taxing exhaustible resources:
evidence from California oil production
Nirupama S. Rao
NYU Wagner
June 2011
Taxes play an important role in the
production and consumption of fossil fuels
• Recent price escalation renewed discussion of
new federal sur-taxes.
– Several 2008 bills and proposals
– 2005 Markey-Emanuel, 1990 bills
• 21 U.S. states and many other nations levy
severance taxes on oil.
• In 2009, the U.S. collected $38B in excise taxes
on fossil fuels.
• Potentially important for climate change
policy
Taxing exhaustible resources is
different than taxing other production
• Key distinction: the intertemporal sum of
services from a given stock is finite.
– Producers face a ‘pump today or pump tomorrow’
decision, linking production each period.
– There is an opportunity cost to pumping today.
– Need to take the re-timing of production into
account when assessing the deadweight loss.
Preview of Results
• Estimate an after-tax price elasticity of 0.24.
– Micro-data are key to estimating the supply response.
• Only minor exit response.
– Producers do not shut wells to avoid taxes.
• Estimates suggest that the dynamic inefficiency
cost of a 15% temporary excise tax is between 1%
and 5% of original producer surplus.
– Each dollar of revenue costs between $1.13 and $1.66
in lost producer surplus.
Impact of Temporary Tax
• Some wells will shut down early
• Wells that continue to pump will tilt their
extraction paths forward, reducing producer
surplus.
– Delay reduces the PDV of profits
– Operators will extract sub-optimally
200
180
Extraction Rate
160
140
120
100
Original Extraction Path
80
60
Extraction Path After Introduction of
Temporary Tax
40
20
0
0
5
10
15
20
25
Time
30
35
40
45
Past Work on Supply Elasticities
• Estimates from previous studies referred to as
generally unreliable and implausible by a
recent CRS study.
• Relying on aggregate data and time-series price
variation, they find small and sometimes
implausible negative elasticities.
Aggregate Time-Series Variation May
Not Identify the Appropriate Elasticity
• Underlying price and taxes may be of different
persistence.
– If taxes are viewed as less (more) persistent, price
elasticities will overstate (understate) tax response
since in reality adjustment is costly.
• Longterm price considerations are important.
– Current price changes inform producer
expectations of longterm price.
1980 “Windfall Profit” Tax provides rich
variation in after-tax price
• Instituted following the de-control of oil prices.
• Levied on only domestically produced oil.
– Transport costs are minimal, meaning full incidence
born by domestic oil producers.
• Raised a substantial amount of revenue.
– At its height in 1982 generated $44B in gross revenue,
or almost half as much as the corporate income tax.
Timing of Decontrol and WPT
Windfall Profit Tax
New (06/79)
1979
Very Heavy
(09/79)
Heavy (01/80)
Old (02/81)
1980
1981
Old--Phaseout
WPT was not a profit tax!
τ it (Pit − Bi ) if Pit > Bi
WPT tax it = 
0 otherwise
W
(1− τ Corp
)(Pit − τ it (Pit − Bi )) if Pit > Bi
t
ATPit = 
Corp
(1− τ t )Pit otherwise
WPT payments were deductible from
corporate taxable income
Tax rates differed across wells
Tier I:
Old Oil
Tier II:
Stripper Oil
Tier III:
New
Heavy
Well produced in 1978
τ = 0.70
Produces < 10 bbl/day for 12
consecutive months
τ = 0.60
Zero Production in 1978
API Gravity<16
τ = 0.30*
τ = 0.30
*Tax rate for new wells were further reduced to 0.225 starting in 1982.
Tax rates based on well characteristics
• Differences across wells that are correlated
with production and tax rates bias estimates.
– Operating costs could vary by specific gravity, or
easier to pump formations may have been tapped
earlier.
– Makes flexibly controlling for well characteristics
important.
Before- and After-Tax Price
Well 120005: Livermore Field
Old Oil, API Gravity of 23; Stripper starting 10/1982
(70% tax rate until 10/1982, then 60% until 1986)
Before- and After-Tax Price
Well 120005: Livermore Field
70
60
50
40
30
Real Posted Price
After-Tax Price Before Corp. Tax
After-Tax Price
Well Mean
Residual--Well FE
Residual--Well, Time FE
20
10
0
-10
-20Jan-77 Jan-78 Jan-79 Jan-80 Jan-81 Jan-82 Jan-83 Jan-84 Jan-85
Before- and After-Tax Price
Well 120005: Livermore Field
70
60
50
40
30
Real Posted Price
After-Tax Price Before Corp. Tax
After-Tax Price
Well Mean
Residual--Well FE
Residual--Well, Time FE
20
10
0
-10
-20Jan-77 Jan-78 Jan-79 Jan-80 Jan-81 Jan-82 Jan-83 Jan-84 Jan-85
Before- and After-Tax Price
Well 120005: Livermore Field
70
60
50
40
30
Real Posted Price
After-Tax Price Before Corp. Tax
After-Tax Price
Well Mean
Residual--Well FE
Residual--Well, Time FE
20
10
0
-10
-20Jan-77 Jan-78 Jan-79 Jan-80 Jan-81 Jan-82 Jan-83 Jan-84 Jan-85
Before- and After-Tax Price
Well 1300071: Brentwood Field
New Oil, API Gravity of 40.7; Never Stripper
(30% tax rate until 1982, then decreases to 22.5%)
Before- and After-Tax Price
Well 1300071: Brentwood Field
80
70
60
50
Real Posted Price
After-Tax Price Before Corp. Tax
After-Tax Price
Well Mean
Residual--Well FE
Residual--Well, Time FE
40
30
20
10
0
-10
-20Jan-77 Jan-78 Jan-79 Jan-80 Jan-81 Jan-82 Jan-83 Jan-84 Jan-85
Data
• Production data comes from the California
Department of Conservation.
– Describes monthly production, specific gravity,
pumping method, location, status and completion
date for all wells in California, 1977-2008.
– 30M observations describing 141K wells
• Price data comes from Platt’s Oil Price Handbook
and Oilmanac
– Monthly posted price for major producers by field.
• Price control provisions and tax rates from the
1977-1986 Code of Federal Regulations
Summary Statistics
Mean
Oil Production (barrels)
Oil Production if Producing
After-tax Price ($)
WPT Tax Rate
Purchase Price
API Gravity (degrees)
Number of Wells
Observations
443.3
666.1
18.3
0.21
41.1
18.2
75,342
6,517,140
Standard Deviation
Overall
Within-Well
3071.1
2858.5
3745.0
3460.5
4.1
3.5
0.24
0.19
10.1
9.78
6.8
1.4
Quantity Regression Specification
• Specifications of the general form:
Qi t = β0 + β1(1 − τ Ct )( B i t + (1 − τiWt )( Pt − B i t )) + β2agei t + σ t + δi + εi t
where Qit is the quantity extracted.
• Specification is in the spirit of other studies.
Identification
• Identification achieved from within-well
deviations in after-tax price driven by the
lifting of price controls and the introduction of
the WP tax, less common time-varying factors.
– Nets out heterogeneity in productivity that is
constant over the life of the well.
– Nets out secular common factors, such as
expectations of future price, that vary over time.
Key Assumptions for Identification
• The specification will not identify the response
to after-tax price if well productivity is
affected by unobserved factors that are:
– Time-varying and well-specific
– (Correlated with the rescindment of price controls
or the introduction or the WP tax)
Table 2: Well-Specific Output: Panel Data Estimates
After-tax Price
Well Age
(1)
8.730
(1.082)
-1.269
(0.069)
Well Age Squared
Well FE
Time FE
API Gravity Decile FE
API Gravity Time Trends
After Tax-price Elasticity
Number of Wells
Observations
Y
Y
N
N
0.237
(0.029)
(2)
8.741
(1.082)
-1.228
(0.081)
-(0.0003)
(0.0002)
Y
Y
N
N
0.238
(0.029)
(3)
7.659
(0.979)
6.531
(1.885)
(4)
9.598
(0.765)
-1.258
(0.050)
Y
Y
Y
Y
0.208
(0.027)
Y
Y
N
N
0.261
(0.021)
75,342
75,342
75,342
73,548
6,517,140 6,517,140 6,517,140 6,350,820
Table 3: Well-Specific Output, Panel Data Estimates
Flowing vs. Pumped Wells
After Tax-price
Elasticity
p-value
95% Confidence
Intervals
After-tax Price
Well Age
Number of Wells
Observations
(1)
Baseline
(2)
Pumped
(3)
Flowing
0.237
(0.029)
0.000
0.356
(0.024)
0.000
-0.101
(0.088)
0.253
[0.180, 0.295] [0.083, 0.108] [-0.274, 0.072]
8.730
(1.082)
-1.269
(0.069)
11.520
(0.784)
-1.570
(0.055)
-12.180
(10.649)
-0.377
(0.866)
75,342
6,517,140
72,797
5,698,198
13,198
818,942
Response Along the Extensive Margin:
The Shut-in Decision
• Shut-in likely means a permanent output gap.
– Shutting-in risky
– Re-opening costly
• If the quantity response is driven by exit, then
the the efficiency cost of the tax is higher.
• Wells with high fixed or variable costs or little
remaining life are most likely to shut.
Shut-In Regression Specification
• Specifications of the general form:
Si t = β0 + β1(1 − τ Ct )( B i t + (1 − τiWt )( Pt − B i t )) + β2agei t + σ t + δi + εi t
where Sit is 1 is the well is shut-in.
• Estimated via conditional logit
– Requires variation in shut-in for each well.
Table 4: Well Shut-in Decisions
(1)
CL
Shut-in
Var.
After-tax Price
-0.0052
(0.0008)
Well Age
0.0126
(0.0007)
Well FE
Y
Time FE
Y
API Gravity Decile FE
N
API Gravity Time Trends
N
After Tax-price Semi-0.095
Elasticity
(0.0148)
Number of Wells
29,297
Observations (millions)
2.69
(2)
CL
Shut-in
Var.
(3)
CL
Drop
NPR
(4)
OLS
Shut-in
Var.
(5)
OLS
Full
Sample
-0.0064 -0.0060 -0.0043 -0.0015
(0.0002) (0.0009) (0.0004) (0.0002)
0.0455
0.0121 0.0014 0.0005
(0.0008) (0.0007) (0.0000) (0.0000)
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
N
N
Y
N
N
N
-0.117
-0.111
-0.080
-0.027
(0.0037) (0.0169) (0.0078) (0.0034)
29,297
29,297 29,297 75,342
2.69
2.69
2.69
6.52
Reconciliation with Previous Estimates
Study
Griffin (1985)
Hogan (1989)
Jones (1990)
Sample
Period
1971Q1–
1983Q3
1966–1987
1983Q3–
1988Q4
Elasticity
Estimate
-0.05 (0.02)
0.09 (0.03)
0.07 (0.04)
Dahl and Yücel
(1991)
1971Q1–
1987Q4
-0.08 (0.06)
Ramcharran (2002)
1973–1997
0.05 (0.02)
Reconciliation with Previous Estimates
• Previous positive significant estimates are 60
to 80 percent smaller.
• Time-series aggregate data yield smaller
estimates:
– DOE pre-tax price series suffers from significant
measurement error.
– Aggregation subsumes heterogeneity in well
productivity.
– Tax and price variation may differ in persistence.
Table 6: Alternative Specifications Using National
Average Pre-Tax Price Series
WTI Price
Well Age
Time
Well Dummies
Time Dummies
After Tax-price
Elasticity
p-value
Number of Wells
Observations
(1)
Baseline
8.730
(1.082)
-1.269
(0.069)
Y
Y
0.237
(0.029)
0.000
75,342
6,517,140
(2)
Within Well
0.320
(0.148)
-0.124
0.101
Y
N
0.021
(0.010)
0.030
75,342
6,517,140
(3)
Pooled
0.365
(0.153)
-0.148
0.081
N
N
0.024
(0.010)
0.017
75,342
6,517,140
(4)
Time-Series
11,223
(10,036)
48,874
(4,468)
N
N
0.017
(0.015)
0.263
75,342
108
Table 7: Alternative Specifications Using After-Tax Price
(1)
Baseline
After-Tax Price
Well Age
Time
Well Dummies
Time Dummies
After Tax-Price
Elasticity
p-value
Number of Wells
Observations
(2)
Pooled
8.730
-19.676
(1.082)
(1.015)
-1.269
(0.069)
0.315
(0.081)
Y
N
Y
N
0.237
-0.535
(0.029)
(0.028)
0.000
0.000
75,342
75,342
6,517,140 6,517,140
(3)
(4)
(5)
Time-Series Pooled Time-Series
-58,302
13.432
158,262
(39,283)
(4.946)
(44,607)
0.098
-3.476
-56,305
(0.007)
(0.362)
(2,164)
N
N
N
N
N
N
-0.036
0.149
0.208
(0.024)
(0.055)
(0.059)
0.138
0.000
0.000
20,699
108
1,090,659
108
Conclusion
• Production is responsive to changes in after-tax
price.
– Little action on the shut-in margin.
– Substantial evidence of retiming.
• Exhaustibility complicates the DWL calculation.
• Caveats:
– Estimate based on California wells more than 20 years
ago.
– Estimate does not include changes in exploration or
development.
Policy Implications & Applicability
• California has considered levying a 9.9% oil
excise tax.
– Reduce near-term CA production
– Trigger minimal shut-in response
• Consider exemptions for very low production wells
– Increase reliance on (Venezuelan) imports
Policy Implications & Applicability
• National carbon pricing
– $20 per ton CO2 => $4.77 per barrel of oil
– 1.5% decrease in production? Not necessarily.
• Would apply to domestic and imported oil.
• Tax could affect pre-tax prices, leading to a
smaller decrease in U.S. production.