Effects of U.S. Tax Policy on Greenhouse Gas Emissions (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Chapter 2 Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions INTRODUCTION The last chapter described the background for the present study. This chapter examines the connection between tax policy and greenhouse gas (GHG) emissions along with the approach the committee followed. We begin by ex- plaining the mechanism by which tax policy can affect greenhouse gas emis- sions and then explain how the committee chose specific provisions of the Inter- nal Revenue Code (IRC) for in-depth analysis. We also give an overview of the concept of tax expenditures, because our methodology for choosing provisions depended on estimates of tax expenditures. Another important facet of the anal- ysis is the regulatory environment and how regulations outside of the tax code affect the code’s impacts on emissions. We next provide a brief description of economic modeling and how the absence of available literature on this topic led the committee to commission new economic modeling to analyze the greenhouse gas impacts of select tax code provisions. In the last part of the chapter, we provide a lengthier discussion of how we chose the four models we used, give some details of how those mod- els represent the economy, and then discuss some of the parameters of the mod- els. A more complete discussion is contained in Appendix A. HOW TAX CODE PROVISIONS IMPACT GREENHOUSE GAS EMISSIONS Taxes are one of the many factors that affect the level and composition of economic activity, which in turn determines the level of anthropogenic emis- sions of CO2 and other greenhouse gases. The purpose of this study is to untan- gle the effects of different individual tax provisions from other determinants of output and to estimate how through these changes the tax code affects U.S. and global GHG emissions. 23 

24 Effects of U.S. Tax Policy on Greenhouse Gas Emissions The percentage depletion allowance provides an illustrative example of the task before the committee. The tax code provides two ways for firms to re- cover capital expenditures made to establish a mineral property such as an oil or gas well. Under cost depletion, each year the owner of the well can deduct from taxable income a portion of the costs of developing the well. Over the life of the well the total amount deducted cannot exceed the total amount invested. Per- centage depletion allows owners a deduction based on a percentage of that year’s gross receipts from the well, and the total lifetime amount can exceed the cost of developing the well. Currently, the code allows producers of oil and gas to deduct 15 percent of the gross income of a property when calculating taxable income. The special provision is subject to several limitations (such as being available only to independent producers, only up to 1,000 barrels per day, and limited to 50 percent of net income). According to the Joint Committee on Taxa- tion (JCT), this deduction reduced tax revenues by as much as $0.9 billion in fiscal year 2011.1 The question asked of the committee is, how much did this change the emissions of CO2 and other GHGs? At first glance, the impact of this policy on GHG emissions appears straightforward. By allowing a larger deduction for depletion than the actual economic costs, the percentage depletion allowance subsidizes oil and gas pro- duction. It thereby lowers prices for petroleum products and increases petroleum use and the resulting GHG emissions. A first estimate could look like the fol- lowing. One could attempt to estimate using econometric methods how much production of oil increased because of the percentage depletion allowance. Mul- tiplying that amount by the emissions per unit of oil production would then yield a first-order estimate of the amount of additional emissions resulting from the tax preference. In reality, estimating the impact would be much more difficult than ap- pears at first glance. In practice, the information on covered oil production would be difficult to obtain. The Internal Revenue Service does not release the necessary data from corporate income tax returns. Nonintegrated firms may have production and output records, but such records are unlikely to reveal how much additional oil the firms produce on account of the incentive. Second, the provision applies to natural gas as well as oil, and it would be necessary to in- clude the impacts on gas in the calculations. Third, the substitution between oil and gas and other fuels would need to be considered. Fourth, since oil is bought and sold in world markets, the role of global supply and demand would need to be considered. Further complicating the issue, climate change results from GHG emis- sions anywhere around the globe and thus depends upon total world consump- tion of fuels. From the standpoint of a U.S. inventory of greenhouse gas emis- sions, we may be interested in how a tax provision affects U.S. fuel 1 The JCT estimates revenues foregone of $0.9 billion, while the Treasury Department estimates $1.2 billion. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 25 consumption. But for climate change, we need to consider how any provision would ripple outside the United States and affect consumption abroad as well. The complexities do not end there. If depletion allowances subsidize oil and natural gas production, but the produced gas displaces coal use, then even though the subsidy is contributing to the production of a fuel that will release GHGs when burned, it may actually be lowering total emissions from what they otherwise might have been, because coal is more GHG intensive than natural gas. This point is particularly important for provisions that favor non-carbon- emitting sources like wind. They contribute to lower emissions only to the ex- tent they on balance displace GHG-emitting technologies. So the GHG reduc- tion benefits depend on what they displace or induce. Calculations of emissions impacts must also incorporate the upstream and downstream impacts of any changes. Biofuels provide a useful example. It is tempting to suppose that biofuels would have zero or very small net GHG emis- sions, because they come from plant material that will be regrown at the next harvest. Recent research has shown that this supposition is far from the mark (see the discussion in Chapter 5). Production, transport, and conversion of bio- fuels all require energy input, usually from fossil fuels. Fertilizers used for crop production require energy to produce, and nitrogen fertilizer directly produces nitrous oxide emissions, itself a powerful greenhouse gas. Further, conversion of non-crop land to biofuels crops may reduce the existing stock of carbon. The effects of broader provisions in the tax code introduce yet further complications. Accelerated depreciation or provisions affecting housing directly affect decisions on investment and housing, but just how those are related to greenhouse gas emissions is far from obvious. Such provisions may affect the overall level of national output and through that channel affect emissions. Tax code provisions also affect the mix of goods and services produced among in- dustries and thus would alter emissions per unit of total production if some sec- tors are more emissions-intensive than others. Additionally, some of the provi- sions have effects on the pattern of housing choices and on land use and real estate development. It is tempting to ignore these economy-wide ripple effects as unimportant. But tax changes, if applied to a very large economic base, may have large impacts on GHG emissions. We discuss such potential impacts of broad-based tax preferences in Chapter 6. APPROACH AND METHODOLOGY OF THIS STUDY Given limited time and resources, the committee needed to decide which of the many provisions in the tax code to analyze. We reasoned that the largest effects were likely to result from some combination of provisions with large tax revenue effects and those targeted toward activities that were closely related to greenhouse gas emissions. Of course, without actually making the calculations of the impacts, the committee could not know how to combine these two factors to develop a firm ranking. Further, the complexity of interactions of regulations, markets, and technology often means this assumed relationship does not neces- 

26 Effects of U.S. Tax Policy on Greenhouse Gas Emissions sarily hold. However, we faced the need to limit the size of task to available time and resources. This study could not proceed in the way that many National Academies’ studies do, with the committee’s task largely confined to evaluating and synthe- sizing an existing body of literature to answer the questions presented in the statement of task. While there is an extensive literature on the response of eco- nomic activity to various tax incentives, including those related to energy provi- sions, with only a few exceptions do existing studies carry through to the impact on greenhouse gas emissions. Thus, the committee faced a task of estimating the GHG impacts of the tax code provisions virtually from scratch. Because of the complexities discussed above, the committee chose to use economic and energy models to evaluate the impacts of different provisions. These many complexities have a critical implication for the strategy that the committee pursued in answering its charge. It is clear that the only way to assess the implications of the tax code for greenhouse gas emissions is to put together a comprehensive or integrated framework that assesses each of these effects and how they interact, moving from the specific tax provision to how that provision affects markets and then to how these market changes affect the many decisions across the economy that produce greenhouse gas emissions. The comprehensive framework used in this study is integrated energy-economic and greenhouse gas modeling. Models are numerical representations of a sector or an economy that attempt to capture the most important economic relationships and behavioral responses that generate these effects under consideration. The models the com- mittee chose to use for this study incorporate estimates of the response of firms and individuals to tax incentives and incorporate the impacts of these behavioral changes on GHG emissions. The committee then directed the National Academies to contract with re- search teams who have developed and used these integrated energy-economic models to calculate the effects of changes in different tax provisions on GHG emissions. This chapter’s section on modeling approaches and comparative strengths of models considered for use in supporting analyses and Appendix A review the models considered, those for which it was possible to develop timely analysis, and a comparison of the relative strengths and weaknesses of different models. There are literally dozens of integrated energy-economic models that are currently in use around the world. These differ in terms of their structures, algo- rithms, databases, empirical estimations, sectoral details, and economic compre- hensiveness. Appendix A provides more details about the models used in this study, including their structure and capabilities, documentation, and alternatives that the committee considered. In addition, because the tax provisions span such a wide range of economic sectors—ranging from very narrow ones directly af- fecting only a small component of the energy sector to those that affect nearly every decision in the economy—no single model could investigate all of the provisions we targeted for analysis. Nevertheless, for most of the provisions we wanted to investigate for this study, we identified at least one model capable of 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 27 analyzing it. Moreover, there was some overlap among models so that we could compare results across models. MAJOR TAX PROVISIONS SELECTED FOR ANALYSIS Given the size and complexity of the U.S. tax code, it is clearly impossible to analyze more than a small fraction of its provisions. The committee’s state- ment of task delimits the task as follows: The committee “will undertake a study to identify tax provisions that have substantial effects on the emission rates of carbon dioxide and other greenhouse gases.” Given the charge, the committee interpreted its task as evaluating specific provisions of that code that are closely related to incentives for activities that would likely affect emissions or ones that have been discussed as a target for change and could have large indirect effects on emissions. After reviewing the existing literature and the major provisions, the committee selected tax provi- sions based on three criteria: 1. The provisions would include the IRC sections most likely to have a significant impact on greenhouse gas emissions. 2. The provisions could be analyzed using available economic models. 3. The analysis would provide guidance on how to undertake a similar analysis of proposed tax legislation in the future for those provisions we could not investigate for this report. This section reviews the method and the rationale the committee used to decide which provisions to analyze as part of this study. Excise Taxes Excises are taxes on the sale of specific goods or when certain activities are undertaken. For example, the federal government and most state govern- ments charge a per-gallon excise tax on gasoline for highway use. The federal government also imposes taxes on air transport of people or property. As shown in Table 2-1, 61 percent of all federal excise revenues came from taxes on trans- portation fuels or transportation activities. Given the large portion of excise rev- enue from these fuel taxes, the committee decided to consider the top three of these—the taxes on highway gasoline and diesel fuels, the airline passenger tax, and the taxes on aviation fuels. Ultimately, modeling capabilities allowed analy- sis only of the tax on motor fuels for highway use and the tax on aviation fuel, shown in bold in Table 2-1. Moreover, we found a large literature discussing the motor fuels excise taxes’ impacts on fuel consumption and vehicle miles trav- eled along with one study that attempts to assess the tax’s impact on greenhouse gas emissions. We discuss the literature and the findings from our commissioned modeling analysis of energy-sector excise taxes in Chapter 4. 

28 Effects of U.S. Tax Policy on Greenhouse Gas Emissions TABLE 2-1 The 10 Largest Excise Tax Collections for Fiscal Year 2010 Amount Collected Chapter Excise Tax FY 2010 (billions of $)2 Where Discussed Gasoline 25.1 Chapter 4 Tobacco, domestic 15.9 N/A Diesel fuel, except for trains and 8.6 Chapter 4 intercity buses Transportation of persons by air 7.6 N/A Liquor, domestic 3.7 N/A Beer, domestic 3.2 N/A Use of international air travel facilities 2.4 N/A Truck, trailer, and semitrailer chassis and 1.9 N/A bodies, and tractors Liquor, imported 1.3 N/A Telephone and teletypewriter services 1.1 N/A Aviation fuel 0.4 Chapter 4 Bold entries were analyzed as part of this study. Source: Internal Revenue Service, Statistics of Income Bulletin Historical Table 20: Fed- eral Excise Taxes Reported to or Collected by the Internal Revenue Service, Alcohol and Tobacco Tax and Trade Bureau, and Customs Service, Fiscal Years 1999-2010. Tax Expenditures Every year the staffs of the congressional Joint Committee on Taxation (JCT) and the Treasury Department (Treasury) prepare an annual compendium of tax expenditures, the latter for the administration’s budget by the Office of Management and Budget.3 The Congressional Budget Act of 1974 defines tax expenditures as “revenue losses attributable to provisions of the Federal tax laws which allow a special exclusion, exemption, or deduction from gross income or which provide a special credit, a preferential rate of tax, or a deferral of tax lia- bility.” 2 Internal Revenue Service Statistics of Income Division. Federal Excise Taxes Re- ported to or Collected by the Internal Revenue Service, Alcohol and Tobacco Tax and Trade Bureau, and Customs Service, by Type of Excise Tax (Table 20). All amounts are nominal dollars. 3 JCT’s estimates are reported in JCT ESTIMATES, usually published in January each year. The Treasury estimates are published in Office of Management and Budget’s annu- al budget analysis (Analytical Perspectives). The two use different methodology to pre- pare the estimates. For a discussion of the differences, see JCT ESTIMATES, pp. 23-24. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 29 Traditionally, the government has seen this foregone revenue as equivalent to subsidizing the activity directly through a budget expenditure. JCT and Treasury organize each set of estimates by budget function activity, and report estimates for the current fiscal year and the next 4 years. These analyses include provisions that are scheduled to expire, permanent provisions, and some provi- sions that have expired but nonetheless continue to have revenue impacts. From the long lists of tax expenditures, the committee sought to analyze those likely to have the largest impacts on greenhouse gas emissions. As men- tioned in Chapter 1, combustion of fossil fuels accounts for roughly 90 percent of anthropogenic greenhouse gas emissions4 (U.S. Environmental Protection Agency, 2012). The committee therefore looked first at IRC provisions that di- rectly affect energy production or consumption. For energy-related tax expendi- tures, the committee ranked each provision by the size of the estimated revenue change for the year 2010.5 This ranking revealed that the 10 policies that the Treasury estimates re- sult in the largest amounts of estimated foregone revenue combined account for over 90 percent of all estimated tax expenditures due to energy-specific code sections. Because of this, the committee initially considered those 10 as a first step to winnow the field. Table 2-2 lists these 10 largest energy-specific tax ex- penditures. Bolded entries indicate provisions that the committee was able to analyze with models as part of this study. Chapters 3 and 5 analyze the impact of these energy-sector tax policies on GHG emissions. Broad-based Tax Expenditures In addition to provisions that directly affect the energy sector, the commit- tee examined broad-based provisions that may indirectly affect emissions. Table 2-3 shows the 10 largest broad-based tax expenditures for fiscal year 2010. After examination and discussion with modeling teams, the committee chose three provisions for the study: subsidies to housing, subsidies to health care insurance, and accelerated depreciation of machinery and equipment. 4 Percent share of energy-related emissions computed by committee and National Re- search Council staff. 5 The 2010 figures were the most current that were available at the time the committee began its deliberations. The ordering and composition of the 10 largest tax expenditures estimates did not change appreciably through the study duration with the notable excep- tion of the Volumetric Ethanol Excise Tax Credit, which expired in 2012. Many of these energy-related tax subsidies were created to support fossil fuel industries, because they were seen as key to war efforts during World Wars I and II. Fossil fuel industries contin- ued to receive the bulk of energy tax expenditures until the turn of the twenty-first centu- ry when provisions favoring renewable fuels began to grow. As of this writing, the tax subsidies now favor renewable energy sources over fossil fuels in terms of total cost to the Treasury. 

30 Effects of U.S. Tax Policy on Greenhouse Gas Emissions TABLE 2-2 The 10 Largest Energy Tax Policies (by dollars of foregone revenue) FY 2010 Chapter (billions of Where Provision 2010$)6 Discussed Alcohol Fuel Credit and Excise Tax Exemption 5.75 Chapter 5 Credit for Electricity Production from Renewable 3.90 Chapter 3 Sources (Including the cash grant in lieu of tax credit) Credit for Energy Efficiency Improvements 3.19 Chapter 3 to Existing Homes Excess of Percentage over Cost Depletion 0.98 Chapter 3 for Oil and Gas Wells Special Tax Rate on Nuclear 0.90 Chapter 3 Decommissioning Reserve Funds Temporary 50-Percent Expensing for Equipment 0.76 N/A Used in the Refining of Liquid Fuels Biodiesel Producer Tax Credit 0.51 Chapter 5 Expensing of Exploration and Development Costs 0.40 N/A for Oil and Gas Credit for Investment Renewable Energy 0.30 N/A Infrastructure Tax Credit and Deduction for Clean-burning 0.24 Chapter 3 Vehicles Preferential Tax Treatment of Certain Publicly 0.50 (JCT) N/A Traded Partnerships with Qualified Income Derived from Certain Energy-related Activities Credits for Advanced Energy-manufacturing 0.18 N/A Facilities Some policies are codified in multiple IRC provisions. Bold entries were analyzed using models as part of this study. Sources: Office of Management and Budget, Fiscal Year 2012 Analytical Perspectives, Budget of the U.S. Government. Staff of the Joint Committee on Taxation, Estimates of Federal Tax Expenditures for Fiscal Years 2010-2014. 6 The figure we report here for Alcohol Fuel Credits includes the refundable excise tax credit mentioned in Treasury’s footnote 2. Likewise, the figure for the electricity produc- tion from renewable resources includes the cash grants offered by the Treasury in lieu of the production credit mentioned in Treasury’s footnote 1. All amounts are nominal dol- lars. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 31 TABLE 2-3 The 10 Largest Broad-based Tax Expenditures FY 2010 Chapter (billions of Where Provision 2010$) Discussed Exclusion of Employer Contributions for Medical 160.1 Chapter 6 Insurance Premiums and Medical Care Deductibility of Mortgage Interest on 79.1 Chapter 6 Owner-occupied Homes Earned Income Tax Credit 59.6 N/A Exclusion of 401(k) Plans 52.2 N/A Accelerated Depreciation of Machinery and 39.8 Chapter 6 Equipment (normal tax method) Exclusion of Employer-sponsored Retirement Plans 39.6 N/A Step-up Basis of Capital Gains at Death 39.5 N/A Making Work Pay Tax Credit 38.9 N/A Deferral of Income from Controlled Foreign 38.1 N/A Corporations (normal tax method) Capital Gains (except agriculture, timber, iron ore, 36.3 N/A and coal) Source: Office of Management and Budget, Fiscal Year 2012 Analytical Perspectives, Budget of the U.S. Government. While the broad-based tax provisions do not affect the energy sector or GHG emissions directly, they may have an effect on overall emissions, because they affect output in large portions of the economy. There are two routes by which broad-based tax provisions can affect GHG emissions. First, they may change the mix of goods and services produced from high (or low) GHG- intensive sectors to low (or high) GHG-intensive sectors, thereby affecting over- all emissions. Second, they may affect the overall size of the economy and therefore change emissions simply because the economy is larger or smaller. At $160 billion per year, the exemption for employer-provided health in- surance premiums is the largest federal tax subsidy. Even though this provision is not directed toward energy per se, because health spending is such a large part of the overall economy, a change in its size or energy use could have a signifi- cant impact on GHG emissions. Similarly, the housing subsidies are available to a large number of taxpay- ers. Moreover, the links between housing and emissions are more straightfor- ward than the links for health care. Because the subsidy lowers the cost of hous- ing, it makes it easier for families and individuals to own more or larger houses. 

32 Effects of U.S. Tax Policy on Greenhouse Gas Emissions Residential housing directly accounts for one-fifth of all U.S. GHG emissions and is involved indirectly through development patterns that may reduce hous- ing density, thereby increasing emissions from automobiles (U.S. Environmental Protection Agency, 2012). A change in the amount of housing may therefore have sizeable impacts on GHG emissions. Accelerated depreciation allows firms to deduct their investments in ma- chinery and equipment on a faster schedule than the standard tax lifetime. This tax preference lowers the cost of capital and encourages firms to invest more, thereby resulting in some combination of increased amounts and faster turnover of the affected capital. There are multiple ways in which energy use and green- house gas emissions would be affected by this provision. A larger capital stock will increase output, and, other things equal, increase emissions. Newer capital may be more energy efficient and a bigger share of it will decrease average en- ergy intensity, or reduce energy use and emissions to the extent it replaces older inefficient capital. Finally, a lower cost of capital would favor the growth of capital-intensive sectors of production, which could increase or reduce energy use and emissions depending on whether capital and energy are complements or substitutes in production. The effects of changing broad-based provisions are analyzed in Chapter 6. REGULATORY INTERACTIONS Although other laws and regulations are not specifically included in the committee’s charge from Congress and the Treasury, they are a critical element because they interact with tax laws to influence GHG emissions. In some cases, regulations can reduce the overall impact of tax provisions on GHGs, while in other cases they may increase the impacts. For example, the federal government has instituted a Renewable Fuel Standard (RFS) mandate that requires transpor- tation motor fuel sold in the United States to contain a minimum volume of re- newable fuel. Separately, until 2012 sections 40 and 40A of the IRC offered tax credits to producers of renewable fuels as an incentive to increase production and use of these fuels. Since the RFS mandate sets a minimum annual produc- tion quantity of fuel, producers would produce at least that much fuel each year even if the tax incentive did not exist. Thus, the tax incentive must operate with- in the constraints set by the minimum production mandate. Another example of regulatory interaction is the Corporate Average Fuel Economy (CAFE) standards. These require new automobiles to attain a given average miles per gallon for each manufacturer. As these standards become in- creasingly tight, the impact of gasoline taxes will generally be reduced because fuel consumption starts from a lower base. However, as fuel economy standards are tightened, studies need to incorporate the indirect impact or rebound effect, which in this context is a phenomenon whereby the increased fuel efficiency induces car owners to drive more and therefore use more fuel. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 33 In addition to federal regulations such as the RFS mandate and CAFE standards, the committee needed to consider the effects of direct subsidies as well as state and local regulations, subsidies, and tax policies. To the extent practicable, all existing laws and policies of different governments were incor- porated into the baseline assumptions used in the computational models in order to include their effects when estimating the impacts of federal tax provisions. AVAILABLE LITERATURE ON EFFECTS OF TAX PROVISIONS ON GREENHOUSE GAS EMISSIONS An important first step in analyzing the potential impacts of the tax code on emissions and climate change was a review of the existing literature. The committee and several commissioned consultants undertook a systematic review of studies to analyze the impact of taxes and subsidies on GHG emissions. Alt- hough there is a vast literature on tax expenditures, there is virtually no empiri- cal research on the impacts of the U.S. tax code on GHG emissions. We found some studies that perform one or more of the steps in the analysis outlined above, but only one that provides empirical estimates of the impacts for most of the energy-sector tax expenditures considered in this report. This literature search confirmed that there are many studies of the demand and supply for gasoline; of the effects of subsidies on the diffusion of renewa- bles; of the effects on GHG emissions of tax subsidies for biofuels; of the impact of accelerated depreciation on economic growth; and of the impact of housing and health care deductions on tax revenues and the composition of economic activity. There is also a vast array of econometric studies of demand and supply elasticities for fuels and electricity. The literature is large even if we focus on some of the major survey articles (e.g, Bohi, 1981, 1984; Dahl C., 1993, 2002, 2012; Dahl and Duggan, 1988; Ko, 2001; Espey, 2004; Taylor, 1975, 1977; Wade, 2003). Little to no literature exists, at any level of analysis, for other tax provi- sions of interest. We found no published empirical studies of the impact on en- ergy use of the nuclear decommissioning tax preference. The committee found a single unpublished working paper (Hitaj, 2012) that provides an econometric analysis of how production tax credits for renewable electricity affect wind ca- pacity in the United States. However, to estimate the impact of the credits on total GHG emissions, the estimates in this paper would need to be supplemented by estimates of how changes in wind generation affect electricity markets and other energy markets as a whole.7 7 Palmer et al. (2010) in a background paper for the Resources for the Future-National Energy Policy Institute (RFF-NEPI) study use the National Energy Modeling System (NEMS) model to analyze the CO2 implications of the production tax credits (PTC) and the investment tax credits (ITC). The RFF-NEPI study analyzes the impacts of regula- 

34 Effects of U.S. Tax Policy on Greenhouse Gas Emissions Similar problems exist in the case of credits for energy-efficient improve- ments to homes. There is a considerable literature on the impact of price incen- tives on the purchase of energy-efficient appliances and home improvements, but none that examines the impact on the investments targeted by these credits.8 It is also necessary to estimate the impact of energy-efficient improvements on energy demand, the consequent effects on energy markets, and the correspond- ing changes in GHG emissions. Again, we found only the one previously men- tioned study that does this. The only area where there is a substantial literature is on the impact of gasoline taxes on fuel consumption and GHG emissions. This literature includes studies using one of the computational models employed in this study (Krupnick, 2010)9 as well as econometric studies of vehicle ownership and gaso- line demand (see Bento, 2009, and Gillingham, 2011). Many of these studies do not, however, incorporate important features of the U.S. tax and regulatory sys- tems, such as biofuels taxes and subsidies, CAFE standards, and regulatory mandates for ethanol and other biofuels. Nor do they include the general equilib- rium effects. Instead, these studies focus on estimates of supply and demand elasticities and how a change in the tax-inclusive price of fuels and electricity will affect demand and supply. While any estimate of the impact of changes in a tax incentive must begin with these, that is just a first step toward estimating GHG-emission impacts. Researchers at the University of Nevada, Las Vegas’s Center for Business and Economic Research (CBER) authored the one study the committee reviewed that contained a comprehensive analysis of the effect of government energy sub- sidies. This study analyzed the effects of many tax expenditures on CO2 emis- sions, but did not include other GHGs (Allaire and Brown, 2011). CBER mod- eled the effects of prices and taxes on energy markets using a simplified supply- and-demand framework. We found this study to be a useful first-order analysis, and compared its results to those from the work we commissioned using models with more detailed representations of energy markets, technologies, and regula- tions to ensure that we captured a full picture of the impacts of tax changes. ECONOMIC EFFICIENCY AND ECONOMY-WIDE MEASURES TO REDUCE GHG EMISSIONS Another set of studies puts the economic and GHG impacts of tax policy in the broader context of climate-change objectives. There is a substantial litera- ture that investigates approaches to achieving national and global climate- tions to promote energy efficiency and renewable energy use on CO2 emissions, but does not examine the effects of tax expenditures. 8 The credits are authorized by sections 25C and 25D of the tax code. 9 The U.S. Energy Information Administration’s (EIA) National Energy Modeling System, described in full detail in Chapter 3. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 35 change objectives (such as concentrations or temperature limits) in the most efficient way.10 The strategy in these studies is to determine the least-cost ap- proaches to achieving a given objective. Generally, these studies use either mar- ketable GHG emissions permits or taxes on GHG emissions as the actual mech- anisms to achieve the targets cost-effectively. Both analytical studies and empirical studies find that the most efficient approach to reducing emissions is through uniform or economy-wide coverage of taxes or regulations.11 The efficient trajectories are ones in which the margin- al costs of emissions reductions are equalized in every sector of the economy (and indeed in different countries for global efficiency). This condition would imply, under certain standard economic assumptions, that the price of emissions (in the case of tradable allowances) or tax per unit of CO2-equivalent emissions (in the case of GHG emissions taxes) is uniform in every sector. A uniform car- bon price provides appropriate incentives for consumers, producers, entrepre- neurs, and innovators to adjust their activities so as to reduce emissions and en- courage low-emissions technologies in the most efficient manner.12 One of the important features of uniform carbon pricing is that the policy directly targets GHG emissions rather than indirectly targeting capital goods, processes, or products that are only indirectly linked to emissions. Studies have also found that the cost per unit of emissions reductions is higher—often much higher—when sector-specific tax expenditures, subsidies, or regulations are used than when economy-wide measures are employed (EIA, 2011; NRC, 2010; Clarke, et al, 2009; Krupnick, 2010). Put differently, a key finding of economic studies of climate-change poli- cy is that the most reliable and efficient way to achieve given climate-change objectives is to use direct tax or regulatory policies that create a market price for CO2 and other greenhouse gas emissions. The economic advantage of targeted policies was emphasized in the Na- tional Research Council report America’s Climate Choices, in its overall sum- mary: Emission reductions can be achieved in part through expanding current lo- cal, state, and regional-level efforts, but analyses suggest that the best way to amplify and accelerate such efforts, and to minimize overall costs (for any given national emissions reduction target), is with a comprehensive, 10 See William J. Baumol; Wallace E. Oates, The Theory of Environmental Policy, 2nd Edition, Cambridge University Press, 1988; Robert Stavins, “Transactions Costs and Trada- ble Permits,” Journal of Environmental Economics and Management, 29, 1995, 133-148. 11 See for example, Intergovernmental Panel on Climate Change, Third Assessment Re- port, Mitigation, Chapter 6, and specifically p. 413, for a discussion of different instruments. 12 A useful survey of the literature with an analysis of different approaches is in Inter- national Energy Agency, Energy efficiency policy and carbon pricing, Information Paper, Energy Efficiency Series, International Energy Agency, Paris, 2011. 

36 Effects of U.S. Tax Policy on Greenhouse Gas Emissions nationally uniform, increasing price on CO2 emissions, with a price trajec- tory sufficient to drive major investments in energy efficiency and low- carbon technologies. (p. 3) This was further elaborated:13 Most economists and policy analysts have concluded, however, that put- ting a price on CO2 emissions (that is, implementing a “carbon price”) that rises over time is the least costly path to significantly reduce emissions and the most efficient means to provide continuous incentives for innova- tion and for the long-term investments necessary to develop and deploy new low-carbon technologies and infrastructure. A carbon price designed to minimize costs could be imposed either as a comprehensive carbon tax with no loopholes or as a comprehensive cap-and-trade system that covers all major emissions sources. (p. 58) The use of uniform carbon pricing has been the organizing principle of the European Union’s Emissions Trading Scheme (ETS), which has operated effec- tively for almost a decade. A recent study by Resources for the Future (RFF) was particularly useful for the committee’s analysis, because it provided quantitative estimates of the relative efficiency of different policy instruments (Krupnick, 2010). Moreover, the study used one of the models employed by the committee (NEMS model) and the same consultant to undertake the modeling (OnLocation, Inc.). The RFF study examined the resource costs of reducing CO2 emissions over the 2012-2030 period using different regulatory approaches. (In nontech- nical terms, resource costs refer to the losses in real income of the country.) It estimated that the average resource cost is $12 per ton of CO2 reduced (2007 USD) for a carbon tax that is set at $18 per ton of CO2 in 2012 and rises to $67 per ton of CO2 in 2020. (All figures in this paragraph are in 2007 USD.) Such a tax is estimated to reduce CO2 emissions 40 percent from 2005 levels by 2030. The economic cost of a cap-and-trade that auctions the allowances is the same (ignoring complications about uncertainty of policy). In contrast, the cost per ton of CO2 reduced of all the other policies examined was higher. For example, the resource cost per ton of CO2 reduced by implementing the building codes for new residential construction described in the Waxman-Markey bill of 2009 was estimated to be $51 per ton. The resource cost of reducing emissions by the PTC and ITC was estimated at $34 per ton of CO2. Therefore, each of the subsidies 13 The report particularly cited C. Fischer and R. G. Newell, “Environmental and tech- nology policies for climate mitigation” (Journal of Environmental Economics and Man- agement 55(2):142-162, 2008); and T. H. Tietenberg, Emissions Trading: Principles and Practice (Washington, D.C.: Resources for the Future, 2006). 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 37 and sector-specific policies were more costly than a uniform national policy of raising carbon prices.14 The same finding has been emphasized in several reports and literature re- views. These have consistently found that the approach of uniform pricing of GHGs is the most reliable and efficient way to achieve different climate-change objectives (EIA, 2011; NRC, 2010; Clarke, et al, 2009). MODELING APPROACHES AND COMPARATIVE STRENGTHS OF MODELS CONSIDERED FOR USE IN SUPPORTING ANALYSES The committee’s review of the literature determined that there is a very thin body of published work on which to base an analysis of the impacts of the tax code on greenhouse gas emissions. In reality, there are virtually no studies that analyze the impact in an empirical analysis that reflects the full complexity of the U.S. energy system and regulatory environment. Given the lack of exist- ing work, the committee decided to direct the National Academies to undertake studies with existing best-practice empirical models. The next section discusses the committee’s approach in detail. To undertake the new modeling efforts for this study, the committee de- cided the most sensible approach was to work with existing modeling groups rather than attempting to create one or more new models. Many models already exist that link energy markets with environmental components. We selected models with long track records of use for academic research, for public policy analysis, and by private enterprises, each of which has been carefully and re- peatedly scrutinized by the research community through a lengthy history of peer-reviewed publications. Additionally, we wanted to avoid any conflicts of interest and therefore did not choose models that committee members were ac- tively managing. We also had to work within the constraints of the budget and time line of the National Academies’ contract with the Department of the Treas- ury and thus needed to identify models and contractors that could deliver results on time and within budget. As mentioned in this chapter’s introduction, no one model or type of mod- el can adequately analyze all of the provisions we wished to investigate. Our review of existing models suggested we would need to utilize one each of three types of models suitable to our task: 14 Note that these estimates refer to the analysis “with no market failure.” The study al- so examined the cost-effectiveness of policies “with market failures.” These included particularly the possibility that households might apply an inappropriately high discount rate to energy-efficiency investments. The study found that one specific technology (geo- thermal heat pumps) had favorable costs relative to other technologies, but the committee did not investigate this proposal. 

38 Effects of U.S. Tax Policy on Greenhouse Gas Emissions 1. Models focused on energy markets and representing them in consid- erable detail; 2. Models focused on agricultural markets that included detailed repre- sentation of how biofuels policies would affect those markets; and 3. Economy-wide models that often include specific but less detailed representation of energy markets, agricultural markets, or both, along with details of a few tax policies. Appendix A offers further detail on the options within each class of model and their principal features. In the following section, we explain our choice of models used and their chief characteristics. Energy-sector Models We considered three energy-focused models.15 Ultimately we chose to use two of these. For the bulk of our analysis, we employed the U.S. Energy Infor- mation Administration’s (EIA) National Energy Modeling System (NEMS), run by a professional consulting firm.16 We also used an energy-sector model devel- oped at the Center for Business and Economic Research at the University of Nevada, Las Vegas (CBER). Details of the NEMS Model The committee looked to NEMS for the bulk of our analysis here for sev- eral reasons. The U.S. Energy Information Administration of the U.S. Depart- ment of Energy designed, implements, and continues to maintain NEMS. EIA publicly publishes details of the model structure and assumptions and updates the model annually to incorporate energy market data from the prior year.17 EIA also uses NEMS to produce its Annual Energy Outlook, an analysis and projec- tion of energy market trends, typically over a 25-year period. Because of these efforts by EIA, NEMS’s capabilities and shortcomings are well understood within the energy-modeling and energy-economics communities. NEMS’s representation of energy markets focuses on four interactions: (1) energy supply-energy conversion-energy demand, (2) domestic energy system- economy, (3) domestic energy market-world energy market, and (4) economic 15 These were the MARKAL model, which is more modeling framework that can be developed for specific applications depending on interests and data availability; the NEMS model, developed and used by the U.S. Energy Information Administration in its annual energy outlook and available for other uses; and the CBER model, developed at the University of Nevada, Las Vegas. 16 OnLocation, Inc., of Vienna, Virginia. 17 EIA provides documentation on the NEMS model on its Web site: http://www.eia. gov/analysis/model-documentation.cfm. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 39 decision making over time. There are many important assumptions that drive the NEMS model, the two most important being U.S. economic growth and oil pric- es. Other assumptions include macroeconomic and financial factors, world ener- gy markets, resource availability and costs, behavioral and technological choice criteria, cost and performance characteristics of energy technologies, and de- mographics. NEMS consists of four supply modules (one for each major fuel), two conversion modules, four end-use demand modules, one module to simulate energy-economy interactions, one module to simulate international energy mar- kets, and one module that provides the mechanism to achieve a general equilib- rium among all the other modules. These details make NEMS an ideal, widely available model for analyzing the energy-focused provisions we sought to study. Importantly, though, CO2 is the only GHG that NEMS includes. Given the mod- el’s focus on energy markets, this is understandable, but means it could not rep- resent potentially important effects of biofuels policies on methane or land use. Details of the CBER Model We also asked UNLV’s Center for Business and Economic Research to extend a prior analysis it performed using its simplified supply-and-demand model of the energy sector (Allaire and Brown, 2011). In CBER, demand and supply are represented as functions of energy prices (including the prices of sub- stitutes and complements).18 The key parameters in the CBER model are the price elasticities of supply and demand. The elasticities for the CBER model are derived by expert judgment of the authors from reviews of the economic litera- ture. An important advantage of the CBER model is that it was developed with an objective of investigating the effects of subsidies, mostly tax expenditures, on CO2 emissions in the energy sector. The developers invested considerable effort in representing most of the energy-specific tax code provisions. Indeed, their earlier paper is the only comprehensive study of the impact of the tax code on CO2 emissions. An additional strength of the CBER study is that consumer behavioral re- sponses and broad technological trends are represented to the extent these can be estimated from historical data. However, the model has important shortcomings compared to all the other models used in this study. First, it is a “static” model, meaning that everything takes place as an equilibrium in a single stylized time period. This means that it cannot capture the dynamics of capital turnover. It 18 The researchers involved solved the equations in their model using a standard math- ematical solver package. The equations can be found in the following report: M. Allaire and S. Brown (August 2012). U.S. Energy Subsidies: Effects on Energy Markets and Carbon Dioxide Emissions. Retrieved 2013, from http://www.pewtrusts.org/uploaded Files/wwwpewtrustsorg/Reports/Fiscal_and_Budget_Policy/EnergySubsidiesFINAL.pdf. 

40 Effects of U.S. Tax Policy on Greenhouse Gas Emissions also provides no detailed representation of the regulatory environment necessary for examining some provisions, such as those involving the Renewable Fuels Standard for biofuels, and so does not capture many important effects of those regulations on energy markets. Lastly, CBER’s model includes only one GHG, CO2. While CO2 is the largest GHG by total emissions, this means the model can not capture potentially important changes in methane or land use resulting from biofuels policies. Details of the FAPRI Model The committee’s focus on the agricultural sector stemmed largely from tax provisions related to biofuels. There are several models that have been used to examine the economics of biofuels and biofuel policy.19 Analysis of these poli- cies is extremely complex. Challenging aspects of modeling biofuel policy include (1) the complex interactions with agriculture and agricultural policy, including competing de- mands for crops and by-products supplies of animal feeds; (2) the complex poli- cy requirements of the Renewable Fuel Standard (RFS2, as described below); (3) investment and production tax credits that differentially treat different biofu- el production pathways and feedstocks; (4) international linkages in agriculture and energy markets; (5) land-use change and competition for land between agri- culture and other uses of land; and (6) the implications of land-use change for GHG emissions. Given consideration of budget and ability of different groups to produce timely analysis, we decided that Missouri University’s implementation of the Food and Agriculture Policy Research Institute’s model, FAPRI-MU, best fit the requirements.20 This model has the combination of detailed agriculture and crop 19 These include macroeconomic models such as Emissions Predictions and Policy Analysis (EPPA) and Global Trade Analysis Project (GTAP) models (e.g., Gurgel, Reilly, and Paltsev, 2007; Gurgel et al., 2011; Tyner et al., 2010; Decreux and Valin, 2007). Agricultural optimization models including the Forest and Agricultural Sector Optimization Model (FASOM) (Adams et al., 1996, as in Beach and McCarl, 2010; Beach, et al., 2013); simulation models such as MiniCAM (Wise et al., 2009); and econ- ometric-based simulation models such as the Food and Agricultural Policy Research Institute (FAPRI) model (Babcock and Carriquiry, 2010). 20 FAPRI provides documentation on its model on the FAPRI-MU website: http://www. fapri.missouri.edu/outreach/publications/umc.asp?current_page=outreach. See FAPRI-MU Report #12-11, Model Documentation for Biomass, Cellulosic Biofuels, Renewable and Conventional Electricity, Natural Gas and Coal Markets; FAPRI-MU Report #09-11, FAPRI-MU Stochastic U.S. Crop Model Documentation; FAPRI-MU Report #05-11, New Challenges in Agricultural Modeling: Relating Energy and Farm Commodity Prices; FAPRI-MU Report #09-10, FAPRI-MU U.S. Biofuels, Corn Processing, Distillers Grains, Fats, Switchgrass, and Corn Stover Model Documentation; FAPRI-MU Report #07-08, 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 41 markets specification, linkage to international markets, inclusion of regulatory constraints relevant for analysis of biofuel provisions, and tax parameters that allow an analysis of changes in the various parameters. It also has a full repre- sentation of the intricacies of renewable fuel credits, with multiple fuel- production pathways representing both conventional and second-generation pro- cesses tied with links to global markets for crude petroleum and refined fuels. However, FAPRI-MU does not explicitly consider land use or the carbon implications of land-use change, a potentially large and important emissions pathway. Land-use changes are highly uncertain, with wide-ranging results found in the literature (Plevin et al., 2010; Searchinger et al., 2008; Melillo et al., 2009; Keeney and Hertel, 2009; Hertel, 2011; Tyner et al., 2010; and Mosnier et al., 2012). Instead, greenhouse gas implications are assessed by ap- plying a fixed GHG coefficient per unit of fuel for different biofuel production pathways. Some coefficients include a factor to estimate land-use change impli- cations on emissions. Default estimates are those of the U.S. Environmental Protection Agency (2010) that include CO2, N2O, and CH4 implications of land- use change. Details of the IGEM Model In principle, economy-wide general equilibrium models are potentially ca- pable of analyzing many of the tax provisions, but they are limited in that they lack the granularity needed for some of the detailed provisions. For example, an economy-wide model that represents the electricity sector as single production function cannot accurately represent the effect of a provision directed at individ- ual technologies such as wind, solar, or nuclear. And, similarly a model that simplifies the agricultural sector as producing a single product is less able than a detailed model of the agricultural sector, with a full range of crops and livestock, to trace how a biofuels policy may affect corn production and land-use change. So the economy-wide models are most useful for considering the broad-based tax provisions and for capturing general equilibrium effects of tax policies, but are less useful for capturing sectoral details. The committee considered six economy-wide models that could analyze broad tax provisions.21 Given the aims of the study and the constraints, the Model of the U.S. Ethanol Market; and FPARI-UMC Report #12-04, Documentation of the FAPRI Modeling System. 21 These included the MIT Emissions Predictions and Policy Analysis (EPPA) model (Paltsev et al., 2009) and/or the MIT U.S. Regional Energy Policy (USREP) model (Rausch et al., 2010) very similar to EPPA but with greater detail on the United States. The Applied Dynamic Analysis of the Global Economy (ADAGE) model (Ross, et al., 2009) developed at RTI International and widely used by the Environmental Protection Agency for analysis of greenhouse gas policies (e.g., EPA, 2009). The Multi-Region National (MRN) model developed at Charles River Associates (Berstein, et al., 2007). The Global Trade Analysis Project (GTAP) model developed at Purdue University 

42 Effects of U.S. Tax Policy on Greenhouse Gas Emissions committee decided that the Intertemporal General Equilibrium Model (IGEM)22 was best positioned to address the questions involved. The committee chose IGEM because it is a well-established economy-wide model containing detailed specification of the U.S. tax code, because several agencies of the U.S. govern- ment have used it for many years for energy and environmental modeling, and because the firm that runs it was able to deliver modeling results in a timely fashion. IGEM is a multisector general equilibrium model that represents the econ- omy following modern neoclassical economic theory. It captures relationships between industry outputs and final consumption goods as represented in input- output tables for the U.S. economy and includes an expanded social accounting system with estimates of factor returns from each production sector and the dis- position of goods to final-demand sectors (households, government, investment, and exports). Thus, an important feature of IGEM is that it captures the effects of tax provisions in one industry on the output and emissions in other related sectors. For example, a provision in the steel industry will have effects on the automobile industry, and that effect can in principle be captured in a general equilibrium model. IGEM also can be used to illustrate how a tax change that alters incentives to work, save, and invest would affect the overall level of eco- nomic activity. The IGEM model differs from most other models, as it includes a time se- ries of data for the U.S. economy, and most parameters are econometrically es- timated from these data. Models differ in how they deal with the dynamics of economic adjust- ment, that is, how investment decisions react to changing prices and interest rates and to expectations of future conditions. IGEM is a deterministic forward- looking model, as are all other models used in the present study. In other words, it assumes that individual consumers and firms look forward and anticipate fu- ture conditions with perfect foresight, making decisions today based on those expectations. Agents are said to have perfect foresight because their expecta- tions are realized exactly. This means that the model assumes that firms and consumers know the trajectory of oil prices, GDP growth, and other important factors. Forward-looking models clearly overestimate the capability of agents to look forward, but the implications of this assumption for estimating the impacts of tax policy on GHG emissions is unclear. IGEM has a representation of the capital stock in which capital is fully malleable or adaptable to changing circumstances. This is equivalent to assum- ing that firms can quickly and economically retrofit their capital—such as build- ings, machinery, or equipment—in response to changing market and regulatory conditions. (Hertel, et al., 2010). And the Intertemporal General Equilibrium Model (IGEM) of the United States (Goettle et al., 2007), developed at Dale Jorgenson Associates. 22 Dale Jorgenson Associates provides documentation on the IGEM model on its web- site: www.igem.insightworks.com/. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 43 IGEM has a very detailed disaggregation of production sectors, but within the constraints of the Standard Industrial Classification system. In this classifica- tion, for example, electricity generation is a sector, but there is no further dis- tinction between renewable power sources (solar, wind, nuclear, hydro) and nonrenewable sources of generation (coal, gas, or oil). Fuels, capital, labor, and intermediate goods are inputs into the electricity sector, and some amount of electricity is produced. Implicitly, substitution of capital for fuels could be inter- preted as an increase in one of the non-fossil-fuel technologies. Rates of tech- nical change are econometrically estimated based on the historical data, and the- se are represented as time trends on input requirements rather than explicit technologies. IGEM has a more extensive representation of the U.S. tax system than most other economy-wide models. This feature was important for consideration of the broad provisions. IGEM does not contain as much detail of energy or ag- riculture markets as the models we chose specifically to model biofuels and en- ergy-focused provisions, and hence it was not able to simulate many of those narrow provisions. On the other hand, the model does include a comprehensive suite of GHGs including CO2, CH4, N2O, and so-called high global warming potential gases, or HGWP. We discuss further details of the IGEM simulations used in this study in Chapter 6. GOVERNMENTAL POLICIES TO PROMOTE INNOVATION AND LOW-CARBON TECHNOLOGIES The economic analysis of subsidies and tax expenditures for standard goods and services like shoes and pizzas finds that they lead to economic distor- tions and reduce national output and economic welfare. When there are no mar- ket failures, subsidies may distort prices and outputs away from their market- determined levels. This is a standard finding about competitive markets (Debreu, 1959). However, investments in innovation and new technologies suffer from a market failure because of the inability of inventors to appropriate the full value of their activities (Arrow, 1962). Policies such as rights to intellectual property and government support for research and development can improve innovative performance and lead to increases in national output. Particularly in cases where markets do not reflect true social costs, as is the case with emissions of GHGs, research support can play a vital role in promoting low-carbon technologies. These analyses lead to the important conclusion that subsidies and other policies to support new technologies are a critical component of a strategy to slow climate change and do not have the same inefficiencies that are found in subsidies of standard goods and services. These points were emphasized in the recent report on U.S. climate policies by the National Academies. This report concluded as follows with respect to governmental support of energy technolo- gies: 

44 Effects of U.S. Tax Policy on Greenhouse Gas Emissions Major technological changes in the U.S. energy system and other sectors will be needed to reduce GHG emissions significantly, and this will re- quire an infusion of financial and human resources to support each phase of the process…. Resources that are critical for technology innovation in- clude money for R&D and people with the requisite training, skills, and creativity to innovate. (NRC, 2010) INDUCED TECHNOLOGICAL INNOVATION, THE TAX SYSTEM, AND ECONOMIC MODELS A final issue arising in evaluating the impact of taxes on GHG emissions is their impacts on technological change and innovation. This is particularly important for the long run. For example, a tax or subsidy that favors renewable technology X would lower the relative price of X and raise the production and use of X. In anticipation of a larger market, firms would devote more resources to research, development, and commercialization of X. Additionally, there might be learning by doing, through which higher levels of cumulative investment and production lower the production costs of the technology. These processes are often called “induced innovation.” Evaluating the rate and direction of technological change has proven a ma- jor challenge. It involves predicting the outcomes of research and development that have not yet been completed, or perhaps even funded. Methods necessarily extrapolate from historical experience and assume that past results are an indica- tor of future performance. Studies on induced innovation have developed two leading theoretical frameworks. The “research framework” of induced innova- tion arose in an attempt to understand why technological change appears to have been largely labor saving (Hicks, 1932; Nelson, 1959; Arrow, 1962). More re- cently, this approach was further developed as the new growth theory (Romer, 1990; Aghion and Howitt, 1999). Under the research approach, higher levels of investment in knowledge will expand society’s production possibilities and in- crease the long-run growth rate of the economy. Over several years, Jorgenson and co-authors have adopted the research approach and currently incorporate it as part of their econometrically estimated production structure in IGEM (see, for example, Jorgenson and Wilcoxen, 1991). The alternative approach to modeling-induced innovation is the “learning model.” This approach has become particularly widely used by energy- economic models in recent years as models increase the granularity of the tech- nological description down to individual technologies. The learning approach is included in the NEMS model. For example, most electricity technologies are assumed to have a cost decrease of 1 percent for every doubling of cumulative capacity in the late stages of development, while the learning rate ranges from 5 to 20 percent in the early stages (EIA, 2010). A more detailed discussion of learning in the NEMS model is provided in Chapter 3. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 45 After reviewing the existing literature, the committee concludes that the impact of induced innovation is clear: Tax provisions that lower the cost or raise the production of a specific technology will generally tend to lead to induced innovation and improve that specific technology. However, while the sign of the impacts is clear, the size of the impact and the impact on GHG emissions is highly speculative. Moreover, the mechanism by which these would work will differ greatly depending upon which of the different innovation models is as- sumed to be driving the impact from technological change. While clearly an important topic, estimating the impacts of the tax code on innovation is a problem beyond the capability of current empirical models. The influence of learning is highly controversial, and there are no reliable structural models of learning in the energy sector. There is a substantial literature on the impact of changing energy prices and other prices on patents (see, for example, Popp, 2002, 2004; Popp and Newell, 2012; Popp, Hascic, and Medhi, 2011). There is also work on the impact of energy prices on productivity (see, for ex- ample, Newell et al., 1999). However, existing empirical studies of induced in- novation in the energy sector do not extend to the effects of changes in the tax code. Additionally, they do not estimate the general equilibrium impact of in- duced innovation on outputs in different sectors or to the impact on GHG emis- sions. A few studies examine the impact of policies such as carbon taxes or cap and trade on economy-wide technological change and innovation (see Popp, 2004; Bosetti et al., 2006). However, these studies are generally at a highly ag- gregated level and do not focus on specific provisions of the tax code. Given these difficulties, the committee did not undertake a separate study of the effects of the tax code on emissions through induced innovation. Nor did the committee attempt to separate out the influence of endogenous technological change or learning from the other forces at work, the most important being sub- stitution. Part of the reason for the decision not to pursue this area is that esti- mating induced innovation empirically has proved very challenging. Additional- ly, this would have required another set of model runs, along with major model modifications, and each would have required time and budgets well beyond what we had available. Finally, there are no widely accepted models available that adequately represent the daunting complexities that arise with endogenous changes in technology. Because of the importance of technological change for emissions reductions, particularly over the long run, we point to work on in- duced innovation as an area of particular interest for further research and im- provement of energy-economic models. TAX BASELINE FOR MODELING AND ANALYTICAL APPROACH Baseline Assumptions In addition to choosing which models to use and which provisions to ana- lyze, the committee needed to select a baseline to use as a comparison for evalu- ating the effects of changing tax policies on greenhouse gas emissions. A base- 

46 Effects of U.S. Tax Policy on Greenhouse Gas Emissions line is a set of economic, tax, and regulatory assumptions that is used as a start- ing point for analyses. Given a baseline, a model can change certain assumptions and determine the impact on important outcomes, such as output or emissions. In the present study, the major changes were those involving provisions of the tax code. The committee determined that the most suitable baseline for its task was the tax code and regulatory system of 2011, with all provisions extended indefi- nitely. Additionally, the committee adopted for its economic assumptions, such as GDP growth rate and global petroleum prices, the assumptions used by the EIA in its Annual Energy Outlook 2012. The major advantage of this approach is that it represents an actual tax and regulatory system, and that could be used as a starting point for comparative and counterfactual calculations. Under this baseline, each model would first estimate energy use and emissions under the baseline assumptions. For comparative purposes, provisions would then be re- moved from the tax code and all model outputs, including emissions, would be re-estimated with the difference attributed to the provision under consideration. Note that not all models were able to incorporate exactly the same baseline as- sumptions, and the differences are noted below in the discussion of the individu- al models and provisions. Treatment of Revenue Changes Another important issue that required attention was to determine how to treat changes in government revenues that would follow changes in tax provi- sions. Removing provisions from the tax code will change the revenues coming to the government. Eliminating a tax expenditure would raise revenue, while removing an excise tax would reduce revenue. Modeling the full effects of a change in the tax provisions requires assumptions about what the government does with increased revenues and how it pays for reduced revenues. Among the several alternative approaches discussed below, the committee adopted three in the calculations. Assume Revenue Changes Have No Effects The simplest approach is to assume that the revenue changes have no ef- fect on the economy. This approach avoids making any assumption about how the government would adjust its budget in response to revenue changes, and examines the economic and emissions effects of a policy change absent any fis- cal changes. This approach is implicitly embodied in the partial equilibrium models. These models simply take the revenues and send them into the rest of the econ- omy, but no impacts on output, income, or prices in the rest of the economy are included. Excluding fiscal impacts can lead to misleading results, particularly for policy comparisons. For example, a large increase in a very efficient tax can 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 47 look more costly than a tiny increase in a very inefficient tax, but it might be a more efficient policy because it raises more revenue. We note the disadvantage of not treating revenues correctly, but for some provisions there is no alternative to partial equilibrium models. Allow the Budget Deficit to Change in General Equilibrium Calculations The simplest approach is unsatisfactory given that changes in government revenues must affect either borrowing or spending. Economy-wide models that include the dynamics of taxes, expenditures, and debt can in principle consider the treatment of changing revenues, since they consider all markets in the econ- omy, estimating emissions in a general equilibrium, where demand and supply are in balance in every market within the economy. Ignoring revenue changes would mean that the government budget no longer adds up in simple accounting terms. Therefore, in a dynamic general equilibrium model, the model must make some assumption about how the rest of the government budget adjusts to ac- commodate the revenue change. Offset Revenue Changes Using Lump-sum Payments The committee directed the IGEM team to recycle revenues in two differ- ent ways: through lump-sum changes in the tax system and through proportional changes in tax rates. A lump-sum change is one that changes taxes or government transfer payments by a fixed amount that does not depend on household behavior. For example, it could take place through a fixed-dollar refundable tax credit availa- ble to all households. This approach is computationally simple. However, in reality, the gov- ernment generally does not change taxes in a lump-sum manner. Instead, it col- lects revenue via distortionary taxes such as the income tax or taxes on compa- nies. Economists refer to “distortionary” taxes as those that alter relative prices of goods, capital, or labor, thereby causing people to change their economic de- cisions from what would otherwise have been optimal choices in order to reduce their tax liability. Income taxes distort the choice between labor and leisure (or market work and home production) by reducing the relative return from an extra hour of work. They distort the choice between present and future consumption by altering the rates of return on different assets. Selective excise taxes distort consumer choices by raising the relative prices of taxed goods. When household and firms change their behavior to reduce their tax liabil- ity (for example, by working less, saving less, or buying fewer taxed goods), the result is a reduction in economic well-being. Therefore, raising government rev- enue from taxes that distort behavior imposes costs on the private sector that exceed the amount of revenue collected; this is labeled an “excess burden” of taxation. Assuming that all revenue losses are offset by lump-sum taxes implicit- 

48 Effects of U.S. Tax Policy on Greenhouse Gas Emissions ly ignores the losses (or excess burden) from distortionary taxes, while assuming that all revenue gains are returned to taxpayers with lump-sum subsidies ignores the extra benefit that could arise from reducing distortionary taxes. Offset Revenue Changes by Adjusting Tax Rates An alternative approach is to assume that revenue changes induced by changing a tax provision are offset by raising or lowering marginal tax rates. The committee asked the IGEM contractor to implement this approach by rais- ing or lowering all corporate and individual tax rates by the same proportional amount. For example, if the top marginal tax rate of corporations is 35 percent and individual rates range from 10 to 35 percent, a 10 percent proportional cut would reduce the corporate rate to 31.5 percent and individual rates to a range of 9 to 31.5 percent. This approach is used because it is transparent and easily modeled. Comparing Approaches: Offsetting Revenue Changes via Lump-Sum Payments v. Adjusting Tax Rates While the two approaches to offsetting revenue changes (lump-sum changes and proportional changes in tax rates) were not the only possibilities, they are relatively easy to implement and interpret. The main reason for choos- ing the lump-sum approach is it allows analysts to focus on the effects of the energy tax change by itself as nearly as possible. The main advantage of the tax- rate-change assumption is that it allows us to capture the effects on distortions of scaling up or down the main taxes used to fund general government services in our current tax structure in a manner similar to how revenues are currently raised. Table 2-4 details what provisions were able to be studied in this report, what chapter discusses the analysis for specific provisions, and what models, if any, were used for analysis. MEASURES TO APPRAISE MODEL ESTIMATES The committee was particularly attentive to determining the reliability of the estimates from the modeling teams. Each of the three major modeling teams that were engaged by the committee has a long track record of work in energy modeling for the U.S. government and for scholarly journals. Notwithstanding past work, the committee took steps at successive stages of the study to deter- mine the appropriateness of the modeling approach, the assumptions, and the model outputs as well as to understand the intuition behind the model results. These steps reflected the need to follow a rigorous standard of quality control for a topic of such importance for public policy and for understanding the im- pacts on government revenues and climate change. 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 49 TABLE 2-4 Provisions Modeled for This Study and Where Discussed in This Report Chapter Where Model(s) Employed Provision Discussed in Analysis Credit for electricity production from Chapter 3 NEMS-NAS renewable sources (including the cash grant in lieu of tax credit) Excess of percentage over cost depletion for Chapter 3 NEMS-NAS oil and gas wells Credit for energy efficiency improvements to Chapter 3 Models not Used existing homes for Analysis Special tax rate on nuclear decommissioning Chapter 3 Models not Used reserve funds for Analysis Excise tax on highway motor fuels (gasoline Chapter 4 NEMS-NAS; and diesel) CBER; FAPRI-MU; IGEM Excise tax on aviation fuel Chapter 4 CBER Alcohol fuel credit and excise tax exemption Chapter 5 NEMS-NAS; FAPRI-MU Biodiesel producer tax credit Chapter 5 NEMS-NAS; FAPRI-MU Exclusion of employer contributions for medical Chapter 6 IGEM insurance premiums and medical care Deductibility of mortgage interest on owner- Chapter 6 IGEM occupied homes Accelerated depreciation of machinery and Chapter 6 IGEM equipment (normal tax method) Independent Verification of Model Choices, Baseline Assumptions, and Tax Code Provisions Analyzed As a first step, before entering into contracts with the modeling groups se- lected, the committee sought an external analysis of our choices of models, modeling assumptions, and criteria for tax provisions to analyze. We asked two experts from the economics and tax policy communities to undertake this task. They were Dr. William Pizer, Associate Professor of Economics at Duke Uni- versity, recognized expert on economics and energy modeling, and former depu- ty assistant secretary for environment and energy in the U.S. Department of the Treasury; and Dr. Richard Newell, Associate Professor of Energy and Environ- mental Economics at Duke University and former director of the U.S. Energy Information Administration, the agency that is responsible for official U.S. gov- ernment energy statistics and analysis. 

50 Effects of U.S. Tax Policy on Greenhouse Gas Emissions In March 2012 Pizer and Newell each submitted independent written anal- yses of our methods and preliminary choices. These reports endorsed the com- mittee’s decisions on models and specifications. At the same time, they made several important suggestions for improvements and modifications. On April 10, 2012, they discussed their reports in a joint conference call with the committee. Both consultants agreed that the models that the committee had selected were the most appropriate among available options. Among their specific sug- gestions for modeling strategy were suggestions for the analysis of biofuels sub- sidies. Following their recommendations, the committee added the credit and deduction for clean-burning vehicles and the credit for advanced energy- manufacturing facilities. How to handle the revenue changes that accompany a change in tax policy was of particular concern to both reviewers. Our choice to request that IGEM analyses be performed both with the revenue being offset by changing tax rates and by lump-sum rebates arose in part from our discussions. These exchanges are part of the record of the committee’s deliberations, and their implications are explained in the following chapters. A second part of the committee’s quality assurance involved working closely with the modeling teams to ensure that the baseline assumptions were correctly and consistently incorporated in their models. Individual members of the committee with experience in the types of models being used worked closely with the modeling groups between committee meetings. The committee member biographies in Appendix B describe this expertise in each case. Finally, as the results were reported to us, the full committee carefully scrutinized the simulation results for anomalies and possible errors. In those instances, we asked the contractors to • Elaborate their procedures for estimating errors; • Explain why their model produced an unexpected result; • Provide a procedure to decompose emissions into sources (e.g., for IGEM); • Confirm the input parameters; • Rerun a simulation if questions remained; and • Conduct sensitivity analyses of key model parameters and input as- sumptions. For NEMS, we relied on the contractor, OnLocation, Inc., which has re- programmed the NEMS model and therefore represents an independent valida- tion. Because the CBER model was extremely simple, we tested some of the results using auxiliary calculations. The committee spent most of three of our five committee meetings and several conference calls between meetings analyz- ing the modeling results, deliberating about their reliability, validity, and inter- pretation, and articulating their limitations. The committee recognizes that the quality control procedures that it fol- lowed were just a subset of possible steps in error estimation and validation of complex models. We could not check the tens of thousands of lines of code, data 

Methods for Evaluating Tax Policy Effects on Greenhouse Gas Emissions 51 points, and calculations of the different models; nor could we retest and re- estimate the econometric studies. To do so would have extended the time and scope of the study indefinitely and in any case greatly exceeded our financial resources. Rather, in the following chapters, the committee describes the issues that arise in interpreting the model results as well as the confidence that should be placed in the results. SUMMARY The committee commissioned four modeling groups to conduct new stud- ies investigating the implications of tax code provisions on GHG emissions. The models were diverse in their construction, data, time periods, complexity, re- gional coverage, tax structure, and degree of coverage of the overall economy. The models employed to the maximum feasible extent a common baseline as- sumption for tax, regulatory, and economic scenarios. While none of the models could calculate a full range of estimates for all provisions, most major provi- sions were analyzed, and some were analyzed by multiple models. The models have different strengths and weaknesses. The committee con- cluded that the NEMS model was best suited for most of the narrow provisions directed toward the energy sector. The FAPRI model had distinct advantages for analysis of biofuels tax incentives among those we studied. The IGEM model was the only model able to consider most of the broad-based tax provisions ana- lyzed by the committee and to weigh economy-wide impacts. The CBER model was useful for providing a comprehensive analysis of the GHG impacts of pro- visions as well as for providing comparisons with other models.