Decision makers often operate on short-term horizons, placing more weight on the short term and revising their decisions later as new information arises. What’s more, literature has shown that the implicit discount rate at which decisions are made for the future isn’t constant, but changes as a function of the point in the future for which the decision is made. While the myopic foresight of decision makers has been extensively studied in the energy system modeling literature, the focus has been on a binary representation in which time periods beyond the defined decision horizon are completely ignored, and changes in strategies are a result of new information, rather than different valuation of a specific future cost. This study introduces a temporal inconsistency framework for energy system modeling. In this approach, decision-makers do plan for the full time horizon, but assess future value based on a changing discount rate and, as time is moved forward, reassess this value at each one year step. This approach advances the state of the art by simulating temporally–inconsistent decision making, in which optimal decision made for a given year changes as a function of when those plans were made. We demonstrate this framework by simulating decarbonization scenarios for the Greek power sector with varying levels of path dependency and commitment over 2019–2070. The results highlight that logarithmic discounting can lead to irreversible sub-optimal pathways, whereas exponential discounting produces sub-optimal early decisions, which however can later be revised, moving closer to the time-consistent pathway (up to 10 times less natural gas-fired power generation in 2050 compared to initial plans). • We introduce a temporal inconsistency framework for energy system modeling. • Optimal decisions made for the future change depending on when they are made. • We apply this framework to assess mitigation scenarios for Greece’s power sector. • Results indicate profound impacts of short-term biases and temporal inconsistencies. • These mainly include increased system costs and lock-ins to sub-optimal pathways.
Koutsandreas et al. (Thu,) studied this question.