سرمایه گذاری در برنامه های مخاطره آمیز تحقیق و توسعه در مواجهه با شرایط عدم قطعیت آب و هوا

We analyze how the socially optimal technology R&D investment changes with the risk-profile of the R&D program and with uncertainty about climate damages. We show that how technology is represented in the model is crucial to the results; and that uncertainty in damages interacts with uncertainty in the returns to R&D. We consider R&D that reduces the cost of abatement multiplicatively, and argue that this is a good representation of R&D into non-carbon technologies; and R&D that reduces the emissions-to-output ratio, and argue that this is a good representation of R&D into fossil fuel technologies. For R&D programs into non-carbon technologies, optimal investment is higher in riskier programs. Our empirical model indicates that the optimal investment in a risky program is about 3 1/2 times larger than in a program with certain returns. For R&D programs aimed at reducing emissions in fossil fuel based technologies, our results show that, qualitatively, investment is higher in less risky programs under most uncertain damage scenarios. Our empirical model shows, however, that the risk-profile of fossil fuel based R&D programs generally has little quantitative impact on optimal investment. The exception is that when the probability of a catastrophe inducing full abatement is very high, investment is about twice as high in risky programs compared to programs with certain returns.

Policy makers are concerned with limiting the future cost of climate change. The economics literature has focussed on the optimal abatement path (e.g. Baker, 2005, Gollier et al., 2000, Keller et al., 2004, Kolstad, 1996, Nordhaus and Boyer, 2000, Manne, 1996, Pizer, 1999, Ulph and Ulph, 1997 and Webster, 2002) and the relative merits of different abatement-related policy instruments for climate change or related environmental issues.1 Policy makers in the U.S., however, have shied away from any emissions policy, and instead have focussed on technology policy.2 In the face of uncertainty – about both eventual climate-related damages and technical success – it is unclear how much R&D is desirable and which categories of technologies should be targeted. Baker et al. (2006) analyzed how socially optimal investment in broad classes of technology R&D was impacted by uncertainty in the damages from climate change. In this paper, we extend their model, considering two categories of technologies, and analyze how the optimal technology R&D investment changes with the risk-profile of the R&D program as well as with uncertainty about climate damages. This paper is related to the literature on investment under uncertainty and stochastic dominance, as well as R&D portfolio problems. The classic investment under uncertainty problem considers how optimal investment is impacted by uncertainty in the environment, such as prices or demand (see Dixit and Pyndyck (1994); Caballero (1991). This literature has also considered R&D programs with uncertain returns and has shown that there may be an option value to investing even when the program has an overall negative expected value, and that this option value may increase in the riskiness of the project (Huchzermeier and Loch, 2001 and Roberts and Weitzman, 1981). Much of the stochastic dominance literature is focused on determining what kinds of stochastic shifts in the environment induce more investment (Athey, 2002). Another strand of the literature considers how the choice between risky prospects is related to the level of risk aversion (Gollier, 1995 and Jewitt, 1989). In this paper we combine exogenous uncertainty in climate damages with uncertainty in R&D, for multiple technologies, to determine how the riskiness of the R&D program impacts the optimal level of investment in that program; and how this is impacted by uncertainty in the damages from climate change. We recognize that different R&D programs can have different levels of risk. Some programs are primarily aimed at incremental improvements, and tend to be low-risk: a larger investment leads to larger incremental returns with a great deal of certainty. Other programs are aimed at achieving breakthroughs, and tend to be high risk: an increase in investment increases the probability of success. We define a breakthrough as technological change that will reduce the cost of abatement to near zero, for example a reduction in the cost of very low-carbon energy that makes it widely economically competitive with fossil fuel technologies, or a combination of efficiency gains and sequestration that would allow for near zero-emissions fossil fuel based energy.3 We model an investment in a risky R&D program as inducing a first order shift in the probability distribution over the possible outcomes of that program. We consider how an increase in the riskiness of a program impacts the optimal investment in a first order shift in the program. This problem differs from classic R&D portfolio problems (See Keefer et al., 2004 for a review) because the effect of technological change is not inherent in the technology alone, but must be understood through the technology's effect on the marginal abatement cost curve (MAC) as it interacts with the eventual damage curve. For example, incremental improvements (as opposed to a breakthrough) in solar energy will only have widespread impacts on the economy if climate damages turn out to be severe, thus inducing a high level of abatement. We present two alternative ways of modeling technical change – both common in the literature – and show that our results differ significantly for the two representations. For the first representation, technical change reduces the cost of abatement by a fixed percentage — it pivots the abatement cost curve down. We argue that this most closely represents investments into very low emissions technologies. For the second representation, technical change reduces the emissions-to-output ratio, pivoting the abatement cost curve to the right. We argue that this most closely represents investments into reducing emissions in conventional, fossil fuel based technologies. While the validity of these interpretations is an open empirical question, it is clear that the representation of technical change has serious impacts for the conclusions drawn. When R&D pivots the cost curve down by improving alternative, zero-emissions technologies, optimal investment is significantly higher in riskier programs. Our empirical model indicates that the optimal goal in a risky program is about View the MathML source times more ambitious than in a program with certain returns, and that optimal investment is more than 12 times higher. For R&D programs that reduce the emissions-output ratio, our results show that, qualitatively, investment is higher in less risky programs under most uncertain damage scenarios. Our empirical model shows, however, that the risk-profile of these R&D programs generally has very little quantitative impact on optimal investment. The only exception is that when the probability of full abatement is very high, investment is about twice as high in risky programs compared to programs with certain returns. Optimal investment in alternative R&D programs increases as the program gets riskier for the following reasons. Alternative energy is not widely competitive and is not likely to be after incremental improvements, unless climate damages are very high. Therefore, incremental improvements in alternative energy are likely to have very little impact on the cost of abatement, unless climate damages are high. Therefore, the downside of investing in a risky program as opposed to a certain program – that the program might fail, and society would lose out on the incremental improvements that could have been achieved by a less risky program – is limited. On the other hand, a breakthrough in alternative energy technologies would have a large impact on the economy, reducing abatement costs while leading to very high levels of abatement. Therefore, the upside of a risky program outweighs the downside, unless climate damages are very high. If damages are very high, the logic is reversed. High damages mean that alternative energy will be widely competitive, therefore incremental improvements will have a large impact and therefore the downside of a risky program is larger. The logic is again reversed for conventional technologies, which are competitive when damages are mild or moderate, but not when damages are high. Thus, when the probability of a catastrophe is low alternative energy investments increase and conventional energy investments decrease in the riskiness of the program. When, however, the probability of a catastrophe is very high, the results may be reversed. The rest of the paper is organized as follows: in the next section we present two representations of R&D programs with their interpretations. In Section 3 we present a diagrammatic analysis of how the different R&D programs impact the MAC, and thus, optimal abatement. In Section 4 we perform a simple analysis indicating when investment will tend to increase in risk. In Section 5 we present our computational model, based on the well-known DICE model, and our results. We perform sensitivity analysis over various damage scenarios, and then we perform sensitivity analysis over probability distributions over damages scenarios. We conclude in Section 6.

In this paper we set up a simple model with two contrasting assumptions about technical change. In the first case, we assume that investments in R&D will lead to a proportionate reduction in abatement costs, and we argue that these represent investments into alternative, non-carbon energy programs. We find that optimal investment is higher in risky R&D than in non-risky, except when there is a large chance of a catastrophe occurring. This is because these investments only have a significant payoff if something dramatic happens: either if technical change is large OR climate damages are severe. Alternative technologies are only found in niche markets right now: incremental improvements will incrementally increase the niches, but not have a huge impact on the economy. This means that, if damages are moderate, the only thing that will have a big impact is a breakthrough in the technology which will allow it to be widely competitive. Therefore, pursuing risky R&D (with a chance of a breakthrough as well as a chance of failure) makes sense. However, if damages are severe, society will use a much larger proportion of alternative, non-carbon technologies. In that case any improvements in these technologies will reverberate through society. When incremental improvements have a big impact risks are less worth taking. In the second case, we assume that investment in R&D will reduce the emissions-to-output ratio, and we argue this represents investments into reducing the emissions from fossil fuel technologies. Here, we get opposite results from those above: optimal investment is, quantitatively, largely independent of risk, and, qualitatively, higher in non-risky R&D than risky, except when the probability of a catastrophe is high. This case is different because R&D into fossil fuel technologies have a significant impact when damages are moderate, but little impact when damages are high. Since fossil fuel use is widespread, incremental improvements will have a widespread effect on the economy and a significant payoff, which make risk less worth taking. But, in the event of really serious damages, society will reduce the use of fossil fuel technologies, making incremental improvements in fossil fuel technologies much less important. In this case, the only thing that will help is a breakthrough (reducing emissions from fossil fuels to near zero), therefore risky R&D is worth pursuing. The bottom line is that the optimal investment in R&D depends crucially on the type of technology considered and the uncertainty in both damages and the R&D process. The policy implications of these results depend on beliefs about the chance of a catastrophe. In general, one justification for technology policy is the assumption that private actors tend to avoid high risk projects with far-future payoff dates (Clarke and Weyant, 2002). If the probability of a catastrophe is low (i.e. closer to the economists' view of the world in Nordhaus (1994)), an R&D portfolio aimed at reducing the costs of alternative technologies should optimally be tilted toward riskier projects. This means there may be a role for government involvement in R&D into low probability, game-changing alternative technologies — technologies that will be widely competitive with fossil technologies, if successful. On the other hand, if the probability of catastrophe is high, the government may have a role in supporting research aimed at developing a radical, emission-free fossil fuel based technology. This model is a considerable simplification from real-world technology decisions. We have considered two broad categories of technical change, represented by different shifts to the abatement cost curve. We argue that they are broadly related to classes of technologies, but work needs to be done connecting the impacts of specific technical change on the marginal cost of abatement. In fact, the key conclusion of this paper is that it matters how technology is modeled, and so research must start connecting potential real-life R&D projects with their representations in models. We have not considered the history of any particular technology, nor are we commenting on the likelihood of a breakthrough in any particular technology. Rather, we suggest that work needs to be done on assessing the future potential of technologies, whether fossil or non-fossil based. Additionally, we use a simple, two-period model. Thus, this work is not commenting on the option value that is created by considering sequential investments in R&D programs. On the one hand, there is a value to waiting to learn more about climate change damages (which would put a downward pressure on all R&D investments); on the other hand, there is a value to investing a small amount in a risky project in order to learn more about the probability of success (which would put upward pressure on risky technologies). This work is primarily considering the impact of the interaction between uncertainty in damages and uncertainty in R&D. We are looking at the overall optimal level of R&D investment in the economy in response to climate change. We are not considering ancillary benefits or costs from the technical change. The DICE model includes autonomous technical change, both in growth in production, and in a decrease in the emissions-output ratio. Thus, the R&D considered here is additional to the assumed autonomous technical change. Furthermore, the rationale for government technology policy is the gap between the overall optimal investment and what is provided by the private sector. We have begun to address the question of the overall optimal investment; an investigation of private sector investment is an important area of future research. Finally, this work assumes that future decisions will be optimal, implying that a carbon policy is implemented eventually, after more information is revealed. If this is not true and future abatement is inoptimally low, that will tend to favor riskier ARD and less risky CRD; if future abatement is inoptimally high, the result is reversed.