Association for Public Policy Analysis & Management

This paper explores tenant behavior in rental housing where the landlord pays for heating. I build and estimate a structural model of housing and heating choice to analyze a hypothetical policy intervention of requiring that all tenants pay for their own energy bills. Households respond by re-sorting into households of different sizes and changing energy-use behavior. Exogenous variation in energy prices identifies key parameters in the model and a machine-learning algorithm predicts heating preferences. I find evidence for significant moral hazard, adverse selection on energy elasticity into landlord-pay contracts, and household sorting into larger housing units when the landlord pays for utilities.

The “energy-efficiency gap” is the idea that due to costly information, inattention, or miscalculation, private individuals have failed to adopt energy-efficient appliances, housing materials, and other technology that would shortly pay for itself in energy savings. Under-investment in energy efficiency reduces welfare for private energy users and increases the negative externalities from pollutants generated by energy use. Recent work on energy use in rental housing suggests that asymmetric information between landlords and tenants significantly distorts both energy-use and energy-efficiency investment incentives, leaving room for win-win energy policies (Elinder et al., 2017; Myers, 2018).

I develop a structural model of joint home efficiency and energy use decisions that characterize moral hazard and adverse selection in the residential energy market. Agents select housing size as well as the energy payment regime, endogenously determining the future cost of energy services. Using U.S. data on rent prices, energy prices, and home energy use, I estimate key structural parameters in the model. A home-energy-demand system pins down estimates of the energy costs of housing attributes and winter temperature settings. I exploit exogenous variation over time in the relative prices of natural gas and electricity to identify the rent premium for having landlord-pay utilities. A machine-learning approach predicts underlying household energy preferences using landlord-pay rent premiums as instruments for sample selection. Using estimated utility functions for each renter household, I simulate the effects of a policy requiring that all tenants pay for their own energy bills, allowing tenants to respond by re-sorting into housing units of different sizes.

I find that landlord-pay households spend $12.22 more on energy per month on average than they would if all rental contracts required the household to pay for heating. This represents roughly a 14 percent difference in energy expenditures. Landlord-pay households choose housing units that are 65 square feet larger due to the reduced marginal cost of housing size; however, the sorting effect only accounts for $0.53 per household per month of increased expenditures. The moral hazard effect dominates, contributing to $11.69 of extra expenditures per household per month. Adverse selection into landlord-pay regimes increases the moral hazard effect by over a third—a finding on par with adverse selection effects in health insurance markets (Einav et al., 2013). The model estimates significant heterogeneity in energy-price responsiveness, predicting that 43 percent of landlord-pay households would not change energy-use behavior if made to pay their own bills.