Much has been written about the use of taxes to control pollution. For some time environmentalists and economists have called for greater use of Pigovian taxes to internalize external costs and discourage polluting activities. More recently, arguments in favor of such taxes were strengthened by pointing out the potential for a double dividend: in a world with pre-existing distortionary taxes environmental improvements might be cost free due to the efficiency gains from recycling Pigovian tax revenues to reduce other taxes. A third phase of the literature examines this premise and generally finds that the second dividend may be small, if it exists at all.

Many environmentalists and policy makers have been eager to adopt the double dividend argument to support environmental taxes, but have been largely silent in response to the third phase of the literature. The ambivalent response of the policy community to this important dialogue may be because of one or more of the following reasons:

1. The literature is highly technical and has not received sufficient popular interpretation. For example, Hamond [1997] says, "Much of the research on this subject is technical and not easily translated into the language of public discourse. The bottom line is that, to date, the double dividend is an unresolved empirical issue."
2. The models are based on restrictive assumptions and their generality is not apparent to policy makers. Bovenberg and DeMooij [1994] show that under some circumstances the optimal tax on polluting goods or inputs is lower than the Pigovian tax if the income tax is positive. They conclude that their model shows "that environmental taxes typically exacerbate, rather than alleviate, preexisting tax distortions even if revenues are used to cut preexisting distortionary taxes."
3. The results are perceived as anti-environmental and therefore there has been some resistance to their acceptance within the environmental community. One explanation for this perception is the nature of the discourse. Much of the work in this area abstracts from the environmental benefits of Pigovian taxes and focuses on the "second dividend," the benefits from restructuring the tax system. Goulder [1995] points out, "the validity of the strong form of the double dividend would significantly reduce information hurdles associated with the evaluation of environmental taxes."

This paper addresses these problems by synthesizing the double dividend literature and putting it in a policy context. The first dividend consists of environmental benefits of equating marginal social benefits and costs where an environmental externality exists. The recent literature does not dispute this and sometimes abstracts from it. The second dividend comes from reforming the tax structure. The existence of the second dividend depends on uncorrected distortions in the tax system. The availability of environmental taxes provides an incentive to overcome inertia and find ways to correct distortions. They also offer significant revenue potential.

We are generally interested in the third phase of the literature, in which general equilibrium models are used to understand the efficiency implications of multiple taxes. For example, Goulder [1992] has argued on the basis of simulations that "...carbon taxes cannot be justified on narrow efficiency grounds; ...these taxes tend to cause greater distortions than the distortionary taxes they might partly replace." Bovenberg and DeMooij [1994] show that under some circumstances the optimal tax on polluting goods or inputs is lower than the Pigovian tax if the income tax is positive. Their model contains only one primary input, labor. Consumers can choose between polluting and nonpolluting goods, so a tax on the polluting good causes a distortion (or correction) both at that margin and in the supply of labor. The optimal environmental tax is lower than the Pigovian rate. Bovenberg and DeMooij conclude that their model shows "that environmental taxes typically exacerbate, rather than alleviate, preexisting tax distortions even if revenues are used to cut preexisting distortionary taxes." Fullerton [1997] shows that the optimal environmental tax can be above the Pigovian rate, but that it is the difference between the taxes on dirty and clean goods that will be less than the Pigovian rate. Parry [1995] presents a partial-equilibrium analysis that breaks the effects of the environmental tax into several effects: the environmental effect, the revenue effect, and the tax-interaction effect. The tax-interaction effect represents the welfare loss due to the exacerbation of pre-existing tax distortions, and this effect can be large. If it is larger than the revenue effect, there is no second dividend. If it is larger than the environmental effect, then revenue recycling may be a necessary component of the tax package for the environmental tax to be welfare improving.

Goulder [1995] clarifies the definitions of a double dividend and examines the evidence for their existence. He defines weak and strong forms. The weak form merely requires that recycling revenues from an environmental tax by reducing other distortionary taxes provide more benefits than distributing them in a lump sum fashion. The strong form requires that the benefits from revenue recycling exceed the cost of collecting the tax; the tax reform by itself has a net benefit. If this is the case, uncertainty about the value of the environmental benefits need not be an obstacle to reform.

Alternative model specifications, including model features and parameter specifications, affect the conclusions we draw from the models. Fuchs, Krueger, and Poterba [1998] find that "both labor economists and public finance economists give widely disparate estimates of many important economic parameters such as elasticities of labor demand and labor supply." Differing estimates of parameter values are one explanation for different policy recommendations. For example, labor supply elasticity is a critical parameter in models that examine distortions caused by income taxes.