The Imaclim-R model: infrastructures, technical inertia and the costs of low carbon futures under imperfect foresight

Abstract

This paper analyzes the transition costs of moving towards a low carbon society when the second-best nature of the economy is accounted for. We emphasize the consequences on mitigation costs of considering the interplay between a) technical systems inertia, including slow infrastructure turnover in transportation and construction; and b) imperfect foresight influencing investment decisions. To this end, the hybrid general equilibrium modeling framework Imaclim-R is employed as it allows for transitory partial adjustments of the economy and captures their impact on the dynamics of economic growth. The modeling exercise quantitatively emphasizes the a) specific risks that the interplay between inertia and imperfect foresight leads to high macroeconomic costs of carbon abatement measures; b) opportunities of co-benefits from climate policies permitted by the correction of sub-optimalities in the reference scenarios. The article draws insights for the framing of future climate architectures by studying the role of measures that act complementarily to carbon pricing in the transport sector. In particular, reallocating public investment towards low-carbon transport infrastructure significantly reduces the overall macroeconomic costs of a given GHG stabilization target and even creates the room for long-term net economic benefits from climate policies.

Appendix: Numerical assumptions and variants of scenarios

1.1 A. Numerical assumption on oil and gas supply

The three crucial determinants of the ‘oil supply’ Nexus are the amount of ultimate resources (and their regional distribution), the inertia on capacity deployment and the decision of Middle-East producers acting as “swing producers”.

Most estimates of proved oil reserves converge around 2.2 Tbbl (BP 2011) including past production. To reflect controversies about the amount of reserves to be discovered, we adopt two assumptions for ultimate resources Q _∞ : 3.3 Tbbl and 3.8 Tbbl. The lower bound reflects a conservative assumption on resource additions, in line with estimates from the Association for the Study of Peak Oil (ASPO). The higher bound considers higher resource potentials, corresponding to median estimates by (USGS 2000; Greene et al. 2006; Rogner 1997).

The intensity of constraints on production growth due to geological constraints is captured by the slope parameter b. ¹⁸ For conventional oil, we adopt the econometric estimate from Rehrl and Friedrich (2006): b _C  = 0.061/year. Given uncertainty on large scale production of non-conventional oil, we consider either the same value than conventional oil, b _NC  = 0.061/year, or more pessimistic assumption of a slower deployment with b _NC  = 0.04/year. For Middle-East producers, we impose in addition a cap on the annual increase of production capacity, ΔCap _ME .

The deployment of production capacities in Middle-East countries is decided by the price objective p _obj . A benchmark for oil price setting is a continuous increase towards a medium-term stabilization around 80$/bbl, reflecting the progressive loss of influence of Middle-East producers. Given uncertainties, especially in the geopolitical context, we also consider the possibility that Middle-East producers are able to expand their production capacities to bring oil price at their pre-2004 level, 40$/bbl. This market flooding option is possible only for the more optimistic assumption on reserves. This exercise of the market power ends up when the finiteness of the resource forces a decline of production. For the sake of simplicity, we assume that it happens once a share sh _D of their reserves remains underground, and consider two values (50% or 25%) to reflect the uncertainties on the stock of resource in Middle-East countries.

Finally, the ‘gas supply’ nexus represents indexation of gas markets on oil markets with a 0.68 elasticity of gas to oil price, as calibrated on the World Energy Model (IEA 2007). But, in order to represent the possibility that gas scarcity triggers faster price increases, we consider an alternative where this indexation disappears when oil prices exceed a threshold level p _oil/gas (chosen at 80$/bbl). In this latter case, gas prices are driven by the increased margins for gas producers.

These numerical assumptions are grouped in three variants summarized in Table 3

Table 3 Numerical assumptions for the three variants on oil and gas supply

1.2 B. Numerical assumption on substitutes to oil

The ‘alternatives to oil’ Nexus considers two large-scale substitutes to oil for liquid fuels production: biofuels and Coal-To-Liquid.

The supply curves, S _bio (t,p) give biofuels production, given competition with oil, and are taken from IEA (2006). They assume maximum biofuels production at 14 EJ/year in 2030 and, thanks to technical progress, at 42 EJ/year in 2050. These assumptions are quite conservative with respect to recent estimates about biofuels potential (Chum et al. 2011, Figure 2.23(b)) and we introduce an alternative, more optimistic, assumption allowing 20 EJ/year in 2030 and 60EJ/year in 2050. The diffusion of biofuels is in addition submitted to the constraint of a time delay, Δt _bio , which captures inertia on the deployment of raw products (biomass) and of refining capacity.

Coal-To-Liquid is treated as a backstop technology, which enters the market as soon as liquid fuel selling price exceeds its total cost, p _CTL , including production processes and risk premium. This backstop technology is submitted to capacity constraints in the form of a delay Δt_CTL between investments and production. Given uncertainty on large-scale CTL production, we consider two possibilities, depending whether CTL is a mature technology (low threshold oil price at 120$/bbl and no inertia in the deployment) or it is submitted to constraints slowing down its deployment (high threshold oil price at 200$/bbl and significant time-lag in the deployment)

These numerical assumptions are grouped in two variants summarized in Table 4

Table 4 Numerical assumptions for the two variants of substitutes to oil

1.3 C. Numerical assumptions on demand-side technical change

The ‘Power generation’ Nexus represents investment choices in new power generation technologies according to a minimization of mean production costs. Technical change is then dependent upon the decrease of capital costs, along with the learning process controlled by technology-specific learning rates γ (it measures the percentage decrease of capital costs for each doubling of experience). Learning does not affect standard technologies due to saturation of experience, but potentially contributes to important costs decreases in more recent or prospective technologies, including wind energy and Carbon Capture and Storage. Due to uncertainties on the technical potentials of these technologies, we represent either fast learning through high learning rates (7% for wind vs 13% for CCS) or constrained learning with low learning rates (3% for wind vs 7% for CCS). Note that we consider lower learning rates for wind units than for CCS to represent that the former is a more mature technology, with less remaining progress potential.

In addition, the ‘Power generation’ Nexus represents the constraints that may affect the diffusion of carbon-free power plants by an exogenous maximum market share, with different dynamics for already existing and new technologies. In the former group, we explicitly represent Nuclear and Wind Energy and assume their maximum shares Sh _Nuke and Sh _W as constant-over-time. We adopt rather conservative assumptions on the long-term potential of Nuclear and consider a maximum market share at 40% to capture limitations for social acceptability reason (20% in a more constrained vision). For wind energy, we consider a benchmark case where it is limited to 15% of production to capture implicitly constraints imposed by intermittent production and additional integration costs at higher shares. This assumption is in line with the median estimate of the 164 global scenarios reviewed by the IPCC (Wiser et al. 2011, Figure 7.25). But, a growing body of work has evaluated higher levels of deployment, around 20% or more, provided that cost and policy factors are favourable. To treat this case, we also consider a higher limit on wind’s market share, at 25%. In the latter group, we consider Carbon Capture and Storage (CCS), and the maximum share Sh _CCS increases over time to represent its progressive deployment, ranging from zero at the starting year (t _0,CCS ) up to its long-term market potential Sh _max,CCS . During the early years, inertia limits the deployment of this new technology as captured by a slow increase of the maximum share during a ‘bottleneck period’ of length Δt _CCS, followed by an accelerated increase once the technology is mature.

In the ‘Industry and services’ Nexus, energy prices affect the selection of new production capacities but do not influence existing ones. This putty-clay assumption implies that changes in final energy use are dependent on their lifetime Δtind. This is an important variable, since it conditions both the turnover of productive capital (and hence the speed of technical change) and investments needs. We take 20 years as a benchmark case, whereas 30 years reflects a more constrained context on investment imposing delayed retirement of production capacities.

In the ‘Housing and Buildings’ Nexus, the baseline trends of energy consumption per square meter, α_res(t) are taken from outcomes of the POLES model. They feature a relative decrease of unitary energy demand in developed regions thanks to energy efficiency, while strong increases in developing countries are due to the access to energy services along with wealth increase. In addition, the energy mix is orientated towards electricity and gas at the expense of coal and oil. We consider also more energy-intensive pathways with proportionally higher unitary energy consumption due to lower efficiency gains (for technical constraints or lack of investments) and/or a more prominent access to energy services in developing countries.

In the ‘Freight transportation’ Nexus, the energy intensity of vehicles is driven by an exogenous trend μ _f (t) and a short-term fuel price elasticity ε _f to capture autonomous and endogenous energy efficiency gains as well as short-term modal shifts, respectively. The long-term price response of the fleet then results from the sequence of those short-term adjustments.

The ‘Passenger Transportation’ Nexus represents the crucial determinant of energy efficiency and modal choices. Energy efficiency in private transportation is mainly dependent on the constraints on the diffusion of Electric Vehicles (EV). They are captured by an exogenous maximum Sh on their market share, which ranges from zero in the first year (t _0,EV ) to Sh _max,EV as it achieves its long-term market potential. During the early years, inertia limits the deployment of this new technology, as captured by a slow increase in the maximum share during a ‘bottleneck period’ of length Δt _EV, followed by an accelerated increase once the technology has matured.

Modal allocation of mobility demand is affected by investments in infrastructure, which determine the relative efficiency of the different modes. Instead of the default assumption that investment is allocated proportionally to modal mobility demand, alternative decisions may trigger a re-allocation from road to low-carbon transportation infrastructure (public and rail transport for passengers and rail and water transport for freight).

These numerical assumptions are grouped in two variants summarized in Table 5.

Table 5 Numerical assumptions for the two variants of demand-side technological change