WPS6238 Policy Research Working Paper 6238 Green Industrial Policy Trade and Theory Larry Karp Megan Stevenson The World Bank Development Research Group Environment and Energy Team & Sustainable Development Network Office of the Chief Economist October 2012 Policy Research Working Paper 6238 Abstract This paper studies the reality and the potential for green provides many lessons for green industrial policy. The industrial policy. It provides a summary of the green authors highlight four of these lessons, involving the industrial policies, broadly understood, for five countries. Green Paradox, the choice of quantities versus prices with It then considers the relation between green industrial endogenous investment, the coordination issues arising policies and trade disputes, emphasizing the Brazil- from emissions control, and the ability of green industrial United States dispute involving ethanol and the broader policies to promote cooperation in reducing a global United States-China dispute. The theory of public policy public bad like carbon emissions. This paper is a product of the Environment and Energy Team, Development Research Group, and the Office of the Chief Economist, Sustainable Development Network, in the World Bank. It was produced for the Green Growth Knowledge Platform (www.greengrowthknowledge.org), a joint initiative of the Global Green Growth Institute, Organisation for Economic Co-operation and Development, United Nations Environment Programme, and the World Bank. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are posted on the Web at http:// econ.worldbank.org. The authors, Larry Karp and Megan Stevenson may be contacted at karp@berkeley.edu and m_stevenson@berkeley.edu. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Green Industrial Policy: Trade and Theory Larry Karp Megan Stevenson June 2012 Key words: green industrial policy, trade conflicts, green paradox, asymmetric information, coordination games, participation games JEL codes: F13, F18, H21, H23 Sectors: Environment, Trade From Growth to Green Growth - A Framework* Larry Karp / karp@berkeley.edu Megan Stevenson / m_stevenson@berkeley.edu Dept. Agricultural and Resource Economics University of California at Berkeley * Views and errors remain ours alone, and should not be attributed to the World Bank Group or its member countries. 1 Introduction Green industrial policy, as used in this paper, refers to government attempts to hasten the development of low-carbon alternatives to fossil fuels. There are similarities, but also important differences between the arguments con- cerning green industrial policy and those that apply to industrial policy in general. The same factors that complicate attempts to assess the record of general industrial policy also apply to green industrial policy. We review these arguments and describe some of the difficulties of evaluating actual policies. We then summarize the green industrial policies currently being used in a group of developed and developing countries. Disputes between the US and Brazil and the US and China illustrate the trade issues created by green industrial policy. We then provide several examples of the use of theory in illuminating aspects of green industrial policy. Governments use industrial policy to promote the development of new industries and the creation and adoption of new technologies. Types of industrial policy include tax credits, input, output and R&D subsidies, min- imum use requirements, standards, and trade restrictions. The usual ratio- nale for industrial policy is that it corrects a market failure. There may be a coordination problem, where the development of an industry requires several components, each of which is proï¬?table to develop only if the others are also developed. By directing the development of a key component, the government may give the private sector the incentives to develop the ancil- lary components. There may be learning by doing that is external to ï¬?rms, or network effects, or economies of scale that cannot be captured by a sin- gle ï¬?rm. These situations involve market failures that in principle can be remedied by government intervention. Persistent unemployment, the desire to diversify an economy that is heavily dependent on volatile primary com- modity prices, or the attempt to capture a ï¬?rst mover advantage are other rationalizations for industrial policies. These policies may also be motivated by the desire to support allies in crony capitalism. Many people are skeptical of the wisdom of industrial policy, because they doubt the ability of governments to correctly identify and then to cor- rect market failures. In a 2010 Economist’s online debate1 Dani Rodrik and Josh Lerner summarize the positions for and against industrial policy. They both recognize the abstract argument, based on market failures, in favor of 1 See http://www.economist.com/debate/days/view/541 1 government policy, and both recognize the practical difficulties of success- fully implementing industrial policy, particularly the difficulty of identifying winners, and preventing regulatory capture. Both provide examples of in- dustrial policy whose success or failure support their position. Harrison and Rodriguez-Clare (2010) review the developing country experience with industrial policy. Rodrik (2009) views the empirical evidence concerning industrial policies as “uninformativeâ€?, but considers the informational and bureaucratic constraints on these policies to be mutable. Lin (2010) notes the prevalence of industrial policies across both developed and developing countries, and suggests some tools based on statistical analysis that might help to identify good candidates for future industrial policy. Chang (2006) evaluates the role of industrial policy in East Asia, and concludes that there is scope for industrial policy even in countries at the technological frontier. Green industrial policy is motivated by many of the same considerations that apply to general industrial policy. There are also two important reasons for using green industrial policies that are either absent or at least less com- pelling in the case of general industrial policy: a commitment problem and the endogeneity of future policy. The rewards of general industrial policy depend on the targeted industries’ eventual success in the market. A gov- ernment might be able to influence the development of its domestic industry, but still have limited ability to influence the market in which that industry operates. In contrast, future government policy largely determines the size of the market facing green industries, and therefore determines the proï¬?tability of green investments. Industries may be reluctant to undertake green R&D or to adopt green technologies because they are uncertain whether future policy will render these kinds of investments proï¬?table. Current govern- ments are not able to make credible commitments about the types of policies that will be in effect during the lifetime of these investments. Lacking the ability to guarantee ï¬?rms that current green investments will be proï¬?table, governments may use green industrial policy to promote those investments. The second difference between the motivations for green and general indus- trial policy rests on the endogeneity of future environmental policy. The practicality of future carbon taxes, for example, depends on the future avail- ability of alternative fuels, which depends on current investment. Green industrial policy provides current governments a means of influencing future environmental policy. The belief that society will move or should move toward low-carbon en- ergy sources may be based on the decline in low-cost fossil fuel stocks, the 2 desire to diversify energy portfolios and reduce dependence on foreign suppli- ers, and concern about climate change caused by the build-up of atmospheric stocks of greenhouse gasses. Even politicians who view the movement toward low carbon alternatives as exogenous may support green industrial policies as a means of ensuring that domestic industries will be positioned to take advantage of future opportunities. Politicians may also see green industrial policy as offering an opportunity to increase domestic employment. These types of considerations are common to industrial policy in general. While countries with a dirigiste tradition consider industrial policy to be a natural activity for governments, it tends to be viewed with considerable skepticism in the US. Even there, however, it enjoyed a kind of vogue dur- ing the 1980s and 1990s. Proponents of industrial policy sometimes agree that although governments are not well placed to pick winners, they can nevertheless intervene to promote well functioning markets. However, green industrial policy — like industrial policy of any hue — often does involve pick- ing winners. This choice occurs at a micro and a macro level. As an example of the former, when a government decides to offer a particular company a loan guarantee, it makes a bet that this company will be a winner. The macro choices are more important. Brazil promotes sugar-based ethanol, and the US supports corn-based ethanol; Germany emphasizes solar power and China emphasizes wind power. A broad based policy, such as a carbon tax or technology-neutral perfor- mance standards, allows the market to identify winners. Most economists prefer those types of policies, but political realities limit their applicability. Both taxes and increased regulation are currently in bad odor. They might also be ineffective because of the commitment problem described above. Un- less mandated or subsidized, current investment and adoption decisions de- pend on the anticipation of future policies, over which current governments have limited control. It would be helpful to have retrospective studies that determine whether past green policies have been successes or failures, but any such study would be quite subjective. Brazil’s ethanol policy illustrates the difficulties. This policy has been in effect for decades, and has resulted in a strong ethanol producing industry. On that ground it appears to be a success. However, we do not know what the counterfactual is; we do not know the oppor- tunity cost of the resources devoted to Brazil’s ethanol policy. Brazilian ethanol currently has low production costs because of previous investments in production capacity and infrastructure. If it had to start from scratch 3 in building its ethanol industry today, even given the improved technology relative to the 1970s, would it be a good investment? How would the answer depend on future oil prices? We also lack reliable measures of the extent to which Brazil’s ethanol policy has contributed to environmental problems, in particular deforestation; and there is no consensus about how to price those environmental consequences. Brazil’s ethanol policy also illustrates the policy risk that investors face. Brazil backed away from its ethanol polices when oil prices fell in the 1980s. As situations change, so does policy, exposing investors to policy risk. In- vestment subsidies shift some of that risk from private investors to taxpayers. This kind of risk-sharing is efficient in circumstances where the proï¬?tability of an investment depends on future government policy. Other examples of the potential responsiveness of policy to current circumstances include the oscillations in US subsidies for wind and solar over the past 25 years, and the recent attempt in California to overturn AB32, the green legislation passed several years ago. Some green industrial policies seem less promising. For example, Ger- many, with approximately 60 days of sunlight a year, has invested heavily in solar power. However, the German policy has been in effect only a few years, the time during which investment costs tend to be highest. If Ger- many eliminates nuclear power more quickly than planned, or if its carbon fuel prices spike (e.g. if Russia decides to curtail deliveries of natural gas), the German investment might in the future be viewed as a success. Green industrial policies are seldom the cheapest of way of achieving goals such as the reduction of gasoline consumption or reduction of carbon emissions. However, the environmental purpose of green industrial policy is to create a basis for future beneï¬?ts, not (primarily at least) to achieve short run goals, except possibly as a metric of progress. Thus, the cost of achieving short run goals is not an adequate basis for judging the policies. Under plausible scenarios, the future beneï¬?ts may be large, but those scenarios are speculative. For example, current investments in non-carbon fuel may delay the need to build some high carbon power plants, while also reducing future low carbon fuel costs via learning by doing or by the presence of economies of scale. In this case, the green policies lower future climate related damages directly, and make it easier to implement future green policies. Our paper consists of three parts, a summary of current green industrial policy, a consideration of trade issues that arise from those policies, and a more general discussion applying lessons from the theory of public policy to 4 green industrial policy. First, we review green industrial policy used in ï¬?ve countries, Brazil, China, the US, India, and Germany. These countries have a wide variety of policies that promote the development of green energy, and thus qualify as components of industrial policy. There has been a rapid growth in green energy capacity in these countries and more generally across the world, to- gether with signiï¬?cant decreases in the cost of this energy. Nevertheless, solar and wind power, two of the fastest growing sources of green energy, still make up less than 1% of energy supply in four of the ï¬?ve countries, and less than 2% in the ï¬?fth, Germany. The current green industrial policies have been associated with trade disputes between Brazil and the US involving ethanol, and a broader dispute between the US and China. The Brazil-US ethanol dispute is a fairly typical type of trade disagreement, comparable to the Brazil-US cotton dispute. WTO law entitles signatories to receive the same treatment for their exports as “likeâ€? national products. This requirement of national treatment of like products means that Brazil’s ethanol exports to the US would receive the subsidy (indirectly) given to US producers. To prevent Brazilian exporters from undercutting US producers, and sending US tax dollars to Brazil, the US imposed a near-prohibitive tariff on ethanol imports until the end of 2011, when the subsidy lapsed. An agreement to accord green products a favorable status with low tariffs would help to eliminate this kind of trade friction. Given the lack of progress in the Doha Round of WTO negotiations, the prospect of that change is slight. The US-China dispute is more fundamental. For over a decade China has run a balance of trade surplus, and a particularly large surplus with the US. At the same time China has ï¬?nanced US consumption, in the form of pur- chases of US Treasury notes. These two phenomena are two sides of the same coin. China ï¬?nances its purchase of US debt by running the balance of trade surplus; it maintains that surplus by means of an undervalued exchange rate and by the simultaneous suppression of domestic consumption. US policy- makers would like China to reduce its trade surplus, to increase the demand for US exports and to reduce competition facing US import-competing sec- tors, thereby increasing US employment. However, the sudden cessation of China’s purchases of US paper would make it harder for the US to main- tain its ï¬?scal deï¬?cit. There is little likelihood of reducing this ï¬?scal deï¬?cit in the short run, and given the state of the economy perhaps it would be counterproductive to do so. 5 This conflict between the US and China is not the result primarily of sec- toral conflicts, unlike the situation with Brazil and the US. However, there is the potential that the conflict will manifest at sectoral levels. For example, the US green stimulus generated increased demand for wind and solar com- ponents. The environmental beneï¬?t of wind and solar installation does not depend on the source of the components, but the stimulus effect does. As increased demand is met by imports from China, there are calls to restrict Chinese imports of these products. Again, a WTO-sanctioned agreement to exempt green products from trade restrictions would help remove this kind of pressure. It will probably take years for the US-China trade conflict to unwind. A resolution will require increased consumption (and higher wages) in China and responsible ï¬?scal policies in the US. These changes will be associated with an appreciation of China’s currency against the dollar and a reduction in its trade surplus. Academic economists should discourage trade measures, and especially those that contravene WTO law. We should also discourage arguments that trade restrictions are justiï¬?ed by our partners’ past green industrial policies, an argument sometimes made in the context of the US- Brazil dispute. The theory of public policy provides a number of insights about the ap- plication of green industrial policy. It is widely thought that green industrial policy, by making future abatement cheaper, is a substitute for current reduc- tions in pollution. In some circumstances that hypothesis makes sense, but the “Green Paradoxâ€? explains why it may be false in the case of fossil-based fuels such as oil. The price of these fuels includes scarcity rent, attributable to the ï¬?nite supply of currently known low cost sources of the fuel. That scarcity rent, and thus the equilibrium price of oil, falls if resource owners expect future competition from alternative energy sources to reduce future oil prices. In addition, the marginal damages associated with carbon stocks are likely to increase with the level of those stocks (i.e. damages are convex in the stock). Under these circumstances, the adoption of green industrial poli- cies, (possibly) creating future low cost alternatives to fossil fuels, increases rather than decreases the importance of current regulation of emissions. We also discuss a limitation of the Green Paradox as a basis for policy advice. As noted above, the cost of reducing emissions at a point in time depends on previous investments. A large literature in environmental economics com- pares the choice of taxes and quantity restrictions when regulators and ï¬?rms have asymmetric information about abatement cost shocks. When abate- 6 ment costs depend on previous investment, the asymmetry of information about current costs means that the role of industrial policy is sensitive to the choice of instruments. If the regulator uses an emissions quota (cap and trade), then ï¬?rms’ optimal investment decisions lead to the information- constrained ï¬?rst best level of investment. In that case (barring other consid- erations) there is no role for industrial policy, e.g. investment subsidies. If, however, the regulator uses an emissions tax, the tax creates an investment distortion. That distortion could be offset by means of an industrial policy. Thus, the importance of industrial policy in this setting depends on the type of regulation that is used to control pollution. The manner in which pollution is regulated can create coordination prob- lems at the industry level; those coordination problems can create a role for government intervention in investment decisions — a type of green industrial policy. Consider the case where non-strategic ï¬?rms make lumpy investment decisions that influence future abatement costs. For example, power sup- pliers’ choices are lumpy: they must choose among a small number of types of power plants, and cannot build a plant that is two-thirds of the way be- tween two alternatives. Investment in low carbon power plants lowers the future cost of maintaining emissions below a given ceiling. Optimal policy depends on both the beneï¬?t of emissions reduction and the cost of achieving that reduction. Because green investments affect future abatement costs, those investments also affect future optimal policy, e.g. the stringency of the future emissions ceiling. Rational ï¬?rms understand this relation. If individual ï¬?rms are small relative to the aggregate, their individual invest- ment decisions have negligible effect on the future social cost of abatement. Their individual investment decisions therefore have negligible effect on the future emissions ceiling, so it is rational for ï¬?rms to behave nonstrategically with respect to policymakers. However, the aggregate investment decisions do affect the future abatement costs, and therefore affect the future environ- mental policy. That future environmental policy affects the beneï¬?t, to an individual ï¬?rm, of the investment: the more stringent is the future policy, the more proï¬?table is the investment. Through this mechanism, the aggregate investment decisions affect the proï¬?tability of investment. In this setting, if regulators use a cap and trade policy, the level of investment is ï¬?rst best, and there is no need for industrial policy. However, if the regulator uses a pollution cap and forbids trade in permits, ï¬?rms play a coordination game at the investment stage. There are multiple equilibria to this game, creating a role for industrial policy that influences the level of investment. 7 Finally, we consider the ability of industrial policy to promote cooperation in solving global externalities. The reduction in abatement cost achieved by the application of green industrial policy might increase with the number of other countries that adopt this policy. For example, there may be increasing returns to scale, or network effects arising from this policy. In this setting, even if green industrial policy does promote cooperation, it is likely to do so to a limited extent. 2 Current green industrial policies There has been a surge in renewable energy investment over the last several years, spurred by more active renewables policy across the globe.2 Currently at least 119 countries have some sort of renewable energy policy in place at the national level, more than double the number of countries with such policies in 2005. Solar photovoltaics (PV), wind, solar water heating systems and biofuels have been growing at average rates ranging from 15% to nearly 50% annually. Approximately half of the new electric capacity added in 2010 came from renewable sources. Total investment in renewable energy reached $211 billion in 2010, a 32% increase from the previous year. Renewable energy, including hydropower, geothermal and biomass, provided approximately 16% of global energy use in 2009. Developing countries have recently overtaken developed countries in re- newable energy investment. For the second year in a row, China was the global leader in new investment in renewables, attracting $49 billion in 2010, more than two thirds of investment for developing countries. The United States ranked second in investment in the renewables sector, followed by Germany. Mandatory blending requirements are probably the most influential policy support for biofuels. Other important policy tools include direct subsidies, tax incentives, infrastructure development, low-cost ï¬?nancing, and R&D sup- port. Mandatory blending laws exist in 31 countries at the national level and 29 states and/or provinces. They take the form of either nationwide usage targets, blending standards on all fuels, or simply a required option at the gas station. Biofuel mandates work by ensuring producers that there will be a minimal level of demand for their products. In the absence of subsidies, 2 Much of the data from this section is taken from the Renewable Energy Policy Network for the 21st Century, REN21 (2009, 2011). 8 consumers bear some of the extra costs of this fuel. Unless demand is com- pletely inelastic, producers also bear some of the incidence, in the form of reduced rents. In 2010 the US was the world’s largest producer of biofuels, followed by Brazil; the two together accounted for 88% of the world’s total ethanol pro- duction. Sugar-based ethanol provided 41.5% of Brazil’s light duty transport fuel, and corn-based ethanol provided 4% of the US fuel consumption. Glob- ally, liquid biofuels provide about 2.7% of road transport fuels. The last few years have seen an increasing consolidation of the biofuel industry. Tradi- tional energy companies have been moving into this ï¬?eld and supply chains have become increasingly vertically integrated. The most common policy tool in promoting renewable electricity is the feed-in tariff, which is in place in 61 countries and 26 states/provinces. Al- though details differ among countries, this mechanism involves a guaranteed premium rate at which electricity is purchased from the generating source, as well as long-term supply contracts. The rate is usually speciï¬?c to the energy source and depends on the type of renewable power that the government is trying to promote. The feed-in tariff is often combined with renewable port- folio standards (RPS) enacted on a state or national level; these require that a certain percentage of the electricity sold comes from renewable sources. Long term targets for renewable energy — most oriented towards electricity generation — now exist in 98 countries and typically have targets of 10-30% within the next one to two decades. Amongst the renewables, PV has seen the fastest growth over the last few years. There were more than 5000 utility scale PV plants in 2010, up from 3200 in 2009. Both strong policy support and falling costs have contributed to this rapid growth. Between 2008 and 2010 the average cost of PV has fallen from approximately 4 to 1.4 dollars per watt of capacity, a 65% drop (IPCC 2011). However wind is still the dominant source of non-hydro renewable energy, with a global capacity of 198 gigawatts compared to 40 for solar PV. Wind has been growing by a fairly steady 25% per year average over the last ï¬?ve years. There has been a trend towards increased size of individual wind projects, driven by cost considerations such as grid infrastructure and licensing costs. China leads the world in wind capacity; Germany leads in solar PV. The production cost for wind-generated electricity is 5-9 cents/KwH and for solar-generated electricity 15-30 cents/KwH. However, the cost of a us- able kilowatt hour is higher for both. In many places wind is weakest during 9 the hours of peak energy demand, and because electricity storage is expen- sive there is overproduction of electricity when it is least needed. Solar has the advantage of being produced during peak hours, but production varies signiï¬?cantly as clouds move across the sun. To compensate for this varia- tion, extra energy must be kept in reserve, raising costs. Developing better electricity storage technologies would signiï¬?cantly lower the cost of using renewable electricity. Renewable energy policy is linked to development policy through incen- tives to provide electricity, hot water and cooking fuel to rural and poor communities. The challenges that rural communities face in connecting to grid infrastructure often make renewable energy sources the cheapest and most practical solution. Rural electriï¬?cation is generally heavily subsidized. Financing is rarely given directly to individual households; investment funds are typically allocated through private companies, community groups, NGOs and microï¬?nance organizations. There have been many campaigns in the last ten years to bring solar, small-scale hydro and wind electricity to rural communities and to promote the use of solar water heaters, biogas plants and other projects. Solar water heaters operate through the absorption of the sun’s heat through black pipes and tanks. Biogas plants (used often for cooking fuels) consist of an under- ground digester tank in which bacteria converts organic waste into methane, and a storage tank. Both systems are low-tech and easily implemented. 3 Trade issues Large-scale industrial policy usually has trade ramiï¬?cations, and green in- dustrial policy is no exception. Frictions between the US and Brazil and the US and China illustrate the kinds of trade issues that are likely to become more frequent, as countries adopt low carbon energy policies. China’s green industrial policy has underwritten the ï¬?xed costs of de- veloping an industry that produces components for solar and wind powered energy. The US and other of China’s trading partners have embarked on more modest green industrial policies. In the US in particular, these policies have been presented as a means of achieving three goals: increasing aggregate demand in order to increase employment, nurturing an infant industry that one day will provide economic beneï¬?ts, and reducing dependency on energy imports, chiefly in order to advance geopolitical goals. The environmental 10 beneï¬?t of reducing US dependence on fossil fuels is usually given less weight in arguing for these policies. The import of cheap components from China undermines the ï¬?rst two goals, but reduces the short run cost of lowering dependence on fossil fuels, and thereby beneï¬?ts the environment at least in the short run. There are at least three reasons why, by harming US and European green industries, low cost green imports from China might be inimical to long run environmental goals. These kinds of arguments cannot be dismissed out of hand, but they are not compelling. First, it may be important to develop green industries in several coun- tries in order to maintain future competition. China’s cheap exports of green products might be construed as dumping, a kind of predatory pricing intended to diminish future competition. The charge of predatory pricing is often made, but difficult to prove. This difficulty is especially great in new technology areas, where the price of a product is likely to be lower than average production costs because large sunk costs have not been amortized. Antidumping cases are brought in order to shield domestic ï¬?rms from foreign competition. Even if there were a sound economic reason for bringing an antidumping case, it will become harder to do so when China gains the WTO status of a market economy in a few years. The second environmental rationale for opposing low cost green imports is that policy is endogenous. The existence of an efficient green sector in the west will probably make it politically easier, in the US in particular, to adopt low carbon policies in the future. These policies create a market for green technology. Politicians may ï¬?nd manufacturers who want to ensure a market for their green products more persuasive than they ï¬?nd environmentalists. This argument cuts both ways. An efficient green sector in China also makes it politically easier for China to agree to future limits on carbon emissions. China has been an obstacle to achieving a climate agreement. Low carbon policies will have more support the cheaper these policies are to adopt. The third type of rationale is even more speculative. The economies of scale within a single country are probably limited. Aggregate costs of producing green components might be lower if several countries rather than only one or two have large green sectors. To the extent that this claim is true, it probably has more to do with the increased competition that arises when production is geographically dispersed than from (obscure) technical mechanisms. Trade restrictions are the wrong way to go about promoting this competition. 11 Just as environmental objectives were not the primary rationale for US green industrial policy, they also are not the basis for objecting to low cost green imports from China. Industrial policy can promote a trade surplus in the targeted sectors but would be unlikely to matter much to the overall trade balance. China’s green-sector trade surplus is part of a large aggregate balance of trade surplus, attributed to an undervalued exchange rate main- tained by China’s suppression of domestic consumption. This aggregate trade imbalance, rather than sectoral imbalances, is the source of US-China conflict. Unilateral action at the sectoral level, such as imposing tariffs against China’s green exports, is not likely to contribute to the resolution of the tension between China and its trading partners. The dispute between the US and Brazil regarding ethanol is another ex- ample of a green trade conflict. The GATT, Article III, enshrines the prin- ciple that “likeâ€? products produced at home and abroad receive the same treatment. In order to comply with this principle, the US ethanol tax credit (VEETC), which was eliminated at the end of 2011, was payable to both do- mestic and foreign ethanol suppliers. However, the US also imposed a tariff consisting of a 2.5% sales tax and $0.54 per gallon unit tax. Defenders of this tariff claim that it counteracted generous Brazilian subsidies and prevented Brazilian producers from taking advantage of domestic US subsidies — a right conferred by WTO membership. The idea that a tariff can be justiï¬?ed on the basis that it eliminates the ï¬?scal cost of ensuring that domestic subsidies are WTO-compliant, has an Alice in Wonderland quality. Moreover, the tariff was approximately 33% higher than the tax credit, and Brazil had already eliminated direct subsidies. The claim that trading partners’ past policies justify current trade restrictions seems unlikely to have a legal foundation; that kind of appeal to history would open the floodgates to protectionism. It would be a considerable achievement to accord green products, includ- ing biofuels, special status that exempts them from tariffs, or at least limits the tariffs that they face. China has (at least on paper) adopted such a policy unilaterally. The prospects for such an agreement under the aegis of WTO seem no better than the prospects for the Doha Round in general. If such an agreement were reached, there would still be the option (under Article XX of WTO) of restricting trade for environmental reasons, as was established by the WTO Appellate Body in the dispute involving shrimp and turtles in the late 1990s. That flexibility would be important to prevent, for example, deforestation in developing countries for the creation of biofuel plantations. 12 Largely in response to concern about the US ï¬?scal deï¬?cit, Congress al- lowed the ethanol tax credit to expire at the end of 2011, and with it the tariff (New York Times, January 1 2012). The US requirement to blend increasing amounts of ethanol in gasoline remains. The loss of the tax credit will there- fore not reduce US demand for ethanol, but will reduce blenders’ proï¬?ts and possibly increase the price to consumers. In recent years, Brazil imported small quantities of ethanol from the US, as a result of high world food de- mand for sugar and for competing crops. If these conditions continue, US corn producers will not face increased competition from Brazil, and US corn prices will not fall. In that situation, US farmers will not be harmed by the elimination of the subsidy and the tariff. However, if conditions change (e.g. world demand for sugar falls or Brazil’s sugar harvests increase), US corn farmers may face competition from Brazilian ethanol exports. In that situ- ation, corn producers’ demand for support, and concomitant trade frictions, might reappear.3 4 Lessons from the theory of public policy This section provides four examples of the lessons that the theory of public policy has for green industrial policy. First, there are circumstance under which green industrial policy and current environmental regulation are com- plements, rather than substitutes as is often thought. Second, when there is asymmetric information about abatement costs, and moreover those costs depend on previous investment decisions, the beneï¬?t of green industrial pol- icy depends on whether the regulator uses an emissions tax or cap-and-trade to control emissions. Third, when ï¬?rms have lumpy investment opportuni- ties that affect abatement costs, the type of policy used to control emissions 3 The tax credit VEETC transferred nearly $6 billion from the US treasury to the US blending industry in 2011 (New York Times), and over $30 billion since 2004 (Taxpayers for Common Sense). The industry group, the Renewable Fuels Association, stated “We are not seeking an extension of the ethanol blenders tax incentive. The industry is moving on. VEETC did what subsidies are supposed to do: help build an industry, ensure that it is stable and successful, and then fade away.â€? The Taxpayers for Common Sense states that companies and groups affiliated with the biofuels industry spent $31 million in lobbying in 2011 and are currently seeking federal funding for ethanol infrastructure and an expansion of the Alternative Fuels Tax Credit to subsidize E85. The Renewable Fuels Association appears to be “moving onâ€? by seeking a different form of subsidy, not eschewing the public trough. 13 can alter the nature of ï¬?rms’ interactions and thereby alter the role of green industrial policy. Fourth, positive externalities might cause industrial poli- cies in other countries to increase the beneï¬?t to a single country of using industrial policy; in this circumstance, industrial policies might alter the equilibrium amount of cooperation amongst nations. 4.1 The Green Paradox and industrial policy The price at which an owner of a nonrenewable resource is willing to sell the resource in the current period depends on the price that they believe the resource will command in the future. The difference between equilibrium price and the extraction costs (absent oligopoly power) is the scarcity rent. A higher expected future price increases the scarcity rent,making it more at- tractive to store the resource rather than sell (and consume) it today. Thus, higher expected future prices reduce the availability of supply in the current period, increasing the current equilibrium price and reducing current con- sumption. Conversely, lower expected future prices increase the availability of supply in the current period, decreasing the current equilibrium price and increasing the current equilibrium level of consumption. The Green Paradox suggests that this relation between future prices and current equilibrium consumption might cause green policies to backï¬?re, ac- tually worsening an environmental problem. Here we discuss the Green Paradox, emphasizing its relation to green industrial (as distinct from emis- sions) policies. We close the section by pointing out a complication that may limit the policy relevance of the Green Paradox. The chain of causation underlying the Green Paradox is that future poli- cies, e.g. carbon taxes that begin or increase in the future, lower the future producer price of oil, thereby lowering the scarcity rent and the equilibrium current price (Sinn 2008). Hoel (2008) and Winter (2011) note that green industrial policies have the same effect. Green industrial policies decrease the anticipated future cost of fossil fuel substitutes, decreasing anticipated future demand for fossil fuels and lowering the future fossil fuel price. These changes lower the scarcity rent and shift out the current supply function, increasing current equilibrium supply. To the extent that environmental damages are greater if a given amount of fossil fuels is consumed over a shorter rather than a longer period of time, green industrial policies can be counterproductive. In addition, the green industrial policies make current restrictions on carbon emissions more rather than less important. In this 14 sense, emissions policies and green industrial policies are likely to be com- plements rather than substitutes. To simplify the discussion, we ignore here the oligopoly rent. The usual static supply and demand model provides a useful way of thinking about the spot market equilibrium, except that in this model the supply function consists of both the marginal production costs and the scarcity rent. We can view this static model as representing supply and demand for the near future, e.g. the next 25 years. The relevance of the Green Paradox, in the current context, depends on the assumption that damages are convex in the amount of emissions that occur during a relatively short period, such as a quarter century. We explain the result taking that assumption as given, and then discuss its plausibility. A linear example helps to make the relation between green industrial policy and current emissions policy concrete. In this example, ï?‘ is the amount of the fuel brought to market during the 25 year period. The inverse demand is ï?? = ï?? − ï?‚ï?‘ and marginal cost (inclusive of scarcity rent) is ï?£ï?‘. With a linear relation between ï?‘ and GHG emissions and a linear marginal damage function, the marginal externality cost is also linear, ï?„ï?‘. ï?„ The optimal ad valorem tax in this static setting is ï‚¿ = ï?£+ ï?„ , the tax that causes the private tax-inclusive cost of bringing the commodity to market to equal the social cost. A green industrial policy lowers the scarcity rent and therefore lowers ï?£, increasing the optimal tax. Figure 1 illustrates this model.4 As the slope of the private marginal cost decreases from ï?£ to ï?£0  ï?£, the optimal unit tax decreases from the height of the line segment labelled ï?´ to the height of the line segment labelled ï?´0 . This comparative statics experiment illustrates the complementarity of the industrial policy and the policy that limits current emissions. The example also shows that the complementarity of the two types of policies depends on the assumed convexity of damages. If marginal damages were constant, the optimal tax is also constant, and independent of ï?£. If damages are concave in the stock of greenhouse gas (e.g., beyond a certain level, additional stocks cause little additional damage) current regulation and green industrial policies are substitutes. Most environmental economists think that damages are likely to be con- vex in stocks. We now turn to the plausibility of the assumption that dam- 4 The parameter values used to construct this ï¬?gure are ï?? = 10 ï?‚ = 1 ï?£ = 2 ï?£0 = 08 and ï?„ = 05 15 P 12 (c+D)Q cQ 10 Aâ€?BQ 8 6 t (c’+D)Q 4 t’ c’Q 2 0 0 1 2 3 4 5 6 Q Figure 1: A downward rotation of the supply curve increases the optimal tax ages are convex in ï?‘, which represent emissions, a flow. The usual challenge to that assumption is based on the fact that GHGs are a stock rather than a flow pollutant, and that emissions during a short period (relative to the half life of the stocks) are small relative to the stock size. This observation means that the change in damages arising from a change in emissions can be adequately approximated using a ï¬?rst order Taylor approximation to the damage function. If that claim is correct, then as noted above, green poli- cies that lower future prices and thus lower the current rent have no effect on the current optimal emissions tax. There are, however, reasons to doubt that a constant marginal damage provides a good approximation to the true damage function. First, there is considerable inertia in the climate system, which is as- sociated with the potential for abrupt changes that are not reversible on a policy-relevant time scale. Consider two scenarios; in one, there is an addi- tional 2ï?¸ units of emissions over the next 25 years, and in the second there is an additional ï?¸ units of emissions in both of the next two 25 year peri- ods. In the absence of inertia or the potential for abrupt changes, the world 100 years from now is better off in the ï¬?rst scenario. Moving the ï?¸ units of emissions from the second to the ï¬?rst quarter century gives it longer to decay, so the stock (and thus damages) 100 years from now is lower in the second scenario than in the ï¬?rst. The possibility of abrupt changes favors the second scenario, because there the stock of GHGs is (slightly) less likely to cross a threshold that triggers an abrupt change. If the second scenario is worse for the future than the ï¬?rst, then there is a plausible case that (the 16 present discounted value of) marginal damages are increasing in emissions.5 There is also a more subtle issue, that arises because here we use a static model to represent a situation that is actually dynamic. The optimal emis- ï?„ sions tax ( ï?£+ï?„ in the linear example above) depends on the relative slopes of marginal costs inclusive of scarcity rent, ï?£, and marginal environmental damage, ï?„. In a static model, these slopes have the same units, so their comparison is sensible. The two slopes do not have the same units in a dynamic setting, where one involves the slope of costs with respect to the flow of emissions, and the other involves the slope of costs with respect to the stock of GHGs.6 The intuition underlying the belief that marginal damages are nearly constant and that therefore the green industrial policy is unre- lated to the optimal level of the current emissions policy, is misleading. In a dynamic setting, other considerations, including the decay rate of the stock and the discount rate used to evaluate future costs and beneï¬?ts, also play an important role. The model underlying the Green Paradox, like all models, is built on a number of simplifying assumptions. One of these is almost certainly not correct, and it limits the policy-relevance of the Green Paradox. The Green Paradox, at least in its standard form, takes potential future supply as ex- ogenous, so the only effect of policy is to change the timing of emissions. Of course, potential future supply depends on investments in the exploration and development of oil ï¬?elds and the transportation infrastructure (pipelines) needed to bring the product to market. These investments sometimes in- volve enormous sunk costs: consider Canada’s tar sands and Brazil’s offshore oil deposits. The optimality of making these initial investments depends on expectations of future prices. Once the investments are made, marginal extraction costs are moderate and it is optimal to exploit the ï¬?elds even if the return does not cover the sunk cost. Green industrial policy or credible future emissions taxes both lower future oil prices, lowering the proï¬?tabil- ity of investing in these ï¬?elds, overturning or at least weakening the Green 5 Another way to think about this is that we are at a point where expected damages are extremely convex in the stock, so that a ï¬?rst order Taylor approximation to damages is not adequate for the purpose of formulating policy. 6 The same problem arises if we attempt to use intuition from the static model used to compare taxes and quotas under asymmetric information. In the static model, comparison of taxes and quanties depends on the relative magitude of the slope of marginal abatement cost and marginal damages. These slopes have different units in a dynamic setting, so comparison of policies is necessarily more complicated (Hoel and Karp 2002). 17 Paradox. 4.2 Instrument selection with information asymmetries The cost to ï¬?rms, and thus the cost to society, of reducing pollution at a point in time depends on ï¬?rms’ stock of “green capitalâ€? at that time. Because of the expense of rapidly changing the stock of this capital (i.e. convex adjustment costs), the cost of reducing pollution at a point in time depends on previous investment decisions. In world where the unpriced pollution externality is the sole distortion, a pollution tax or a cap and trade scheme that achieves the optimal level of pollution conditional on ï¬?rms’ level of green capital, also achieves the optimal level of investment in green capital. In this setting, with the optimal emissions tax or quota, there is no need for green industrial policies to encourage ï¬?rms to invest in green capital. This result arises because the tax or quota cause ï¬?rms to internalize the cost of emissions, and their cost-minimizing behavior then leads to socially optimal investment. Here we consider a situation where both taxes and cap and trade are po- litically feasible, and there are none of the investment-related market failures often used to motivate industrial policy. However, there is asymmetric in- formation: ï¬?rms receive a shock that effects their marginal abatement costs, and this shock is private information to the ï¬?rms. The regulator knows only the distribution of these shocks. If the regulator uses a cap-and-trade policy, there is no additional beneï¬?t of also having a green industrial policy that enables the regulator to influ- ence investment. With a cap-and-trade policy the regulator has a single instrument (the cap) and only one target (one “birdâ€?), the level of emis- sions. However, the regulator who uses a pollution tax has two targets (two “birdsâ€?), the level of emissions and the level of investment, but still only one instrument (here, the tax). When the regulator uses an emissions tax, the availability of a second policy, such as green industrial policy that enables the regulator to directly influence the ï¬?rms’ investment decision, improves welfare. The only exception to this claim occurs under quite special circum- stances: when the “two birdsâ€? under the tax are exactly lined up, so that it is as if they were a single bird (Karp and Zhang 2012). This claim is not obvious, and it does not (apparently) have a simple explanation. Section 4.2.1 provides a formal model that conï¬?rms the claim. Readers willing to accept the claim on faith can skip that section without loss of continuity. 18 Thus, with asymmetric information about abatement costs and the pos- sibility of investments that lower abatement costs, the value of green indus- trial policy depends on the form of policy used to control pollution. With asymmetric information regarding abatement cost shocks, the use of a pol- lution tax creates a role for green industrial policy to target investment. In contrast, when the regulator uses an emissions cap, the private investment decision is information-constrained optimal, and there is no role for green industrial policy. 4.2.1 Model details A one period, multi-stage model helps to explain why the tax and the cap and trade have different implications for green industrial policy. In the ï¬?rst stage, the regulator chooses either a tax or an emissions ceiling. Taking the current policy level — the tax or the quota — as given, ï¬?rms then make their investment decision, which affects their next-stage abatements costs. Nature then reveals a current cost shock (e.g. the price of labor or mate- rials needed to abate) and ï¬?rms then make their production and emissions decisions conditional on this information.7 Denote ï?¸ as emissions, ï?« as the level of green capital, and  as the ran- dom variable that affects abatement costs. Denote ï?ƒ (ï?¸ï€» ï?«ï€» ) as the sum of abatement costs plus environmental costs, and denote ï?‰ (ï?«) as the (convex) investment costs. The regulator’s objective is to minimize the expectation over  of the sum of these costs, ï?… [ï?ƒ (ï?¸ï€» ï?«ï€» ) + ï?‰ (ï?«)]. The regulator who uses a tax, ï‚¿ , understands the the ï¬?rm’s optimal in- vestment level depends on the tax, ï?« = ï?« ∗ (ï‚¿ ), but not on , because the ï¬?rm chooses investment before observing . In addition, the ï¬?rm’s emissions level depends on the tax, the level of green capital, and the realization of the 7 It may appear that the results of the model are sensitive to the timing of actions, in particular the assumption the regulator chooses the level of the policy before ï¬?rms make their investment decisions. However, this one period model is merely a device for thinking about a genuinely dynamic model. In that model, the stock of GHGs and of green capital at the beginning of a period depend on previous stocks and previous emissions and investment decisions. Even if, as in the one period example, the regulator moves before ï¬?rms within a period, ï¬?rms in the current period move before regulators in future periods. Thus, the game is really one of alternating rather than sequential moves. Changing the timing within a period does not qualitatively affect the model. 19 random variable, ï?¸ = ï?¸âˆ— (ï‚¿  ï?«ï€» ). The regulator’s ï¬?rst order condition is h ∗ ï‚¿ ï?« ∗  ) i ï?… ï?ƒï?¸ (ï?¸âˆ—  ï?« ∗  ) ï?€ï?¸ (ï?€ï‚¿ + h ∗ ï‚¿ ï?« ∗  ) i (1) ï?¤ï?«âˆ— ï?… (ï?ƒï?¸ (ï?¸âˆ—  ï?«âˆ—  ) + ï?‰ 0 (ï?«)) ï?€ï?¸ (ï?€ï?« ï?¤ï‚¿ = 0 The two terms on the left side of this equation are the two margins that the regulator would like control, i.e. the two “targetsâ€? that she would like to hit. The ï¬?rst term is the marginal (abatement plus environmental) costs resulting from a marginal increase in emissions due to a change in the tax, holding the level of investment constant. The second term is the change in marginal cost resulting from a change in investment, due to a change in the tax. A regulator who has two policies, an emissions tax and an investment tax/subsidy can set each of these margins equal to zero. In contrast, a regulator with a single policy, here the emissions tax, can only set the sum of the two margins equal to 0. Except in very special circumstances, when the two targets are perfectly lined up, having a second policy instrument increases the regulator’s expected payoff. Now consider the situation of a regulator who uses an emissions cap, ï?˜ . We assume that the environmental problem is great enough that the cap is binding for all realizations of . In this case, the regulator controls emissions directly, rather than merely influencing emissions (as under the tax). The ï¬?rst order condition for the regulator’s problem is now ∙ ¸ ∗ ∗ 0 ï?¤ï?«âˆ— ï?… [ï?ƒï?¸ (ï?˜ï€» ï?«  )] + ï?… (ï?ƒï?« (ï?˜ï€» ï?«  ) + ï?‰ (ï?« )) = 0 (2) ï?¤ï?˜ The important difference between the two ï¬?rst order conditions is that the emissions decision depends on the realization of  under the tax, in equa- tion (1); under a cap, the regulator chooses emissions, so ï?¸ = ï?˜ , which is independent of , in equation (2). We now decompose the total cost associated with emissions: ï?ƒ (ï?¸ï€» ï?«ï€» ) = ï??(ï?¸ï€» ï?«ï€» ) + ï?„(ï?¸) where ï?? is the ï¬?rm’s abatement cost (which depends on the allowable level of emissions, the stock of green capital, and the cost shock) and ï?„ is the social damage of emissions, which depends only on the level of emissions. Using this deï¬?nition, we rewrite the ï¬?rst order condition (2) as ï?¤ï?«âˆ— ï?… [ï??ï?¸ (ï?˜ï€» ï?«âˆ—  ) + ï?„ï?¸ (ï?˜ )] + ï?… (ï??ï?« (ï?˜ï€» ï?«âˆ—  ) + ï?‰ 0 (ï?« )) = 0 (3) ï?¤ï?˜ 20 The ï¬?rm’s optimal investment decision, under the quota, requires ï?… [ï??ï?« (ï?¸ï€» ï?«ï€» ) + ï?‰ 0 (ï?«)] = 0 so the second term in equation (3) is zero. Consequently, under the cap and trade, the regulator has a single margin, the ï¬?rst term in equation (3), together with a single instrument. 4.3 Coordination issues The design of environmental policy in general, and green industrial policy in particular, is rife with coordination issues. The private and social returns to one type of investment depend on the level of investment of a different type. China’s investment in wind farms without simultaneously creating a means of transporting the power to high demand areas illustrates a partic- ular kind of coordination failure. India’s lack of policy coordination across states, resulting in the failure to take advantage of arbitrage opportunities that would have reduced the overall cost of achieving a renewable resource target, illustrates a different kind of coordination failure. These kinds of coordination failure are easy to spot, at least ex post, which of course does not make them easy to solve. Because coordination failures come in so many forms, it is hard to imagine a general theory on the subject, or even a set of general guidelines that we can present to policymakers to deal with these issues. Probably the most help that theory can give is to develop simple models that illuminate generic policy issues. The purpose of this section is to illustrate the possibility that a second best policy can create coordination failures. Without theory, it might be difficult to detect the coordination failure, because we only observe the outcome that emerges, not other potential outcomes. However, once we are alert to the possibility of coordination failure, the policy remedy quickly becomes apparent. In the setting here, that remedy is likely to involve industrial policy that nudges an industry to adopt one type of investment strategy rather than another. The coordination problem discussed here arises when ï¬?rms have lumpy investment opportunities, and a regulator uses a second best policy to con- trol pollution (Karp 2008). A two-stage model suffices to explain the issue. In the ï¬?rst stage, non-strategic ï¬?rms decide whether to make a lumpy in- vestment that reduces their future average and marginal abatement costs. 21 For simplicity, assume that ï¬?rms are ex ante identical. In the second stage, the regulator observes the fraction of ï¬?rms that have made the investment, and is able to calculate the industry marginal abatement cost. Based on that calculation, the regulator chooses the optimal allowable level of per ï¬?rm emissions. The important assumptions are that the regulator conditions the emis- sions level on aggregate investment, and that all ï¬?rms receive the same emissions allowance. The time consistency problem associated with un- conditional policies motivates the ï¬?rst assumption; the examples of Brazil’s changing ethanol policy during the 1980s, the US changing subsidies to wind and solar power during the last quarter century, and recent attempts to over- turn putatively binding green legislation in California illustrate the respon- siveness of policy to current circumstances. Informational and incentive problems motivate the second assumption. The informational problem is that the regulator may not be able to identify the abatement costs of indi- vidual ï¬?rms. The incentive problem is that ï¬?rms that chose not to invest will subsequently have higher abatement costs than ï¬?rms that did invest. In the absence of trade, it is ex post optimal to give the high cost ï¬?rms a higher allowance of permits; the anticipation that the regular would behave in that way undermines ï¬?rms’ incentives to make the investment. For the purpose of comparing the effect of policy, consider the following two scenarios. In the ï¬?rst, the regulator chooses a market based policy, cap and trade. In the second, the regulator chooses a command and control policy: ï¬?rms are all restricted to the same level of emissions, despite the fact that the ï¬?rms that invested have lower marginal abatement costs. The optimal level of aggregate emissions permits differs in the two scenarios, but not in an obvious way.8 When ï¬?rms have binary choices, the aggregate level of investment is pro- portional to the fraction of ï¬?rms that invest. Because the policy scenario (market-based versus non-market-based) affects the equilibrium level of per- mits conditional on (most) levels of investment, the policy scenario also af- fects the equilibrium level of investment; that relation is our concern here. An individual ï¬?rm’s incentive to invest equals the proï¬?ts net of investment costs that the ï¬?rm expects to receive if it invests, minus the proï¬?ts that it 8 Even though the opportunity to trade reduces the total cost of achieving any level of abatement (compared to the no trade scenario), the marginal social cost of abatement might be higher or lower under trade. Thus, the optimal level of emissions can be higher or lower under the trade scenario even for the same level of industry investment. 22 expects to receive if it does not invest. The equilibrium price of permits is a decreasing function of aggregate investment. Therefore, when ï¬?rms antic- ipate that the regulator will allow trade, the decisions to invest are strategic substitutes: any ï¬?rm’s incentive to invest is a decreasing function of the frac- tion of other ï¬?rms that invest. In this setting, there is a unique equilibrium level of aggregate investment, which equals the socially optimal level. Here, the cap and trade policy achieves both the socially optimal level of invest- ment and the optimal level of emissions. The emissions policy is ï¬?rst best, and there is no need for an industrial policy to control investment. In contrast, when ï¬?rms anticipate that the regulator will choose a per-ï¬?rm emissions cap with no ability to trade, the investment decisions are strategic complements: a ï¬?rm’s incentive to invest is an increasing function of the fraction of other ï¬?rms that invest. The explanation is that an increase in aggregate investment lowers the industry marginal abatement cost and makes the optimal emissions policy more stringent (lowers the per ï¬?rm quota). As the allowable level of emissions fall, the advantage to a ï¬?rm of investing in the new technology increases. The no-trade emissions policy transforms the equilibrium problem into a coordination game. There are in general two equilibria, in which neither or all of the ï¬?rms invest. In both of these equilibria, ï¬?rms all make the same decisions, so they are all identical; there is no apparent cost of forbidding trade in emissions, because there are no opportunities for arbitrage. This absence of arbitrage opportunities is an endogenous outcome, however. Taking into account the endogeneity of investment, the prohibition against trade results in poten- tially large losses in welfare. In this scenario, a green industrial policy might involve either an investment tax or a subsidy. The constraint against trade causes the social planner’s problem to be convex in the fraction of ï¬?rms that invest, so having either all ï¬?rms or no ï¬?rms invest is constrained socially op- timal. Both of these outcomes are equilibria in the absence of an investment policy; an investment (or an alternative) policy is needed to induce ï¬?rms to coordinate on the constrained socially optimal equilibrium. This example, like the one in Section 4.2, shows that the role of green industrial policy depends on the other policies that are used to control emis- sions. With market based policies (here, cap and trade), green industrial policies have no role; with command and control policies (here, cap without trade) they are potentially important. 23 4.4 Incentives to cooperate There may be increasing returns to scale in green policies. The beneï¬?t to any country of adopting such policies may increase with the number of other countries that adopt them, due to traditional economies of scale, larger markets for their green products, the increased ability to take advantage of re- turns to specialization, or positive spillovers from network effects. Countries are (plausibly) more likely to adopt green policies if they join an interna- tional environmental agreement (IEA) that promotes or mandates emissions reduction. To the extent that the kinds of international positive spillovers from green industrial policies mentioned above exist, it might seem that they would increase the incentive to cooperate in an IEA. A model that has been widely used to study the formation of IEAs helps to understand why that conjecture might be wrong. Barrett (2006) Barrett (2003)shows that lower abatement costs can de- crease the equilibrium number of countries that decide to join an IEA. This conclusion arises in a two stage model where pollution is a public bad, so abatement is a public good. In the ï¬?rst stage, identical countries simulta- neously decide whether to join the IEA or to remain as a non-member. The outcome of this participation game is a noncooperative Nash equilibrium. In the second stage, non-members choose their level of abatement to maxi- mize their individual welfare, and the IEA chooses its level of abatement to maximize the joint welfare of members. Because members’ abatement levels take into account the positive beneï¬?ts on other members, members (typi- cally) abate more than non-members. All countries enjoy the same beneï¬?t from aggregate abatement, but members incur larger abatement costs and therefore have lower payoffs, compared to non-members. In deciding whether to join the IEA, in the ï¬?rst stage, countries under- stand the effect of their membership on the equilibrium abatement decisions of other members (and possibly also of non-members). It is individually ra- tional for a country to join the IEA if and only if, by joining, it increases the aggregate level of abatement by enough to compensate it for the additional cost that it incurs, as a member, of having to increase its own abatement. Trade in emissions permits, like green industrial policies, lowers expected abatement cost. A brief detour to consider the role of emissions trade may be instructive in thinking about the likely role of green industrial policies in promoting cooperation. The more countries that join the IEA, the greater is the probability that members will have different marginal abatement costs, 24 and that there will be opportunities to trade. The expected value of allowing international trade in permits increases with the number of countries in the agreement. Consequently, conditional on trade being allowed, an additional member creates two kinds of beneï¬?ts. First, the additional member increases the equilibrium amount of abatement, creating a direct beneï¬?t to all other countries. Second, the additional member increases the expected value of the option to trade, beneï¬?ting other members of the IEA. Karp (2009) shows that allowing trade reduces equilibrium participation, and is likely to reduce equilibrium global welfare. Membership by an addi- tional country weakly increases equilibrium abatement by more when trade is allowed, compared to the scenario without trade. However, one country leav- ing an IEA decreases equilibrium abatement by less when trade is allowed, compared to the scenario without trade. Therefore, trade has ambiguous incentives on the incentive to participate in the IEA, but in this model the effect opposing membership is larger. A less extreme parametric example (available on request) shows that the type of positive externality that is plausibly associated with green industrial policies, leads to a small net increase in the incentive to participate in an IEA. However, the magnitude of the effect is not likely to result in substantial additional membership. All of the models described above rely on speciï¬?c functional forms. In contrast, Karp and Simon (2011) provide a non-parametric analysis of this type of participation game. Depending on the manner in which green in- dustrial policy alters marginal abatement costs, the policy change might lead to large (or small) increases or decreases in equilibrium participation. Making marginal abatement cost more convex weakly reduces the equilib- rium membership size, and making marginal abatement costs more concave weakly increases membership size. The policy implication of this conclusion is largely negative, because we are unlikely to be able to determine, with any conï¬?dence, the effect of a policy change on the curvature of marginal abatement. The analysis is still useful, as a remedy against drawing strong (and unwarranted) policy conclusions based on parametric models. 5 Conclusion The debate concerning green industrial policy shares many of the features of the debate about industrial policy in general. Neither debate is likely to be 25 resolved either by theoretical or empirical work, but research is nevertheless valuable in helping us to think more clearly about these questions. Despite the considerable overlap, the debate about green industrial policy has some elements that are either absent or less important in the more gen- eral debate. The success of a general industrial policy depends on the ability of the targeted industry to meet market challenges that are largely exoge- nous to the policy. In contrast, the proï¬?tability of an industry targeted by green industrial policy depends to a great extent on the type and magnitude of environmental policy that will be used in the future, e.g. the size of the carbon tax or the stringency of the emissions ceiling. Current policymakers cannot commit to future policy; all agents understand that future policy is likely to be conditioned on circumstances that prevail in the future. Cur- rent investments are important determinants of future abatement costs, and those future abatement costs are important determinants of future policy. The future policy affects the proï¬?tability of current investments. The en- dogeneity of future policy and the inability of current policymakers to make binding commitments regarding future policy, create a rationale for green investment policy. Green industrial policy provides a means of sharing the policy-induced risk, and also of influencing future policy. Proponents of industrial policy often agree that governments should avoid trying to pick winners; but it is hard to envision an industrial policy that does not do exactly that. However useful it would be to have studies that categorize as successes or failures previous attempts at green industrial pol- icy, there are conceptual and measurement problems that would make such studies quite subjective. Green industrial policy, broadly understood, is prevalent in both devel- oping and developed countries. There have been rapid increases in the capacity of green energy sources, and corresponding reductions in costs, but these energy sources are still only a small component of total energy supplies. Green industrial policies are associated with, but may not be the cause of trade disputes. The Brazil-US ethanol dispute is a “typicalâ€? trade disputes and can be dealt with using the WTO dispute settlement mechanism. The US-China trade dispute is “fundamentalâ€?, rather than caused by sectoral policies. However, the underlying friction sometimes manifests as sectoral disputes. These should be dealt with using the WTO mechanism, recog- nizing that the fundamental causes of the friction will take years to resolve. Exempting green products from tariffs, or at least the recognizing that bio- fuels are industrial rather than agricultural products is desirable but unlikely 26 to happen. The breadth of issues arising from green industrial policy makes it unlikely that anything resembling a general theory for that policy can be constructed. Nevertheless, the theory of public policy provides many lessons for green industrial policy. Four examples illustrate the kinds of lessons that we might expect. If the potential stock of fossil fuels were exogenous, then green industrial policies might be a complement to, not a substitute for current environmen- tal regulation. This policy conclusion is weakened or even overturned to the extent that green industrial policy affects exploration and development decisions, and thereby affects the stock of fossil fuels. With asymmetric information about ï¬?rms’ abatement costs, in a setting where the investment in green capital affects abatement costs, the use of emissions taxes creates a rationale for investment policies, whereas the use of a ceiling on emissions does not. When ï¬?rms make lumpy investment decisions that affect their abatement costs, the form of future emissions policy (cap-and-trade or cap- without-trade) determines whether their investment decisions are strategic substitutes or complements. In the latter case there are multiple investment equilibria, creating a role for industrial policy that selects the equilibrium level of investment. If economies of scale or network affects cause the av- erage beneï¬?t of green industrial policy to rise with the number of countries that employ the policy, then the green industrial policy might promote in- ternational cooperation in reducing a global bad. Examples cast doubt on this theoretical possibility, but a more general analysis shows why we should be cautious about deriving general conclusions from parametric examples. 27 References Barrett, S. (2003): Environment and Statecraft. Oxford University Press. Chang, H.-J. (2006): “Industrial policy in East Asia - lessons for Europe,â€? Discussion paper, European Investment Bank, Economic and Financial Studies. Harrison, A., and A. Rodriguez-Clare (2010): “Trade, foreign in- vestment and industrial policy for developing countries,â€? in Handbookd of Development Economics, volume 5, ed. by D. Rodrik, and M. Rosenzweig, pp. 4039 —4214. North Holland. Hoel, M. (2008): “Bush meets Hotelling: the effects of improved renew- able energy technology on greenhouse gas emissions,â€? University of Oslo Working Paper. Hoel, M., and L. Karp (2002): “Taxes Versus Quotas for a Stock Pollu- tant,â€? Resource and Energy Economics, 24, 367 — 384. IPCC (2011): “Special Report on Renewable Energy Sources and Climate Change Mitigation, Summary for Policy Makers,â€? Special Report. Karp, L. (2008): “Correct (and misleading) arguments for using market-based pollution control technologies,â€? http://are.berkeley.edu/ karp/marketbasedJuly08.pdf. (2009): “Incentives to join international environmental agreements with permit trading and safety valves,â€? Giannini Foundation working pa- per. Karp, L., and L. Simon (2011): “Participation games and international en- vironmental agreements: a non-parametric model,â€? Giannini Foundation working paper. Karp, L., and J. Zhang (2012): “Taxes Versus Quantities for a Stock Pol- lutant with Endogenous Abatement Costs and Asymmetric Information,â€? Economic Theory, 49, 371 — 409. Lin, J. (2010): “Industrial policy comes out of the cold,â€? http://www.project- syndicate.org/commentary/lin2/English . 28 REN21 (2009): “Recommendations for Improving the Effectiveness of Re- newable Energy Policies in China,â€? Renewable Energy Policy Network for the 21st Century,Special Report. (2011): “Global Status Report 2011,â€? Renewable Energy Policy Network for the 21st Century, Special Report. Rodrik, D. (2009): “Industrial Policy: Don’T Ask Why, Ask How,â€? Middle East Development Journal (MEDJ), 1, 1—29. Sinn, H. (2008): “Public policies against global warming: a supply side approach,â€? International Tax and Public Finance, 15, 360—94. Winter, R. A. (2011): “Innovation and the dynamics of global warming,â€? Sauder School of Business Working Paper, UBC. 29