Appendix 1: Model Description
The electric sector only version of EPRI’s North American Regional Economy, Greenhouse Gas, and Energy (NA-REGEN) model represents detailed capacity planning and dispatch decisions simultaneously with state-based regions (Fig. 1). Each customizable-length time step (often 5-year intervals) includes capacity investment, retrofit, and retirement decisions as well as dispatch for installed capacity over representative intra-annual hours. The intertemporal optimization structure of NA-REGEN determines investment and operational choices through 2050 while representing regional resource endowments, costs, inter-regional transmission, demand, and regulations. This appendix summarizes the main features and assumptions of the model, especially those relating to the experiments in this paper. Additional detail about the NA-REGEN is provided in EPRI (2018a, b).
NA-REGEN provides customizable state or regional resolution, accounting for differences in policy, transmission, and demand. The regional configuration used for this study is shown in Fig. 1. Although the model includes endogenous investments in inter-regional transmission and segment-level electricity trade across regions, NA-REGEN does not represent intra-regional transmission constraints.
Hourly renewable output and resource potentials are based on analysis and data by EPRI, AWS Truepower, and NASA’s MERRA-2 dataset (Blanford et al. 2018; EPRI 2018a, b), which give synchronous time-series values with load. Figure 15 shows Canadian wind speed data on the MERRA-2 grid (0.625° longitude by 0.5° latitude).
Technological cost and performance assumptions come from EPRI’s Integrated Generation Technology Options report with more frequent updates for technologies like solar and wind. Default capital cost assumptions are shown in Fig. 16, and other cost assumptions are detailed in the REGEN documentation (EPRI 2018a, b).
NA-REGEN uses a bottom-up representation of capacity grouped into technology blocks within a region based on heat rates and dispatches these blocks across a range of intra-annual time segments. Joint variation in load, wind, and solar across regions is captured through the selection of so-called “representative hours” using an approach described in Blanford et al. (2018). This novel feature more accurately captures the spatial and temporal variability of power systems, which are critical for evaluating asset investments and operations especially under higher renewable deployment scenarios. Power plant data come from the ABB Velocity and were last updated in June 2018, which includes projects in the development pipeline like variable renewable projects and the Site C Dam in British Columbia. Announced retirements for plants in the U.S. and Canada are also included. REGEN includes a reserve margin constraint, where the sum of firm capacity must be greater than or equal to the peak residual load plus a reserve margin (which is set at 15% by default). Contributions from renewable resources and dispatchable technologies vary by hour and season, and residual load (i.e., grid-supplied load less variable renewable output) is determined endogenously and varies by region, load shape scenario, and levels of wind and solar deployment.
Note that this analysis uses an electric sector only version of the REGEN model, so end-use sectors and their interactions with load shapes are not modeled explicitly. All hourly time-series data, including load and variable renewable resources, come from 2015. Annual load over time is exogenous for most scenarios and is based on values from U.S. Energy Information Administration’s Annual Energy Outlook 2018 (U.S. Energy Information Administration 2018) for U.S. states and the National Energy Board’s Canada’s Energy Future 2016 (National Energy Board 2016) for Canadian provinces.
Appendix 2: Scenario Assumptions
This appendix provides additional detail about scenario assumptions.
Fuel price trajectories come from the U.S. Energy Information Administration’s Annual Energy Outlook 2018 (U.S. Energy Information Administration 2018). Fuel prices are not responsive to changes in demand for these runs, though such feedbacks are possible using the integrated version of NA-REGEN. Delivered gas prices in the model include region-specific adders, which are calibrated to observed 2016 values and assumed to decline over time (Fig. 17).
Transmission between regions can be endogenously added with an assumed cost of $3.85 million per mile for a notional high-voltage line to transfer 6400 MW of capacity. Note how, due to changes in flows across regions (with associated transmission losses) and different levels of new transmission investments, total national generation may vary across scenarios.
The Canadian federal carbon pricing backstop is implemented as a carbon levy on upstream fuels (Environment and Climate Change Canada 2017). For Canadian nuclear units (primarily in Ontario), announced retirement and refurbishment schedules are incorporated into the analysis. All Pickering units retire by 2025, and Bruce units 1 and 2 retire in 2043. Refurbishment schedules are taken from IESO data (Indepedent Electricity System Operator 2018), and refurbished units are assumed to unavailable during these years.
Appendix 3: Emissions Intensity Calculation for Border Carbon Adjustments
Many measures have been proposed to reduce leakage associated with climate policy. The principal measures are border carbon adjustments, where imports from non-regulated (or under-regulated) jurisdictions are taxed at the emissions price of the regulated region (Hoel 1991). Border carbon adjustments are often relied on as tools when production in unregulated regions have higher emissions intensities and when a regulated region is a net exporter (Fell and Maniloff 2015). Other measures to reduce leakage include trade constraints, output-based allocations, and exemptions for trade-exposed industries. The literature suggests that it is difficult to rank order these anti-leakage measures, since their effectiveness is highly context-dependent and depends on considerations like relative emissions rates, elasticities of substitution, and consumption volumes (Fischer and Fox 2012). Border carbon adjustments can be attractive in leakage reduction and cost-effectiveness terms, which comes at the expense of equity considerations (Böhringer et al. 2012b).
As described in Sect. 2.2, the leakage mitigation sensitivity that imposes a border carbon adjustment uses a tariff rate that is dynamically updated over time based on Canadian carbon price and emissions rate from a new NGCC unit. There are many methods and considerations for calculations emissions intensities on imported power, including whether marginal versus average rates are used (the former more accurately represent emissions rates, but the latter are simpler to calculate), dynamic versus static rates (the former adjust over time based on changing conditions, while the latter are easier to calculate), and temporal granularity for assessment (regardless of the method chosen, emissions intensities could be calculated on a sub-hourly basis, hourly basis, or something more aggregated). Different combinations for the above conditions are possible, but most entail tradeoffs between accuracy and administrative simplicity/cost (Cosbey et al. 2019). An additional question is whether to include export subsidies in addition to import tariffs. The border carbon adjustment used here includes only import tariffs given the uncertain legal status of export subsidies under the World Trade Organization’s General Agreement on Tariffs and Trade (GATT) (Cosbey et al. 2019; de Cendra 2006) and due to the fact that the literature indicates measures with import adjustments only capture most of the BCA leakage reduction benefits (Böhringer et al. 2012a; Fischer and Fox 2012).
A uniform benchmark based on a new NGCC plant emissions rate is administratively simple and, as suggested by the calculations below, aligns with the marginal emissions intensity of U.S. generation. Table 2 shows changes in U.S. emissions and emissions intensity between the reference case without Canadian carbon pricing and the treatment scenario with Canadian policy. The marginal CO2 intensity of power imports is the ratio of incremental U.S. CO2 emissions to incremental U.S. generation between these two scenarios. Increased exports from the U.S. to Canada are mostly be supported by incremental NGCC output under reference gas prices, as shown in the left panel of Fig. 7. An implicit assumption in this scenario is that U.S. exporters do not adjust the composition of generation to lower emissions liabilities associated with trade to the regulated region as firms might if exposed to the import tariff in the scenario with the border carbon adjustment.
Table 2 Change in U.S. emissions, generation, and CO2 intensity over time Note that using the NGCC-based benchmark means that the intensity does vary across time (supporting a dynamic adjustment) and is in some cases higher than the NGCC intensity. Figure 7 shows that a portion of Canadian imports in early years are supported by coal and combustion turbine generation in addition to NGCC, which leads to higher intensities in 2020 and 2025. However, increased NGCC output represents the bulk of the response in the near years, and almost all the response in the out years. Thus, using an intensity based on the emissions rate from a new NGCC unit to set the benchmark captures ex-post values well and is considerably easier to administer than a dynamically updated intensity calculated in real time.
Appendix 4: Additional Results
This section provides additional reporting related to the analysis.
Figure 18 shows Canadian generation by technology over time across the five core policy scenarios under high gas price conditions. Relative to the reference gas price scenario (Fig. 3), Canadian NGCC generation falls, while generation from renewables rises.
Figure 19 shows the U.S. generation mix across the five main scenarios assuming reference gas prices. Canada’s CO2 emissions intensity of generation is roughly a third of U.S. values in 2015. New capacity additions over time are dominated by natural gas units (specifically NGCC without carbon capture), wind, and solar. Figure 20 shows the same scenarios with higher gas prices, which leads to higher wind and solar generation.
The main text focuses on cumulative emissions metrics over time. Figures 21 and 22 show CO2 emissions trajectories by scenario for Canada and the sum of Canada and U.S., respectively.
The generation mixes in the U.S. and Canada exhibit considerable variation when a U.S. carbon tax is adopted, as shown in Fig. 23. The middle panel provides a counterfactual scenario (“U.S. Only Tax”) where the U.S. adopts a unilateral carbon tax with the timing and stringency in Fig. 2. Without carbon pricing in Canada, gas and wind generation increase, and the unilateral U.S. policy leads to higher U.S. imports from Canada (5.3% of U.S. energy for load in 2030 and 2.9% in 2050). When a Canadian tax is simultaneously adopted (“U.S./Canada Tax”), U.S. imports drop and emissions increase, while Canadian generation decreases, especially from gas-fired units. These changes lead to lower Canadian emissions but higher U.S. emissions relative to the U.S. unilateral policy scenario. Figure 24 shows cumulative CO2 emissions changes and leakage under cases with and without U.S. federal carbon pricing. The second bar in Fig. 24 compares emissions in the “U.S./Canada Tax” scenario in Fig. 23 with the corresponding counterfactual without a Canadian carbon tax (“U.S. Only Tax” in Fig. 23).