A Computable General Equilibrium Modelof Energy Taxation
André J. Barbé
Department of EconomicsRice University
International Association for Energy EconomicsJune 16, 2014
Barbé A New Model of Energy Taxation 1 / 29
Motivation
Background:Corporate income tax reform is always a hot topicObama’s 2014 budget eliminates deductions for fossil fuelsAdministration says these deductions favor fossil fuels
Complication:Current energy tax models are missing key issues
Barbé A New Model of Energy Taxation 2 / 29
My contribution
Key problems:Does the budget improve tax neutrality or social efficiency?How important are the issues previous models are missing?
SolutionCreate new energy tax modelInclude all key issues of energy taxation in my modelUse model to determine the social efficiency of the budgetproposal
Barbé A New Model of Energy Taxation 3 / 29
Outline
1 IntroductionWhat are the proposed changes?Why is a computable general equilibrium (CGE) modelthe best way to examine this tax reform?
2 ModelHow does my model improve on previous literature?
3 ResultsTax reform increases social welfare if carbonexternalities are at least $14 per tonMy model’s innovations are important
Barbé A New Model of Energy Taxation 4 / 29
Budget increases taxes on fossil fuel
Budget proposal changes a number of provisions inthe corporate income tax
Supporters say budget eliminates tax preferences forfossil fuels
Tax neutralityA tax system is neutral if it does not favor any type ofeconomic activity over anotherRelated to social efficiency
Barbé A New Model of Energy Taxation 5 / 29
Changes in the budget proposal
Proposed Change Revenue($ Billions)
LIFO inventory accounting 78.3Domestic manufacturing deduction 19.9Intangible drilling costs 13.7Cost depletion 11.1Superfund excise taxes 8.2Dual capacity rules 7.9Other 5.2Total 144.2
Source: Joint Committee on Taxation (2013)
Budget identifies areas that need reform but does notaccomplish reform
Barbé A New Model of Energy Taxation 6 / 29
LIFO inventory accounting changes
Table 1 : Taxation of Inventory Appreciation
Inflationary RealAppreciation Appreciation
Current Law No Tax No TaxBudget Proposal Tax TaxNeutral Tax System No Tax Tax
Current law taxes too little, budget proposal too much
Barbé A New Model of Energy Taxation 7 / 29
Are taxes higher or lower on fossil fuels thanother sectors?
Marginal Effective Tax Rate (METR) on investment is the metricused by the literature to measure tax rates
CBO (2005): 9% to 25% for fossil fuel assets and 24% forall business assetsMackie (2002): 25% to 36% for fossil fuel industries and20% for entire economy
Barbé A New Model of Energy Taxation 8 / 29
3 main issues determine the social efficiencyof energy taxes
1 Substitution between different goods
Taxing fossil fuels at different rates than other goods leads toproductive inefficiency due to substitution away from themore taxed goods
2 Energy resource supply
If energy resources are inelastically supplied, there is littleinefficiency to taxing them
3 ExternalitiesTaxes internalize costs from climate change
Barbé A New Model of Energy Taxation 10 / 29
3 main issues determine the social efficiencyof energy taxes
1 Substitution between different goods
Taxing fossil fuels at different rates than other goods leads toproductive inefficiency due to substitution away from themore taxed goods
2 Energy resource supply
If energy resources are inelastically supplied, there is littleinefficiency to taxing them
3 ExternalitiesTaxes internalize costs from climate change
Barbé A New Model of Energy Taxation 10 / 29
3 main issues determine the social efficiencyof energy taxes
1 Substitution between different goods
Taxing fossil fuels at different rates than other goods leads toproductive inefficiency due to substitution away from themore taxed goods
2 Energy resource supply
If energy resources are inelastically supplied, there is littleinefficiency to taxing them
3 ExternalitiesTaxes internalize costs from climate change
Barbé A New Model of Energy Taxation 10 / 29
Previous models did not include all 3 issues
Partial GeneralEquilibrium Equilibrium
Substitution No SomeEnergy resource supply Yes NoExternalities No Yes
Partial equilibrium (PE) models:Dasgupta, Heal, and Stiglitz (1981)Hotelling (1931)
General equilibrium (GE) models:Babiker et al. (2008)Jorgenson and Yun (2001)
Barbé A New Model of Energy Taxation 11 / 29
Overview of commodity flows in my model
22Composite
GoodsHousehold
22Industries
Rest ofthe World
Government
Labor Energy ResourceCapital
Domestic Goods
Imports
Intermediate Inputs
Exports
Investment
Consumption
Consumption
Barbé A New Model of Energy Taxation 12 / 29
Overview of commodity flows in my model
22Composite
GoodsHousehold
22Industries
Rest ofthe World
Government
Labor Energy ResourceCapital
Domestic Goods
Imports
Intermediate Inputs
Exports
Investment
Consumption
Consumption
Barbé A New Model of Energy Taxation 12 / 29
Overview of commodity flows in my model
22Composite
GoodsHousehold
22Industries
Rest ofthe World
Government
Labor Energy ResourceCapital
Domestic Goods
Imports
Intermediate Inputs
Exports
Investment
Consumption
Consumption
Barbé A New Model of Energy Taxation 12 / 29
Overview of commodity flows in my model
22Composite
GoodsHousehold
22Industries
Rest ofthe World
Government
Labor Energy ResourceCapital
Domestic Goods
Imports
Intermediate Inputs
Exports
Investment
Consumption
Consumption
Barbé A New Model of Energy Taxation 12 / 29
Cost and expenditure functional forms
Fixed coefficient (Leontief)No substitutionCannot capture productive inefficiency at all
Constant elasticity of substitution (CES)Restricts all inputs to have the same elasticity of substitution
TranslogAllows for different elasticities of substitution for each pair ofinputsCan accurately model productive inefficiency
Barbé A New Model of Energy Taxation 14 / 29
Regression
Use regression to find credible parameter values for the costfunction
Cost share of input i for industry x at time t :
sharexit =∑N
j=1 βsubstitutionxij ln (pricexit) + βtrend
xi year + βconstantxi
Problems with estimating this equation:Cost shares are endogenous to input pricesCost share error terms are correlated
Iterated 3-stage least-squares solves both of these problems
Barbé A New Model of Energy Taxation 15 / 29
Data sources for regression
Regression needs data on prices and cost shares foreach industry and input
Data sources for regression describe the USeconomy from 1960-2010
Jorgenson (2007)BEA
NIPA (National Income and Product Accounts)Gross output price index
Barbé A New Model of Energy Taxation 16 / 29
Energy resource supply
An energy resource is required to produce fossil fuels
This resource has isoelastic supply
Determine impact of resource supply by runningsimulation multiple times with different elasticities
Barbé A New Model of Energy Taxation 18 / 29
Externalities
There is disagreement about carbon externalities so Iavoid the debate entirely
Calculate social cost of carbon for which budget hasno net effect on welfare
Social cost of carbon =Equivalent variation
Reduction in emissions
Barbé A New Model of Energy Taxation 20 / 29
Budget is inefficient without externalities
Table 2 : The Effects of the Budget Proposal
Percent Change inSpecification Welfare Capital Stock EmploymentBaseline -0.50 -0.04 -0.01
The budget proposal decreases household welfare, capitalstock, and employment, but also emissions
Social cost of carbon must be at least $14 for the budgetproposal to be welfare neutral
Barbé A New Model of Energy Taxation 22 / 29
Budget decreases productivity
Table 3 : The Effects of the Budget Proposal on Productivity
Percent Change in Productivity ofSpecification Capital LaborBaseline -0.09 -0.06
Costs of the budget proposal come from decreased productivity
Worse allocation of capital, labor, and consumption across uses
Lower productivity means lower income, output, and welfare
Barbé A New Model of Energy Taxation 24 / 29
Fossil fuel production decreases
Table 4 : Effects of the Budget Proposal on Selected Industries
Percent Change in:Capital
Industry Output Stock EmploymentOil and gas extraction -2.65 -7.34 -14.5Petroleum and coal products -2.40 -3.61 -6.94
manufacturingPipeline transportation -0.62 -0.77 -0.71All -0.13 -0.04 -0.01
Decrease in output due to higher taxes is mitigated bysubstitution
Barbé A New Model of Energy Taxation 25 / 29
Fossil fuel production decreases
Table 4 : Effects of the Budget Proposal on Selected Industries
Percent Change in:Capital
Industry Output Stock EmploymentOil and gas extraction -2.65 -7.34 -14.5Petroleum and coal products -2.40 -3.61 -6.94
manufacturingPipeline transportation -0.62 -0.77 -0.71All -0.13 -0.04 -0.01
Decrease in output due to higher taxes is mitigated bysubstitution
Barbé A New Model of Energy Taxation 25 / 29
Fossil fuel production decreases
Table 4 : Effects of the Budget Proposal on Selected Industries
Percent Change in:Capital
Industry Output Stock EmploymentOil and gas extraction -2.65 -7.34 -14.5Petroleum and coal products -2.40 -3.61 -6.94
manufacturingPipeline transportation -0.62 -0.77 -0.71All -0.13 -0.04 -0.01
Decrease in output due to higher taxes is mitigated bysubstitution
Barbé A New Model of Energy Taxation 25 / 29
Fossil fuel production decreases
Table 4 : Effects of the Budget Proposal on Selected Industries
Percent Change in:Capital
Industry Output Stock EmploymentOil and gas extraction -2.65 -7.34 -14.5Petroleum and coal products -2.40 -3.61 -6.94
manufacturingPipeline transportation -0.62 -0.77 -0.71All -0.13 -0.04 -0.01
Decrease in output due to higher taxes is mitigated bysubstitution
Barbé A New Model of Energy Taxation 25 / 29
Capital and labor are a small fraction of costs
Table 5 : Selected Industry Cost Shares (Percent)
Cost share of:Industry Capital and Energy and
Labor MaterialsOil and gas extraction 11 89Petroleum and coal products 12 88
manufacturing
Output decreases very little because capital and labor are only asmall fraction of costs
Barbé A New Model of Energy Taxation 26 / 29
Results are robust
Results are robust to changes in assumptions for:CapitalLaborEnergy resourceImportsInstrumental variablesCost function parameters
Energy resource and substitution affect the size ofthe costs of the budget proposal
Barbé A New Model of Energy Taxation 28 / 29
Conclusions
Budget is not social efficiency enhancing on purelytax criteria
Budget proposal needs a social cost of carbon higherthan $14 per ton to be socially efficient
Important factors:Flexible substitutionExternalitiesEnergy resource supplyGeneral equilibrium modeling
Barbé A New Model of Energy Taxation 29 / 29
Contact Information
André J. BarbéPh.D. CandidateDepartment of Economics, Rice UniversityEmail: [email protected]: barbe.rice.edu
Barbé A New Model of Energy Taxation 30 / 29
LIFO and FIFO govern inventory deductions
Firms can use either LIFO (Last-in, First-out) or FIFO (First-in,First-out) to determine their tax deduction when selling frominventories
This tax deduction is either the amount the firm paid for the:newest unit in inventory under LIFO oroldest unit in inventory under FIFO
LIFO is more desirable if prices are increasing over time
Budget Proposal:All firms must use non-inflation indexed FIFO instead of LIFORevenue estimate: $78 billion over 10 years
Barbé A New Model of Energy Taxation 31 / 29
LIFO change disproportionately affectspetroleum refining
Energy companies own 82% of LIFO reserves on S&P 500Index (Pryzbyla, 2011)
Firms using LIFO own:73% of petroleum refining inventories (Knittel, 2009)23% of all corporate inventories (Knittel, 2009)
Mean of value of LIFO reserves by firm:$2.6 billion and 119% of inventories for oil and gas (Tipton,2012)$13 to $150 million and 2% to 28% of inventories for allother sectors (Tipton, 2012)
Barbé A New Model of Energy Taxation 32 / 29
Cost depletion would replace percentage
Cost Depletion: taxpayer deducts a percent of lease cost equalto percent of resource extracted
Percentage Depletion:Taxpayer deducts a constant percentage of property’s grossincomePercentage varies from 5-22% depending on resourceNot allowed for integrated oil companies
Budget Proposal:Coal, oil, and gas extraction must use cost depletion
Barbé A New Model of Energy Taxation 33 / 29
Percentage depletion is not first-best but maybe second
Non-neutral in a first best world:Percentages based on resource extractedEligibility based on organizational formDeduction not based on actual cost of capital invested
Offset other distortionary features:Severance taxesExcise taxes
Barbé A New Model of Energy Taxation 34 / 29
US taxes foreign source income but creditstaxes paid
Territorial tax system:Does not tax foreign source income
Worldwide tax system:Does tax foreign source incomeCredit for certain foreign taxes
Dual capacity:A foreign tax is creditable if it is not payment for a specificeconomic benefitA dual-capacity taxpayer has some non-creditable taxes
Barbé A New Model of Energy Taxation 35 / 29
Firms could only credit income tax
What foreign taxes are creditable?Facts and circumstances method: tax creditable if not forspecific economic benefitSafe harbor method: credit an amount equal to hostcountry’s generally imposed income tax rate
Budget Proposal:Firms must credit an amount equal to host country’s income taxrate for other industries
Barbé A New Model of Energy Taxation 36 / 29
No consensus on taxation of foreign sourceincome
Is budget method more accurate than facts andcircumstances method?
Should foreign source income be taxed at all?Kleinbard (2007) and Gravelle (2012) say yesDesai and Hines (2004) says no
Barbé A New Model of Energy Taxation 37 / 29
Taxes excluded from METR are significant
Figure 1 : Taxes paid by Fossil Fuel Producers in 1998-2009
Source: Author’s calculation from Bureau of Economic Analysis (BEA)US Input-Output Accounts and NIPA Table 6.18D.
Barbé A New Model of Energy Taxation 38 / 29
Corporate prevalence by industry
Table 6 : Business Receipts in 2007 by Industry and Type of Entity(Percent)
Sector C Corporation Pass-throughMining 66 34Manufacturing 80 20All sectors 62 38
Source: Table 3 of the Internal Revenue Service’sIntegrated Business Dataset
Barbé A New Model of Energy Taxation 39 / 29
Calculating average effective tax rates
Average effective tax rate =tax payments
tax baseWhat payments?
All corporate income and firm production taxesIncludes federal, state, and local taxes
What base?Total value of output base
Forward shifting (tax borne by consumers)Factor income base
Backward shifting (tax borne by capital or labor)
Barbé A New Model of Energy Taxation 40 / 29
Average of all taxes are higher on fossil fuels
Table 7 : Average Effective Tax Rates of All Firm Taxes by Sector,1998-2009
AETR (%)Sector Factor Income Value of OutputOil and gas extraction 19.3 12.0Petroleum and coalproducts manufacturing 20.4 5.1Pipeline transportation 16.5 7.2All fossil fuel 19.6 7.3
All sectors 10.8 5.9
Barbé A New Model of Energy Taxation 41 / 29
Average of all taxes are higher on fossil fuels
Table 7 : Average Effective Tax Rates of All Firm Taxes by Sector,1998-2009
AETR (%)Sector Factor Income Value of OutputOil and gas extraction 19.3 12.0Petroleum and coalproducts manufacturing 20.4 5.1Pipeline transportation 16.5 7.2All fossil fuel 19.6 7.3
All sectors 10.8 5.9
Barbé A New Model of Energy Taxation 41 / 29
But the AETR is incomplete too
Limitations:Firm taxes excludes those paid under the personal incometaxMarginal rates can differ greatly from average rates
Non-uniform rates may be efficiency enhancing:Externalities
Barbé A New Model of Energy Taxation 42 / 29
Most of my model is conventional
1 representative household and 22 firms
Industries are perfectly competitive with constantreturns to scale
Capital, labor, and energy resource are perfectlymobile between sectors
Consumer expenditure function and producer costfunction determine their purchases
Barbé A New Model of Energy Taxation 43 / 29
Translog has good traits but needs a largenumber of parameters
Cost c of output o for industry x at time t :
ln (cxot) =12
∑Ni=1
∑Nj=1 β
substitutionxij ln (pxit) ln (pxjt)
+∑N
i=1 ln (pxit)(βshare trend
xi t + βshare constantxi
)+βcost trend
xo t + βcost constantxo
The p’s are prices and the β’s are the parameters to beestimated
Notable featuresAllows for both Hicks-neutral and biased technologicalchangeLarge number of parameters dealt with by nesting
Barbé A New Model of Energy Taxation 44 / 29
Translog has good traits but needs a largenumber of parameters
Cost c of output o for industry x at time t :
ln (cxot) =12
∑Ni=1
∑Nj=1 β
substitutionxij ln (pxit) ln (pxjt)
+∑N
i=1 ln (pxit)(βshare trend
xi t + βshare constantxi
)+βcost trend
xo t + βcost constantxo
The p’s are prices and the β’s are the parameters to beestimated
Notable featuresAllows for both Hicks-neutral and biased technologicalchangeLarge number of parameters dealt with by nesting
Barbé A New Model of Energy Taxation 44 / 29
Translog has good traits but needs a largenumber of parameters
Cost c of output o for industry x at time t :
ln (cxot) =12
∑Ni=1
∑Nj=1 β
substitutionxij ln (pxit) ln (pxjt)
+∑N
i=1 ln (pxit)(βshare trend
xi t + βshare constantxi
)+βcost trend
xo t + βcost constantxo
The p’s are prices and the β’s are the parameters to beestimated
Notable featuresAllows for both Hicks-neutral and biased technologicalchangeLarge number of parameters dealt with by nesting
Barbé A New Model of Energy Taxation 44 / 29
Nesting
Nesting functions means putting cost functions inside other costfunctions in order to group the most similar products together
Nesting increases the number of equations but reduces thenumber of parameters in each equation
I follow the nesting structure of Jorgenson and Yun and have 9nests
The model has 23 sets (22 industries and 1 household) ofregressions for each nest
Barbé A New Model of Energy Taxation 45 / 29
Cost functions nesting in my model
Final domestic output of industry x
Mx
MOx
MOTx
484442
5323
MSx
MSSx
815452
565551
MPx
72716261
MMx
N312111
Ex
48632422211
LxKx
Barbé A New Model of Energy Taxation 47 / 29
Cost functions nesting in my model
Final domestic output of industry x
Mx
MOx
MOTx
484442
5323
MSx
MSSx
815452
565551
MPx
72716261
MMx
N312111
Ex
48632422211
LxKx
Barbé A New Model of Energy Taxation 47 / 29
Cost functions nesting in my model
Final domestic output of industry x
Mx
MOx
MOTx
484442
5323
MSx
MSSx
815452
565551
MPx
72716261
MMx
N312111
Ex
48632422211
LxKx
Barbé A New Model of Energy Taxation 47 / 29
Predictive power of my model is high
Do not look at individual parameters for significance
Predictions of the model as a whole are what matter
Table 8 : R2 of Cost Function Regressions for All Industries
Min Mean MaxR2 0.971 0.993 0.999
Barbé A New Model of Energy Taxation 48 / 29
Remove tax preferences in abase-broadening reform
Compare long-run equilibrium of US economy under current lawand budget proposal
Tax reformIncrease tax rates on fossil fuel producing sectors as givenin budget proposalSimultaneously lower overall capital tax rate on all sectorsRevenue neutral
Results express the effects of the budget proposal
Externalities are included by calculating the social cost ofcarbon for which the proposal has no net effect on welfare
Barbé A New Model of Energy Taxation 49 / 29
Including the energy resource matters butexactly how does not
Table 9 : The Effects of Budget Proposal Under Various ResourceSupply Assumptions
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 14Elastic resource -0.05 -0.01 15Inelastic resource -0.04 -0.01 13No energy resource -0.09 -0.02 21
Including an energy resource changes results but exactly how itis modeled matters little
Barbé A New Model of Energy Taxation 50 / 29
Including the energy resource matters butexactly how does not
Table 9 : The Effects of Budget Proposal Under Various ResourceSupply Assumptions
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 14Elastic resource -0.05 -0.01 15Inelastic resource -0.04 -0.01 13No energy resource -0.09 -0.02 21
Including an energy resource changes results but exactly how itis modeled matters little
Barbé A New Model of Energy Taxation 50 / 29
Including the energy resource matters butexactly how does not
Table 9 : The Effects of Budget Proposal Under Various ResourceSupply Assumptions
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 14Elastic resource -0.05 -0.01 15Inelastic resource -0.04 -0.01 13No energy resource -0.09 -0.02 21
Including an energy resource changes results but exactly how itis modeled matters little
Barbé A New Model of Energy Taxation 50 / 29
Instruments are weak
Weak instruments can harm inference
Test instruments by testing for for underidentification (Anderson,1951) and weak instruments (Stock and Yogo, 2002)
Table 10 : Summary of Instrumental Variable Tests
Regressions (%)Reject underidentification (5% level) 57Reject weak instruments (0.30 maximal bias) 13
My instruments are weak
Barbé A New Model of Energy Taxation 51 / 29
Results are not sensitive to weak instruments
Table 11 : Effects of Budget Proposal under Different InstrumentalVariables
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 142 period lags for IV -0.03 -0.01 19No instruments -0.05 -0.01 13
Barbé A New Model of Energy Taxation 52 / 29
Results are not sensitive to cost functionvalues
Table 12 : Effects of Reform in Monte Carlo Simulations
Percentile of PercentChange in Variable
Variable 5% 50% 95%Capital Stock -0.05 -0.04 -0.04Employment -0.01 -0.01 -0.01
Percentile of VariableVariable 5% 50% 95%Social Cost of Carbon 19 15 12
Even at the 95% level, no variable changes sign
Barbé A New Model of Energy Taxation 53 / 29
Results are not sensitive to cost functionvalues
Table 12 : Effects of Reform in Monte Carlo Simulations
Percentile of PercentChange in Variable
Variable 5% 50% 95%Capital Stock -0.05 -0.04 -0.04Employment -0.01 -0.01 -0.01
Percentile of VariableVariable 5% 50% 95%Social Cost of Carbon 19 15 12
Even at the 95% level, no variable changes sign
Barbé A New Model of Energy Taxation 53 / 29
Elasticity parameters do not affect efficiencycosts
Table 13 : The Effects of Budget Proposal Under Various Capital andLabor Elasticities
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 14Elastic capital -0.09 -0.01 16Inelastic capital -0.02 -0.01 13Elastic labor -0.05 -0.02 15Inelastic labor -0.04 -0.01 12
The social cost of carbon is very stable because greater taxdistortion increases both its numerator and denominator
Barbé A New Model of Energy Taxation 54 / 29
Substitution is important for the efficiencycosts of the budget proposal
Table 14 : Effects of Tax Reform Under Various Import Assumptions
Percent Change in: Social Cost ofSpecification Capital Stock Employment Carbon ($/ton)Baseline -0.04 -0.01 14No world market 0.00 -0.01 7for fossil fuels
Removing the world market assumption reduces thecosts of the budget proposal
Barbé A New Model of Energy Taxation 55 / 29