Marie Tamba, University of Strathclyde, Economics
Patrizio Lecca, Fraser of Allander Institute, Department of Economics, University of Strathclyde, Peter McGregor, Department of Economics, University of Strathclyde and Kim Swales, Fraser of Allander Institute, Department of Economics, University of Strathclyde
In light of climate change concerns, promoting renewable energy sources (RES) has become a priority for energy policy-makers. The penetration of these technologies in the electricity mix is highly dependent on the costs and risks associated with large-scale investment projects, and more importantly on the speed and path of costs reductions through potential technological improvements. To reflect this constraint, many recent climate and energy policy models, linking the energy system to the rest of the economy, have incorporated technological change as an endogenous feature in the form of learning-by-doing (LBD).
This paper aims to further the understanding of the use of LBD in energy-economy models. To accomplish this, the main objectives of the paper are to provide a review of the energy-economy modelling literature introducing endogenous LBD, to identify the specification of endogenous LBD in these models, to classify the identified specifications into key categories, and finally to illustrate the use of these different specifications into a single applied energy-policy model.
The literature review of several energy-economy models with endogenous LBD first reveals the traditional distinction between modelling approaches, widely known as Top-down vs. Bottom-up (Kahouli-Brahmi, 2008). This differentiation between modelling approaches essentially corresponds to the variations in the level of disaggregation of the economy and the energy system in energy-economy models (Böhringer, 1998). The models reviewed in the paper are classified using this distinction.
The detailed analysis of these models reveals that the same LBD phenomenon has been represented in different ways, i.e. different equation form, different variables affected by learning and different variables used as proxy for experience accumulation.
The different LBD specifications seem to correspond to choices in the modelling approaches: First, Bottom-up models, such as MESSAGE (Messner, 1997), largely adopt the same one-factor or two-factor learning curve specifications, with technology specific investment costs decreasing with experience, generally represented by installed capacity (and R&D investments for the two-factor learning curve). This specification appears in line with the empirical literature relating to LBD for energy technologies. On the other hand, Top-down models show large variations in the way they implement endogenous LBD. Some adopt the simple learning curve specification referred to previously, but other models choose to alter this specification to make learning a function of past improvements. Behind this observation, one can point out the wide range of models which fall under the Top-down category. From Global Integrated Assessment models of climate change to national or regional CGE models, the choice of variables and equation form originates from the economic theory underlying each model.
From this analysis, two major equation forms, as well as two major variable choice options are identified and generalised to be tested in our model.
These specifications are implemented through different simulations in a CGE model for Scotland. The model is highly disaggregated at the energy level, enabling the introduction of LBD only in specific renewable energy sectors. The specifications are tested through the simulation of production subsidy to the marine electricity sector which is also subjected to endogenous learning.
The results of the simulations show large divergences in the long-run aggregated and sectoral impacts of the subsidies when changing the LBD specification. The overall GDP impact resulting from the implementation of a 10% production subsidy covers a wide range from 0.53 to 1.13% across the series of simulations. At sectoral level, efficiency gains in production for the marine sector can more than triple when changing the LBD equation form while sectoral output can more than double. In addition to variation in long-run results, the adjustment paths also prove to vary significantly with the LBD specification. These results also proved highly sensitive to changes parameter values.
This paper identified several specifications in the energy-economy modelling literature that are designed to represent the same phenomenon of learning-by-doing. Testing these specifications through a policy simulation in a CGE model for Scotland reveals large variations in long-run results as well as adjustment paths. These findings have important implications both for the modelling community as well as policy-makers, as the output of such models are often used a basis for policy choices. The penetration of renewable energy sources in our energy systems is likely to be equally dependent on policy support and technological improvements. Therefore, an accurate representation of technological learning in energy-policy models is of prime importance. These findings enhance the need for explicitly stating assumptions in the modelling exercise and for thorough sensitivity analysis.
Böhringer, C. (1998) The synthesis of bottom-up and top-down in energy policy modelling, Energy Economics, Volume 20, Issue 3, pp. 233-248
Kahouli-Brahmi, S. (2008) Technological Learning in energy-environment-economy modelling: A survey, Energy Policy, vol. 36, pp.138-162
Messner, S. (1997) Endogenized technological learning in an energy systems model, Journal of Evolutionary Economics, vol. 7 (3), pp. 291-313
All Papers in this Session
Presented by Patricia a Fortes, CENSE, Faculty of Science & Technology/New University of Lisbon
Presented by Marie Tamba, University of Strathclyde, Economics