Housing is a heterogeneous good (Sheppard, 1999). Hedonic pricing models attempt to capture the internal and external attributes of housing that increase (or decrease) consumer utility (Rosen, 1979). Assuming perfect competition, the implicit price
55 Tables B.4 – B.7 in Appendix B3 shows that the significance levels and signs for the estimated
p(z)=(z1,…, zn) is set per the vector of attributes that increase (or decrease) the value of the good in response to the forces of supply and demand (Rosen, 1979; Sheppard, 1999). Internal elements contained in Z include the energy performance of the building among other attributes, but also may include the cost of signalling such information and the relative contractual costs in an imperfect market. Thus, a hedonic pricing model is estimated using (log-linear) OLS regression to investigate the link between the rental value of housing by landlord type, EPC rating and decision by the landlord to disclose the EPC to the tenant prior to the rental agreement:
ln(RENTi(t)k)= α + θ1LTYPEi(t)|k=1+ γ1EPC
i(t)k+ ρ1LTYPEi(t)|k=1·EPCi(t)k+
ζjDISC
i(t)k 4
j=1
+ δdDISCi(t)k·EPCi(t)k
4
d=1
+Xi(t)k' β +Zi(t)k' β + γStk+ µRi(t)k+ εi(t)k
(8)
Due to the relevant questions appearing only in the last two available surveys, the data is taken from and pooled between the fiscal years t=2012/13 and 2013/2014. The dependent variable lnRENTi(t)k represents the natural log of weekly rent paid for privately rented and social housing. Subscript i(t), variables X, S, R andthe residual ε carry the same denotation as in Equation 7. Similar to Levinson and Niemann (2004) and Choi and Kim (2012), the analysis is extended to include an additional set of local demand and housing attributes (contained in Zi) to further control for heterogeneity in the pricing function.
Equation 8 is estimated separately by rented subsector, represented by subscript k=1,…,3. Thus, the regression estimates correspond to housing rented in the private rental sector (k=1), housing let by the local authority (k=2), and registered social
landlord (RSL) (k=3). This procedure eases the comparison of coefficients for tenants across different types of (private or social) tenure.
Furthermore, it is important to note that in Equation 8, the variable LTYPE is an indicator variable, set equal to 0 if the household rents from a private landlord and 1 if they rent from any other type of landlord. The landlord variable in Equation 8 is aggregated due to the loss of observations after limiting the sample to the last two waves of EHS. This enables a more robust estimation of the interaction term, as it increases the number of observations within each cell.
Lastly, the variable EPC is an indicator set equal to 0 if the EPC rating of the property is above the median EPC rating (i.e. A – C) and equal to 1 if the property is rated D – G. Although the current standards stipulate that private sector housing must attain an EPC rating of E, this is an abnormally low benchmark if one considers that the majority of housing has achieved D or higher and housing installed with a condensing boiler would generally achieve such a rating. Hence the analysis focuses on the effect of achieving at least the median standard attained in the current housing market on prices.
DISC is a categorical variable set equal to 1 if the tenant recalls whether the (social or private) landlord had disclosed the EPC certificate before moving into the property, 2 if the tenant does not recall whether the landlord disclosed the certificate, 3 if the tenant did not know whether she saw the certificate 56 and 4 if the household moved in
56 The household representative was asked to answer the following question: ‘Before you moved into
before 2008. Therefore, the fourth category represents a voluntary information system, since EPCs were officially introduced in the rental sector in 2008.
In line with Hypotheses 3a one would expect coefficients: θ1<0 if households renting from all other types of landlords (in comparison with private ones) pay lower rent, given the property is energy efficient; γ1<0, if energy inefficient housing earns lower rents (i.e. if a ‘price premium’ exists for energy efficiency), given the private landlord discloses the EPC information prior to the tenant moving in; and ρ1<0 if the additional effect of renting housing from all other types of landlords decreases the ‘price premium’ paid for greater energy efficiency services.
In line with Hypotheses, 3b - 3c one anticipates that: ζ1<0 if private landlords earn lower rents if they do not disclose the EPC information (i.e. they miss out on a ‘signalling’ premium), given they own energy efficient housing; and δ1>0 , representing the additional non-disclosure effect for properties that are energy inefficient (i.e. these landlords earn a premium from engaging in ‘cheap talk’).
It is important to note that the English Housing Survey focuses primarily on the structure, hazards and energy efficiency of the building and is limited in its information on internal and external characteristics that may also effect prices. A typical hedonic price function does not include the socio-economic characteristics of the household, however, due to data limitations this paper includes such variables in order to help control for consumer preferences. Reiterating the preceding discussion, estimates derived using a hedonic pricing function may suffer from potential endogeneity (as well as standard issues of functional form, collinearity etc.), therefore the results of previous studies (see, e.g. Levinson and Niemann, 2004; Choi and Kim, 2012) and the current hedonic analysis are exploratory in nature.
3.3.3 Control variables
Additional controls contained in Xi(t) include: the log transformed (equivalised and weekly) income and the square of this variable, the number of children, and
household representative person’s ageare included to control for household demand
and energy requirements (Davis, 2010; Gillingham et al., 2011; Leicester and Stoye, 2013). A series of indicator variables control for the household representative person’s characteristics, including: gender, marital status (Leicester and Stoye, 2013), highest qualification attained (Gillingham et al., 2011), ethnicity (Davis, 2010), and vulnerability. In addition, the model controls for a number of property characteristics including the type; age, size, property type, fuel type (Levinson and Niemann, 2004; Gillingham et al., 2011; Choi and Kim, 2012), and time at current address. Indicators equal to 1 if the households’ utility bills are included in the rent or whether the tenant lives rent free, and 0 otherwise (see, e.g., Gillingham et al., 2011). Finally, indicator variables controlling for villages, regions (see, e.g., Davis, 2010; Gillingham et al., 2011; Leicester and Stoye, 2013), and year are included. All variable label names and descriptions are presented in Table 17 and summary statistics are presented in Table B.1 (Appendix B2).
Additional internal and external elements included in the hedonic price analysis to control for neighbourhood characteristics (i.e. the appearance of the neighbourhood), local demand characteristics (i.e. a set of disaggregated regional indicators) and housing characteristics (i.e. adequacy of parking provision and total number of bedrooms) (Levinson and Niemann, 2004; Choi and Kim, 2012) (Table 17).
Table 17: Variable definitions and labels
Variable name Specification Description
Main independent variables
TENURE 1 = Owner-occupier (OWNER)
2 = Local authority (LA)
3 = Registered social landlord (RSL) 4 = Private rented (PRS)
Tenure status of household. Baseline group is PRS.
LANDLORD TYPE 0 = OWNER, LA or RSL
1 = EMPLOYER ORGANISATION of household member 2 = OTHER ORGANISATION 3 = RELATIVE or FRIEND of household member 4= EMPLOYER INDIVIDUAL 5 = (Standard) PRS LANDLORD
Landlord arrangement. Due to tenure baseline group being set to PRS and landlord type to standard private landlord, the baseline group overall is standard private landlord.
EPC 1 = EPC > MEDIAN (i.e. A-C)
0 = EPC ≤ D
An indicator controlling for housing rated greater than the median (i.e. A-C) or rated up to and including the median (D).
DISC 1 = Landlord disclosed EPC
2 = Landlord did not disclose EPC (UNDISCLOSE EPC)
3 = Tenant is unsure (DK DISCLOSE EPC)
4= Tenant moved prior to 2008 (PRE 2008 EPC)
Disclosure of EPC, non-
disclosure, and uncertainty
surrounding disclosure.
Household characteristics
Ln INCOME Log Income Log of equivalised weekly income
(£)
CHILDREN [0, 9] Number of children living in the
home.
FEMALE 1 = Female
0 = Male
Gender of the household
representative person.
AGE [16, 95] Age (years) of the household
representative person.
MARRIED 1 = Married
0 = Otherwise
Marital status of the household representative.
DEGREE 1 = Degree or higher
0 = Otherwise
Highest qualification attained by the household representative.
NON-WHITE 1 = Non-white 0 = White
Household ethnicity (proxied by the ethnicity of the household representative person).
VULNERABLE 1 = Vulnerable
0 = Otherwise
A vulnerable household is defined by the EHS as any household containing a member aged less than 16, a member aged 60 years or over, a member who reports a long-term physical or mental condition, or a household with the household representative person or partner who is registered as disabled.
Housing characteristics
POST 1944 1 = Built after 1944
0 = Otherwise
Year property was built.
SIZE > 70sqm 1 = Property > 70sqm
0 = Property < 70sqm
Property size measured in square meters.
PROPERTY TYPE 1 = SEMI-DETACHED house or
bungalow 2 = TERRACED
3 = DETACHED house or bungalow 4 = FLAT
Property type.
FUEL-TYPE 1 = Electricity, oil or solid fuel and
communal heating (NON-GAS) 0 = Gas
Including electricity, oil or solid fuel and communal heating.
RESIDENCE 1 = RESIDE > 2+yrs
0 = Otherwise
Time spent at current address.
RENT FREE 1 = Tenant lives rent free
0 = Otherwise
An indicator capturing whether the tenant does not pay rent.
BILLS in RENT 1 = Bills are included in rent
0 = Otherwise
An indicator representing whether the tenant’s bills are included in rent.
Regional indicators
VILLAGE 1 = Village
0 = Otherwise
Classification of whether the property is located in a village.
REGION 1 = Yorkshire and Humber 2 = North West 3 = East Midlands 4 = West Midlands 5 = South West 6 = East 7 = South East 8 = London Regional. Time indicators YEAR 1 = 2009 2 = 2010 3 = 2011 4 = 2012 5 = 2013
EHS survey year.
Additional hedonic controls
AREA APPEARANCE 1=Satisfactory (SATISFACTORY)
2=Some problems (PROBLEMS) 3=Poor (POOR)
Classification of the appearance of the local area/surroundings. DISAGGREGATED
REGIONS
1=City centre (CITY CENTRE) 2=Other urban centre (OTHER URBAN CENTRE)
3=Suburban residential
(SUBURBANRES)
4=Rural residential (RURALRES) 5=Village centre (VILLAGE) 6=Rural (RURAL)
Disaggregated regional indicators.
PARKING PROVISION 1=Inadequate street parking
2=Adequate street parking
(ADEQUATE PARKING) 3=Garage (GARAGE)
4=Other off street parking (OFF STREET)
5=No parking (NO PARKING)
Classification of the provision of parking.