The landmark randomized evaluation of the US Weatherization Assistance Program (WAP) in Michigan found that upfront retrofit costs were about twice the realized energy savings, that engineering-model projections overstated actual savings by more than a factor of three, and that the social rate of return was about -7.8% per year (Fowlie, Greenstone, and Wolfram 2018). The authors' concluding section calls for a body of credible ex post evidence — based on realized consumption rather than engineering-model projections — on the true returns to energy efficiency investments in the residential and other sectors, and notes that WAP does not appear to pass a conventional cost-benefit test although the program's full set of goals, including benefits the paper explicitly flags as unmeasured (health effects associated with reduced draftiness and comfort beyond indoor temperature), may not be reflected in such tests. Later literature has substantially settled, in the confirmatory direction, the external-validity and projection-realization-decomposition fronts of that call: realized savings well below engineering predictions replicate across US states and many countries, and an Illinois decomposition attributes the wedge mainly to engineering-model bias and contractor workmanship. The surviving open core is two directions serving the same generic call: rigorous causal quantification of the non-energy benefits that the conventional cost-benefit test did not capture for low-income weatherization — sufficient to determine whether the negative-return conclusion survives fuller benefit accounting — and extension of credible ex post causal evidence to settings and sectors beyond the Michigan low-income housing retrofit program.

Background

A long-running debate in energy and environmental economics concerns the energy efficiency gap: the observation that households and firms appear to forgo investments in energy efficiency (insulation, efficient appliances, better equipment) that seem privately profitable, which, if real, would justify subsidies on both private and environmental grounds (Allcott and Greenstone 2012; Gerarden, Newell, and Stavins 2017). Whether the gap is real, and how large it is, matters directly for climate and energy policy, because large public programs rest on the premise that efficiency investments pay off.

The Weatherization Assistance Program (WAP) is the United States' largest residential energy efficiency program: it funds comprehensive retrofits of low-income homes, typically including attic and wall insulation, air sealing, and heating-system replacement. Fowlie, Greenstone, and Wolfram (2018) evaluated WAP with a randomized encouragement design: roughly 34,000 low-income Michigan households were randomly assigned to receive, or not receive, intensive outreach encouraging them to apply for free weatherization worth about $5,000. Because assignment to outreach was random while participation remained voluntary, the outreach serves as an instrumental variable (a source of random variation in treatment that isolates causal effects), and the resulting estimates are local average treatment effects: effects on the compliers, the households induced to participate by the outreach. The companion take-up study documented that even this intensive campaign raised participation from under 1% to only about 6% (Fowlie, Greenstone, and Wolfram 2015).

The findings were stark. Participation reduced energy use by about 20% among compliers (natural gas by about 21%), while the National Energy Audit Tool (NEAT) — the engineering audit model used to size the installed measures — projected roughly a 46% reduction in natural gas use and 15% in electricity use for the weatherized households whose audits were examined. The published version's statement that model-projected savings were more than three times the actual savings refers to the ratio of projected to realized total energy (MMBtu) savings, not to the 46%-versus-21% percentage pairing for natural gas. Upfront costs (about $4,581 per home on average in the authors' NBER working-paper version of the study, NBER Working Paper 21331, incurred during the 2009 American Recovery and Reinvestment Act (ARRA) stimulus expansion of WAP, when labor and materials costs may have been inflated) were about twice the present value of realized energy savings. The implied private internal rate of return (IRR, the discount rate at which the investment breaks even) was negative (-2.2% in the working-paper version), and the social return, adding the value of avoided emissions, was about -7.8% per year in the published version (-9.5% in the working-paper version); the cost per ton of CO2 avoided exceeded the then-prevailing social cost of carbon several-fold. A direct test of the rebound effect — the increase in consumption of energy services, such as warmer rooms, that arises when efficiency lowers the marginal cost of comfort — found no significant indoor-temperature increase, so rebound could not explain the shortfall. The evidence comes from one state, one low-income population, and the stimulus period. The paper's conclusion defines the direction of this problem: it calls for a body of credible ex post (after-the-fact, consumption-based rather than model-projected) evidence on the true, rather than projected, returns to energy efficiency investments in the residential and other sectors, and notes that WAP does not appear to pass a conventional cost-benefit test although its full set of goals may not be reflected in such tests. The paper's own accounting explicitly excludes comfort conditional on indoor temperature and flags possible health effects of reduced draftiness — in either direction — as unmeasured non-energy benefits (NEBs).

Later work has confirmed the headline pattern well beyond Michigan. A systematic review of 39 evaluations covering 23 residential retrofit programs observed between 1984 and 2021 (about 141,000 retrofitted households) found a mean measured reduction in energy consumption of only about 7.2%, no program achieving deep savings of 50% or more, and reported savings shrinking as the internal validity of the study design strengthened (Giandomenico, Papineau, and Rivers 2022). Using Illinois WAP billing data and a machine-learning event-study design (a panel method comparing consumption trajectories before and after retrofit), Christensen, Francisco, Myers, and Souza (2023) decomposed the performance wedge between projected and realized savings — whose components had previously been lumped together — into bias in the engineering models (up to 41%, concentrated in wall insulation), heterogeneity in contractor workmanship (43%), and the rebound effect (6%). The same pattern of realized savings falling 30-50% or more short of engineering predictions appears in a Phoenix retrofit program (Liang et al. 2018), a nationwide Irish program evaluated with whole-home energy-use data (Coyne and Denny 2021), a Swiss subsidy scheme (Hondeborg et al. 2023), a Canadian program where average realized gas savings were about 60% of predictions, with whole-envelope retrofits realizing only about half of predicted savings (Papineau, Rivers, and Yassin 2025), and a Mexican field trial of new energy-efficient housing that found no detectable effect on electricity use or thermal comfort against predicted savings of up to 26% (Davis, Martinez, and Taboada 2020); disentangling behavioral from capital components, Zivin and Novan (2016) found upgrade savings substantially below engineering predictions in US weatherization programs.

Evidence on other investment classes points the same way, though more sparsely: a large Mexican appliance-replacement program delivered refrigerator savings of about one-quarter of predictions while air-conditioner replacement increased consumption, at a cost above $500 per ton of CO2 avoided (Davis, Fuchs, and Gertler 2014); California's new-construction building energy codes saved significantly less than projected (Levinson 2016); and the economic value of efficiency savings depends materially on the hours in which they occur, which raises the measured value of some investments (Boomhower and Davis 2020). For firms, survey evidence still reports substantial perceived barriers rather than causal savings estimates (Bremer, den Nijs, and de Groot 2024).

What remains genuinely open is the non-energy-benefit component and the completion of the welfare assessment it feeds. US evidence on WAP health and well-being effects is survey-based and non-experimental (Tonn et al. 2021; Tonn, Rose, and Marincic 2023); the strongest causal health evidence comes from a Dutch natural experiment in public housing — retrofits reduced children's respiratory medication use by about 3.8% in a staggered difference-in-differences design comparing retrofitted and not-yet-retrofitted tenants over ten years (Roberdel et al. 2026) — a different institutional and climatic context from the US low-income program; and energy efficiency is capitalized into home prices only in observational housing-market data (Aydin, Brounen, and Kok 2020). No design has yet delivered causal estimates of health, comfort, or property-value benefits for the US low-income weatherization population that could be entered into the social cost-benefit ledger to test whether the negative-return conclusion survives. Beyond the source's own ledger — which monetized operational energy savings and avoided emissions only — the accounting can be further complicated by the embodied emissions of retrofit materials, whose upfront greenhouse-gas burden from producing and installing the measures is on the order of 1.0-9.4 Mg (metric tons) of CO2-equivalent per weatherized northern-US home (Pfadt-Trilling and Fortier 2025); this is a later-literature extension rather than part of the source's cost-benefit framework, and reviews of WAP evaluation practice emphasize that its benefit streams remain incompletely measured (Anthony et al. 2025).

Problem Statement

Under the overarching question posed by Fowlie, Greenstone, and Wolfram (2018) — do energy efficiency investments deliver, and how large is the energy efficiency gap once measured against credible ex post evidence rather than engineering projections? — the authors' conclusion calls for a body of credible ex post evidence on the true, rather than projected, returns to energy efficiency investments in the residential and other sectors, and notes that the program's full set of goals may not be reflected in a conventional cost-benefit test. Two directions serving that call remain open after the intervening literature:

  1. Non-energy benefits and the full social return to low-income weatherization. Provide credible causal estimates of the non-energy benefits that the source's conventional cost-benefit test did not capture for low-income weatherization in the US context — health outcomes and comfort beyond the indoor-temperature measure already collected, the benefit domains the source explicitly flags as unmeasured; property-value effects may additionally be included where they can be credibly identified — of a quality that permits them to be entered into the social cost-benefit ledger of the Weatherization Assistance Program, and determine whether including them (together with the energy savings already established) reverses or confirms the conclusion that these investments have negative private and social rates of return. This requires an identification strategy that handles the severe selection into program take-up. Because the sign of the determination depends on the accounting conventions used, the determination is to be made against the ledger that generated the source's negative-return estimates, with valuation conventions (energy prices, discount rates, and measure lifespans, all of which the source's own accounting varies) stated explicitly enough to be contested and the sensitivity of the conclusion to alternative conventions shown rather than left to the analyst's choice of parameters.

  2. Credible ex post evidence in other settings and sectors. Extend the same credible-ex post causal standard to settings and investment types beyond the Michigan low-income housing retrofit program — consistent with the source's call for evidence on true returns in the residential and other sectors — so that the magnitude of the energy efficiency gap is measured, rather than assumed, beyond the single evaluated program.

Both directions serve the source's single generic call: credible ex post evidence on the true returns to energy efficiency investments, toward a complete welfare assessment of energy efficiency investments as a policy class. A complete resolution requires both directions, each with its own stopping point: direction 1 is resolved by causal non-energy-benefit estimates for the US low-income weatherization population entered into the ledger that generated the source's negative-return estimates, together with an explicit determination of whether including them reverses or confirms the negative-return conclusion; direction 2 is resolved by credible ex post causal estimates of realized returns for at least one additional identified setting or investment class beyond the Michigan low-income housing retrofit program. The related questions — whether the Michigan results generalize across regions, climates, income groups, and non-stimulus conditions, and how the projection-realization wedge decomposes between engineering-model bias, contractor workmanship, and rebound — are recorded in Background as substantially advanced by subsequent literature and are not restated as open here.

References

  1. Fowlie, Meredith, Michael Greenstone, and Catherine Wolfram (2015). Are the Non-Monetary Costs of Energy Efficiency Investments Large? Understanding Low Take-up of a Free Energy Efficiency Program. American Economic Review 105(5):201-204. DOI: 10.1257/aer.p20151011. URL: https://doi.org/10.1257/aer.p20151011
  2. Fowlie, Meredith, Michael Greenstone, and Catherine Wolfram (2018). Do Energy Efficiency Investments Deliver? Evidence from the Weatherization Assistance Program. Quarterly Journal of Economics 133(3):1597-1644. DOI: 10.1093/qje/qjy005. URL: https://doi.org/10.1093/qje/qjy005
  3. Allcott, Hunt, and Michael Greenstone (2012). Is There an Energy Efficiency Gap? Journal of Economic Perspectives 26(1):3-28. DOI: 10.1257/jep.26.1.3. URL: https://doi.org/10.1257/jep.26.1.3
  4. Gerarden, Todd D., Richard G. Newell, and Robert N. Stavins (2017). Assessing the Energy-Efficiency Gap. Journal of Economic Literature 55(4):1486-1525. DOI: 10.1257/jel.20161360. URL: https://doi.org/10.1257/jel.20161360
  5. Zivin, Joshua Graff, and Kevin Novan (2016). Upgrading Efficiency and Behavior: Electricity Savings from Residential Weatherization Programs. The Energy Journal 37(4):1-24. DOI: 10.5547/01956574.37.4.jziv. URL: https://doi.org/10.5547/01956574.37.4.jziv
  6. Davis, Lucas W., Alan Fuchs, and Paul Gertler (2014). Cash for Coolers: Evaluating a Large-Scale Appliance Replacement Program in Mexico. American Economic Journal: Economic Policy 6(4):207-238. DOI: 10.1257/pol.6.4.207. URL: https://doi.org/10.1257/pol.6.4.207
  7. Liang, Jing, Yueming Qiu, Timothy James, Benjamin L. Ruddell, Michael Dalrymple, Stevan Earl, and Alex Castelazo (2018). Do energy retrofits work? Evidence from commercial and residential buildings in Phoenix. Journal of Environmental Economics and Management 92:726-743. DOI: 10.1016/j.jeem.2017.09.001. URL: https://doi.org/10.1016/j.jeem.2017.09.001
  8. Levinson, Arik (2016). How Much Energy Do Building Energy Codes Save? Evidence from California Houses. American Economic Review 106(10):2867-2894. DOI: 10.1257/aer.20150102. URL: https://doi.org/10.1257/aer.20150102
  9. Tonn, Bruce, Erin Rose, Beth Hawkins, and Michaela Marincic (2021). Health and financial benefits of weatherizing low-income homes in the southeastern United States. Building and Environment 197:107847. DOI: 10.1016/j.buildenv.2021.107847. URL: https://doi.org/10.1016/j.buildenv.2021.107847
  10. Davis, Lucas W., Sebastian Martinez, and Bibiana Taboada (2020). How effective is energy-efficient housing? Evidence from a field trial in Mexico. Journal of Development Economics 143:102390. DOI: 10.1016/j.jdeveco.2019.102390. URL: https://doi.org/10.1016/j.jdeveco.2019.102390
  11. Boomhower, Judson, and Lucas Davis (2020). Do Energy Efficiency Investments Deliver at the Right Time? American Economic Journal: Applied Economics 12(1):115-139. DOI: 10.1257/app.20170505. URL: https://doi.org/10.1257/app.20170505
  12. Aydin, Erdal, Dirk Brounen, and Nils Kok (2020). The capitalization of energy efficiency: Evidence from the housing market. Journal of Urban Economics 117:103243. DOI: 10.1016/j.jue.2020.103243. URL: https://doi.org/10.1016/j.jue.2020.103243
  13. Coyne, Bryan, and Eleanor Denny (2021). Retrofit effectiveness: Evidence from a nationwide residential energy efficiency programme. Energy Policy 159:112576. DOI: 10.1016/j.enpol.2021.112576. URL: https://doi.org/10.1016/j.enpol.2021.112576
  14. Giandomenico, Lauren, Maya Papineau, and Nicholas Rivers (2022). A Systematic Review of Energy Efficiency Home Retrofit Evaluation Studies. Annual Review of Resource Economics 14:689-708. DOI: 10.1146/annurev-resource-111920-124353. URL: https://doi.org/10.1146/annurev-resource-111920-124353
  15. Hondeborg, Dianne, Benedict Probst, Ivalin Petkov, and Christof Knoeri (2023). The effectiveness of building retrofits under a subsidy scheme: Empirical evidence from Switzerland. Energy Policy 180:113680. DOI: 10.1016/j.enpol.2023.113680. URL: https://doi.org/10.1016/j.enpol.2023.113680
  16. Christensen, Peter, Paul Francisco, Erica Myers, and Mateus Souza (2023). Decomposing the Wedge between Projected and Realized Returns in Energy Efficiency Programs. Review of Economics and Statistics 105:798-817. DOI: 10.1162/rest_a_01087. URL: https://doi.org/10.1162/rest_a_01087
  17. Tonn, Bruce, Erin Rose, and Michaela Marincic (2023). Cascading benefits of low-income weatherization upon health and household well-being. Building and Environment 242:110470. DOI: 10.1016/j.buildenv.2023.110470. URL: https://doi.org/10.1016/j.buildenv.2023.110470
  18. Bremer, Leon, Sacha den Nijs, and Henri L.F. de Groot (2024). The energy efficiency gap and barriers to investments: Evidence from a firm survey in The Netherlands. Energy Economics 133:107498. DOI: 10.1016/j.eneco.2024.107498. URL: https://doi.org/10.1016/j.eneco.2024.107498
  19. Pfadt-Trilling, Alyssa R., and Marie-Odile P. Fortier (2025). The greenhouse gas burden of weatherizing northern US homes and estimated lifetime energy savings. Cell Reports Sustainability 2:100550. DOI: 10.1016/j.crsus.2025.100550. URL: https://doi.org/10.1016/j.crsus.2025.100550
  20. Papineau, Maya, Nicholas Rivers, and Kareman Yassin (2025). Household benefits from energy efficiency retrofits: Implications for net zero housing policy. Energy Economics 143:108245. DOI: 10.1016/j.eneco.2025.108245. URL: https://doi.org/10.1016/j.eneco.2025.108245
  21. Anthony, Ryan, Brian An, Marilyn A. Brown, Michelle Graff, and Daniel C. Matisoff (2025). Frontiers of weatherization: Evolving programmatic needs and expanding evaluation tools. Energy Research and Social Science 130:104442. DOI: 10.1016/j.erss.2025.104442. URL: https://doi.org/10.1016/j.erss.2025.104442
  22. Roberdel, Vincent P., Ioulia V. Ossokina, Jos van Ommeren, and Theo A. Arentze (2026). Effect of energy-efficient homes on residents' health: evidence from a natural experiment in the Netherlands. The Lancet Public Health 11(4):e245-e252. DOI: 10.1016/s2468-2667(26)00023-x. URL: https://doi.org/10.1016/s2468-2667(26)00023-x
  23. Fowlie, Meredith, Michael Greenstone, and Catherine Wolfram (2015). Do Energy Efficiency Investments Deliver? Evidence from the Weatherization Assistance Program. NBER Working Paper 21331, National Bureau of Economic Research (authors' version of the 2018 Quarterly Journal of Economics article). DOI: 10.3386/w21331. URL: https://www.nber.org/papers/w21331