Energy efficiency and household behavior: the rebound effect in the residential sector
| Author | Erdal Aydin,Nils Kok,Dirk Brounen |
| DOI | http://doi.org/10.1111/1756-2171.12190 |
| Published date | 01 August 2017 |
| Date | 01 August 2017 |
RAND Journal of Economics
Vol.48, No. 3, Fall 2017
pp. 749–782
Energy efficiency and household behavior:
the rebound effect in the residential sector
Erdal Aydin∗
Nils Kok∗
and
Dirk Brounen∗∗
This article investigates the rebound effect in residential heating, using a sample of 563,000
households in the Netherlands. Using instrumental variable and fixed-effects approaches, we
address potential endogeneity concerns. The results show a rebound effect of 26.7% among
homeowners, and 41.3% among tenants. We corroboratethe findings through a quasiexperimental
analysis, using a large retrofitsubsidy program. We also document significant heterogeneityin the
rebound effect, determined by household wealth and income, and the actual energy use intensity.
The findings in this article confirm the important role of household behavior in determining the
outcomes of energy efficiency improvement programs.
1. Introduction
Energy consumption in the durable building stock has once again returned to the agenda
of policy makers. Around the world, regulatory measures are introduced to reduce and mitigate
the harmful effects of climate change that result, in part, from the carbon externality of energy
consumption in buildings. Evidence on the effect of stricter building codes on the energy con-
sumption of newly constructed dwellingsis inconclusive (Jacobsen and Kotchen, 2013; Levinson,
2016), and codes as a policy instrument alone may thus be insufficient to meet broader energy
reduction targets for the built environment (Majcen, Itard,and Visscher, 2013). Irrespective of the
effectiveness of policies that aim to increase the thermal quality of the building stock, a critical
debate focuses on how households respond to improvements in the energy efficiency of their
homes.
∗Maastricht University; e.aydin@maastrichtuniversity.nl, n.kok@maastrichtuniversity.nl.
∗∗Tilburg University; d.brounen@uvt.nl.
We thank the participants at the 2013 AREUEA International Meeting in Jerusalem, the 2013 ERES Conference in
Vienna, the 2014 IAEE International Conference in New York, the seminars at Tilburg Sustainability Center (TSC) in
2013 and Paris Dauphine University in 2014, the 2014 Energy Efficiency Gap workshopat ZEW, Mannheim, the 2015
Rebound workshop at RWTH Aachen University, as wellas the Editor and two anonymous reviewers, for their helpful
comments. The Dutch Ministry of Interior and Kingdom Relations (BZK) provided financial support for this research.
Kok is supported by a VIDI grant from the Netherlands Organization for Scientific Research (NWO).
C2017, The RAND Corporation. 749
750 / THE RAND JOURNAL OF ECONOMICS
Indeed, research has shown that as a consequence of the associated changes in consumer
behavior, technological improvements may lead to lower energy savings than expected (Jevons,
1906; Khazzoom, 1980; Wirl, 1997). The mechanism underlying this behavioral change relates
to neoclassical economic theory: when the energy efficiency of a particular energy service is
improved, households realize a reduction in the effective price of that service. Consequently,
improved energy efficiency leads to an increase in the demand of energy service. This implicit
price mechanism generates a so-called rebound effect, as it partially offsets the initial efficiency
gains.
Although the existence of the rebound effect is widely acknowledged, the real debate lies
in the identification and the size of the effect (Greening, Greene, and Difiglio, 2000; Gillingham
et al., 2013). The discussion on the extent of the rebound effect has led to different views on the
role of energy efficiencypolicies in addressing climate change (Borenstein, 2015). Thus far, due to
the uncertainty regarding its actual size, the rebound effect has been disregarded in ex ante impact
assessments of energy conservation measures (e.g., building regulations and energy efficiency
subsidy programs), leading to perhaps misguided expectations about the role of these measures in
saving energy (Jacobsen and Kotchen, 2013; Fowlie, Greenstone, and Wolfram, 2015). This is of
importance, as realized savings ultimately determine the success of energy efficiency policies in
reducing energy consumption and carbon emissions. Incorporating the rebound effect into policy
evaluations can thus help to develop cost-effective energy conservation policies.1
In this study, we address some of the limitations in the current literature that focuses on the
identification of the rebound effect. This is the first study that is based on a large, representative
sample of dwellings, using a continuous energy efficiency measure. We analyze a detailed panel
data set that covers both the individualengineering predictions and the actual energy consumption
of 560,000 households in the Dutch housing market. Exploiting the widespread diffusion of home
energy performance certificates (EPCs), which are mandatory in all member states of the European
Union (EU), we investigatethe elasticity of actual energy consumption relative to the engineering
predictions of energy performance. In addition to the use of an extensive data set, we benefit
from different identification strategies to identify the magnitude of the rebound effect and its
heterogeneity among households.
First, we address the issue of potential random measurement error in the engineering pre-
dictions of energy efficiency, as it might lead to a downward bias in ordinary least squares (OLS)
estimates. It is possible that the engineering predictions of energy efficiency include a random
measurement error, because of assumptions made in the calculation procedure and potential
mistakes made during the inspection. In order to eliminate the effect of this type of error, we
apply an instrumental variable (IV) approach. Our identifying assumption is that engineering
models do not include a nonrandom measurement error. Although we discuss the plausibility of
our identifying assumption in detail in the following sections, we should note that our parameter
estimate will be sensitive to any systematic mistake in engineering models. Using the year of
construction and the stringency of building codes at the time of construction as instruments, we
document that, on average, the rebound effect for residential heating is 41.3% for tenants and
26.7% for homeowners.
Another concern about the identification of a rebound effect relates to the potential en-
dogeneity problem originating from unobserved household heterogeneity. Although we control
for a large set of observed household characteristics such as income, size, employment status,
gender, and age, unobserved household characteristics correlated with the energy efficiency of
the dwelling and the household’s energy demand may still exist. Exploiting the panel structure
of our data set, we estimate a fixed-effects model by tracking the movements from one address
to another address for the same households over time. The address change generates variation
1It is important to note that, as the rebound effect is a reoptimization as a response to implicit price changes, it can
be regarded as welfare improving according to neoclassical economic theory.On the other hand, its extent has important
implications on the outcomes of energy conservation policies.
C
The RAND Corporation 2017.
AYDIN, KOK AND BROUNEN / 751
in the energy efficiency level of the home, while keeping the characteristics of the household
fixed. Using this approach, we document that fixed-effects (FE) results are comparable to the
cross-sectional IV estimates.
Third, we corroborate our findings through a quasiexperimental analysis. Although a fixed-
effects estimator controls for the potential differences in unobserved household characteristics,
it is not able to eliminate the influence of unobserved home characteristics that might also
lead to a bias in the estimated rebound parameter. We estimate the rebound effect based on a
subsample of dwellings that benefited from an energy efficiency subsidy program initiated bythe
Dutch government. Using this quasiexperimental setting enables examining the rebound effect,
while controlling for unobserved home and household characteristics at the same time. For the
households that participated in the energy efficiency subsidy program, wes howthat the efficiency
improvements in their homes led to a rebound effect of around 55%.
Wethen explore the heterogeneity of the rebound effect, which may help to better understand
our findings. The identification of heterogeneity in the rebound effect mayalso contribute to better
assessment of potential outcomes from energy efficiency policies. As discussed by Borenstein
(2015), the size of the rebound effect might differ across households that are targeted by energy
efficiency regulations. For instance, low-income households, which are more likely to reside in
poorly insulated homes, might be more responsiveto efficiency improvements as these households
are expected to be more cost-sensitive (having higher price elasticity). In that case, efficiency
policies that are specifically targeting inefficient dwellings will result in a rebound effect that is
higher than average.
We separately estimate the models for cohorts of households with different income and/or
wealth levels. We document that the rebound effect is strongest among lower income groups—
these households are likely to be further from their satiation in consumption of energy services,
including thermal comfort (Milne and Boardman, 2000). This result can also provide guidance
regarding policy expectations for differentregions of the world (or within countries) with different
income levels.
Another source of heterogeneity might be the variation in the energy use intensity of house-
holds. As the cost of heating is higher for households that are more energy dependent, these
households may display a stronger response to energy efficiency changes. Identification of this
heterogeneity may also help to predict how the size of the rebound effectmay vary for other resi-
dential energy services that require different amounts of energy input. Using a quantile regression
approach, we examine whether the magnitude of the rebound effect depends upon the actual en-
ergy use intensity of households. Wefind that the rebound effect is larger among consumers with
relatively high energy consumption. This result implies that the magnitude of the rebound effect
is also determined by the energy requirement of the demanded energy service, which can partially
explain the variation in the rebound effect documented for different energy-usingproducts (such
as automobiles, air conditioners, lighting, etc.).
The results of this article have some implications for policy makers. There is much excite-
ment about the potential for energy savings, and thus reductions of carbon emissions, from the
residential and commercial building sectors. Some estimates indicate that it is the built environ-
ment where such savings come at a financial return rather than just at a capital cost (Enkvist,
Naucl´
er, and Rosander, 2007). However, in the current debate on energy efficiency, program
evaluations on, for example, the effects of subsidies and rebates are often based on engineering
calculations of energy savings. Although the behavioralresponse of consumers through a rebound
effect should be “no excuse for inaction” (Gillingham et al., 2013), it needs to be incorporated
in models of projected energy savings through energy efficiency measures that governments and
public policy outfits often employ. Using these adjusted, more realistic models may increase
the effectiveness of policies targeting energy efficiency measures. This holds for governments
in EU member states when it comes to, for example, the deployment of mandatory disclosure
schemes through energy performance certificates, but also more generally for countries outside the
C
The RAND Corporation 2017.
To continue reading
Request your trial