ABSTRACT
Despite the importance of the questions raised by modeling the long-term future, one still must ask whether modeling long-term outcomes is worthwhile if answers to those questions are elusive – especially if they remain elusive over decades. George Box said, ‘all models are wrong, but some are useful’ (p. 424) (Box & Draper, 1987). Despite its multiple contributions, Nordhaus’ model’s predictions about our current electricity system over predict use of nuclear power; what makes it useful? Some contemporary discussants of Nordhaus’ paper had a clear perspective, with William Poole questioning the reliability of the model, specifically because of the uncertainty surrounding technological change. James Duesenberry gave the example that oil firms limit their planning to short time horizons because technological change makes planning beyond that ‘ineffective’. Moreover, the record of retrospective analyses of past predictions in energy is quite damning, including those of energy demand (Craig, Gadgil, & Koomey, 2002; Winebrake & Sakva, 2006), supply technologies (Hultman & Koomey, 2007), and end-use devices (Dale, Antinori, McNeil, McMahon, & Sydny Fujita, 2009). We know from elicitation studies that experts are typically over-confident in their predictions, that is, they underestimate the appropriate confidence interval around their predictions (Lin & Bier, 2008; Speirs-Bridge et al., 2010). In the realm of climate change, where Nordhaus has been a prolific contributor, concerns about the utility of models have also been raised (Cullenward, Schipper, Sudarshan, & Howarth, 2011; Rosen & Guenther, 2014).
