Published on April 29th, 2018 | by Zachary Shahan0
Why Cleantech Forecasting Often Sucks, & Some Solutions (#Electrifying Webinar)
April 29th, 2018 by Zachary Shahan
I was in a meeting last year with a bunch of cleantech experts. One of them asked a top analyst in the room something akin to, “Hey, why does your forecast for EV adoption level off like this around 2030?” This was an aggressive adoption forecast among conventional forecasters and analysts, but not the kind of earth-shattering forecast you get from a Tony Seba or such.
The response was that when they looked at the market — the overall electric vehicle (EV) ecosystem — in detail and projected how things would change, they saw some potential bottlenecks to growth with regard to infrastructure and other matters (I forget if batteries were mentioned). The expert who posed the question, Dr. Sgouris Sgouridis, shook his head. He said, “No, no, no — this is too conservative. The growth will keep going up exponentially.” (Or something like that.)
Of course, I was intrigued by his comments and they matched the impression we’ve developed over time here on CleanTechnica — conventional forecasters (even the best ones) seem to repeatedly include assumptions in their models that underestimate cleantech adoption. The most accurate forecasts I’ve seen are from “radical” environmental organizations like Greenpeace and WWF. Apparently, they either have the best forecasters or are not blocked by certain assumptions or methodology that others have blocked others.
Sgouris started to explain that he had just finished a paper on fundamental methodological flaws in related forecasting. Here’s a detailed summary from the introduction of the paper he co-authored:
“A sustainable energy transition that eliminates atmospheric carbon emissions within a few decades is a critical component of climate change mitigation strategies. Identifying the optimal mix of energy resources to reach the carbon mitigation targets is one of the aims of integrated assessment models (IAMs), models that combine climate and economic considerations. IAMs study the co-evolution of the climate and economic systems several decades into the future. By comparing the costs of climate change against the costs of preventative actions, they intend to inform climate and energy policy. IAMs offer a long-term representation of the economic system and estimate the costs of reducing or substituting GHG emitting activities in order to comply with a desired level of GHG atmospheric concentration, known as the marginal abatement cost (MAC)—equivalent to the efficient carbon price (Kuik et al. 2009). This paper reviews the properties of the constant elasticity of substitution (CES) function and discusses how IAMs that use it—referred to as general equilibrium IAMs (GE-IAMs)—for modeling the energy system may generate results that are inconsistent with empirically observed dynamics of energy transitions and are biased against alternative energy resources. While such issues have been casually recognized by the modeling experts, as the quotes following indicate, no comprehensive critical review of them has been undertaken to help those not in the CES/IAM community evaluate these issues. Our paper intends to fill this gap and provide some additional suggestions.
“Barreto and Kemp (2008) consider that energy models fail to take into account the empirically solid observations of S-curve innovation and technology diffusion patterns, but they did not identify specifically the inconsistencies arising from this omission. Pearce and Weyant (2008) point out the great sensitivity of IAMs’ cost estimates on “structural characteristics and assumptions embedded in the model” even when the key assumptions are the same. Truong (2009) notes a potential flaw that can lead to overestimation of the MAC indicating that “the CES production function […] can have unclear biophysical implications […] imply[ing] a pattern of the output elasticity […] which can be inaccurate and unrealistic from a biophysical or technological viewpoint.” He also points to the need to separate technical from economic efficiency in monetary models as their confusion may lead to physically impossible results. Frei et al. (2003) note that “structural change, i.e. the ‘natural’ emerging and phasing out of technologies, is incompatible with the neoclassical concept of smooth (i.e. differentiable) substitution” that is assumed in a CES. Carrara and Marangoni (2017) recognizing that CES models “represent an implicit constraint to renewable penetration” quantitatively evaluate such effects. Pietzcker et al. (2017) attempted to address identified deficiencies in the representation of variable renewable sources in IAMs and noted that CES “create[s] a preference for base-year calibration shares […that] can lead to physically implausible aggregation.” Finally, Rosen and Guenther (2015) highlight that the complexity and uncertainty surrounding economic and technological evolution make it difficult to defend simplistic conclusions of economic costs drawn by IAMs as their assumptions do not capture the synergistic benefits of technology development, climate mitigation, and increased reliance on renewable sources.
“Building on these critiques, we investigate the structural issues in GE-IAMs, arising from a CES that can bias (i) the estimation and shape of the MAC and (ii) the technological composition of future energy scenarios. The second case includes the exaggeration of the energy efficiency potential and the longevity of the fossil fuel status quo with only minor variation in the technology shares of future energy supplies. We show that while the MAC is highly sensitive to the parameter values (elasticity of substitution, initial shares, and assumptions on economic costs), the technology composition is also strongly dependent on the model structure per se.”
Too deep in the weeds? Got you giddy? Either way, you should love today’s #Electrifying webinar with Sgouris.
In this #Electrifying webinar, Dr. Sgouris Sgouridis and CleanTechnica Director Zachary Shahan chat about EV forecasts, forecasting and modeling more generally, disruptive technology, S-curves, and more.
Posted by CleanTechnica on Sunday, April 29, 2018