تحقیق و توسعه پیشرفته خورشیدی : ترکیبی از تجزیه و تحلیل اقتصادی با استنباط های کارشناسان برای اطلاع رسانی به سیاست آب و هوا

The relationship between R&D investments and technical change is inherently uncertain. In this paper we combine economics and decision analysis to incorporate the uncertainty of technical change into climate change policy analysis. We present the results of an expert elicitation on the prospects for technical change in advanced solar photovoltaics. We then use the results of the expert elicitations as inputs to the MiniCAM integrated assessment model, to derive probabilistic information about the impacts of R&D investments on the costs of emissions abatement.

In this paper we combine expert elicitations and economic modeling to analyze the potential for R&D into solar photovoltaics (PV) to impact climate change. When it comes to the question of what to do about climate change, the role of technical change is large. Estimates of the costs of control and of eventual damages both depend heavily on assumptions about technical change (Edenhofer et al., 2006 and Popp, 2006). In order to craft good climate change policies — whether they are emissions policies or technology policies — we need to understand how technical change responds to policy, and how emissions respond to technical change. Technical change can come through two channels — investment in R&D and learning by doing. We focus primarily on R&D, but our analysis of how improvements in technology will impact costs is also relevant to learning by doing. We note that a wide variety of policies can impact investment into R&D, from policies that directly allocate government funds to R&D, to R&D-tax incentives, to carbon taxes, to adoption incentives. For this paper, we focus on R&D investment directly, and leave the analysis of the government's role in R&D investment to future research. We focus on how R&D impacts technical change, and how technical change impacts the cost of reducing carbon emissions. Specifically, we study the impacts of technical change on the entire abatement cost curve, which measures the costs of abatement, defined as a reduction in greenhouse gas emissions, at each level of abatement between zero and 100%. We note two gaps in the current literature. First, there is very little work that directly addresses the fact that the results of investment in R&D are inherently uncertain. This topic is just starting to be studied, most notably by others in this special issue (Blanford, 2007, Bohringer and Rutherford, 2006 and Bosetti and Drouet, 2005). Second, there is virtually no work that discusses how particular technologies are likely to impact the abatement cost curve. Yet, for decisions made under uncertainty, it is the shape of the whole curve, and not just a point estimate, that determines results. Thus, we need to understand the impact of technology on the abatement cost curve, for many different levels of abatement. The difficulty here is that not all technical change is alike, and not all R&D programs are alike. Different types of technologies will impact the abatement cost curve in different ways. For example, an incremental improvement in a non-carbon transportation technology may have a very small effect on the cost of abating a small amount of emissions, because of infrastructure and network effects. If climate change damages turn out to be very severe, however, then even small improvements in non-carbon transportation technologies may be very important. On the other hand, consider improvements in coal-fired electricity generation. An incremental improvement is likely to have a large impact if damages are low and abatement is minor; but virtually no impact if damages are extreme and a no-carbon world is desirable. Another distinction between R&D programs is their levels of risk. Some programs provide a possibility of a breakthrough, but also a large chance of failure. Other programs are less risky, aiming only to improve the technology incrementally. In Baker et al. (2007) we described a general framework for quantifying the uncertainty in climate change technology R&D programs and their associated impacts on emissions. Here, we present an implementation of that framework, focusing on advanced solar technology. Fig. 1 illustrates the flow of the data in this framework; the actions placed within the box are discussed in this paper; the actions outside the box are applications of the outputs of this paper. The first step of the project, discussed in Section 2, is collecting probabilistic data on advanced solar PV technologies through expert elicitations. The products of the elicitations include explicit definitions of success for each technology, and probabilities of success for given funding trajectories. In Section 3 we determine how the technologies would impact the abatement cost curve, if they achieve success as defined. For this step we use MiniCAM, a technologically detailed Integrated Assessment Model (IAM), to determine the impact of each technology on the Marginal Abatement Cost (MAC) Curve. In Section 3.4 we discuss the parameterization of each technology's impact on the MAC. In Section 4, we combine the probabilities with the impacts on the MAC to derive multiple representations of the probabilistic impacts of R&D. As shown in the figure, these can then be combined with probability distributions over climate damages in technology policy models. We conclude in Section 5.

5. Conclusions We have collected data on the relationship between R&D investment and technological outcomes for advanced solar PVs; and implemented a framework for understanding the probabilistic impacts of R&D funding on the cost of abatement. We have presented information on this relationship in a number of ways, and discussed how it can be implemented in a variety of models. Subject to the limitations discussed below, this analysis leads to four conclusions. First, it appears that even very large advances in PVs will have a relatively small effect on the abatement cost curve. However, when these advances are combined with improvements in storage, the impacts are considerably more significant, and highly non-linear. That is, improvements in PVs or storage alone have little impact; a breakthrough in solar (to around 3 cents/kWh) when combined with low cost storage has considerable impact compared to a more moderate improvement to 5 cents/kWh. Even more striking, PVs with a cost of 5 cents/kWh with free storage have a larger impact than PVs with a cost of 2.9 cents/kWh but facing a capacity limit due to costly storage. This implies that capturing the interactions between technologies is crucial; and that capturing the impact on the cost of abatement goes beyond just improvements in cost. Second, if we focus on cost reduction in PVs alone (the 20% limit case), we see that these will tend to have the most significant impact (in terms of changing optimal abatement as well as reducing costs) at moderate levels of abatement, up to about a 20% reduction below our baseline, equivalent to stabilizing at an atmospheric CO2 level of about 550 parts per million. The reason these improvements to PVs do not have a significant impact on the MAC at high levels, is that solar is likely to be implemented up to the capacity limit at those high levels, even without the improvements. Thus, the improvements will lead to a cost reduction for the given amount of solar, but will not lead to a higher implementation of solar. All of these results depend on our assumptions about the other technologies available in the economy, most notably the availability (and acceptability) of nuclear; and the non-availability of carbon capture and storage. Third, there is significant disagreement among experts about the efficacy of R&D expenditures, especially on reducing the costs of manufacturing. This suggests directions for future work: a) facilitated interaction between experts such as a workshop setting, allowing experts to share information and come closer to agreement about underlying assumptions; b) making explicit assumptions about industry activity and funding; c) using a longer and higher funding trajectory, making it easier to separate differences of opinion about the relative maturity of the technologies from differences over their ultimate promise; and d) a more detailed analysis that looks at technologies at a finer level, incorporating judgments from scientists working on each component of the individual technologies. Additionally, we note that our probability assessments are based on just three experts. Studies have shown that the incremental benefit of adding another expert decreases significantly after 3–4 experts (Winkler and Clemen, 2004). Nevertheless, it is possible that our sample is not representative of the population of solar scientists as a whole. Another concern is that our project descriptions, particularly in the definitions of the budgets, are too limited. Thus, these results should be seen as preliminary. Finally, even with disagreement among the experts, some regularities appear. The order of investment is rather robust, with higher expected values for the “other inorganic” projects, followed by the lower risk organic project, with third generation somewhere in the mix. Thus, if we face a problem with a given budget, the resulting portfolio would be the same for each of the experts. At a theoretical level, an important finding from this study is that our analytic methodology appears viable. Expert elicitation, which has been powerfully applied in R&D portfolio management in various industries, can be applied even at the industry or public policy level by deriving the impact of success in terms of economic curves, in this case the MAC. This approach requires special attention to the definitions of technical success and the way that technical success is translated to impact in the economic model. This methodology provides information about the supply of technical change. This can be combined with information on the demand for (or benefits to) technical change in models of sequential decision making under uncertainty to determine robust climate change policies. We have generated MACs using MiniCAM, a particular integrated assessment model, with its own unique set of assumptions and foibles.8 Beyond this, the data we have collected can be used with other IAMs, to better confirm the qualitative and quantitative impact that advances in solar are likely to have on the MAC. Additionally, this is only one part of a much larger analysis. We are performing expert elicitations on a much larger group of energy technologies. We expect that analyzing the interactions between technologies will lead to broader and more robust results. By quantifying both technical uncertainty and the impact of potential progress on the MAC, we can facilitate economic analysis of investment in energy technology to mitigate climate change.