangot × Continuous Leverage -0.974 -1.923
(1.056) (1.235)
FirmFEi Yes Yes Yes Yes Yes Yes
6 Sq Mile Geog FE Yes Yes Yes Yes Yes Yes
N 3509 3509 3509 1196 1196 1196
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table AVI (Cont.)
Contangot 1.087*** 1.079*** 1.238*** 1.196* 0.973*** 1.329**
(0.331) (0.213) (0.375) (0.671) (0.360) (0.556)
Contangot × Leverage Q2 -0.004 0.397
(0.417) (0.835)
(0.636) (0.991)
(0.633) (0.708)
Contangot × Leverage Q5 -0.936* -0.923** -1.240 -1.007**
(0.476) (0.438) (0.745) (0.474)
(1.056) (1.235)
N 3502 3502 3502 1189 1189 1189
R2 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.086*** 1.088*** 1.238*** 1.204* 0.972*** 1.329**
(0.331) (0.227) (0.375) (0.673) (0.360) (0.556)
(0.485) (0.837)
(0.601) (0.988)
(0.695) (0.709)
Contangot × Leverage Q5 -0.616 -0.612 -1.246* -1.005**
(0.428) (0.406) (0.746) (0.474)
(1.056) (1.236)
N 3499 3499 3499 1186 1186 1186
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table AVI (Cont.)
Panel D: Top 4 High-leverage Firms Dropped from the Sample
Contangot 1.086*** 1.098*** 1.180*** 1.206* 1.007*** 1.235**

(0.331) (0.228) (0.380) (0.674) (0.370) (0.559)

(0.487) (0.808)

(0.607) (0.954)

(0.711) (0.706)
Contangot × Leverage Q5 -0.566 -0.573 -1.171 -0.962**

(0.419) (0.395) (0.740) (0.471)

(1.102) (1.297)
N 3470 3470 3470 1157 1157 1157
R2 0.54 0.54 0.54 0.54 0.54 0.54
Panel E: Top 5 High-leverage Firms Dropped from the Sample

Contangot 1.024*** 1.098*** 1.182*** 0.938 1.007*** 1.286**
(0.356) (0.228) (0.567) (0.784) (0.370) (0.567)
(0.504) (0.894)
(0.620) (1.043)
(0.706) (0.898)
Contangot × Leverage Q5 -0.527 -0.595 -1.009 -1.070**
(0.449) (0.418) (0.846) (0.474)
(1.183) (1.366)
N 3445 3445 3445 1132 1132 1132
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table A.VII: Production Decisions and Debt Renegotiations, Univariate Tests: The Role of Outliers
This table reports the results of a univariate analysis that compares firms' well completion rates around debt renegotiation dates. Each panel
replicates the univariate analysis reported in Table VI of the paper after respective number of highest market-leverage firms (top of the
leverage distribution) is dropped from the sample. In each panel the remaining firms are reassigned to new leverage quintiles that reflect the
remaning set of firms. The sample contains wells of firms with asset-based lending that have borrowing base collateral redeterminations
scheduled during February, March, April, and May of 2015. The unit of observation is well j for firm i in month t . The variable of interest
equals one if production was initiated in a given month and zero otherwise for months before and after completion. Therefore the statistics
reported can be interpreted as the share of spudded wells that were completed in a given month. Panel A reports the results using the book
leverage measure, while Panel B utilizes firm coverage ratio to capture leverage. High-leverage firms are firms in the top quintile of the
respective leverage measures as of September 2014, the last quarter prior to the beginning of the super-contango period. * indicates
Probability of Well Starting Production Well Startst=-1 -

Time 0 = month of debt renegotiation Well Startst=0
-3 -2 -1 0 1 2 3+
High Leverage 0.23 0.18 0.19 0.08 0.09 0.04 0.05 0.11***
N 132 241 241 241 241 241 241
Low Leverage 0.15 0.18 0.11 0.08 0.03 0.03 0.11 0.04**
N 581 595 595 595 595 595 595
DifferenceHigh - DifferenceLow 0.07**
p-value 0.032
High Leverage 0.20 0.17 0.20 0.07 0.09 0.04 0.05 0.13***
N 140 249 249 249 249 249 249
Low Leverage 0.15 0.18 0.10 0.08 0.03 0.03 0.11 0.02
N 566 580 580 580 580 580 580
p-value 0.001
High Leverage 0.20 0.17 0.21 0.06 0.09 0.04 0.05 0.15***
N 137 246 246 246 246 246 246
Low Leverage 0.15 0.18 0.10 0.08 0.03 0.03 0.11 0.02
N 566 580 580 580 580 580 580
p-value 0.000
High Leverage 0.20 0.18 0.21 0.04 0.09 0.05 0.03 0.17***
N 137 234 234 234 234 234 234
Low Leverage 0.15 0.18 0.10 0.08 0.03 0.03 0.11 0.02
N 566 578 578 578 578 578 578
p-value 0.000
-3 -2 -1 0 1 2 3+
High Leverage 0.20 0.18 0.23 0.05 0.08 0.03 0.03 0.19***
N 138 221 221 221 221 221 221
Low Leverage 0.15 0.18 0.10 0.08 0.03 0.03 0.11 0.02
N 565 577 577 577 577 577 577
p-value 0.000

Table A.VIII: Production Decisions and Debt Renegotiations: The Role of Outliers
This table reports the results of the difference-in-differences ananlysis similar to the one reported in Table VII of the paper after respective number of
highest leverage firms is droppped from the sample. In each panel the firms are reassigned to new leverage quintiles that reflect the remaning set of firms.
The sample contains wells of firms with asset-based lending. The unit of observation is well j for firm i in month t . The dependent variable is a dummy
equal one if production was initiated in a given month and zero otherwise. For firms with borrowing base collateral redeterminations scheduled during
February, March, April, and May of 2015 we create a set of firm-level dummies capturing the number of months relative to the firm-specific credit
agreement renegotiation date (month 0). These dummies are always equal to zero for wells of control firms that are not subject to credit renegotiations.
Standard errors clustered at the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at the 1%, 5%, and 10% level
respectively.
Dependent Variable = Well Start Panel A: Top 1 High-leverage Firm Dropped from Panel B: Top 2 High-leverage Firms Dropped from
(1 if well starts producing in month, 0 the Sample the Sample
otherwise) High Leverage Low Leverage All High Leverage Low Leverage All
Month=-2 to Renegotiation Dt -0.048 0.020 0.025 -0.036 0.024 0.028
(0.058) (0.041) (0.041) (0.059) (0.041) (0.041)
Month=-1 to Renegotiation Dt -0.038 -0.007 -0.003 -0.001 -0.018 -0.015
(0.042) (0.032) (0.031) (0.054) (0.030) (0.029)
Month=0 to Renegotiation Dt -0.118*** 0.001 0.003 -0.105** 0.001 0.003
(0.043) (0.038) (0.037) (0.044) (0.038) (0.038)
Month=1 to Renegotiation Dt -0.083* -0.033 -0.033 -0.063 -0.035 -0.035
(0.043) (0.031) (0.030) (0.043) (0.031) (0.031)
Month=2 to Renegotiation Dt -0.091** -0.012 -0.011 -0.069* -0.014 -0.014
(0.035) (0.032) (0.032) (0.038) (0.032) (0.032)
Month≥3+ to Renegotiation Dt -0.066 0.080 0.080 -0.049 0.081 0.081
(0.046) (0.056) (0.056) (0.044) (0.057) (0.057)
High Leveragei × Month=-2 to Renegotiation Dt -0.070 -0.061
(0.064) (0.066)
High Leveragei × Month=-1 to Renegotiation Dt -0.026 0.024
(0.047) (0.058)
High Leveragei × Month=0 to Renegotiation Dt -0.115** -0.102*
(0.052) (0.053)
High Leveragei × Month=1 to Renegotiation Dt -0.039 -0.016
(0.044) (0.044)
High Leveragei × Month=2 to Renegotiation Dt -0.073* -0.049
(0.040) (0.043)
High Leveragei × Month≥3+ to Renegotiation Dt -0.143** -0.127*
(0.068) (0.068)
N 15,222 18,395 20,108 15,286 18,275 20,052
R2 0.055 0.049 0.051 0.054 0.050 0.051

Table AVIII (Cont.)
Dependent Variable = Well Start (1 if well starts producing in month, 0 otherwise)
Panel C: Top 3 High-leverage Firms Dropped from Panel D: Top 4 High-leverage Firms Dropped
the Sample from the Sample

(0.058) (0.041) (0.041) (0.057) (0.041) (0.041)
Month=-1 to Renegotiation Dt -0.008 -0.018 -0.015 0.002 -0.020 -0.017
(0.054) (0.030) (0.029) (0.054) (0.030) (0.029)
Month=0 to Renegotiation Dt -0.126*** 0.001 0.003 -0.139*** 0.001 0.004
(0.038) (0.038) (0.038) (0.035) (0.038) (0.038)
Month=1 to Renegotiation Dt -0.071 -0.035 -0.034 -0.065 -0.035 -0.033
(0.043) (0.031) (0.031) (0.044) (0.031) (0.031)
Month=2 to Renegotiation Dt -0.077** -0.014 -0.014 -0.072* -0.014 -0.013
(0.038) (0.032) (0.032) (0.038) (0.032) (0.032)
Month≥3+ to Renegotiation Dt -0.056 0.081 0.081 -0.071* 0.082 0.081
(0.045) (0.057) (0.057) (0.041) (0.057) (0.057)
(0.065) (0.065)
High Leveragei × Month=-1 to Renegotiation Dt 0.017 0.028
(0.058) (0.058)
High Leveragei × Month=0 to Renegotiation Dt -0.122** -0.137***
(0.048) (0.046)
(0.044) (0.045)
(0.043) (0.043)
High Leveragei × Month≥3+ to Renegotiation Dt -0.134* -0.150**
(0.068) (0.066)
MonthFEt Yes Yes Yes Yes Yes Yes
6 Sq Mile Geog FEj Yes Yes Yes Yes Yes Yes
N 15,259 18,275 20,025 15,187 18,263 19,941
2
R 0.055 0.050 0.052 0.056 0.050 0.052
1 2 3 1 2 3

Table AVIII (Cont.)
Dependent Variable = Well Start
(1 if well starts producing in month, 0 otherwise)
Panel E: Top 5 High-leverage Firms Dropped from
the Sample
High Leverage Low Leverage All
Month=-2 to Renegotiation Dt -0.037 0.023 0.027

(0.061) (0.041) (0.041)
Month=-1 to Renegotiation Dt 0.024 -0.021 -0.017
(0.051) (0.030) (0.029)
Month=0 to Renegotiation Dt -0.130*** 0.001 0.005
(0.034) (0.038) -0.038
Month=1 to Renegotiation Dt -0.071 -0.035 -0.033
(0.043) (0.031) (0.031)
Month=2 to Renegotiation Dt -0.083** -0.014 -0.013
(0.035) (0.032) (0.032)
Month≥3+ to Renegotiation Dt -0.070* 0.082 0.082
(0.041) (0.057) (0.057)
High Leveragei × Month=-2 to Renegotiation Dt -0.061
(0.067)
High Leveragei × Month=-1 to Renegotiation Dt 0.050
(0.056)
High Leveragei × Month=0 to Renegotiation Dt -0.129***
(0.046)
High Leveragei × Month=1 to Renegotiation Dt -0.027
(0.044)
(0.040)
High Leveragei × Month≥3+ to Renegotiation Dt -0.148**
(0.067)
FirmFEi Yes Yes Yes
MonthFEt Yes Yes Yes
6 Sq Mile Geog FEj Yes Yes Yes
N 15,110 18,256 19,857
R2 0.057 0.050 0.053

Table A.IX: Leverage and Production Decisions: The Role of Outliers
This table reports the results of a difference-in-difference analysis that evaluates how the time to project completion varies across firms based on leverage.
Each panel replicates the DiD regression analysis reported in Table IV of the paper after respective number of highest market-leverage firms (top of the
leverage distribution) is droppped from the sample yet the definitions of the leverage quintiles remain the same . The unit of observation is at the well j
firm i , year t level. The dependent variable is the number of months between the date a well is spudded (started) and the date when it is completed and
production begins. The pre-contango period (Contangot = 0) is composed of wells started in September, October, and November of 2013, the year prior to
the super-contango period. The contango period (Contangot = 1) is composed of wells started in September, October, and November of 2014, just before
the oil market entered contango in December of 2014. Leverage quintiles are based on a firm's market leverage (total debt divided by debt plus equity
market cap) measured as of September 30, 2014. Columns (1) thorugh (3) report the results for the full sample. Columns (4) thorugh (6) report the results
for the sub-sample of ABL firms where the township fixed effects are estimated using all available projects via SUR setting. Standard errors clustered at
the firm level are reported in round parenthesis. ***, **, and * represent statistical significance at the 1%, 5%, and 10% level respectively.
Dependent Variable=Month to Production Full Sample ABL Firms

Contangot 1.090*** 1.077*** 1.238*** 1.201* 0.967*** 1.329**
(0.331) (0.210) (0.375) (0.675) (0.343) (0.556)
(0.631) (0.714)
(0.553) (0.992)
(0.428) (0.811)
Contangot × Leverage Q5 -1.031** -1.020** -1.361* -1.133**
(0.496) (0.448) (0.744) (0.458)
(1.056) (1.235)
N 3509 3509 3509 1196 1196 1196
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Table AIX (Cont.)

Contangot 1.090*** 1.077*** 1.238*** 1.201* 0.967*** 1.329**
(0.331) (0.210) (0.375) (0.675) (0.343) (0.556)
(0.631) (0.714)
(0.553) (0.992)
(0.428) (0.811)
(0.496) (0.448) (0.744) (0.458)
(1.056) (1.235)
N 3502 3502 3502 1189 1189 1189
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Contangot 1.090*** 1.077*** 1.238*** 1.201* 0.967*** 1.329**
(0.331) (0.210) (0.375) (0.675) (0.343) (0.556)
(0.631) (0.714)
(0.553) (0.992)
(0.428) (0.812)
(0.496) (0.448) (0.744) (0.458)
(1.056) (1.236)
N 3499 3499 3499 1186 1186 1186
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Table AIX (Cont.)

Contangot 1.089*** 1.076*** 1.180*** 1.200* 0.966*** 1.235**
(0.332) (0.210) (0.380) (0.676) (0.343) (0.559)
(0.632) (0.714)
(0.554) (0.994)
(0.428) (0.812)
Contangot × Leverage Q5 -0.892* -0.880* -1.235 -1.007**
(0.520) (0.475) (0.743) (0.456)
(1.102) (1.297)
N 3470 3470 3470 1157 1157 1157
2
R 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.089*** 1.076*** 1.182*** 1.200* 0.967*** 1.286**

(0.332) (0.210) (0.390) (0.676) (0.343) (0.567)
(0.632) (0.715)
(0.554) (0.995)
(0.429) (0.813)
Contangot × Leverage Q5 -0.996* -0.985* -1.386* -1.159**
(0.541) (0.502) (0.741) (0.451)
(1.183) (1.366)
N 3445 3445 3445 1132 1132 1132
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table A.X: The Role of Other Firm Characteristics
This table reports results of the difference-in-difference analysis reported in Table IV conditional on other firm characteristics and their interaction with the contango
dummy. Panel A utilizes continuous control variables (profitability, size and Tobin's Q), while Panel B exploits dummies for the lowest quintile of respective firm
characteristics. Columns (1) through (5) report the results for the full sample. Columns (6) through (10) report the results for the sub-sample of ABL firms where the
township fixed effects are estimated using all available projects via SUR setting. Standard errors clustered at the firm level are reported in parentheses. ***, **, and *
represent statistical significance at the 1%, 5%, and 10% levels respectively.
Contangot 0.589 2.288* 0.786 0.803** 0.165 1.133 3.372 2.415* 0.553 2.515
(0.541) (1.192) (0.588) (0.305) (1.717) (0.933) (2.113) (1.242) (0.503) (3.171)
Contangot × Q5 High-leverage Di -0.921** -1.225*** -0.907** -0.992** -0.909* -1.103** -1.142*** -1.624** -0.881* -1.401**
(0.399) (0.391) (0.429) (0.381) (0.467) (0.431) (0.409) (0.649) (0.465) (0.628)
Contangot × Profitabilityi 10.827 12.660 -3.620 5.649
(10.130) (13.472) (20.067) (19.934)
Contangot × Log Assetsi -0.121 0.025 -0.283 -0.090
(0.121) (0.153) (0.244) (0.341)
Contangot × Tobin's Qi 0.234 -0.145 -1.067 -1.069
(0.423) (0.547) (0.857) (0.919)
Contangot × Capexi 3.980* 3.898 3.426 3.221
(2.291) (2.854) (2.208) (3.614)
Firm and 6 Sq. Mile Geog FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 3557 3557 1244 1244 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54
Panel B: Controlling for Firm Characteristic: Quintile Approach
Contangot 1.062*** 1.069*** 1.061*** 1.317*** 1.302*** 0.941*** 0.997** 0.984*** 0.970*** 0.979**
(0.224) (0.217) (0.241) (0.208) (0.246) (0.345) (0.383) (0.347) (0.342) (0.380)
Contangot × Q5 High-leverage Di -1.052** -1.005*** -1.022*** -1.247*** -1.267*** -1.204** -1.095** -0.811* -1.077** -1.030**
(0.416) (0.373) (0.377) (0.379) (0.416) (0.488) (0.450) (0.421) (0.423) (0.507)
Contangot × Q1 Profitability Di 0.144 -0.015 0.340 0.663
(0.372) (0.487) (0.512) (0.572)
Contangot × Q1 Log Assets Di 0.204 -0.020 -0.230 -0.296
(0.658) (0.696) (0.516) (0.469)
Contangot × Q1 Tobin's Q Di 0.068 0.081 -0.528 -0.683
(0.333) (0.337) (0.418) (0.549)
Contangot × Q1 Capex Di -0.806* -0.809* -1.970*** -1.959***
(0.440) (0.455) (0.342) (0.520)
N 3557 3557 3557 3557 3557 1244 1244 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54

Table B.I: Leverage and Production Decisions (Bootstrap Standard Errors)
This table reports the results of a difference-in-difference analysis from Table IV using bootstrapped standard errors. The
cluster-bootstrapped standard errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based
statistical significance at 1%, 5%, and 10% levels respectively.
Dependent Variable =
= Months to Production Panel A: Full Sample Panel B: ABL Firms
Contangot 1.09 1.077 1.262 1.202 0.967 1.264

A A A B A
[0.216] [0.109] [0.186] [0.464] [0.212] [0.293]A

[0.287] [0.562]

[0.306] [0.588]

[0.297] [0.652]
Contangot × Leverage Q5 -1.014 -1.002 -1.303 -1.071

A A B
[0.308] [0.313] [0.532] [0.353]A

B B
[0.552] [0.678]

6 Sq. Mile Geo j Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 1244 1244 1244
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Table B.II: The Role of Other Firm Characteristics
This table reports the results of a difference-in-difference analysis from Table V using bootstrapped standard errors. The cluster-bootstrapped standard errors are
reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10% levels respectively.

Contangot 0.589 2.288 0.786 0.803 0.165 1.133 3.372 2.415 0.553 2.515
[0.320]
C
[0.779]
A
[0.291]
A
[0.200]A [1.536] [0.579]
C
[1.622]
B
[0.733]
A [0.360] [2.612]
Contangot × Q5 High-leverage Di -0.921 -1.225 -0.907 -0.992 -0.909 -1.103 -1.142 -1.624 -0.881 -1.401
A
[0.292] [0.339]A [0.322]A [0.279]A [0.386]B [0.353]A [0.297]A [0.437]A [0.311]A [0.484]A
Contangot × Profitabilityi 10.827 12.660 -3.620 5.649

[6.927] [9.732] [10.685] [14.01]
Contangot × Log Assetsi -0.121 0.025 -0.283 -0.090

[0.077] [0.124] [0.186] [0.248]
Contangot × Tobin's Qi 0.234 -0.145 -1.067 -1.069

[0.211] [0.311] [0.516]
B
[0.544]B
Contangot × Capexi 3.980 3.898 3.426 3.221

[2.595] [3.845] [3.282] [3.803]
N 3557 3557 3557 3557 3557 1244 1244 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54
Panel B: Controlling for Firm Characteristic: Quintile Approach
Contangot 1.062 1.069 1.061 1.317 1.302 0.941 0.997 0.984 0.970 0.979
A A A A A A A A A A
[0.120] [0.123] [0.117] [0.133] [0.148] [0.201] [0.208] [0.201] [0.165] [0.213]
Contangot × Q5 High-leverage Di -1.052 -1.005 -1.022 -1.247 -1.267 -1.204 -1.095 -0.811 -1.077 -1.030
A A A A A A A B A A
[0.327] [0.309] [0.337] [0.337] [0.330] [0.478] [0.338] [0.351] [0.310] [0.333]
Contangot × Q1 Profitability Di 0.144 -0.015 0.340 0.663

[0.280] [0.373] [0.464] [0.465]
Contangot × Q1 Log Assets Di 0.204 -0.020 -0.230 -0.296

[0.474] [0.501] [0.366] [0.453]
Contangot × Q1 Tobin's Q Di 0.068 0.081 -0.528 -0.683

[0.224] [0.328] [0.424] [0.548]
Contangot × Q1 Capex Di -0.806 -0.809 -1.970 -1.959

A
[0.225] [0.253]A [0.165]A [0.547]A
N 3557 3557 3557 3557 3557 1244 1244 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54 0.54

Table B.III: Production Decisions and Debt Renegotiations
This table reports the results of a difference-in-difference analysis from Table VII using bootstrapped standard errors. The cluster-bootstrapped
standard errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10%
levels respectively.
Dependent Variable = Well Start (1 if well starts High Leverage Low Leverage
High Leverage Low Leverage
producing in month, 0 otherwise) All + Non-ABL + Non-ABL
Only Only
Firms Firms
(1) (2) (3) (4) (5)
Month=-2 to Renegotiation Dt 0.020 -0.062 0.017 -0.064 0.017
[0.019] [0.038]
C [0.020] [0.052] [0.020]
Month=-1 to Renegotiation Dt -0.003 -0.022 -0.007 0.001 -0.008
[0.018] [0.041] [0.018] [0.071] [0.018]
Month=0 to Renegotiation Dt -0.005 -0.135 -0.006 -0.117 -0.007
[0.015] [0.035]A [0.014] [0.101] [0.014]
[0.014]
B
[0.033] A
[0.014] B [0.118] [0.014]
B

[0.014] [0.033]
A [0.014] [0.118] [0.014]
Month≥3+ to Renegotiation Dt 0.074 -0.092 0.074 -0.146 0.074
[0.017]
A
[0.033]A [0.017]A [0.135] [0.017]
A

A
[0.043]
[0.045]
A
[0.039]
[0.038]
B
[0.035]
High Leveragei × Month≥3+ to Renegotiation Dt -0.162
A
[0.037]
Firm i Fixed Effects Yes Yes Yes Yes Yes

Month t Fixed Effects Yes Yes Yes Yes Yes
6 Sq Mile Geog j Fixed Effects Yes Yes Yes Yes Yes
N 20,297 15,051 18,755 1,569 18,728
R2 0.052 0.056 0.049 0.080 0.049

Table B.IV: High Collateral Impact vs. Low Collateral Impact
This table reports the results of a difference-in-difference analysis from Table XII using bootstrapped standard errors. The cluster-bootstrapped standard
errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10% levels
respectively.
Dependent Variable = Well Start High Collateral Impact Low Collateral Impact
(1 if well starts producing in month, High Low High Low
All All
0 otherwise) Leverage Leverage Leverage Leverage
(1) (2) (3) (4) (5) (6)
Month=-2 to Renegotiation Dt 0.023 -0.124 0.019 0.020 0.029 0.020
[0.027] [0.051]B [0.028] [0.027] [0.056] [0.026]
Month=-1 to Renegotiation Dt -0.004 -0.046 -0.010 0.012 0.004 0.011
[0.026] [0.054] [0.027] [0.025] [0.058] [0.025]
Month=0 to Renegotiation Dt 0.011 -0.213 0.006 -0.003 -0.020 0.000
[0.025] [0.05] A [0.025] [0.023] [0.051] [0.023]
A
[0.021] [0.048] [0.022] [0.020] [0.051] [0.020]
Month=2 to Renegotiation Dt 0.016 -0.179 0.014 -0.027 -0.017 -0.027
[0.022] [0.046]A [0.022] [0.018] [0.045] [0.018]
Month≥3+ to Renegotiation Dt 0.106 -0.187 0.106 0.059 0.048 0.059
A A A B
[0.027] [0.047] [0.027] [0.023] [0.046] [0.023]A

A
[0.058] [0.059]
[0.059] [0.058]
A
[0.054] [0.054]
High Leveragei × Month=1 to Renegotiation Dt -0.146 0.041
[0.050]A [0.052]
High Leveragei × Month=2 to Renegotiation Dt -0.188 0.018
[0.049]A [0.045]
High Leveragei × Month≥3+ to Renegotiation Dt -0.288 -0.009
[0.052]A [0.049]
N 11160 8868 10194 9137 6183 8561
2
R 0.046 0.052 0.040 0.071 0.075 0.071

Table B.V: Leverage and Production Decisions: Alternative Measures of Leverage
This table reports the results of a difference-in-difference analysis from Table AII using bootstrapped standard errors. The cluster-bootstrapped standard
errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10% levels
respectively.
Contangot 1.204 1.039 0.175 2.016 0.84 1.589
[0.194]A [0.121]A [0.499] [0.494]A [0.183]A [0.760]B

[0.336] [0.839]

A
[0.267] [0.533]

[0.332] [0.562]
Contangot × Leverage Q5 -0.881 -0.710 -0.712

A
[0.312] [0.260]A [0.294]B
Contangot × Continuous Leverage 1.505 -1.589

(2.140) (1.469)
[0.926] [1.284]
Firm and 6 Sq. Mile Geog Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 1244 1244 1244
R2 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.18 1.08 1.011 -0.075 0.978 0.632
A
[0.103]A [0.118]A
A
[0.186] [0.578] [0.192] [0.262]A

[0.288] [0.833]B

[0.297] [0.636]

[0.353] [0.717]B
Contangot × Leverage Q5 -1.122 -1.021 -1.097

[0.309]A [0.273]A [0.312]A
Contangot × Continuous Leverage -0.001 0.004

[0.003] [0.019]
Firm and 6 Sq Mile Geog Fixed Effects Yes Yes Yes Yes Yes Yes
N 3557 3557 3557 3557 3557 3557
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Table B.VI: Production Decisions and Debt Renegotiations: Alternative Measures of Leverage
This table reports the results of a difference-in-difference analysis from Table A.IV using bootstrapped standard errors. The cluster-bootstrapped
standard errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10% levels
respectively.
Panel A: Book Leverage Panel B: Interest Coverage Ratio
Dependent Variable = Well Start (1 if well starts
producing in month, 0 otherwise)
Month=-2 to Renegotiation Dummy -0.046 0.004 0.005 -0.055 0.015 0.016

[0.032] [0.010] [0.009] [0.075] [0.016] [0.020]
Month=-1 to Renegotiation Dummy -0.020 -0.012 -0.011 0.019 -0.018 -0.016
A A
[0.049] [0.001] [0.016] [0.001]A [0.017] [0.001]A
Month=0 to Renegotiation Dummy -0.165 -0.008 -0.008 -0.142 -0.006 -0.004

A
[0.027] [0.020] [0.022] [0.009]A [0.006] [0.006]
[0.008]A [0.004]
A
[0.012]
A
[0.004]
A
[0.010]
A
[0.006]
A

A
[0.008]A [0.004]A [0.012] [0.004] [0.010]C [0.006]A
Month≥3+ to Renegotiation Dummy -0.105 0.066 0.066 -0.091 0.072 0.072

[0.007]A [0.001]A [0.020]
A
[0.012]A [0.034]B [0.010]A
High Leveragei × Month=-2 to Renegotiation -0.050 -0.067
A
[0.013] [0.066]
High Leveragei × Month=-1 to Renegotiation -0.001 0.041
[0.018] [0.012]A
[0.036]A [0.012]
A

A
[0.022] [0.013]A
A
[0.007] [0.023]A
High Leveragei × Month≥3+ to Renegotiation -0.167 -0.160
[0.013]A [0.006]A
N 14,765 19,041 20,297 14,939 18,867 20,297
R2 0.056 0.050 0.052 0.057 0.050 0.052

Table B.VII: Leverage and Production Decisions: Placebo Tests, Prior Time Periods
This table reports the results of a difference-in-difference analysis from Table A.V using bootstrapped standard errors. The cluster-
bootstrapped standard errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at
1%, 5%, and 10% levels respectively.
Dependent Variable = Dependent Variable =
Months to Production Ln(Months to Production)
Panel A: 2012-2013 Placebo Tests
Dummy2013 0.047 -0.184 0.104 0.014 -0.028 0.026
B
[0.370] [0.090] [0.216] [0.078] [0.023] [0.046]
Dummy2013 × Leverage Q2 -0.055 0.007
[0.349] [0.085]
Dummy2013 × Leverage Q3 -0.430 -0.109
[0.385] [0.092]
Dummy2013 × Leverage Q4 -0.353 -0.059
[0.496] [0.082]
Dummy2013 × Leverage Q5 0.245 0.473 0.068 0.108
[0.437] [0.328] [0.088] [0.069]
Dummy2013 × Continuous Leverage -0.862 -0.158
[0.652] [0.157]
N 2803 2803 2803 2803 2803 2803
2
R 0.58 0.58 0.57 0.59 0.59 0.57
Panel B: 2011-2012 Placebo Tests
Dummy2012 0.222 0.116 -0.274 0.013 0.015 -0.040
[0.446] [0.298] [0.169] [0.043] [0.051] [0.054]
Dummy2012× Leverage Q2 0.248 0.080
[0.374] [0.070]
Dummy2012 × Leverage Q3 -0.767 -0.119
A
[0.799] [0.045]
Dummy2012 × Leverage Q4 0.500 0.111

B
[0.682] [0.056]
Dummy2012 × Leverage Q5 0.525 0.631 0.091 0.088
C
[0.505] [0.521] [0.066] [0.045]
Dummy2012 × Continuous Leverage 1.611 0.227
A
[0.527] [0.152]
N 1928 1928 1928 1928 1928 1928
2
R 0.55 0.55 0.53 0.56 0.55 0.54

Table B.VIII: Leverage and Production Decisions: The Role of Outliers
This table reports the results of a difference-in-difference analysis from Table A.VI using bootstrapped standard errors. The cluster-bootstrapped standard
errors are reported in square brackets, A, B, and C represent the cluster-bootstrapped-based statistical significance at 1%, 5%, and 10% levels respectively.

Contangot 1.091 1.079 1.238 1.205 0.973 1.329
[0.216]A [0.109]A [0.186]A [0.464]A [0.198]A [0.366]A

[0.291] [0.714]
[0.306] [0.550]

[0.297] [0.555]
[0.308]C [0.313]A [0.567]B [0.340]A
B B
[0.552] [0.917]
N 3509 3509 3509 1196 1196 1196
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table B.VIII (Cont.)
Contangot 1.087 1.079 1.238 1.196 0.973 1.329
[0.223]A [0.114]A [0.195]A [0.490]B [0.212]A [0.355]A
[0.309] [0.725]
[0.318] [0.574]
[0.321] [0.580]
[0.387]C [0.337]C [0.578]B [0.347]A
[0.591] [0.877]B
N 3502 3502 3502 1189 1189 1189
R2 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.086 1.088 1.238 1.204 0.972 1.329
A A A A A
[0.205] [0.118] [0.209] [0.549] [0.209] [0.375]A
[0.324] [0.712]
[0.313] [0.666]
[0.317] [0.642]
[0.307]B [0.272]
B
[0.577]
B
[0.386]
A

[0.664] [0.955]
N 3499 3499 3499 1186 1186 1186
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table B.VIII (Cont.)
Contangot 1.086 1.098 1.18 1.206 1.007 1.235

[0.223]A [0.124]
A
[0.200]
A
[0.500]
A
[0.210]
A
[0.359]
A

[0.311] [0.732]

[0.334] [0.577]

[0.316] [0.578]

C B B A
[0.330] [0.272] [0.579] [0.382]

[0.681] [0.980]
N 3470 3470 3470 1157 1157 1157
R2 0.54 0.54 0.54 0.54 0.54 0.54

Contangot 1.024 1.098 1.182 0.938 1.007 1.286
[0.217]A [0.121]
A
[0.197]
A
[0.533]
A
[0.215]
A
[0.379]
A

[0.344] [0.787]
[0.333] [0.623]
[0.319] [0.659]
[0.329]C [0.273]
B
[0.612]
C
[0.386]
A

[0.644] [1.051]C
N 3445 3445 3445 1132 1132 1132
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table B.IX: Production Decisions and Debt Renegotiations: The Role of Outliers
This table reports the results of a difference-in-difference analysis from Table A.VIII using bootstrapped standard errors. The cluster-bootstrapped standard
Dependent Variable = Well Start Panel A: Top 1 High-leverage Firm Dropped from Panel B: Top 2 High-leverage Firms Dropped from
(1 if well starts producing in month, 0 the Sample the Sample
otherwise) High Leverage Low Leverage All High Leverage Low Leverage All
[0.032] [0.004]
A
[0.004]A [0.020]C [0.006]A [0.011]B
Month=-1 to Renegotiation Dt -0.038 -0.007 -0.003 -0.001 -0.018 -0.015
A [0.042] [0.014] [0.007] A [0.015]
[0.006] [0.001]
Month=0 to Renegotiation Dt -0.118 0.001 0.003 -0.105 0.001 0.003
[0.016]
A [0.004] [0.010] [0.016]A [0.015] [0.009]
A A C A A
[0.013] [0.011] [0.017] [0.021] [0.006] [0.003]A
A [0.011] A [0.013] [0.017]
[0.005] [0.002] [0.005]A
A B A [0.046] A
[0.010] [0.032] [0.021] [0.008] [0.001]A
[0.054] [0.005]
A

[0.040] [0.026]
B
[0.051] [0.027]A
[0.049] [0.032]
B
[0.035] [0.028]C
A A
[0.018] [0.014]
N 15,222 18,395 20,108 15,286 18,275 20,052
R2 0.055 0.049 0.051 0.054 0.050 0.051

Table B.IX (Cont.)
Dependent Variable = Well Start (1 if well starts producing in month, 0 otherwise)
Panel C: Top 3 High-leverage Firms Dropped from Panel D: Top 4 High-leverage Firms Dropped
the Sample from the Sample

A A
[0.057] [0.006] [0.018] [0.024] [0.025] [0.007]
Month=-1 to Renegotiation Dt -0.008 -0.018 -0.015 0.002 -0.020 -0.017
A
[0.037] [0.007] [0.024] [0.004] [0.031] [0.023]
Month=0 to Renegotiation Dt -0.126 0.001 0.003 -0.139 0.001 0.004
B
[0.053] [0.004] [0.005] [0.031]A [0.017] [0.017]
A
[0.032] [0.006] [0.011]A [0.035]C [0.012]A [0.001]A
A C C B
[0.027] [0.008] [0.019] [0.043] [0.022] [0.006]
C
[0.032] [0.023]A [0.010]A [0.019]C [0.028]A [0.020]A
A
[0.003] [0.033]C
High Leveragei × Month=-1 to Renegotiation Dt 0.017 0.028
[0.068] [0.047]
[0.019]A [0.027]A
A
[0.008] [0.036]
B
[0.040] [0.021]
A A
[0.032] [0.017]
MonthFEt Yes Yes Yes Yes Yes Yes
6 Sq Mile Geog FEj Yes Yes Yes Yes Yes Yes
N 15,259 18,275 20,025 15,187 18,263 19,941
R2 0.055 0.050 0.052 0.056 0.050 0.052
1 2 3 1 2 3

Table B.IX (Cont.)
Dependent Variable = Well Start
(1 if well starts producing in month, 0 otherwise)
Panel E: Top 5 High-leverage Firms Dropped from
the Sample
High Leverage Low Leverage All
Month=-2 to Renegotiation Dt -0.037 0.023 0.027

B
[0.017] [0.025] [0.053]
Month=-1 to Renegotiation Dt 0.024 -0.021 -0.017
A
[0.002] [0.007]A [0.038]
Month=0 to Renegotiation Dt -0.13 0.001 0.005
[0.019]A [0.014] [0.040]
B
[0.035] [0.005]A [0.015]B
C
[0.045] [0.019] [0.013]
Month≥3+ to Renegotiation Dt -0.070 0.082 0.082
A A C
[0.025] [0.026] [0.043]
[0.061]
High Leveragei × Month=-1 to Renegotiation Dt 0.050
[0.129]
B
[0.058]
[0.050]
B
[0.027]
High Leveragei × Month≥3+ to Renegotiation Dt -0.148
B
[0.064]
FirmFEi Yes Yes Yes
MonthFEt Yes Yes Yes
6 Sq Mile Geog FEj Yes Yes Yes
N 15,110 18,256 19,857
2
R 0.057 0.050 0.053

Table B.X: Leverage and Production Decisions: The Role of Outliers
This table reports the results of a difference-in-difference analysis from Table A.IX using bootstrapped standard errors. The cluster-bootstrapped standard

Contangot 1.09 1.077 1.238 1.201 0.967 1.329
A A A B A A
[0.230] [0.115] [0.163] [0.483] [0.179] [0.355]
[0.302] [0.554]
[0.318] [0.604]
[0.342] [0.616]
B A B A
[0.474] [0.366] [0.610] [0.349]
C B
[0.588] [0.926]
N 3509 3509 3509 1196 1196 1196
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table B.X (Cont.)

Contangot 1.09 1.077 1.238 1.201 0.967 1.329
A A A B A A
[0.210] [0.119] [0.205] [0.477] [0.203] [0.338]
[0.295] [0.558]
[0.303] [0.581]
[0.319] [0.650]
[0.395]A [0.364]A [0.558]B [0.374]A
B
[0.655] [0.848]
N 3502 3502 3502 1189 1189 1189
2
R 0.54 0.54 0.54 0.54 0.54 0.54

Contangot 1.09 1.077 1.238 1.201 0.967 1.329
A
[0.210] [0.110]A [0.176]A [0.509]B [0.211]A [0.361]A
[0.287] [0.602]
[0.322] [0.619]
[0.306] [0.699]
[0.391]A [0.323]
A
[0.631]
B
[0.398]
A

C
[0.565] [0.904]C
N 3499 3499 3499 1186 1186 1186
R2 0.54 0.54 0.54 0.54 0.54 0.54

Table B.X (Cont.)

Contangot 1.089 1.076 1.18 1.2 0.966 1.235
A A A B A A
[0.244] [0.113] [0.189] [0.494] [0.221] [0.321]
[0.332] [0.566]
[0.331] [0.558]
[0.324] [0.621]
C
[0.457] [0.373]B [0.580]B [0.392]A
C
[0.618] [0.853]
N 3470 3470 3470 1157 1157 1157
2
R 0.54 0.54 0.54 0.54 0.54 0.54
Contangot 1.089 1.076 1.182 1.2 0.967 1.286

A
[0.214] [0.114]A [0.202]A [0.494]B [0.205]A [0.374]A
[0.306] [0.573]
[0.304] [0.570]
[0.321] [0.656]
B
[0.442] [0.350]A [0.629]B [0.424]A
[0.666] [1.071]
N 3445 3445 3445 1132 1132 1132
R2 0.54 0.54 0.54 0.54 0.54 0.54

SSRN Id2939603 PDF

Uploaded by

Copyright:

Available Formats

You might also like

SSRN Id2939603 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN Id2939603 PDF

Uploaded by

Copyright:

Available Formats

Drilling and Debt

Journal of Finance, forthcoming

The Wharton School, University of Pennsylvania and NBER

University of Virginia, Darden School

University of Virginia, Darden School

August 27, 2019

Electronic copy available at: https://ssrn.com/abstract=2939603

In this paper, we use new detailed project-level data to document a previously

thus, also highlight a previously unexplored hidden cost of collateral-based financing.

Electronic copy available at: https://ssrn.com/abstract=2939603

leverage increases (see, e.g., Stein, 2003). 2

We exploit an empirical setting that allows us to make significant progress on each of

endogeneity of potential differences in investment opportunities across high-leverage and low-

(Modigliani and Miller, 1958).

Electronic copy available at: https://ssrn.com/abstract=2939603

set of granular township (6-by-6-mile) geography fixed effects.

frictions. These pre-scheduled renegotiations are not a consequence of covenant violations,

imposes significant hurdles on non-debt-related interpretations of our results.

effects). It cannot be explained by firms’ heterogeneity in profitability, size, and market-to-book.

contango and during contango.

Electronic copy available at: https://ssrn.com/abstract=2939603

about 1.2% of equity value.

similar conclusions: in the month prior to a pre-scheduled renegotiation or amendment, high-

acceleration (not delay) of investments.

The short-term transparent nature of the decision to accelerate production in relation to

Electronic copy available at: https://ssrn.com/abstract=2939603

binding-covenant-based explanations. 6 The acceleration of well completions should have a

indicate impending liquidity challenges at the time of credit renegotiations.

months EBITDA numbers.

Electronic copy available at: https://ssrn.com/abstract=2939603

ability to secure/expand access to finance.

undeveloped oil reserves. The geographic proximity of undeveloped reserves to producing

frictions at renegotiation and expanding access to finance.

on a firm’s collateral value and borrowing base.

Electronic copy available at: https://ssrn.com/abstract=2939603

inconsistent with the traditional risk-shifting and debt-overhang channels. It speaks to

literature argues that covenants and renegotiations lead to Pareto-improving reallocation of

renegotiations and in the absence of covenant violations.

based financing, linked to the real investment activity of firms.

production technology characterized by high-cost production adjustments (Anderson, Kellogg,

Electronic copy available at: https://ssrn.com/abstract=2939603

I. Institutional Background and Data

investment responses to changes in expected project value.

A. Institutional Setting and Oil Price Contango

Electronic copy available at: https://ssrn.com/abstract=2939603

as of February 2015 to one as of September 2014.

contango through 2015 and into early 2016.

The severe contango episode of 2014-2015 played an important role in companies’

expectations (Society of Petroleum Evaluation Engineers, 1995).

Electronic copy available at: https://ssrn.com/abstract=2939603

Consistent with these arguments, we exploit the sharp shift to contango in

December 2014/January 2015 as an exogenous shock to U.S. producers’ incentives to complete

production in favor of opening the available reserves later.

B. Shale Oil Drilling Overview

In 2003, a surprise technological breakthrough combined horizontal drilling with

complete new wells.

Electronic copy available at: https://ssrn.com/abstract=2939603