The use of revenue from a carbon tax or cap-and-trade program substantially affects who gains and who loses as a result of the policy; indeed, it is often more important than the effect of the carbon tax itself. In a previous RFF discussion paper, we (together with RFF’s Dallas Burtraw and Richard Morgenstern, and Jared Carbone of the Colorado School of Mines) examined how the initial distribution of the costs across income groups in the United States varied under three different ways to use the revenue generated by a $30 per-ton carbon tax: returning the money via cuts in taxes on labor income, cuts in taxes on capital income, or a equal-sized rebate for every person. (See the blog entry about this paper for more detail).
Our new discussion paper, written with the same team of researchers, follows up by looking at the distribution of costs across states, for the same set of policies. Our methodology is very similar to the earlier paper: we used a dynamic overlapping-generations model to predict price and quantity changes for a range of consumption goods and income sources, and combined that with government data on income from different sources and spending on different goods in each state to estimate how those price changes affect each state. Using RFF’s Haiku electricity market model, we estimated the effects on the price and quantity of electricity consumed in each state, which varies substantially based on characteristics of the local electricity market.
First, we tackled the effect of higher prices for “direct energy goods” (gasoline, electricity, fuel oil, and natural gas, for example). Higher prices for these goods come almost entirely from the carbon tax itself, and thus are nearly identical across the different revenue-recycling methods. Higher prices for these goods have the largest effect in the South, Midwest, and Appalachia, generally speaking (Figure 1). While gasoline price increases account for a substantial fraction of the cost of the policy, that effect varies relatively little across states. Instead, most of the variation in cost across states comes from differences in electricity price increases and levels of electric consumption (electricity in the West and Northeast is, in general, cleaner).
Turning to the effects of revenue use, the total cost of the policy is lowest when carbon tax revenue is used to cut capital taxes and highest under a lump-sum rebate, with labor tax cuts falling in between. But that pattern varies somewhat across states. States with large shares of capital income (such as Florida, Wyoming, and Connecticut) do well under capital tax cuts, while poorer and younger states do worse (Figure 2). Labor tax cuts have fairly even effects across states, so the map for this case looks very similar to a map of how the costs of direct energy goods affect each state (Figure 3). The lump-sum rebate tends to favor poorer states, such as those in the South, Midwest, and Appalachia, thus reversing the pattern: these states are hit hardest by price increases for direct energy goods, but benefit enough from the lump-sum rebate to do better than most other states in this case (Figure 4).
[portfolio_slideshow size=post-main exclude=8385]
Note that these estimates represent mean values for each state, and our previous work showed that costs can vary widely across income groups at the national level. For example, while the lump-sum rebate has the highest overall cost, it nonetheless makes the bottom three income quintiles better off (even without considering any environmental benefits). The same pattern seems likely to hold within any given state. Thus, just because a state as a whole is worse off under a particular policy, that doesn’t necessarily mean that a majority of households in that state are worse off (and certainly not that any particular household is worse off), or vice-versa. For example, Florida does quite well when the carbon tax revenue is used to cut capital taxes, because Florida as a whole gets a large share of income from capital, but relatively young or poor Floridian households would probably fare badly under such a policy.
Comparing the results of our two papers shows that the variation in costs across states is less than the variation across income quintiles. One possible explanation is that income differences simply matter more than geographic differences. Another might be that the relevant geographic variation is between urban, suburban, and rural areas within each state, rather than across states. Washington, DC, which has no suburban or rural areas, does comparatively very well in all but the lump-sum rebate case. If we view DC as a rough estimate for all core urban areas, then that suggests that denser areas will do well under a carbon tax. But regardless of the reason, our results suggest that income is a better predictor than state of residence for how a household will be affected by these policies.