How Flexible is US Shale Oil Production? Evidence … Flexible is US Shale Oil Production? Evidence...
Transcript of How Flexible is US Shale Oil Production? Evidence … Flexible is US Shale Oil Production? Evidence...
-
How Flexible is US Shale Oil Production?Evidence from North Dakota
Hilde C. Bjrnland1,2 Frode Martin Nordvik1
Maximilian Rohrer1
1Centre for Applied Macro and Petroleum economics (CAMP),BI Norwegian Business School
2Norges Bank
Oil, Middle East, and the Global Economy,USC April 1, 2016
-
North Dakota oil boom
Figure: Night light map of US reveals large cluster of lights insparsely populated North Dakota.
-
Oil production: short-run supply elasticity
I Most (if not all) existing studies find close to zero short-termelasticity for oil production (see f.ex. Kellogg et al. (2015),Ramcharran (2002), Dahl and Yukel (1991)).
I Study production from conventional extraction technologies.
I Use mainly aggregate data to estimate elasticities.
-
Elasticity depends on technology
I However, the elasticity of output depends on the technologyof the marginal producer.
I The marginal producer used to be a conventional oil producer.
I The new marginal producers use shale oil extractiontechnology (fracking).
I However, little (if any) evidence exist of the short-term supplyelasticity for producers that use this technology.
-
What we do and how we contribute
I We construct a novel data set that includes monthlyproduction volumes, individually, for 16,252 wells for timeperiod 1974-2015 in North Dakota.
I Allows us to identify wells according to extraction technology,and estimate supply elasticity for wells that are using frackingtechnology.
I We study elasticity along two margins:I Intensive margin: Elasticity of producing wells.
I Dep. variable is barrels produced by well i at time t. Identifywells according to well technology, conventional andunconventional and compare.
I Extensive margin: Study firms timing of well completions.I Use monthly nr. of well completions as dependent variable.I Allows us to ask if firms are optimizing the timing of when to
exercise the real option to produce.
-
Preview of findings I
I Large differences between conventional and unconventional oilwells. More specifically:
Output elasticity:
I 1. Output elasticity wrt price of unconventional wellproduction 3-4 times that of conventional wells. An averageBakken well increases output about 3-5 percent to a onest.dev. increase in the short term spot-future spread.
I 2. Respond to shifts in the slope of the short term futurescurve (short term expected price trend), not price levelchanges.
I 3. Find evidence of behaviour consistent with in-the-groundstorage.
-
Preview of findings II
Well completions:
I 2. Monthly completions of unconventional wells respond tothe slope of future curves (expected price trend), but not toprice level changes. Find no evidence of this behaviour forconventional.
I Nr. of well completions is reduced by 7 percent for a onestandard deviation increase in the 12 month future spread.
I This behaviour by producers is consistent with in-the-groundstorage, leaving wells uncompleted awaiting a higher expectedprice.
-
Agenda
I Well-level panel analysis: elasticity of producing wells
I Data and comparative characteristics
II Empirical model
III Results
I Firm-level panel analysis: elasticity of well entry/completion
I Regression plot
II Empirical model
III Results
I Conclusion
-
Figure: Total well completions over time
details
-
Figure: Average production profiles, by well type
details
-
Why would we expect differences?
I Extensive margin:
I A front-loaded extraction profile increases incentives tooptimize the timing of well completion.
I Intensive margin:
I Only when prices enter the range of marginal costs mightproducers adjust the flow rate of the well. As a marginalproducer, this happens more freqently for shale oil wells.
-
Empirical model: Supply elasticity - producing wells
Qit = 1Qit1 + 2Ft+1 + 3[Ft+1 Pt ] + Xit + it
I Qit is monthly percentage change in nr. of barrels produced bywell i.
I Ft+1 is the percentage change in the nearest term futures price.
I [Ft+1 Pt ] is a shift in the short-term spot-futures spread (shift inexpected price trend).
I Xit are well-specific time-varying controls, f. ex well age.
I Model includes well fixed effects.
I Errors, it allowed to be clustered (correlated) on time and well-level.
-
Price curve example
Figure: Futures price curve, 12 months
-
Unconventional wells more elastic
Table: Baseline results for producing wells
Dep var: Qit Conv. wells Unconv. wells Bakken
Qit1 -.35*** -.32*** -.32***Well age -.00 .00*** .00***Well age q .00 -.00*** -.00***Ft+1 .058** .01 -.00[Ft+1 Pt ] -.04 .17** .20**N 718,142 484,645 353,381Nr. of wells 4,517 11,735 10,471Well FE Yes Yes Yes
Notes: * 10 percent sign, ** 5 percent sign, *** 1 percent sign. Standard errors are clustered at the time andwell level.
-
Storing oil in the ground?
Figure: Mean conventional well output left and mean unconventional welloutput right, against 12-month spread. Time: 6.2008-11.2015
-
A closer look at Bakken post-2008
Table: Bakken producing wells
Dep var: Qit Bakken Bakken Bakken
Qt1 -.32*** -.32*** -.32***Well age .00*** .00*** .00***Well age q -.00 -.00*** -.00***[Ft+1 Pt ] 0.20*** 0.19** 0.18**[Ft+3 Pt ] -0.63[Ft+6 Pt ] -0.37*[Ft+9 Pt ] -0.27*N 353,381 353,381 353,381Nr. of wells 10,471 10,471 10,471Years 2008-2015 2008-2015 2008-2015Firm FE Yes Yes Yes
Notes: * 10 percent sign, ** 5 percent sign, *** 1 percent sign. Standard errors are clustered at the time andwell level.
-
Extensive margin
When do firms put on new wells?
-
Extensive margin: Well completions
1. We also want to study the timing of well entry/completions.
2. Use monthly nr. of well completions as dependent variableinstead of monthly output from existing wells.
3. A well completion is equivalent to exercising the real option toproduce.
-
Slope of future curve and monthly well entry, 2008-2015
Figure: y-axis shows the expected price change 12 month ahead,and x-axis shows the monthly nr. of wells being put into
production.
-
Main empirical model: Firms monthly completions
Wit = 1Wit1 + 2Ft+j + 3[Ft+j Pt ] + it
I Wit is monthly percentage change in nr. of wells completed byfirm i.
I Ft+j is the percentage change in the futures price of oil at timet+j.
I [Ft+j Pt ] is the spot-futures term spread at horizon t to t+j.I Model also includes firm-fixed effects, firm-specific trend and
seasonal dummies.
I Allow standard errors to be clustered (correlated) on time andwell-level.
-
Firm panel: Well completions
Table: Baseline results
Dep var: Nr. wells wells wells wells wells
log(wells)t1 -0.46*** -0.46*** -0.46*** -0.46***Ft+3 0.03Ft+6 0.02 .00*Ft+12 0.01[Ft+3 Pt ] -0.36***[Ft+6 Pt ] -0.20*** .00[Ft+12 Pt ] -0.15***Firms 165 165 165 165N 12,677 12,677 12,677 12,677Years 2008-2015 2008-2015 2008-2015 2008-2015Firm FE Yes Yes Yes YesTime trend Yes Yes Yes Yes
Notes: Blue font represent conventional wells. * 10 percent sign, ** 5 percent sign, *** 1 percent sign.Standard errors are clustered at the firm and time-level.
-
Aggregate data: Well completions
Table: Results for aggregate nr. of unconventional wells
Dep var: Nr. wells wells wells wells
log(wells)t1 -.21** -.22** -.23***Ft+3 .25Ft+6 .21Ft+12 .12[Ft+3 Pt ] -1.60*[Ft+6 Pt ] -1.00 **[Ft+12 Pt ] -.68**N 90 90 90Years 2008-2015 2008-2015 2008-2015Time trend Yes Yes Yes
Notes: * 10 percent sign, ** 5 percent sign, *** 1 percent sign. Standard errors are clustered at the firm andtime-level.
-
Conclusion
I We document that unconventional oil output is moreprice-responsive than conventional.
I New finding in the literature.
I Go further than the empirics in Kellogg et al. (2015), whoonly study conventional, while half of US oil output isunconventional.
-
1. First year shale well decline rate: 5.2 percent pr. month.
2. First year conventional well decline rate: 0.12 percent pr.month.
3. Difference is 43 to 1.
Densities
-
Figure: relative production profiles, by well type
back
-
I The geological layer in the Bakken formation where shale oil isfound is thin, and needs horizontally drilled wells, combinedwith hydraulic fracturing, to be produced.
I The conventional oil boom in ND was a vertical boom,while the unconventional/fracking boom is a horizontalboom (Miller et al. (2008)).
I This means fracking technology can be plausibly identified bythe horizontal wells, and conventional wells by the verticalwells.
I Gives us the opportunity to explore differences in welltechnology wrt supply dynamics.
back
Appendix