Energy Efficiency Premium Issues and Revealing the Pure Label Effect.

• Author: Khazal, Aras

1. INTRODUCTION

   Climate change continues to be one of the most topical concerns of political discussion and decision-making. Measures have been taken in various parts of commercial trade to reduce energy consumption and carbon emission. The energy efficiency of buildings is a key factor for increased environmental sustainability in the housing sector, and certain regulatory actions have been made in this regard. In the EU, Energy Performance Certificates (EPCs) were introduced due to the Energy Performance of Buildings Directive in 2002. The EPC is a legal document that summarizes the energy efficiency of buildings, rated from A (low expected energy consumption) to G (high expected energy consumption). The EPC policy is aimed to provide information and awareness regarding the energy efficiency of buildings and to create economic incentives for actors to invest in environmentally friendly improvements of dwellings, since improvement of a dwelling's energy performance is expected to yield higher transaction prices and rents (Mudgal et al., 2013, p. 18). Through the CONSEED (CONSumer Energy Efficiency Decision making) project, funded by Horizon 2020, the biggest EU Research and Innovation programme to date, evidence is found that consumers are willing to pay more for energy efficient properties when the energy costs are disclosed. Research through the ongoing Horizon 2020 funded PENNY (Psychological, social and financial barriers to energy efficiency) project sets out to analyze behavioral factors, purchasing decisions, institutional conditions and broader implications regarding energy efficiency and policies in order to support the development of energy efficiency strategies, policies and programmes across Europe.

   The capitalization of energy efficiency in transaction prices and rents has been subject to much research in recent years. For example, in commercial office markets, Eichholtz et al. (2010), Wiley et al. (2010), Fuerst and McAllister (2011a), and Kok and Jennen (2012) found evidence of capitalization in both rental and sales prices, whereas Fuerst and McAllister (2011b) concluded that no relation exists between EPCs and rental prices. In residential transaction markets, Murphy (2014), Wahlstrom (2016), Harsman et al. (2016), and Fregonara et al. (2017) discovered no evidence of capitalization, whereas Brounen and Kok (2011), Ramos et al. (2015), de Ayala et al. (2016), Davis et al. (2015), Wilhelmsson (2019), Jensen et al. (2016), and Hogberg (2013) found a premium for more efficient dwellings. Aydin et al. (2017) also discovered that energy efficiency is associated with higher prices but that the signaling effect of the EPCs is not significant, as opposed to Jensen et al. (2016) who find the signaling effect to be significant. In residential rental markets, Salvi et al. (2010), Hyland et al. (2013), Cajias and Piazolo (2013), and Kholodilin et al. (2017) indicated that energy-efficient homes are associated with higher rents compared with inefficient homes. A small number of existing EPC studies exist on the Norwegian real estate market. In the residential sales market, Olaussen et al. (2017) and Olaussen et al. (2019) discovered no evidence that the disclosure of energy efficiency has any effect on prices, whereas Khazal and Sonstebo (2020) discovered a positive effect on residential rental prices. Table A1 in the Appendix offers an overview of the EPC literature with methods and major findings.

   It is challenging to assess the price implications of the EPC policy implementation. Confounding factors for the impact on prices could include sample selection (size, location, and time window) and endogeneity problems originating from omitted variables correlated with energy efficiency. Only a few studies thoroughly address this issue; for example, Wilhelmsson (2019) applied propensity score matching, spatial dependency, and controls for outliers and selection bias, and Aydin et al. (2017) applied instrumental variable (IV) and repeat sales approaches to overcome potential bias. Hence, it is still necessary to investigate the EPC policy effects on prices using different strategies: both to identify confounding factors behind the price impact of energy efficiency and to reveal the pure price effect of disclosing the dwelling's energy efficiency through label adoption.

   In this paper, we apply different identification strategies using highly representative samples for both the rental and sales markets. First, we thoroughly control for the locational effects by employing pooled cross-sectional high-dimensional fixed effects (HDFE) and panel fixed effects (FE) approaches. Next, we apply the IV approach using the total number of new EPCs (NNC) as an instrument to assess the pure label effect on prices of disclosing the dwelling's energy efficiency through label adoption. Finally, we investigate the signaling impact of label adoption over time by recursively estimating the IV regression over the sample periods.

   We find that the valuation of energy efficiency is subject to endogeneity originating from unobserved locational factors, and that dwellings with lower energy efficiency are associated with more locational bias in the rental market, while this bias is higher for the energy efficient dwellings in the sales market. Further, we find that the lower the energy efficiency, the less bias comes from unobserved quality in the sales market. Overall, improving the energy efficiency of the dwelling with one letter on the EPC rating has similar effects for both rental and sales objects, with a price impact of about 0.8-1.0%.

   Investigating the signaling effect of label adoption, we discover that the IV approach produces biased estimations if the location is not controlled for. However, after considering this issue, both the IV and panel FE demonstrate that a positive signaling effect exists on prices in both markets. In the rental market, the IV approach provides evidence that the main source of endogeneity is due to locational heterogeneity rather than unobserved quality. The recursive estimations reveal that labeling has immediate, short-run, and long-run price effects and that different effects are observed in different submarkets. The findings also highlight the possibility that different conclusions might be drawn due to sample selection issues related to time periods and submarkets, and that methodological and data limitations are essential factors that must be considered when assessing the effects of the EPC implementation.

   The rest of the paper is structured as follows. In the next section, we offer a description of the rental and sales data used in the analyses. Section 3 comprises the methodology and analyses, and Section 4 concludes.

2. DATA AND DESCRIPTIVE STATISTICS

   In Norway, the EPC policy was implemented on July 1, 2010, and the certificate follows the EU standard. At any time, the document can be updated with new information submitted by the homeowner or landlord, and the certificate is valid for 10 years. Although the EU Commission requires that certificates be included in all advertisements in commercial media when a building is announced for sale or rent, a majority of dwellings are still not labeled in the Norwegian rental market. The reasons for the insufficient labeling by landlords could be the diffusion of information from the managing body and the inefficient penalty system (Khazal and Sonstebo, 2020). However, in the Norwegian sales market, where the pricing mechanism is characterized as an English auction, most dwellings are labeled.

   Hyland et al. (2016) find evidence of bunching in the Irish residential property market, where energy efficiency assessors tend to over-evaluate the energy performances of dwellings at the thresholds between labels--indicating a potential measurement error in the labels. The authors speculate that this happens because the assessors want to generate repeat business and improve their reputation. However, Norwegian homeowners and landlords have the option of cost-free self-assessment when obtaining an EPC, and the certificate is automatically produced based on the information input. Hence, there is no third-party agent with hidden intentions and little incentive for the behavior found in Ireland. The certification system is trust-based, but it is illegal to provide wrongful information and the homeowner has the legal responsibility. Additionally, insurance is often bought to ensure that all information regarding the dwelling, including the EPC, is correct. Hence, measurement error in the EPCs seems to be less likely in the Norwegian market.

   2.1 The Rental Market Data

   In this study, we use two datasets for residential dwellings, presented in Tables 1 and 2, both provided by Norway's largest and most frequently used online advertisement site, Finn.no. The first dataset comprises nearly 670,000 rental advertisements over the period from 2011 to 2019, containing posted monthly rents, EPC information, dwelling characteristics, several property amenities, and location at the county, municipality, and zip code levels. Because we are unable to follow each unit over time, the rental data are treated as pooled cross-sectional data. The average rent for the sample period is EUR 1,031, and about 16% are labeled dwellings, distributed somewhat equally among the specific labels. The dataset is highly representative, comprising observations from all counties, 98% of municipalities, and 74% of zip codes in Norway. Boligbygg Oslo KF is a Norwegian municipal corporation for housing that every quarter, since the autumn of 2005, have conducted a separate survey aimed at those who have advertised their apartment for rent in the last quarter. They are able to link the answers from the survey with the information in the ad. In the survey they ask a number of questions that provides the opportunity to correct the difference between the advertised and actual (contract) rental...