Read an Excerpt
The Design and Implementation of US Climate Policy
The University of Chicago PressCopyright © 2012 National Bureau of Economic Research
All right reserved.
Chapter OneDistributional Impacts in a Comprehensive Climate Policy Package
Gilbert E. Metcalf, Aparna Mathur, and Kevin A. Hassett
Distributional considerations figure importantly in the design of comprehensive climate policy legislation. The allowance allocation in the American Clean Energy and Security Act of 2009 (H.R. 2454), popularly known as the Waxman-Markey bill, that was passed by the House of Representatives in June 2009, suggests the care and attention paid to distributional considerations in crafting the bill. Both the Kerry-Boxer bill and the Cantwell-Collins proposals in the Senate also paid close attention to distributional considerations.
This chapter uses data from the 2003 Consumer Expenditure Survey to allocate the burden of carbon pricing from possible cap-and-trade legislation under different assumptions about the relative importance of uses-and sources-side heterogeneity as well as differing assumptions about relative factor price changes. It builds on previous research using the Consumer Expenditure Survey by generalizing the incidence assumptions beyond the assumption of full-forward shifting of the carbon price. It also improves on the measurement of capital income burden allocation by using capital income distribution data from the 2004 Survey of Consumer Finances (SCF) to augment the data in the Consumer Expenditure Survey.
The approach detailed in this chapter provides a method for carrying out a back-of-the-envelope calculation of the distributional impact of carbon pricing using readily available data that allows for sensitivity analysis of assumptions on sources-and uses-side incidence of carbon pricing. We find that accounting for sources-side impacts of carbon pricing yields less regressive impacts on households looking across the income distribution.
Households differ on a number of dimensions that policymakers may care about. When designing a climate policy bill, policymakers have made it clear that many of these dimensions are important and affect the allocation of allowances as well as the mechanisms of allowance use. Households differ by income, regional location, primary heating source, and predominant mode of electricity generation among other things. We focus in this chapter on measuring the impact of carbon-pricing policies on households looking across the income distribution.
In carrying out distributional analyses, a number of considerations come into play. First is the question of how best to sort households to distinguish them by some measure of relative well-being. Income is often used for this ranking and this analysis sorts households by annual income. This brings a potential bias to the analysis to the extent that annual income is a poor proxy for lifetime well-being. As discussed elsewhere (see, for example, Fullerton and Metcalf ) many low-income households are not poor in a lifetime sense. They may have transitorily low income or may be at a low income-earning stage of their careers. In both these cases consumption-to-income ratios may be unusually high and may provide a misleading picture of the distributional impact of consumption-related taxes (like energy taxes) or carbon-pricing policies. As a check for the importance of our income measurement we also provide results where we use current consumption as a proxy for lifetime income under the assumption that households engage in consumption smoothing.
A second issue is that the economic impact of carbon pricing depends importantly on how prices adjust to the new equilibrium with carbon pricing. This is particularly important for a policy that creates and distributes financial assets in excess of $100 billion by the middle of this decade (see Congressional Budget Office 2009). A number of computable general equilibrium economic analyses have argued that carbon pricing will predominantly be passed forward to consumers in the form of higher energy prices. See, for example, Bovenberg and Goulder (2001) and Metcalf et al. (2008).
Based on analyses focusing on uses-side incidence impacts of carbon pricing, a number of economists have carried out distributional analyses of carbon pricing using the Consumer Expenditure Survey, including Bull, Distributional Impacts in a Comprehensive Climate Policy Package 23 Hassett, and Metcalf (1994), Dinan and Rogers (2002), Metcalf (1999), Parry (2004), and Hassett, Mathur, and Metcalf (2009). The Consumer Expenditure Survey is particularly useful for this analysis given its high level of detailed disaggregation on household spending patterns. But these analyses are useful only to the degree that the assumption of full-forward shifting (e.g., impacts on uses side only) is correct.
In the following analysis we refer to forward shifting and backward shifting when we wish to analyze the distributional impacts of carbon pricing according to how households spend their income (uses side) or earn their income (sources side). The terminology of forward and backward shifting has a long-standing place in public economics, albeit an imprecise meaning. Whether a tax is shifted forward (leading to higher consumer prices) or shifted back (leading to lower factor returns) depends on the normalization employed in the general equilibrium framework. Since the normalization choice in a general equilibrium model has no real effects, forward or backward shifting cannot have real effects either (see Fullerton and Metcalf  for more on this point). When we later refer to forward or backward shifting, we use this to refer to heterogeneous impacts of carbon pricing based on how different households spend or earn their income.
A recent study by Metcalf et al. (2008) found that for a given price normalization forward shifting of carbon pricing ranged widely depending on the fuel in question, the proposal under consideration, and the particular year of analysis. Carbon pricing on coal was nearly fully passed forward into higher prices, reflecting in large part the low Hotelling resource rents for coal. Shifting for natural gas ranged from a low of 14 percent to a high of over 200 percent. The latter occurs as demand rises for natural gas in the intermediate term as gas substitutes for coal in the production of electricity. Finally, forward shifting for crude oil ranged from a low of 2 percent to a high of nearly 90 percent depending on the year and tax scenario.
If taxes are not passed forward to consumers in the form of higher product prices, then they are passed back to factors of production in the form of lower wages, returns to equity, and reduced resource rents. Changes in resource rents can also affect government revenues since much fossil fuel extraction in the United States occurs on publicly owned land (e.g., the Powder River Basin coal reserves in Wyoming and the Outer Continental Shelf oil and gas drilling). We ignore that complication in this analysis in part because the impact of taxes on government revenue from land-leasing activities is poorly understood.
This chapter uses burden-shifting insights from computable general equilibrium (CGE) models along with the Consumer Expenditure Survey to measure the burden of carbon pricing. A goal of the analysis is to demonstrate the ability to use the survey with a broader range of assumptions to obtain a rough-and-ready guide to the distributional impacts of carbon-pricing proposals without having to run full-blown CGE analyses.
1.3 Measuring Carbon Price Burdens
Our goal in this chapter is to provide a simple rough-and-ready measure of the burden impact of carbon pricing that builds on the insights of more complex economic analyses. This is in the tradition of a number of studies that use detailed data sets such as the Consumer Expenditure Survey (CEX) along with results and insights from sophisticated economic models to allocate the burden of government policies to different economic groups.
As noted earlier, previous studies using the CEX have assumed that carbon pricing is fully passed forward into higher consumer prices based on the carbon content of goods and services. Input-Output tables from the Bureau of Economic Analysis are used to trace through carbon content and thus carbon-pricing impacts. If carbon prices are passed back to factors of production, then we need to use income information in the CEX to distribute the carbon-pricing impacts. We distribute the burden of carbon pricing that falls on owners of capital in proportion to capital income shares as a proxy for capital ownership shares.
Carbon-pricing burdens may also fall on owners of fossil fuel resources. To the extent these resources are privately owned, carbon pricing may lead to a reduction in returns to owning property with fossil fuel resources. Some of this property is held by sole proprietors and partnerships while other tracts are owned by corporations. Lacking detailed information on resource ownership, we assume that resource ownership is distributed across households in the same manner as capital.
Turning to allowances, we can allocate the value of allowances to households either according to consumption or income patterns depending on how allowances are distributed. The Waxman-Markey bill sets aside roughly 30 percent of allowances in the early years for distribution to customers of electricity and natural gas utilities to compensate them for higher electricity and gas prices. We allocate the value of those allowances to households based on their electricity and natural gas expenditures, respectively. Allocations to industry are assumed to benefit owners of capital. Allocations to households are distributed to households.
In general we follow the distribution approach of Rausch et al. (2010) for distributing the value of allowances. One place where we differ is in the allocation of allowances to the US government for deficit reduction. Under the assumption that reductions in the deficit reduce pressure to decrease government spending, we allocate the allowances for deficit reduction based on government spending that would otherwise have to be cut. Our assumptions on the benefits of government spending across the income distribution are taken from the Tax Foundation (2007).
Rather than assume a particular burden-sharing outcome, we report results for four different scenarios to illustrate the importance of the burden-sharing assumption on distributional outcomes. The four scenarios we consider are reported in table 1.1. The first scenario assumes full-forward shifting of carbon pricing to final consumers (i.e., burden is based on heterogeneity in household expenditure patterns). The next three scenarios allow for a greater role in sources-side effects with different assumptions about relative price changes between capital and labor. These approaches are based on a particular normalization (price of non-carbon-based consumption goods held fixed). As noted previously, forward and backward shifting is imprecise (and potentially misleading) terminology though long used in public finance. More precisely we focus on distributional impacts based on uses-side impacts and sources-side impacts. Scenario 1 focuses on uses-side heterogeneity only. The remaining three scenarios allow for greater amounts of sources-side heterogeneity and also allows for differential impacts on wage and capital (and resource) income.
1.4 Issues in Using the Consumer Expenditure Survey
The Consumer Expenditure Survey has been used by a number of researchers investigating the burden impacts of carbon pricing because of its rich detail on consumption patterns of US households. It also contains information on the demographic makeup of households as well as some income information. The CEX has a single capital income measure that researchers have used to allocate taxes to owners of capital in scenarios assuming some degree of backward shifting. The survey question for this data asks whether households received any regular income from dividends, trusts, estates, or royalties. A separate question asks about interest income from bank accounts, money market funds, CDs, or bonds. Researchers have used the dividend income amount (or dividends and interest) as a proxy for capital holdings under the assumption that capital income is proportional to capital holdings.
The problem with using CEX-reported capital income is that it may misrepresent capital holdings across income groups. There are two possible reasons. First, the CEX focuses primarily on spending and the income data quality may not be as high quality as the spending data. Second, if holdings of growth stocks are disproportionately held by higher income groups, then the CEX capital income measure will be biased toward more capital holdings in lower income groups. Table 1.2 suggests that the first problem is significant with the CEX showing more capital income in the lower income deciles than the SCF.
Using data from the 2004 SCF, Wolfe (2010) estimates that 85 percent of net worth capital is held by households in the top quintile and 92 percent of nonhousehold wealth by this quintile. The CEX in contrast reports only 70 percent of capital income accruing to the top quintile. Using CEX capital income distributions will skew any carbon-pricing distribution toward greater progressivity to the extent that any of the burden is placed on owners of capital.
One advantage of using the SCF is that it disproportionately samples wealthy families. Each survey consists of a core representative sample combined with a high-income supplement, which is drawn from the Internal Revenue Service's Statistics of Income data file. Further, the survey questionnaire consists of detailed questions on different components of family wealth holdings. For these reasons, the SCF is widely acknowledged to be the best at capturing both the wealth at the top of the distribution and the complete wealth portfolio of households in the middle. Since the wealth distribution is highly skewed toward the top, most other surveys (like the CEX) that have poor data on high-income families tend to underreport measures of income and wealth.
The problem of distributional bias is not as significant for labor income as for capital income. Table 1.3 reports labor income shares across deciles from the CEX and SCF. The distributions are more closely aligned than those for capital income.
In this analysis we distribute the burden of carbon pricing that is shifted to owners of capital based on the distribution of capital income from the SCF (table 1.2).
For purposes of our analysis, we consider the effect of a carbon tax set at a rate of fifteen dollars per metric ton of carbon dioxide. We trace the effect of this carbon tax on the prices of consumer goods produced by the industries through the use of Input-Output matrices available from the Bureau of Economic Analysis. Once we obtained the effect of the tax on prices of consumer goods, we used data from the Consumer Expenditure Survey (CEX) to compute carbon taxes paid by each household in the survey. For a detailed discussion of this methodology as well as the computed price increases, see Metcalf (1999) and more recently, Hassett, Mathur, and Metcalf (2009).
We extend the analysis in this chapter by considering the incidence on the sources-side as well. Using capital and labor income shares from the Survey of Consumer Finances (SCF), we are able to compute the carbon tax burdens on capital and labor income for households in the CEX. Hence the total burden on any household is computed as the sum of the burden on the consumption side, as well as on the income side.
The final step in the calculations shown in tables 1.4, 1.5, 1.6, and 1.7 is the allocation of the allowance revenues under the three proposals. Every proposal allows some level of rebates to households that are based on their energy use, their labor and capital income shares, or whether they are low income. The final burden is lowered by the level of rebates allowed under the three proposals.
As noted earlier, the distributional tables are based on a carbon-pricing policy that yields a carbon price of fifteen dollars per ton CO2. This is consistent with permit price estimates in the 2015 to 2020 period for either H.R. 2454 (Waxman-Markey) or the Kerry-Boxer bill in the Senate. In the analyses in which allowance revenues are returned to households, we assume full return of revenue to households allocating permit value using the assumptions in Rausch et al. (2010).
Table 1.4 shows results for a cap-and-trade program in which we ignore the rebate of permit revenue to households. This scenario focuses on carbon pricing itself without the confounding effects of allowance allocations. The left panel of the table sorts households by annual income while the right panel sorts households by annual consumption, a proxy for lifetime income under the assumption that households engage in consumption smoothing.
Excerpted from The Design and Implementation of US Climate Policy Copyright © 2012 by National Bureau of Economic Research. Excerpted by permission of The University of Chicago Press. All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.