Top-down vs. Bottom-up Models

[ Home ] [ Up ]

 AggregationEndogenizationTechnologyIntangible costs / market barriers

Introduction

This section discusses some of the major influences that account for much of the discrepancy in predicted costs of mitigation.  Economic simulation models used to predict mitigation costs can be divided into two general categories – top-down models and bottom-up models.  Over time these two general model types are converging and adopting characteristics of one another, but for the sake of assessing currently available cost studies, it is important to distinguish between the two model types and identify the predominate characteristics of each.  The costs reported by the studies vary significantly.  For example, many bottom-up studies that estimate mitigation costs for the United States identify significant reductions from baseline emissions that can be achieved with zero net average costs [1].  The majority of the cost estimates for reduction in U.S. CO₂ emissions by year 2010, a near-term and thus more costly goal, range between –0.6% and +0.5% of Gross National Product (GNP) [1].  On the other hand, the majority of top-down studies exploring similar reductions of CO₂ from the baseline project costs ranging from a 0.6 to a 4.0% reduction of GNP [1].  The remaining discussion identifies some major differences between the two types of models that explain these discrepancies.

Level of Aggregation

Top-down models are aggregate models of the whole economy that analyze behavior based on economic indices of prices and elasticities [1, 7, 12]. The focus is on financial flows across the whole economy.  These models assess the interaction and feedback between energy policies and the macroeconomic performance of the economy [7].  The models include details about economic activity, such as consumer demand but do not generally look in detail at energy consumption. 

Bottom-up analyses, on the other hand, are disaggregated, looking at energy consumption in detail and examining the technological options for energy saving and fuel-switching, especially on the consuming side of the economy [1, 12, 7].  Analysts use information on available technologies and efficiency information on existing technologies to model the direct costs and benefits of incremental investments in efficiency and fuel-switching [7].  The information from individual sectors is then aggregated to calculate an overall cost of CO₂ emissions reductions [12].  This gets to the issue of whether or not no-regrets policy options exist, which will be discussed in detail later.

Endogenization

The issue of endogenization is related to the difference in the level of aggregation, in that it is easier to endogenize behaviors in model equations at the highly aggregated level.  Bottom up models tend to be driven by exogenous scenarios of macroeconomic assumptions, which limits their ability to consider feedback effects [7].  It is difficult to link bottom-up models to macroeconomic models because the level of detail is much higher than in most macroeconomic models and the bottom-up models are run in terms of energy units, whereas the econometric models operate in monetary units [7].  Top-down models on the other hand often include variables that account for macroeconomic factors and are easier to link to macroeconomic modes to obtain feedback results [7].  The significance of this discussion is that models that endogenize behavior are able to predict actual outcomes, whereas models that exogenize behavior are better suited to simulate effects of changes in historical patterns [1].  Climate change models are forward-looking and thus, more useful if they are able to predict outcomes. 

How technology is represented

The third important difference between top-down and bottom-up models is how technology is represented.  The bottom-up models capture technology in the engineering sense and identify a given technique with a given technical performance at a given direct cost [1].  The technology term in macroeconomic models is measured by “the share of the purchase of a given input in intermediate consumption (e.g., steel products to make cars) and by the allocation of the sales revenue among the cost of intermediary inputs, returns to labor, and returns to capital” [1]. 

Methodologically, the bottom-up modelers explicitly represent the evolution of technologies through aggregate ratios of energy intensity based on exogenous engineering or economic studies, or they attempt to endogenize the evolution of technologies by using information on the efficiencies and breakdown of existing equipment and the likelihood of replacement, retirement, or retrofit [10].  Top-down modelers generally tend to rely on two aggregate parameters, the autonomous energy efficiency index (AEEI) and the elasticity of substitution (ESUB) [1, 10].  The AEEI is a function of time and suggests the rate at which new technologies will penetrate the economy and change the energy intensity of the economy [1].  It is a proxy for all the non-price factors that affect the tendencies of technology change, which in turn affects the rate at which the aggregate energy intensity of the economy will change [7].  The ESUB is a function of the relative price of inputs and allows for a measurement of the degree to which capital or labor can be substituted for energy [1, 7].  The problem with these variables is that it is difficult to capture the appropriate values in historical data, given the multiple factors that influence changes in energy consumption over time.

Jaccard and Bailie make another interesting point about bottom-up cost estimates when discussing the methodology used in a residential sector bottom-up simulation model they ran.   When calculating the cost of a technology, the cost is the life-cycle cost at the societal discount rate [7].  The standard bottom-up modeling approach is to first determine the life-cycle costs of different technologies at a given discount rate and then to choose the technology mix based on those costs.  A more realistic behavioral assumption used by the ISTUM-R model bases the technology mix on realistic behavior patterns and lets the private cost perspective determine the technology choices.  It then uses that technology mix to calculate abatement costs from society’s perspective [7].  This cost may be positive or negative, depending on whether incentive existed to lead consumers to purchase goods with higher or lower lifetime costs. 

Assumptions about intangible costs and market barriers

The final major distinction that accounts for the discrepancies in cost estimations between these two types of models is their assumptions about consumer surplus, intangible costs and market barriers.  Many top-down economists tend to rely more heavily on the AEEI and ESUB variables than on independent predictions made regarding changes in consumption behavior [1, 7, 10].  These economists recognize that intangible costs and benefits can affect consumption decisions and thus they assume that if a technology that is cost-effective fails to penetrate the market, it is because of these intangible costs and benefits.  Thus, they argue that the AEEI and ESUB are bettor predictors of consumption behavior because they are derived from historical data and are therefore based on actual market outcomes [1, 7].

Bottom-up modelers, on the other hand, see economic potential for society to adopt new technologies more quickly than in the past, which would mean higher ESUB and AEEI values [1].  These analysts argue that it is unrealistic to assume that the intangible costs and benefits are always the exact magnitude necessary to explain the divergence between technological analysis and observed market outcomes [7].  According to these modelers, market barriers and imperfections account for the difference in predicted energy efficiency expenditures and daily market decisions.  Examples include lack of information about efficient choices combined with high transactions costs associated with learning about efficiencies; lack of access to capital for efficiency investments; and risk aversion [7].  Bottom-up proponents argue that efforts to reduce these barriers will lead to energy efficiency measures and low or negative cost CO₂ emission reductions [7].