|
Article Excerpt
1. INTRODUCTION
Economists have long argued that market based instruments are more efficient than regulations as a means of addressing the social damages arising from polluting activities. By market-based instruments we mean policies that force firms to "internalize" the cost of polluting activities. In the context of climate change arising from greenhouse gas (GHG) emissions, the polluting activity is the release of carbon dioxide and other greenhouse gases. (1) Carbon taxes and cap and trade systems are two examples of market based instruments that create a cost to emissions. A carbon tax does this directly by taxing the carbon content of fuels while a cap and trade system imposes a cost by requiring the surrender of valuable permits in proportion to the carbon content of fossil fuels. (2)
U.S. greenhouse gas emissions equaled 7,147 million metric tons of C[O.sub.2] equivalent (MMC[O.sub.2]e) in 2005, an increase of 17 percent over 1990 levels. Carbon dioxide emissions account for the vast bulk of emissions and equaled 6008.6 MMC[O.sub.2] in that year. A consensus is emerging in the United States that climate change is a critical issue requiring a reduction in GHG emissions. Several bills have been proposed in the current Congressional legislative session (110th Congress, 2007-2008) to control greenhouse gas emissions. (3) And May 2007, President Bush called for the United States along with other major greenhouse gas emitting countries to "set a long-term goal for reducing greenhouse gases" (Stolberg (2007)). The recent releases of reports by Intergovernmental Panel on Climate Change Fourth Assessment Report's Working Groups provide additional evidence to support the role of anthropogenic warming.
A major concern with either a carbon tax or a cap and trade program to reduce emissions is that the burden of the costs arising from the policy will fall disproportionately on poorer households--or in the terminology of incidence analysis, the policies will be regressive. (4) Metcalf (1999, 2007) argues that even if a carbon tax is regressive, a carbon tax reform (combining a carbon tax with a revenue neutral reduction in some other tax) can be distributionally neutral or even progressive if desired.
In this paper, we focus on a related but different point. We measure to what extent a carbon tax is regressive in a lifetime income framework. (5) We also decompose the burden of the tax into direct and indirect components. The direct component measures household burdens from their direct consumption of fuels (gasoline, home heating, and electricity) while the indirect component measures the increase in the cost of other goods resulting from the higher fuel costs used in their production. We look at three different years, 1987, 1997, and 2003 to see how the incidence pattern would change had a carbon tax been in effect in these three time periods.
Our results suggest that in general, carbon taxes appear more regressive when income is used as a measure of economic welfare than when consumption (current or lifetime) is used to measure incidence. Measures based on current energy consumption can be misleading as retired persons have relatively low incomes and high energy consumption shares. Further, the direct component of the tax, in any given year, is more regressive than the indirect component. In fact, for 1987, the indirect component of the tax is actually mildly progressive as the higher income groups tend to pay a larger fraction of their income in carbon taxes. This is consistent with the observation that many luxury goods such as air travel are energy intensive while, in contrast, direct energy consumption is a necessity.
Studying the intertemporal distribution, we find that between 1987 and 2003, direct taxes have become marginally less regressive while indirect taxes have become marginally more regressive. As a result the distribution of the total tax burden has not changed much over time.
Carbon taxes are also thought to have uneven regional effects. We report the average carbon tax paid per household across regions and find that the regional variation is at best modest. By 2003 variation across regions is sufficiently small that one could argue that a carbon tax is distributionally neutral across regions. Not surprisingly much of the variation that we do observe arises from the direct carbon tax rather than the indirect tax. In other words, differences in electricity generation, driving patterns and weather conditions lead to the variation rather than the choice of energy intensive commodities in different regions.
In the next section, we explore different methods used to measure incidence and motivate the lifetime measure of consumption employed in Bull et al. (1994). Section 3 details our data and methodology. Section 4 presents results for the economic incidence of the tax. Section 5 explores the geographic incidence of the tax. Section 6 concludes.
2. MEASUREMENT OF INCIDENCE
Tax incidence measures the ultimate impact of a tax on the welfare of members of society. The economic incidence of a tax may differ markedly from the statutory incidence due to price changes. For a carbon tax, the short run economic incidence is likely to differ markedly from the statutory incidence. While the statutory incidence of an upstream tax on gasoline may be on the refinery owner, the economic incidence is likely to be on final consumers as the tax is shifted forward to consumers in the form of higher prices. Measuring the incidence of a tax requires numerous assumptions and we begin the analysis by setting out our assumptions and methodology.
First, we must determine the appropriate unit of observation, which could be an individual or a household. For this study, we use the household as a unit. Second, we must choose the appropriate time frame of analysis. As we discussed in the introduction, the choice of the time frame for the analysis is extremely important. Early tax incidence analysis used current income as the base i.e. it compared the tax liability over a short period to income earned over that period. Following Friedman (1957) and the permanent income hypothesis, there was a realization that consumption decisions are made over a longer time horizon. Hence income should be measured as the present discounted value of lifetime earnings and inheritances. Failing to do so creates substantial measurement problems, particularly at the low end of the income distribution. For example, elderly people drawing down their savings in retirement will look poor when in fact, they may be comfortably well off in a lifetime context. In other words, many low-income people are not necessarily poor. (6) Caspersen and Metcalf (1993) report cross tabulations on income and consumption that show that a large fraction of households are in consumption deciles substantially above their income deciles.
Poterba (1989) follows the approach of using current consumption as a proxy for permanent income, since if consumer behavior is consistent with the permanent income hypothesis, then consumers would set current consumption proportional to permanent income. However, as we mentioned earlier, using data from the 1987 Consumer Expenditure Survey, Bull et al. (1994) show that consumption, instead of being smooth, closely tracks current income over the lifecycle. Moreover, energy consumption also shows a marked lifetime pattern.
This could be problematic for the incidence measurement. Suppose that as people grow old their energy consumption becomes a larger share of their total consumption, and suppose, as well, that over a lifetime the energy tax has a proportional incidence, then using current consumption to measure lifetime income, the energy tax would appear regressive.
As an alternative to current consumption, we use an adjusted lifetime measure for consumption that is intended to correct for long-run predictable swings in behavior. This measure was first employed in Bull et al. (1994). Ideally, a lifetime measure of incidence would be constructed by taking the ratio of lifetime energy taxes to lifetime earnings. Unfortunately, the lack of any sufficiently long longitudinal panel data set precludes such an approach. (7)
To proxy for lifetime consumption, we therefore use the age profile of people sampled in a particular year by the Consumer Expenditure Survey (a more detailed description of the calculation is presented in Bull et al. (1994)). In particular, we first classify people into different subsamples based on their educational level, since the pattern of income and consumption will be quite different for people with vastly different human capital stocks. For each sub sample, we then calculate a "typical" path of consumption through the averages for the age groups. For a given person in the sub sample we know the ratio of their current consumption to the average for their age group. We then compute their lifetime consumption by multiplying this ratio by the present value of the typical lifetime path.
For example, suppose an individual is a 35-year-old PhD whose energy consumption is 80 percent of the average for her age and education group. Let's say the present discounted value of total lifetime energy consumption for a person with a PhD is $80,000. Then for this individual, the imputed lifetime energy consumption is $64,000.
This procedure allows...
|
