The California Solar Initiative (CSI) was enacted in 2007 and is the largest state rebate program for solar in the US. It provides a lump-sum payment for residential solar in addition to the federal investment tax credit (ITC), a non-refundable tax credit for a share of the installation cost of residential and utility-scale solar that stood at 26% in 2020. Under third-party ownership, a household leases a system or buys its power from a solar company that owns the panels, and the rebate is paid to that company. Pless and van Benthem (2019) estimate elasticities for home-owned and third-party-owned residential solar; this entry covers third-party-owned systems. Using a panel of solar installations in California from 2010 to 2013, they exploit price variation caused by sharp changes in the incentive schedule over time and between investor-owned utilities, and they estimate demand by instrumental variables. They also estimate that more than the full subsidy, 153%, passes through to the prices third-party customers pay. Hahn, Hendren, Metcalfe, and Sprung-Keyser (2026) use these estimates to calculate the MVPF of a marginal expansion of the subsidy for residential solar beyond the 26% ITC, implemented nationally in the US in 2020.
MVPF = 3.8
Hahn, Hendren, Metcalfe, and Sprung-Keyser (2026) report every component per $1 of program spending. In their 2020 baseline, a residential system costs $2.05 per watt net of the 26% ITC, based on the installed costs reported by NREL (2022), and each watt produces 1.44 kWh a year, following Barbose, Darghouth, O’Shaughnessy, and Forrester (2020) and Ramasamy et al. (2022), over a 25-year life. Because the elasticities they use are with respect to the installation price rather than the subsidy, they scale each induced effect by the pass-through rate, the share of the subsidy that reaches the price paid for the system. For third-party-owned systems, they use a pass-through rate of 152.8%, the only estimate of pass-through for such systems in their sample.
As for home-owned systems, Hahn et al. (2026) calculate the elasticity by taking the implied derivative at a price of $3.70 per watt from the paper’s instrumental-variables demand estimation for third-party-owned systems and multiplying it by the ratio of the mean price, $3.70, to mean installations, 0.518. The resulting price elasticity for third-party-owned residential solar is −1.04.
The net cost has five components: the $1 mechanical transfer to inframarginal households that would have installed panels anyway, federal ITC spending on induced installations, state rebate spending (zero, since the baseline assumes no state rebates), lost tax revenue on utility profits, and the climate fiscal externality. The federal credit is $0.72 per watt, or $0.35 per dollar of the $2.05 net cost, which Hahn et al. (2026) multiply by the product of the elasticity and the pass-through rate: $0.558. Utilities lose profits as customers switch to rooftop solar. After a 20% rebound, a utility loses sales of about 1.15 kWh a year for each watt installed, and at government revenue of $0.006 per kWh of utility profit, discounted over 25 years and scaled in the same way, the lost tax revenue is $0.108. The climate fiscal externality, the US government’s share of the avoided climate damages through future tax revenue, is 1.92% of the global environmental benefit (including learning-by-doing, net of the rebound and lifecycle emissions); it lowers the cost by $0.061. Per $1 of program spending, the fiscal externalities from existing subsidies and taxes total $0.667, so the net cost is $1.606.
The willingness to pay (WTP) that Hahn, Hendren, Metcalfe, and Sprung-Keyser (2026) estimate has five components: the mechanical transfer, environmental benefits, a rebound effect, learning-by-doing effects, and lost utility profits. The transfer is $1.528 per $1 of program spending, equal to the 152.8% pass-through rate that Hahn et al. (2026) use for third-party-owned systems. The monetized environmental benefit of solar output in the US in 2020, using the marginal emission rates of AVERT, the Avoided Emissions and Generation Tool of EPA (2024), is $0.125 per kWh for global pollutants and $0.025 per kWh for local pollutants, and the grid, forecast with the REPEAT Project of Jenkins and Mayfield (2023), is allowed to change over the panels’ 25-year life. Hahn et al. (2026) discount the stream of benefits per watt over that life, divide by the cost per watt, and multiply by the product of the elasticity and the pass-through rate. Net of lifecycle emissions from manufacturing, maintaining and decommissioning the panels, which NREL (2013) puts at 40 grams of CO2e per kWh generated, this gives global environmental benefits of $1.604 and local environmental benefits of $0.235 per $1 of program spending. The rebound, 20% of the local and global benefits excluding lifecycle emissions, offsets $0.393.
Learning-by-doing, with a learning rate of 0.319, from Way, Ives, Mealy, and Farmer (2022), and cumulative global production of solar panels of 713,918 MW in 2020, from IRENA (2023), adds $1.982 of future environmental benefits and $1.371 of future price reductions. Utilities lose 1.10 cents of profit per kWh of solar output in 2020; scaled down by the 20% rebound, discounted over the panels’ life and multiplied by the product of the elasticity and the pass-through rate, this lost profit lowers the WTP by $0.200. Summing these components, the WTP is $6.128 per $1 of program spending. In their baseline, Hahn et al. (2026) do not add the induced households’ savings on electricity bills to the WTP, since a household’s decision to install already reflects them.
Combining these components, the MVPF is $6.128 / $1.606 = 3.815.
In the paper’s own setting, using externality values for California in 2013, the last year of its sample, Hahn, Hendren, Metcalfe, and Sprung-Keyser (2026) estimate an in-context MVPF of 3.275. With a social cost of carbon of $76 per ton in 2020 instead of $193, the MVPF falls to 2.886; with $337, it rises to 5.544. Across the five estimates in their main sample, the category average MVPF for residential solar subsidies is 3.862; across the estimates for which they have standard errors, the average is 3.282, with a 95% confidence interval of [1.967, 33.888].
Barbose, Galen, Naïm Darghouth, Eric O’Shaughnessy, and Sydney Forrester (2020). “Distributed Solar 2020 Data Update [Slides].” Department of Energy. DOI: https://doi.org/10.2172/1735556
EPA (2024). “Avoided Emissions and geneRation Tool (AVERT).” US Environmental Protection Agency. https://www.epa.gov/avert
Hahn, Robert W., Nathaniel Hendren, Robert D. Metcalfe, and Ben Sprung-Keyser (2026). “A Welfare Analysis of Policies Impacting Climate Change.” American Economic Review, 116(7): 2368–2421. DOI: https://doi.org/10.1257/aer.20250166
IRENA (2023). “Renewable Power Generation Costs in 2022.” International Renewable Energy Agency. https://www.irena.org/Publications/2023/Aug/Renewable-Power-Generation-Costs-in-2022
Jenkins, Jesse D., and E. Mayfield (2023). “Rapid Energy Policy Evaluation and Analysis Toolkit.” Princeton University Zero Lab. https://repeatproject.org
NREL (2013). “Life Cycle Greenhouse Gas Emissions from Solar Photovoltaics.” US Department of Energy. DOI: https://doi.org/10.2172/1056745
NREL (2022). “Solar Installed System Cost Analysis.” US Department of Energy. https://www.nlr.gov/solar/market-research-analysis/solar-installed-system-cost
Pless, Jacquelyn, and Arthur A. van Benthem (2019). “Pass-Through as a Test for Market Power: An Application to Solar Subsidies.” American Economic Journal: Applied Economics, 11(4): 367–401. DOI: https://doi.org/10.1257/app.20170611
Ramasamy, Vignesh, Jarett Zuboy, Eric O’Shaughnessy, David Feldman, Jal Desai, Michael Woodhouse, Paul Basore, and Robert Margolis (2022). “U.S. Solar Photovoltaic System and Energy Storage Cost Benchmarks, With Minimum Sustainable Price Analysis: Q1 2022.” Department of Energy. DOI: https://doi.org/10.2172/1891204
Way, Rupert, Matthew C. Ives, Penny Mealy, and J. Doyne Farmer (2022). “Empirically Grounded Technology Forecasts and the Energy Transition.” Joule, 6(9): 2057–2082. DOI: https://doi.org/10.1016/j.joule.2022.08.009