Datasets usually provide raw data for analysis. This raw data often comes in spreadsheet form, but can be any collection of data, on which analysis can be performed.
Table 1. Descriptive statistics. (Notes: the R&D rec. and the R&D high variables are dummy variables equal to 1 if the associated expert's cost estimate refers to the recommended and high R&D scenarios, respectively. The average values of the variables therefore represent the share of cost estimate referring to that specific R&D scenario within our sample.)
Characterization of the anticipated performance of energy technologies to inform policy decisions increasingly relies on expert elicitation. Knowledge about how elicitation design factors impact the probabilistic estimates emerging from these studies is, however, scarce. We focus on nuclear power, a large-scale low-carbon power option, for which future cost estimates are important for the design of energy policies and climate change mitigation efforts. We use data from three elicitations in the USA and in Europe and assess the role of government research, development, and demonstration (RD&D) investments on expected nuclear costs in 2030. We show that controlling for expert, technology, and design characteristics increases experts' implied public RD&D elasticity of expected costs by 25%. Public sector and industry experts' cost expectations are 14% and 32% higher, respectively than academics. US experts are more optimistic than their EU counterparts, with median expected costs 22% lower. On average, a doubling of public RD&D is expected to result in an 8% cost reduction, but the uncertainty is large. The difference between the 90th and 10th percentile estimates is on average 58% of the experts' median estimates. Public RD&D investments do not affect uncertainty ranges, but US experts are less confident about costs than Europeans.