Utilities Included: Split Incentives in Commercial Electricity Contracts.

• Author: Jessoe, Katrina

1. INTRODUCTION

   Separating the party who pays for energy from the one making decisions about electricity use has long been cited as creating incentives for energy over-consumption or underinvestment in energy efficiency in both the commercial and residential sectors. In the U.S., roughly 20 percent of commercial building occupants rent space with electricity bundled into their monthly rent. Under this contract structure commercial tenants face zero marginal cost of consuming electricity, creating an incentive to over-consume. The remaining 80 percent of tenants pay their own monthly utility bills, which will dampen the incentive for building owners to invest in energy efficiency if owners cannot capitalize on a rent premium for energy efficiency upgrades. These misalignments between tenant and landlord incentives may lead to overconsumption of energy and overproduction of pollution that Pigouvian taxes are not well suited to correct (Jaffe and Stavins, 1994; Gillingham and Palmer, 2014). Given that the commercial sector accounts for over 35 percent of end-use electricity consumption in the U.S., the welfare costs from excess energy use may be substantial. Yet little evidence exists about the magnitude of these "split incentive" principal-agent problems in the commercial sector. (1)

   In this paper, we estimate an important component of the change in electricity use from switching commercial customers on electricity-inclusive rent contracts to tenant-paid utility contracts, a distinction we refer to as "contract type". We do this by evaluating how the relationship between electricity use and temperature (the temperature response gradient, henceforth "TRG") differs by contract type. We illustrate how the structure of the rental contract may create two distinct split incentives, one on the intensive and another on the extensive margin of demand for energy services, which lead to different empirical predictions relating to the TRG. When considering the intensive margin, the TRG will be less steep under a tenant-paid than an electricity-inclusive, or "owner-paid" contract. This occurs because, for a given level of energy efficiency capital, firms on a tenant-paid contract pay a positive marginal price for electricity use while those on an owner-paid contract face a marginal price of zero. The second split incentive relates to owner incentives to invest in energy efficient durables. Under a tenant-paid contract, owners have little incentive to invest. Since (all else equal) lower investment in energy efficiency leads to a steeper TRG, firms located in buildings on tenant-paid contracts should exhibit a steeper TRG relative to owner-paid contracts. We refer to this as the extensive margin effect. These two split incentives impact the TRG in opposite directions in relation to contract type, allowing us to empirically test which split incentive (if any) dominates in our setting.

   Results suggest that the intensive margin effect dominates among the largest firms. Tenant-paid contracts induce considerable energy savings among these customers during the hottest summer months. For the largest decile of firms, switching from an owner-paid to tenant-paid utility contract would reduce electricity use by roughly 3 percent over the course of a year and up to 14 percent in the summer months. The annual savings among large consumers are comparable to popular energy conservation measures such as home energy reports, which produce average savings of approximately 2 percent (Allcott, 2011). Furthermore, the savings occur at times when the value of electricity is likely to be high: during the hottest days of the year. Our finding that the largest customers are most responsive to contract type corroborates recent evidence from the residential sector in Sweden (Elinder et al., 2017). In contrast, contract type does not measurably impact consumption decisions for the smallest 90 percent of commercial customers. These results are consistent with profit-maximizing firms facing adjustment costs in electricity consumption, such that a relatively small absolute value of bill savings, among smaller firms, would not warrant conservation behavior.

   Our empirical approach exploits the differential effect of an exogenous weather shock on electricity use across firms on an owner versus tenant-paid contract. To do this we make use of cross-sectional variation in local weather exposure within a calendar billing month generated from the staggering of electricity billing periods across customers. We combine these weather data with monthly bills from 1,074 commercial firms serviced by a Connecticut electric utility between October 2007 and May 2011, and property-level information on fixed observables including whether the tenant or landlord pays the electric bill. This panel data set allows us to examine the differential impact of local weather shocks on electricity use across contract types, controlling for potential selection into contract type based on firm or fixed building attributes. A "levels" comparison of electricity use across contract type would be biased if firms on owner- and tenant-paid contracts differ in ways that are correlated with energy use. Focusing instead on the TRG across contract type captures the sensitivity of electricity consumption to fluctuations in temperature and allows us to control for a rich set of variables that may correlate with selection into contract type, thereby permitting identification under weaker assumptions.

   Our identifying assumption is that selection into contract type is unrelated to unobservable electricity demand drivers that are correlated with the TRG. We present three pieces of empirical evidence that support this assumption. First, motivated by recent work demonstrating that the electricity response to temperature shocks meaningfully differs across certain building attributes, we directly control for the possibility that the TRG is heterogeneous in observable building attributes (Novan et al., 2017). After controlling for interactions between temperature and attributes such as building age and industry type, our results are unchanged. Second, we use a change to a Connecticut metering regulation. This change was legislated after the end of our sample period and altered building owners' ability to select into contract type. It provides us with an opportunity to examine the TRG of firms located in buildings that switched contract types shortly after the change, and to test whether they exhibit a differential response gradient. They do not. Third, we assess the effect of potential correlations between any remaining unobservable characteristics and the treatment, as described in Oster (2017). This places bounds on the potential bias from selection on unobservables. Each of these tests exposes our identifying assumption to an opportunity to fail, and the results of each test support our main conclusions.

   Given the size of the firms responsive to contract type, the estimated treatment effect translates into significant costs from misaligned incentives. Using very conservative assumptions, we find that if incentives were aligned among the largest decile of commercial customers nationwide, total energy savings would be roughly one and a quarter times the savings from solving the split incentives problem for the entire U.S. residential electricity sector. The magnitude of the treatment effect and the relative size of large commercial firms are the primary factors leading to this result. Though the number of commercial customers affected by the split incentives problem is small relative to residences, these customers use much more energy. Thus, addressing the commercial split incentive problem requires a fraction of the contact points, while likely leading to greater energy savings. Our estimates imply greenhouse gas reductions of between 615-1200 thousand tons of C[O.sub.2] per year, or (to give a sense of scale) roughly 3.3 to 6.6 times the average annual savings from yearly Weatherization Assistance Program retrofits. These savings may be achievable at a relatively low cost. When we compare the cost of retrofitting units with sub-meters (to allow switching to tenant-paid utility bills) with the estimated annual bill savings amongst the largest customers, the payback period is less than one year. One caveat to note is that our results are identified based on the TRG. A switch from owner to tenant-paid contracts, and the accompanying disincentive for landlords to invest in energy efficiency, may also alter the level of electricity use. While many building-level energy efficiency investment choices available to landlords affect the TRG, we cannot rule out that these "level" effects may lead to a consumption increase and mitigate or even overwhelm the TRG savings we identify relating to the temperature gradient.

   Despite the robustness of our results, identification of our main effect arises from the behavior of 110 large firms, 19 of whom are on owner-paid contracts. Our need to lean on a small sample raises the possibility that our study may be under-powered and that our results may be driven by outliers. While our research setting prevents us from completely nullifying these concerns, we present evidence that allows us to reject the hypothesis that our findings are driven by a single outlier among the largest owner-paid firms in our sample.

   This work makes four main contributions to the academic literature and environmental policy discussion. First, compared to the residential setting where a growing literature points to both the potential and limitations of energy efficiency and contracting solutions (Gillingham et al., 2012; Hassett and Metcalf, 1999; Fowlie et al., 2015; Elinder et al., 2017), little is known about the commercial setting. We provide a commercial counterpart to existing residential estimates on the split incentives problem. Second, our identification strategy makes several...