زمان بندی سیاست های نوآوری در هنگام انتشار کربن محدود شده: تجزیه و تحلیل اعمال شده تعادل عمومی

This paper studies the timing of subsidies for emissions-saving research and development (R&D) and how innovation policy is influenced by a carbon tax. We develop a dynamic computable general equilibrium (CGE) model with both general R&D and specific emissions-saving R&D. We find two results that are important when subsidizing emissions-saving R&D in order to target inefficiencies in the research markets. First, the welfare gain from subsidies is larger when the carbon tax is high. This is because a high carbon tax raises the social value of the emissions-saving technology and that this increase in value is not fully appropriated by the private firms. Secondly, the welfare gain is greater when there is a falling time profile of the rate of subsidies for emissions-saving R&D, rather than a constant or increasing profile. The reason is that knowledge spillovers are larger in early periods.

In the coming decades, industrialized countries will have to face large reductions in emissions of greenhouse gases (GHGs) to curb anthropogenic interference with the global climate system. One expects that a variety of policy measures will be needed to achieve the cuts in GHGs. Most likely these will include a cap-and-trade system that gives a price to carbon emissions, like the European Trading Scheme, and a policy to improve clean energy technology like power production with carbon capture and storage (CCS). The most cost-efficient single policy to reduce emissions is an environmental policy that directly targets emissions, such as a tax or cap-and-trade system.1 However, in the presence of induced technological change and market failures in R&D, a combination of environmental and innovation policies may be more cost effective.2Jaffe et al. (2005) argue that market failures associated with environmental pollution interact with market failures associated with the innovation and diffusion of new technologies. These combined market failures provide a strong rationale for a portfolio of public policies to foster emissions reductions as well as the development and adoption of environmentally beneficial technology. Key arguments for subsidizing innovation activities are external knowledge spillovers from previous R&D and love of capital variety in demand, combined with inefficiencies arising from imperfect competition in the capital variety market.3 The market inefficiencies imply that the private returns from R&D are lower than the social returns, which leads to underinvestment in R&D. This underinvestment is the rationale for subsidizing R&D activity. There are several papers with R&D driven technological change that include specific policies to target inefficiencies in innovation markets in an environmental context.4 A typical policy is providing a constant subsidy for R&D. However, the optimal subsidy rate would not necessarily be constant over time. Hart (2008) finds that the gap between social and private returns from R&D investments may vary across time along a transition path. He implements optimal second-best carbon taxes, which may be higher than the Pigouvian level outside the balanced growth path in order to encourage investment in emissions-saving technology at the expense of ordinary production technology. Hart (2008) does not study the implication this has for first-best environmental R&D subsidies. Our contribution to the literature is two-fold. First, we analyze the timing and impact of emissions-saving R&D subsidies when future emissions are limited, e.g. by a binding international agreement implemented via a carbon tax. In particular, we ask two questions. First, we ask whether the welfare gains from emissions-saving R&D subsidies are larger when the future carbon tax is high. Raising the carbon tax causes an increase in the social value of R&D in emissions-saving technologies. This may only partly be captured by private firms and thus change the welfare gains from subsidizing R&D. Secondly, we ask if welfare gains are influenced by how emissions-saving R&D subsidies are distributed across time. If knowledge spillovers vary over time, there is reason to believe that R&D subsidies should vary over time. This timing problem of R&D subsidies has not received much attention in the literature, to our knowledge. Our second contribution to the literature is that we develop a computable general equilibrium (CGE) model with endogenous technological change based on the Romer (1990) approach with decreasing returns to new ideas (Jones, 1995). The model is dynamic with forward looking agents where technological change stems from new patents produced by profit-maximizing R&D firms. Emissions of GHGs are accounted for in the model, and a carbon tax influences the direction of technological change towards either general R&D or emissions-saving R&D. In addition, we take into consideration the high reliance of small, open economies on externally given international prices, competition, and growth. The case of a small, internationally exposed economy is exemplified by the Norwegian economy. Our model falls into the tradition of several CGE model developments with induced technological changes during the recent decade; see Gillingham et al. (2008) and Wing (2006) for overviews. However, most of the CGE models with induced technological change and carbon policies are based on ad hoc modelling of the innovation process. The model closest to ours is Otto et al. (2007).5 In their model for the Netherlands the innovation process is explicitly modelled as R&D based growth of the Romer (1990) type with imperfect competition in the markets for new technologies embodied in variety capital, as in our model. Contrary to ours, they treat productivity growth as entirely domestically driven. In this paper we find two results that are important when subsidizing emissions-saving R&D in order to target inefficiencies in the research markets. First, the welfare gain from emissions-saving R&D subsidies is larger when the carbon tax is high. This is because a high carbon tax increases the underinvestment in emissions-saving R&D. When the underinvestment increases a subsidy to R&D gives more welfare. This result indicates that there is an interaction between the carbon tax and the gains from subsidizing emissions-saving R&D. This is not opposed to the conventional wisdom of de-linking climate policy from R&D policy, i.e. that in a first-best world, there should be a carbon tax to target the environmental externality and an R&D subsidy to correct for innovation externalities.6 Rather, the result means that the effect of innovation policy is influenced by the carbon tax, since the tax influences the market imperfections related to R&D. Greaker and Rosendahl (2008) find a similar result in a three-stage game between the government, polluting industries, and suppliers of abatement technology. They find that an R&D subsidy is a strategic complement to a stringent environmental policy. The reason is that when environmental policy is more stringent, there is an increase in inefficiencies arising from imperfect competition in the supply of abatement technology. However, they do not take into account intertemporal inefficiencies, e.g. knowledge spillovers. Secondly, we find that the welfare gains are greater when there is a falling time profile for the subsidy rate for emissions-saving R&D, rather than a constant or increasing rate, when the economy is subject to emissions restrictions. This means that it is better policy to subsidize emissions-saving R&D more heavily initially than to distribute policy incentives evenly across time. The reason is that the knowledge spillovers are greatest in early periods due to the decreasing returns to new knowledge in the R&D production. Gerlagh et al. (2009) also study the optimal timing of R&D subsidies; however, they study inefficiencies in the R&D market related to limited patent lifetime. In our model, patent lifetime is infinite and not the source of the underinvestment in R&D. Further, externalities from knowledge spillovers are not included in their study. They find that the externality related to finite patent lifetime makes the optimal R&D subsidy fall over time. The reason is that the value of abatement increases rapidly in the beginning as the carbon tax increases. The early innovators get a smaller share of the benefits from this increase than late innovators, since patent lifetime is finite. In another study, Kverndokk and Rosendahl (2007) find that optimal technology subsidies also decrease over time, since newly adopted technologies have higher spillovers than older technologies. Their technology externalities, however, come from learning effects, as opposed to R&D externalities in our paper. When learning effects are present, the technology has to be implemented in order to reduce the future cost of the technology. This is not the case in R&D-driven models, as the development of the technology can be separated from implementation. In such models, it is efficient to conduct R&D in early periods and implement the technology later, when costs are driven down. Goulder and Mathai (2000) show that the presence of R&D is an argument for delaying the implementation of abatement technology; the presence of “learning by doing”, on the other hand, may be an argument for immediate abatement action. Section 2 presents the main structure of the CGE model and the key market imperfections related to R&D. Section 3 describes model details and the quantification and simulation procedures.The policy effects of the emissions-saving R&D subsidies and sensitivity tests are presented and discussed in Section 4, while Section 5 concludes.

Facing higher costs of GHG emissions, an important policy question is how to stimulate emissions-saving R&D. Stimulation of innovation gives welfare gains, since private firms do not reap the full benefits of their R&D activities. An appropriate policy instrument to correct for the externalities in innovation is an R&D subsidy. We ask whether the benefits of such a subsidy are influenced by the carbon tax, and whether emissions-saving R&D should be subsidized more heavily in early periods. First, we find that the welfare gain from subsidizing emissions-saving R&D increases with the carbon tax. This means that when the price of carbon increases, e.g. through international carbon markets or domestic reduction programs, it is more important to have an active innovation policy to promote emissions-saving R&D. The reason is that a carbon tax raises the social value of the emissions-saving technology and that this increase in value is not fully appropriated by the private firms. Secondly, we find that the welfare gain is greatest from a falling time profile of subsidy rates for emissions-saving R&D, rather than a constant or increasing profile. This means that it is a better policy to take R&D action now than to distribute policy incentives evenly across time. The reason for this is that the knowledge spillovers are larger in early periods. Further research is needed to analyze the mechanisms that influence the interaction between the carbon tax and underinvestment in emissions-saving R&D, both theoretical and empirical. Specifically, it may be fruitful to research optimal subsidy profiles in theoretical models under different carbon tax scenarios, with a focus on the development of social and private rates of return from R&D. The maturity of different technologies may also be interesting to explore. It may be that new technologies have larger knowledge spillovers than older technologies, where most advances have already been exploited. This would give another argument for an active R&D policy toward new emissions-saving technologies. There is potential to add to the model many new features that are empirically significant and relevant to the effects of innovation policies. First of all, the modeling of knowledge spillovers may be made richer. Including spillovers between general technological development and the specific emissions-saving technological development could diminish the gains from early subsidies to emissions-saving R&D, as the technology would benefit from early increases in the general knowledge stock. Further, the assumptions about labor supply have crucial implications for innovation policies. The present model assumes one national labor market with exogenous, unaltered labor supply across all the policy alternatives. The welfare potential of innovation policy is restricted by limited resources, in particular the inflexible labor resources. Expansion of the R&D industries is likely to attract mainly high-skilled labor. In reality, therefore, the allocation effect of increasing innovation in one R&D industry would be to crowd out the other R&D industries even more than in the present framework.