Click for next page ( 166


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 165
Appendix A Modeling Approaches to the Effects of Tax Policy on GHG Emissions INTRODUCTION This technical appendix is intended to assist those interested in examining the modeling approaches selected by the committee from among the alternatives in greater depth and detail and in understanding better their differences, strengths, and limitations. It compares the structures of available models and provides sources of documentation of the National Energy Modeling System for the National Academy of Sciences, Food and Agricultural Policy Research Insti- tute at the University of Missouri (FAPRI-MU), Intertemporal General Equilib- rium Model (IGEM), and Center for Business and Economic Research (CBER) models, whose results are described in the report. Modeling approaches relevant to the questions before the committee range from those that compare the cost of alternative investment decisions, taking into account how taxes affect the investment return, to models of the entire economy that project how a tax or tax expenditure affects domestic and international mar- kets and the overall level of economic activity. In between are models that focus on some limited set of markets that are directly affected by a tax provision or that interact strongly with markets that are directly affected. Narrow tax provi- sions that affect very specific investment decisions such as those directed toward wind turbine or solar photovoltaic installations require a detailed investment- level evaluation. The effect of the provision on cost of the project may depend on eligibility criteria, profitability of the investor if the value is as a credit against taxes, expected prices of inputs, and expected inflation. Whether the tax provision is actually used and has an impact on emissions will depend on how it alters the cost of the project and investment decisions. Most sector or economy-wide models do not represent the structural de- tails of investment decisions at the detailed level, but instead simplify the incen- tive effect as either an effective tax rate on investment, as a change in the 165

OCR for page 165
166 Effects of U.S. Tax Policy on Greenhouse Gas Emissions levelized cost of the technology, or as a supply shift. Thus, a detailed model of investment decisions may be a needed first step in using a broader market model that must model an intricate tax provision as, for example, a simpler effective tax rate on a particular technology, sector, or factor input. The markets directly affected by the narrow provisions identified by the panel were energy markets, and for biofuel-related tax expenditures, agricultural markets. Energy and agri- cultural markets are themselves fairly complex, with a variety of existing regula- tory policies that affect them and that potentially interact with tax incentives. The broad provisions the committee identified, such as accelerated depreciation or those related to housing and health care, require an economy-wide approach, or at least a scope beyond just energy and agricultural markets. In short, no sin- gle model was likely to have detail on agricultural markets and energy markets, while also capturing economy-wide effects of broad policies. The committee’s review focused on three types of models. First was a set of economy-wide models, often with some detail on energy or agricultural sec- tors. A second set of models has been developed with a strong focus on agricul- tural markets and the effect of biofuel policies on them, with varying degrees of detail on how biofuels would also affect energy markets. A third set of models has focused on energy markets in considerable detail. In principle, the set of economy-wide models are potentially capable of analyzing many or most of the tax provisions, but they are limited in that they generally 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 easily represent the effect of a provision directed just at wind, solar, or nuclear. And similarly, a model that simplifies the agricultural sector as producing an aggregate crop or livestock product is less able to trace how a biofuels policy may affect corn production and land-use change. For each of the tax provisions to be examined by one or more models, the first step is to determine the effect of the provisions on parameters available in the model, based on a detailed model of investment decisions, if needed. ECONOMY-WIDE MODELS The committee identified six economy-wide models with the capability to examine at least some of the tax provisions. These included (1) the MIT Emis- sions Predictions and Policy Analysis (EPPA) model (Paltsev et al., 2005, 2009); and/or (2) 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; (3) 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 (EPA) for analysis of greenhouse gas (GHG) policies (e.g., EPA, 2009); (4) the Multi-Region National (MRN) model devel- oped at Charles River Associates (Bernstein et al., 2007); (5) the Global Trade Analysis Project (GTAP) model developed at Purdue University (Hertel et al.,

OCR for page 165
Appendix A 167 2010); and (6) the Intertemporal General Equilibrium Model (IGEM) of the United States (Goettle et al., 2007) developed at Dale Jorgenson Associates. These models have some similarities and differences. All are multisectoral general equilibrium models that represent the economy following modern neo- classical economic theory, meaning that consumers and producers are both as- sumed to utility- or profit-maximize, given constraints. All are constructed on some version of input-output tables for the United States (or other regions of the world if included) and an expanded social accounting matrix that includes esti- mates of factor returns from each production sector and the disposition of goods to final-demand sectors (households, government, investment, and exports). Thus, a general strength of these models is that any ripple effects on other sec- tors, final demands, and exports and imports of a tax provision that affects one sector will be included in estimates of their greenhouse gas emissions effects. If changes in the energy sector affect the level of steel production or what type of energy it uses, these models will, in principle, include this effect. While some of these models cover the whole world, and some just the United States, they in- corporate some estimate of effects of trade via export demand and import supply functions. They can also estimate how a tax may affect the overall level of eco- nomic activity via incentives for labor and investment. Beyond these general theoretical similarities there are various similarities among some of them in terms of specific databases, solution approaches, and rep- resentation of dynamics, but this is also where some differences emerge. EPPA, ADAGE, and GTAP were developed using the same global economic database that is maintained and updated in the Global Trade Analysis Project at Purdue University. The USREP and MRN models were developed using the IMPLAN state-level economic database for the United States (Minnesota IMPLAN Group, 2008), and as with USREP, there is a companion global model to the MRN built on the GTAP database. The EPPA, USREP, MRN, and ADAGE models all utilize the General Algebraic Modeling System’s (GAMS) Mathematical Programming System for General Equilibrium Analysis (MPSGE) model development and solv- er software (Rutherford, 1999). The GTAP model uses its own solution algorithm that allows more flexibility in the functional forms than does the GAMS/MPSGE approach. All of these are simulation models where the model developers have surveyed the literature for the value of critical parameters in the model. The key parameters are elasticities of substitution among inputs, and these generally deter- mine how a change in a tax provision will ultimately affect markets. Here the IGEM model differs from the other models, as it includes a time series of data for the U.S. economy, and the parameters of response are econometrically estimated from these data. It also utilizes its own solution algorithm. Different appraisals of the literature, and differences from the econometric estimates of the IGEM model, mean these models will respond differently to tax provisions. The models also differ in how they deal with investment and savings. The GTAP model is a static model. It is developed for the base-year data, and simu- lates how the economy would have been different in that year under different conditions, in our case, for example, the difference with or without a tax provi-

OCR for page 165
168 Effects of U.S. Tax Policy on Greenhouse Gas Emissions sion. In GTAP, capital is fixed and investment is based on a fixed savings rate. The EPPA and USREP models, in their standard formulations, are recursive dynamic models, meaning that investment is determined by a fixed savings rate. Unlike a static model, investment in one period becomes new capital in the next. ADAGE, MRN, and IGEM are forward-looking models, meaning that the sav- ings rate is determined by the model. Higher returns to capital in the future lead to more savings by households today. This requires agents to look forward and anticipate the future returns to capital, making decisions today based on those expectations. Agents are said to have perfect foresight in these models because they are solved so that their expectations are realized exactly. The models also deal with existing capital differently. The EPPA and USREP models have explicit capital vintaging by model sector, so once a facto- ry or power plant of a particular vintage is put in service in a particular sector it is difficult to adjust it until it is fully depreciated. This is sometimes referred to as a putty-clay model: Capital is malleable when the investment is put in place, but once it is built, its features are fixed. ADAGE, MRN, and IGEM have sim- pler structures, generally some version of putty-putty, where capital is fully mal- leable and can be redeployed to other sectors. In some cases the current capital stock characteristics are fixed until it fully depreciates, but any future investment remains fully malleable over its lifetime. These features can be important in the assessment of tax provisions: In provisions with sunset laws, one can observe forward-looking behavior, as there is sometimes a rush of investment to take advantage of a provision before it ex- pires. Hence, both the forward-looking aspects of models and the capital vintag- ing can be important. Investment installed in anticipation of a sunset provision (e.g., a wind turbine) will remain as a viable contributor to the capital stock for 20 years or more. However, one can observe that often these provisions have been extended just as they are to expire, and so agents observing these are likely to have imperfect expectations—based on past experience they may give some chance that the provision will continue indefinitely. Agents also do not have a crystal ball into the future of energy prices, broader carbon policy, or other envi- ronmental policy. Explicitly modeling imperfect expectations or a fully stochas- tic solution to these models is not numerically feasible, and so forward-looking models certainly overestimate the capability of agents to look forward, even as recursive dynamic models may underestimate this ability. There are also differences in the sectoral disaggregation and attention in these models. The EPPA, USREP, MRN, and ADAGE models have retained a relatively high level of aggregation for most sectors of the economy—a total of 6-10—but focused detail on the energy sector, including explicit representation of electricity-generation alternatives, transportation fuels, and vehicle options. The GTAP model has focused heavily on detail in the agriculture sector. IGEM’s model strategy is greater general disaggregation of production sectors, but within the constraints of the Standard Industrial Classification sys- tem. In this classification, for example, electricity generation is a sector, but there is no further distinction of solar, wind, nuclear, or hydro from generation

OCR for page 165
Appendix A 169 using 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 interpreted as an increase in one of the non-fossil-fuel technologies. Rates of technical change are econometrically es- timated based on the historical data, and these are represented as time trends on input requirements rather than explicit technologies. Along with the features discussed above, the IGEM model has the most extensive representation of the general tax system, a feature that was important for consideration of the broad provisions. IGEM is not as rich in energy or agri- culture detail, and hence was not able to simulate many of the narrow provisions of interest to the committee, but its strength was potentially in the broad provi- sions. Its treatment of the effects of accelerated depreciation was a particular strength. Although IGEM is not as strong in analyzing the health and housing provisions, the committee found no model capable of treating these better for our purposes. With regard to treatment of housing provisions, the IGEM and other models above generally treat housing as household investment, and the rental value is a substitute for other household goods. For analysis of energy implications of change in investment of housing, it is more realistic to treat the investment level in housing as a complement to energy use rather than as a sub- stitute. With less investment in housing, we would expect to see smaller houses, larger households and thus fewer houses, and so generally lower energy re- quirements and lower GHGs. As discussed in the main text, as far as we could ascertain there are no models capable of assessing the GHG impacts of specific tax policies that treat household capital and energy as complementary inputs, and so our ability to assess accurately the effects of tax provisions on housing was limited. IGEM simulated the impact of the tax treatment of health insurance as an untaxed portion of wages (supplied in the form of health insurance) rather than a subsidy to health care provision. Thus, it did not have any impact on the pricing of health care or the purchase decisions among goods by households. One might expect that if health insurance was more expensive because it was taxed like other income, that households would consume less health care and more of other goods, and so to the extent health services had different GHG im- plications than other goods we would see an effect. IGEM treatment of the hous- ing and health care tax provisions is better suited to estimating how these exclu- sions distort investment decisions, and thus the level of overall economic activity, rather than how these provisions distort choices among goods and ser- vices. It is difficult to assess how IGEM features might lead to different results than other models of this type, especially without simulating them. Even with comparable simulations, any particular outcome comes about as a result of hun- dreds of different parameters and a variety of complex differences in model structure. In general, we expect, and some comparisons have shown, that the forward-looking behavior results in much greater flexibility, as does the mallea- bility of capital.

OCR for page 165
170 Effects of U.S. Tax Policy on Greenhouse Gas Emissions ENERGY-FOCUSED MODELS There were three principal energy-related models considered by the com- mittee. These were the MARKAL model (MARKAL) (Loulou, et al., 2004), which is more of a modeling framework that can be developed for specific ap- plications depending on interests and data availability; the CBER model (Allaire and Brown, 2011), developed at the University of Nevada, Las Vegas; and the NEMS model, developed and used by the U.S. Energy Information Administra- tion (EIA) in its annual energy outlook and available for other uses through On Location, Inc. MARKAL energy models have been developed for individual energy sec- tors such as transportation or power generation, or for the entire energy system for many different countries. MARKAL models are engineering cost-based models, where technological options are presented as fixed coefficient produc- tion technologies. They generally take prescribed energy or service demands as fixed and explicitly minimize the cost of meeting those demands. They are tech- nology rich. Preferences among technologies are fully determined by out-of- pocket (pecuniary) costs. This assumption may be reasonable for large-scale investments in power plants, where a relatively few well-described options exist and local conditions that are poorly known are unlikely to affect the cost ranking of different options. However, the ability of the approach to realistically simu- late outcomes rests on its ability to fully describe regional considerations that might affect choices, and all the options available. Transportation demand for fuel is an example of some of the limits of this approach. Broad vehicle options are represented with different costs and energy-use requirements, and change in the choice among these options is the only possible response to changing fuel price. But a MARKAL model has difficulty representing the near-continuous nature of actual or potential options in terms of transmission, engine, size, weight, and energy-consuming options that are available, or could be offered if there was demand. Miles driven are prescribed and unaffected by fuel price. Decisions about maintenance or how one drives that might affect fuel use are also assumed to be unaffected by price. Preferences for amenities of vehicles and willingness to trade these against vehicles with greater fuel efficiency are not represented. Thus, this class of models tends to “over optimize,” represent stark technology options, and not consider any behavior or preference other than strictly minimizing vehicle and fuel cost without consideration of other features of the technology that may deliver welfare value to consumers. In this type of model, the purchase of expensive, relatively low-mileage luxury vehicles would appear irrational, whereas an econometric approach based on observed behavior would show resistance to giving up luxury for fuel economy. The CBER model is a polar opposite of MARKAL-type models. In the CBER model, demand and supply for energy are represented as continuous functions. The key parameters in the CBER model are price elasticities of sup- ply and demand. Elasticities for CBER are deduced from reviews of economet- ric literature. The strength of such an approach is that implicitly regional varia-

OCR for page 165
Appendix A 171 tion in the cost of technology, consumer’s willingness to make trade-offs, and broad technological differences are captured to the extent these differences were captured in the historical data. The astructural nature of the model means that it may not capture dynamics of capital turnover well, and it is somewhat captive of the history in terms of the available technology and the structure of policy. For example, the introduction of the Renewable Fuels Standard (RFS) for biofuels substantially changes the fuel market, a relatively recent development that is not reflected in historical supply-demand data on which CBER-type models rely for estimates of elasticities. An advantage of the CBER model is that it was developed with an objec- tive of investigating the effects of tax policy on GHG emissions in the energy sector, and is the only significant effort in this regard to have been done prior to the committee’s work. Given that focus, the developers have invested effort in representing most of the energy-specific tax code provisions. That said, because the model lacks the specific structure around which the tax code affects deci- sions—tax code changes are translated into simple shifts in demand for supply of a fuel or electricity—the outcomes are highly dependent on the analysis and availability of data, or lack thereof, that goes into determining how supply or demand are shifted by a tax code provision. In contrast, all of the other models we evaluated were more structural—a change in the tax code could be repre- sented as a change in the cost of capital, for example. The NEMS model is an eclectic combination of the previous two types of models with considerable detail on the structure of markets and policies affect- ing them and, in general, far more sectoral demand and energy-supply detail. The power sector is similar to a MARKAL-type model, but other sectors of the economy are represented by demand and supply elasticities. It has benefitted from a large and ongoing investment in it to keep up with changes in policy and technology options because results of each EIA annual energy outlook are care- fully scrutinized. While it can be run in conjunction with a macroeconomic model with iteration to achieve energy-sector macroeconomic feedbacks, that was not an option that was feasible in applications considered here. Thus, aggre- gate economic and industrial activity were held constant. And those feedbacks between the energy sector and the rest of the economy are not as fully integrated as in the macroeconomic models discussed above. In the end, the committee was able to commission additional runs from the CBER model and runs of the NEMS model, conducted by On Location, Inc. The strength of the CBER approach is in the breadth of tax provisions it was able to consider. Its weakness was lack of structural detail—it does not explicitly repre- sent existing policies such as the RFS2, capital as different from other inputs, or the relationship of the energy sector to the macroeconomy. The advantage of the NEMS is in both technological and market structural detail on the energy sector. Its weakness is lack of macroeconomic feedback, and more limited detail in the agricultural sector that is important for biofuels tax provisions. However, it does include important structural and policy details of fuels markets and the RFS2 provisions.

OCR for page 165
172 Effects of U.S. Tax Policy on Greenhouse Gas Emissions AGRICULTURE-SECTOR MODELS The committee’s interest in agriculture-sector models stemmed largely from tax provisions related to biofuels. A large variety of approaches have been used to examine the economics of biofuels and biofuel policy. These include macroeconomic models such as EPPA and GTAP (e.g., Gurgel et al., 2007, 2011; Tyner et al., 2010; Decreux and Valin, 2007) broadly reviewed above, agricultural optimization models including FASOM (Adams et al., 1996, as in Beach and McCarl, 2010; Beach et al., 2010), simulation models such as MiniCAM (Wise et al., 2009), and econometric-based simulation models such as FAPRI (Babcock and Carriquiry, 2010). There are several challenging aspects of modeling biofuel policy, including (1) the complex interactions with agricul- ture and agricultural policy, including competing demands for crops and by- products supplies of animal feeds; (2) the complex policy requirements of the Renewable Fuel Standard (RFS2, as described below) and investment and pro- duction tax credits that differentially treat different biofuel production pathways and feedstocks that are the focus of this report; (3) international linkages in agri- culture and energy markets; (4) land-use change and competition for land; and (5) the carbon implications of land-use change. The macroeconomic models are generally more coarsely resolved and have less detail on the agriculture sector and less explicit representation of policy; however, the GTAP model represents agriculture in considerable detail. A limit, as discussed above, is that it is purely static. FASOM has some similarities to a MARKAL-based energy model that chooses least cost production activities. Given the importance of regional differ- ences, it is explicitly spatial, and it represents demands as continuous functions, where key parameters are demand elasticities. FAPRI at its base is similar to the CBER model, but incorporates considerably more structural detail on policy as it affects markets, and approximates dynamic adjustment of markets over time. While it is not maintained by a federal agency as is NEMS, the FAPRI model produces an annual agricultural baseline projection that incorporates changing policy in the agricultural sector. While the FASOM and FAPRI models were initially purely agricultural-sector models, given the interest in biofuels, they have added much more detail on gasoline and petroleum markets and thus are in better position to analyze these interactions. The FAPRI-MU model as applied in this study is a system of demand and supply functions for 16 crops, 15 crop-based products, and 17 different types of livestock and livestock-based products (Meyers et al., 2010; Devadoss et al., 1993). Some equations are econometrically estimated, others are not. In particu- lar, the rapid changes in biofuel markets make direct estimation of biofuel- related equations based on observed behavior problematic. A good example is E85 demand, which can be very important in projections, but has not accounted for more than a very small amount in the past. In contrast, FAPRI-MU updates estimates for equations in some other components where there are more histori- cal data.

OCR for page 165
Appendix A 173 The model’s focus is on the United States, but the rest of world is either collapsed into a single rest-of-world supply-and-demand response as in the case of animal products, composed of a similar rest-of-world aggregate response but with key countries identified, as in the case of ethanol, or represented with ag- gregate rest-of-world supply-and-demand aggregates, as in the case of main crops. This longstanding agricultural model has been recently augmented to in- clude detailed modules on oil markets (Thompson et al., 2011) and U.S. biofuels markets (Thompson et al., 2008). The strength of the model is in its detailed representation of agricultural markets, including global markets, modeling of the complex Renewable Identification Number fuel credits with multiple fuel pro- duction pathways representing both conventional and second-generation pro- cesses, and links to global petroleum and refined fuel markets (Thompson et al., 2010). The modeling approach does not explicitly consider land use or the car- bon implications of land-use change. These are highly uncertain responses with wide-ranging results in the literature (Plevin et al., 2010; Searchinger et al., 2008; Melillo et al., 2009; Keeney and Hertel, 2008; Hertel, 2011; Tyner et al., 2010; Mosnier et al., 2012). Instead, greenhouse gas implications are assessed by applying a fixed GHG coefficient per unit of fuel for different biofuel pro- duction pathways. Default estimates are those of the U.S. EPA (2010) that in- clude CO2, N2O, and CH4 implications of land-use change. DOCUMENTATION OF THE MODELS USED BY THE COMMITTEE Intertemporal General Equilibrium Model (IGEM). Documentation is avail- able at http://www.igem.insightworks.com/docs/190.html. National Energy Modeling System (NEMS). Documentation is available on the website of the U.S. Energy Information Administration at http://www. eia.gov/analysis/model-documentation.cfm (various modules) and http:// www.eia.gov/forecasts/nemsdoc/integrating/pdf/m057(2012).pdf (integrating mo-dule, 08/2012). Center for Business and Economic Research Model (CBER). The equations used in the CBER analysis can be found in Appendix A of the report by Allaire and Brown: Allaire, M., and Brown, S. (August 2012). U.S. Energy Subsidies: Effects on Energy Markets and Carbon Dioxide Emissions. Re- trieved 2012, from http://www.pewtrusts.org/uploadedFiles/wwwpewtrus tsorg/Reports/Fiscal_and_Budget_Policy/EnergySubsidiesFINAL.pdf. Allaire and Brown and Brown and Kennelly used the MATLAB program to solve the system of equations within the CBER model. Any mathemati- cal solver package, including open-source solvers, can be used to solve the equations and thus replicate results of the CBER model. Food and Agriculture Policy Research Institute Model (FAPRI). Documen- tation is available at http://www.fapri.missouri.edu in the following re- ports:

OCR for page 165
174 Effects of U.S. Tax Policy on Greenhouse Gas Emissions 1. FAPRI-MU Report #12-11, Model Documentation for Biomass, Cellu- losic Biofuels, Renewable and Conventional Electricity, Natural Gas and Coal Markets 2. FAPRI-MU Report #09-11, FAPRI-MU Stochastic U.S. Crop Model Doc- umentation 3. FAPRI-MU Report #05-11, New Challenges in Agricultural Modeling: Relating Energy and Farm Commodity Prices 4. FAPRI-MU Report #09-10, FAPRI-MU U.S. Biofuels, Corn Processing, Distillers Grains, Fats, Switchgrass, and Corn Stover Model Documenta- tion 5. FAPRI-MU Report #07-08, Model of the U.S. Ethanol Market FPARI- UMC Report #12-04, Documentation of the FAPRI Modeling System Consultant Reports Detailing Results of Modeling Efforts 6. The original modeling using the models described above was undertak- en by four independent consultants. Each of those consultants produced reports to the committee detailing the results of their modeling efforts. Readers can download those reports at the National Academies Press website, http://www.nap.edu/catalog.php?record_id=18299.