Energy 47 (2012) 348e357
Contents lists available at SciVerse ScienceDirect
Energy journal homepage: www.elsevier.com/locate/energy
Closing the Energy Efficiency Gap: A study linking demographics with barriers to adopting energy efficiency measures in the home Marcos J. Pelenur*, Heather J. Cruickshank Centre for Sustainable Development, Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK
a r t i c l e i n f o
a b s t r a c t
Article history: Received 2 April 2012 Received in revised form 24 September 2012 Accepted 25 September 2012 Available online 23 October 2012
This paper presents a study which linked demographic variables with barriers affecting the adoption of domestic energy efficiency measures in large UK cities. The aim was to better understand the ‘Energy Efficiency Gap’ and improve the effectiveness of future energy efficiency initiatives. The data for this study was collected from 198 general population interviews (1.5e10 min) carried out across multiple locations in Manchester and Cardiff. The demographic variables were statistically linked to the identified barriers using a modified chi-square test of association (first order RaoeScott corrected to compensate for multiple response data), and the effect size was estimated with an odds-ratio test. The results revealed that strong associations exist between demographics and barriers, specifically for the following variables: sex; marital status; education level; type of dwelling; number of occupants in household; residence (rent/own); and location (Manchester/Cardiff). The results and recommendations were aimed at city policy makers, local councils, and members of the construction/retrofit industry who are all working to improve the energy efficiency of the domestic built environment. Ó 2012 Elsevier Ltd. All rights reserved.
Keywords: Cities Energy efficiency Barriers Demographics Multi-response contingency tables RaoeScott chi-square test
1. Introduction Currently, the existing UK built environment accounts for approximately 50% of total energy demand [1], and 45% of its anthropogenic CO2 emissions [1,2]. As such, improving the energy performance of buildings is a significant opportunity to help the UK government meet its long-term energy and emission goals. This opportunity is further strengthened with the estimate that between 70 and 80% of the existing poor performance buildings will still be functional in 2050. However, empirical data shows that the adoption rate of energy efficiency measures are often short of their full potential, a phenomenon termed the ‘Energy Efficiency Gap’ [3,4]. For OECD countries, this neglected energy conservation due to the Energy Efficiency Gap is estimated at 30% of the total potential energy savings [5]. The Energy Efficiency Gap is a complex issue, where technical, institutional, market, organisational, and behavioural barriers all play a significant role and are interconnected [5]. As such, the effectiveness of energy efficiency interventions may be improved by approaching this problem from an interdisciplinary perspective.
* Corresponding author. Tel.: þ44 (0)1223 333321; fax: þ44 (0)1223 765625. E-mail addresses:
[email protected] (M.J. Pelenur),
[email protected] (H.J. Cruickshank). 0360-5442/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.energy.2012.09.058
Therefore, the aim of this research was to apply social science insights to an energy engineering problem; specifically, to investigate which UK population demographic variables were associated with empirically identified barriers affecting the adoption of domestic energy efficiency measures. The goal of this research was to statistically correlate demographics with barriers, in order to improve the design and effectiveness of future energy efficiency initiatives. In order to satisfy this aim, this paper applied a modified chisquare test (first order RaoeScott corrected to compensate for multiple response data) to test the association between demographics and barriers. The subsequent effect size for each demographic level was estimated with the analysis of the corresponding odds-ratio table. The energy efficiency barriers used in this paper were empirically identified in a previous study through thematic analysis of 198 general population interviews (1.5e10 min) carried out in the urban areas of Greater Manchester and Cardiff [6]. Manchester and Cardiff were selected as case sites because they both represent large urban centres, accounting for approximately 6% of the UK population [7], [p.119] and were also identified as cities of interest by the EPSRC funded RETROFIT 2050 research project [8]. They are both cities of interest for retrofitting because “both have long industrial histories, both have suffered decline in recent decades and both are seeking to overcome this decline, regenerating themselves into modern, vibrant cities” [8].
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
The results in this paper revealed that strong associations exist between all of the barriers and most of the demographic variables. Through analysis, insights are presented as guidance to improve the design and effectiveness of future energy efficiency interventions. By improving our understanding of the link between demographic variables and barriers affecting the adoption of energy efficiency measures in the home, this research is relevant to policy makers, local councils, and members of the construction/retrofit industry who are working towards improving the energy efficiency of the built environment. 2. Background In order to tackle climate change, the UK government passed the Climate Change Act in 2008 which set an ambitious target to cut the country’s greenhouse gas (GHG) emissions from all sources by 80% (160 MtCO2e) by 2050, compared to 1990 levels [9]. The UK has remained committed to this target despite a change in national government in 2010, and in 2011 adopted the 4th Carbon Budget (path to 2030) recommended by the Committee on Climate Change (CCC) [10]. Specifically relevant to the domestic built environment, the CCC expects emissions from homes to fall by over a third of 2011 levels by 2022 [11]. The case to reduce energy and GHG emissions from the built environment is further strengthened by the estimate that over 70e80% of the existing 22 million homes in England will still be occupied and functioning in 2050 [12,2]. Therefore, it is apparent that the existing domestic built environment will need to be significantly upgraded in order to help the UK meet its longterm energy and GHG emission goals. However, the adoption rate of energy efficiency measures is often short of its full potential, a phenomenon termed the ‘Energy Efficiency Gap’ [3,4]. 2.1. The Energy Efficiency Gap The Energy Efficiency Gap is described as the gap that exists between the current or expected future energy use, and the optimal current or future energy use [4]. In OECD countries, this energy conservation loss due to the Energy Efficiency Gap is estimated at 30% of the total potential energy savings of the measures [5]. Understanding the reasons that give rise to the Energy Efficiency Gap is a well researched topic across a wide range of disciplines. From a standard economic perspective, Jaffe and Stavins [4] sought to understand why compact fluorescent light bulbs, improved thermal insulation materials, and energy-efficient appliances were not more widely adopted. In their research, they argued that the Energy Efficiency Gap is due to market failures, such as a lack of transparent information about the benefits of energy efficiency, and non-market failures, such as the transaction costs of adopting new technology or the use of inaccurate discount rates by consumers making energy efficient retrofit decisions. Along a similar economic analysis Weber [5] identified four main types of obstacles for energy efficiency measures: institutional barriers (public government); market barriers; organisational barriers; and behavioural barriers. Again, similar to Jaffe and Stavins, Stern [13] suggested that the barriers to rational behaviour are: financially hidden costs/benefits; conflicting market signals or imperfect information; and motivation factors. Complementing this economic interpretation, there is also considerable relevant extant psychological and sociological scholarship to help understand the reasons for the Energy Efficiency Gap. In their review paper for the Living with Environmental Change programme, Upham et al. [14] brought together an extensive list of psychological and sociological studies to help explain public attitudes to environmental change. Two relevant implications from the paper are that individuals’ attitudes and actions were not always
349
consistent (defined as the ‘value-action’ gap), and that an individual’s behaviour in one context may be inconsistent with their behaviour in another context [14]. Such insights may help explain why households do not adopt energy efficiency measures even though they are economically justified. From a more detailed behaviour perspective, the Behaviour Change Knowledge Review referenced over 60 relevant socio-psychological models, theories and frameworks [15] which can be used to help understand the Energy Efficiency Gap. Most models presented in the paper use the variables of ‘attitudes’, ‘norms’, and ‘agency’ to explain behaviour, while others also include ‘habit’ and ‘emotion’ [15]. To test the relative importance of such psychological variables to energy use, as well as socio-demographic variables, Abrahamse and Steg administered questionnaires and examined the energy use of 189 Dutch households [16]. They found that household energy consumption was mainly determined by socio-demographic variables, whereas energy savings (viz., changes in behaviour) were mainly determined by psychological factors [16]. Supporting these results, Faiers et al. argued that policy makers should consider a broad range of factors, such as individuals’ cognitive abilities, values, attitudes, and social networks in the context of understanding consumer domestic energy use [17]. Without this broad understanding, it is possible that well intentioned energy policy strategies backfire, leading to more energy use rather than less (termed the “Rebound Effect”) [18]. From a sociological perspective, Lutzenhiser introduced the idea of an integrated Energy Culture model to understand behaviour, which considers social norms and culture alongside the more traditional econometrics [19]. More recently, Stephenson et al. [20] applied the Energy Culture framework to consumer energy behaviour, specifically examining the interactions between cognitive norms (e.g. beliefs, understandings), material culture (e.g. technologies, building form), and energy practices (e.g. activities, processes). Such interactions highlight the complexity of the Energy Efficiency Gap, and reinforce the benefit of interdisciplinary research on the barriers preventing the adoption of energy efficiency measures in the home. 2.2. Barriers to energy efficiency The barriers used for the analysis in this paper were identified from interviews in an earlier study [6], and are summarised in Section 3. Table 1 shows the collected demographic variables and identified barriers which are discussed individually in the following sub-sections. 2.2.1. [B1] Beliefs/information This barrier related to information (or lack of) about energy efficiency measures, and beliefs which affect the adoption of such
Table 1 List of barriers and collected demographic variables. Barrier
Demographic variables
[B1] Beliefs/information [B2] Cost [B3] Family/partner/housemate [B4] Institutional [B5] Landlord-tenant/housing associations [B6] None (no barriers) [B7] Personal behaviour [B8] Property itself
[D1] [D2] [D3] [D4] [D5]
Sex Age Household income Marital status Education level
[D6] Type of dwelling (flat, detached, etc.) [D7] Number of bedrooms in household [D8] Number of occupants in household [D9] Residence (own, rent, live with family/friends) [D10] Location (Manchester/Cardiff)
350
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
measures. Examples cited in the interviews: a lack of expertise/ knowledge of what to do; unclear or lack of trustworthy information from government; and mistrust of energy companies or contractors. This type of barrier was also suggested by Stern [13] and described as a cognitive norm factor affecting energy use [19,20]. 2.2.2. [B2] Cost The upfront cost of energy efficiency measures was the most frequently mentioned barrier, and is the economic barrier which receives the most attention in similar studies [6]. However, there is some complexity within this barrier, for example in the idea of discounted costs versus perceived benefits. Specifically, if the benefits of the measure are not correctly valued by the home occupant, then reducing the price of the measure may not necessarily increase its up-take. 2.2.3. [B3] Family/partner/housemate This barrier is often ignored by the pure technical/economic perspective, but was frequently mentioned in interdisciplinary scholarship [21e23], and represented a significant proportion of responses in this study. This barrier included inter-occupant opposition towards energy efficiency measures, specifically from husbands/wives/partners, as well as apathy from other family members, particularly children. The challenge of reaching consensus in multi-tenanted homes was also captured within this barrier. 2.2.4. [B4] Institutional Institutional barriers related to the perception from some interviewees that the government and/or energy companies were the main barrier towards adopting energy efficiency measures. Examples cited in the interviews: government incentives are incorrectly targeted; energy companies are unwilling to sincerely promote energy efficiency; and that consumer choice is being actively hampered by government and energy companies. This type of barrier was also suggested by Weber [5]. 2.2.5. [B5] Landlord-tenant/housing associations This barrier referred to the split-incentive between landlords/ housing associations and their tenants. Specifically referring to the dilemma that landlords do not want to invest in energy efficiency measures, since they do not benefit from the corresponding reduced energy bills. Similarly, tenants are unwilling to invest in energy efficiency measures for homes they do not own. 2.2.6. [B6] None (no barriers) This is the non sequitur barrier of none (representing no barriers). It captured interviewees who felt there were no barriers for them to adopt energy efficiency measures. 2.2.7. [B7] Personal behaviour This is a complex barrier which conflated behaviour and attitudes, two variables which are often separated from a psychological perspective but combined in this study as a social norm [19,20,23]. Specifically Stephenson et al. [20] defines ‘energy practices’ as activities and processes relating to energy use. Based on the interview transcriptions, this barrier was subdivided into the following areas: a feeling by interviewees that they have already done everything possible; current lifestyle choices; and interviewees who consciously do not want to adopt energy efficiency measures. Lifestyle in itself included the themes of laziness; lack of time; convenience; and forgetfulness, as well as ideas that highlight the rebound effect, such as: keeping up with appearances, and a desire for more gadgets.
2.2.8. [B8] Property itself This barrier encompassed the sub-barriers of the physical property, and conservation & heritage. The property sub-barrier referred to the limitations that the property structure itself imposes on residents, for example the space available in the home, its age, or unsuitable loft space roofs. The conservation & heritage sub-barrier captured the case in which owners were unable to install energy efficiency measures because of planning issues, specifically if they were either listed buildings or in a conservation area. 3. Method The purpose of this paper was to investigate the association between demographic variables and barriers identified from general population interviews in Manchester and Cardiff (demographic variables and barriers shown in Table 1). The interview questions were first piloted twice in Cambridge to ensure that the wording of the questions were easily understood by participants, yet open enough to allow for varied responses. See Pelenur and Cruickshank [24] for a discussion of the pilot study results. The first question asked, “Is there anything you would like to change about how your household uses energy? If yes, what? And why? And how would your household go about making the change? What are the drivers?” The second question asked, “What are some of the barriers preventing your household from making the change?”. It is relevant to note that the questions were worded to encourage the interviewee to adopt a ‘household’ perspective, rather than simply an ‘individual’ viewpoint. This was a deliberate phrasing of the question so that the results would not only include individual behaviours, but also insights about how relationships between family members or multi-tenanted homes affect consumption. The interviews were conducted in three different locations in each city, so that as many divergent viewpoints as possible could be captured. The Manchester set of interviews took place over three days of a bank holiday weekend in April 2011, and elicited 100 interviews in three locations. On day one, the interviews were conducted in front of an ASDA store in the neighbourhood of Hulme; on day two, in the Trafford Centre (up-scale retail); and on day three, in the city centre. The Cardiff set of interviews followed the same pattern and also took place over three days of a bank holiday in May, and elicited 98 interviews in the following three locations: day one, in front of a Super ASDA; day two, in front of a suburban Tesco; and day three, in the city centre. The varied locations were selected in order to reduce the sample bias from a single location and the interviews were conducted over bank holiday weekends to increase the range of demographics capable of participating (since the majority of people were off work). After transcribing the interviews, the barriers were identified through thematic analysis, which involved recognising important information and encoding it prior to a process of interpretation [25]. The identified codes were then used to organise the data in order to identify and develop the barriers [26], as such, identified barriers represented specific patterns found in the data [27]. An advantage of this approach was that it provided a method for the systematic detailed analysis of qualitative data; although, a disadvantage is that the barriers identified should to be tested against a larger statistically representative sample in order for them to be more generalisable. See Pelenur and Cruickshank [6] for a detailed description of the sampling strategy, coding methodology, and qualitative analysis of the barriers. For this study, the barriers attributed to each participant were linked to their demographics, and statistically analysed by examining contingency tables formed between each demographic variable and all the barriers. For the analysis, all the demographic
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
variables were treated as categorical, i.e. levels were assigned to each variable, in order to allow for consistent tests of association (the levels for each variable are shown in the Figures of Section 4). The threshold values which defined each level were selected in order to ensure as closely as possible that “no more than 20% of the expected counts [in the resulting contingency tables] are less than 5, and all individual counts are 1 or greater” [28], [p. 734]. The test of association for contingency tables with categorical data is normally done with a Pearson’s chi-square test of association; however, the data for this study were multi-response, specifically because the interviewees were asked to list ANY barriers towards the adoption of energy efficiency measures; rather than a SINGLE barrier. Therefore, a traditional Pearson chi-square was inappropriate since one of the categorical variables had multiple responses (Barriers). Specifically, there was within-participant dependence among responses which invalidated the independence of observation assumption underpinning a Pearson chi-square test [29]. Table 2 is an example of this multi-response data, shown as a contingency table between the demographic variable of Sex and the identified barriers. Table 2 highlights the within-participant dependence problem. Note that since some interviewees mentioned multiple barriers, the total number of responses for females was 135 with only 88 females interviewed. As such, a single by multiple (SM) response test was required. 3.1. Categorical multiple response survey data Despite the historic use of multiple response surveys, tests for association for this type of data have only recently been proposed during this last decade. As listed by Thomas and Decady [30], current tests fall into two classes: the first, ‘bootstrapping’ a suitable test statistic when its distribution is not known exactly [31]; and the second, approximating chi-squared tests [32]. The latter are of particular interest to the current study because of their familiarity and close relation to the classical Pearson chi-square test [30], which is widely understood and recognised as a standard statistical test. In order to retain intuitive familiarity with the results, this paper used a modified first-order corrected RaoeScott chi-square 2 ðALÞ, proposed by Agresti and Liu statistic [33,34], denoted as XSM [35], and Thomas and Decady [30,36]. As well as testing for association, the strength of the resulting association was also examined with a corresponding odds ratio table, as per Thomas and Decady [30]. 2 ðALÞ is calculated for a r c data Briefly, the test statistic XSM table by summing up the individual Pearson chi-square statistics for each of the c marginal r 2 tables relating the single response variable to the multiple response variable with df ¼ c(r1) [35]. Agresti and Liu found that this approach yielded a chi-square test statistic numerically similar to those produced by other asymptotically correct procedures [35]. Thomas and Decady [36], and 2 ðALÞ can be Bilder et al. [37] also independently showed that XSM regarded as a member of the familiar RaoeScott corrected chisquared family of tests for complex surveys [30]. As such, the 2 ðALÞ statistic is a simple and parsimonious approach to apply to XSM categorical multiple response data. For example, in Table 2, the Table 2 Contingency table of [D1] versus barriers. Sex [D1]
Barriers B1
B2
B3
B4
B5
B6
B7
B8
Female Male Total
12 4 16
34 31 65
18 7 25
4 11 15
10 18 28
13 17 30
19 14 33
25 15 40
Total number of responses
Total number of participants
135 117 252
88 85 173
351
eight marginal Pearson statistics are 4.11, 0.09, 5.22, 3.85, 3.07, 0.82, 2 ðALÞ 0.73, and 2.82, each having df ¼ 1. Therefore, the resulting XSM statistic is the summation of these values and equals 20.71 with df ¼ 8, p ¼ 0.008. This approach can be re-written as a single equation by considering a single by multiple response data table with n participants, and with row and column variables which consist of lists of items of length r and c. In this table, let mij denote the number of participants (out of a total n) who selected item i on the first variable and item j in the second. The number of participants responding in row i of the table is denoted by ni for the single response variable i ¼ 1,.,r, and the marginal count, m.j, j ¼ 1,.,c denotes the number of participants selecting item j in the multiple response variable column (irrespective of row selection) [30]. Thomas and Decady [30] and Bilder et al. [37] showed that with the 2 ðALÞ can be calculated simply as: above notation and definition, XSM 2 ðALÞ ¼ XSM
2 m ni m:j =n ij ni m:j =n 1 m:j j¼1
r X c X i¼1
The probability pij that a participant will respond positively to item i of the row variable, and item j of the column variables is b ij ¼ mij =n respecdefined and estimated as pij ¼ E(mij)/n and p tively, where E(.) denotes expectation [30]. Similarly, the one-way marginal probabilities pi and p.j are defined as pi ¼ E(ni)/n, b :j ¼ m:j =n. Therefore, the hypothb i ¼ ni =n and p:j ¼ Eðm:j Þ=n, p p esis for row by column marginal independence is expressed as:
H0 : pij ¼ pi p:j H1 : At least one equality does not hold The above hypothesis of marginal independence is equivalent to the odds ratio hypothesis Fij ¼ 1ci,j so that as well as applying the 2 ðALÞ test statistic, the data can also be displayed in terms of XSM odds ratios, which can be used to examine the strength of association [30]. 4. Results In total, 198 semi-structured interviews were conducted. From this total, 25 interviews could not be used because the topic of barriers was not discussed. Table 3 shows the summary
Table 3 Summary demographics. Demographic variables Sex Female Male Age Under 30 30e45 45e60 Greater than 60 Education Degree or more High school/Trade Number of bedrooms 1e2 3 Greater than 4 Number of occupants 1e2 3 or greater
Percent of participants 51% 49% 19% 34% 31% 15% 56% 44% 37% 36% 27% 52% 48%
Demographic variables Location Manchester Cardiff Income £40k and less Greater than £40k Refused Marital Status Single/widowed Married/common law Type of dwelling Flat/apartment Terrace (end or mid) Semi/detached house Type of tenure Own Rent/live with family/friends
Percent of participants 43% 57% 50% 36% 14%
45% 55% 24% 26% 51% 58% 42%
352
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
demographics from the remaining participants (percentages do not always sum to 100 due to rounding). Following the demographics, a frequency graph of the barriers is shown in Fig. 1. Based on the null hypothesis for marginal independence with 2 ðALÞ, the following demoa ¼ 0.05 and the test statistic of XSM graphic variables shaded in Table 4 were found to be significantly associated with the identified barriers. As shown in Table 4, most of the demographic variables were significantly associated with barriers, with the exception of [D2] Age, [D3] Household income, and [D7] Number of bedrooms in household (a typical proxy for size [38]). As well as testing for association, it is informative to examine the strength of such associations using the odds-ratio (F); where an odds-ratio of greater than 2 is considered a strong association [30]. For clarity, the odds-ratio results and contingency table proportions are visually shown with shaded mosaic plots in Section 5. Mosaic plots essentially visualise the data as “tiles” representing the cells of the table, such that the area of each tile is proportional to the cell frequency. Specifically, the vertical height of each row represents the barrier frequency proportion and the horizontal width of each column represents the demographic variable frequency proportion [39]. The “tiles” in all the mosaic plots are also shaded for significant odds ratio results. Specifically, heavily shaded tiles represent an odds ratio of greater than 2, while lightly shaded tiles represent an odds ratio between 1.5 and 2. 5. Discussion It is immediately striking from Table 4 that occupant age, household income, and the number of bedrooms in the household were not significantly associated with any specific energy efficiency barrier. This is an interesting contrast to other studies which found that age and house size were factors associated with energy consumption [16,40] (see Guerin et al., Lenzen et al. [40,41] for a review of studies investigating predictors of household energy consumption). So even though these demographic variables were associated with energy consumption, the results from this study imply that they were not associated with any barrier to adopting domestic energy efficiency measures. As such, specifically targeting energy efficiency interventions to certain age demographics is not likely to increase the effectiveness of overcoming any specific barrier. This result is not meant to undermine education and awareness campaigns about energy efficiency for the elderly and young, but instead implies that such campaigns should be broadened to include all age groups, rather than specific age segments. From the physical built environment perspective, it is also interesting to note that the number of bedrooms in a home (a proxy for size) was not a statistically significantly variable. Although
Table 4 Shaded demographic variables are significantly correlated with barriers (p < 0.05). Demographic variables
2 ðALÞ XSM
df
p value
[D1] Sex [D2] Age [D3] Household income [D4] Marital status [D5] Education level [D6] Type of dwelling [D7] Number of bedrooms in household [D8] Number of occupants in household [D9] Residence (own, rent, live with family/friends) [D10] Location (Manchester/Cardiff)
20.71 29.38 17.14 15.51 20.57 29.21 25.40 19.64 66.89 37.92
8 24 16 8 8 16 16 8 8 8
0.008 0.206 0.377 0.050 0.008 0.018 0.063 0.012 w0.000 w0.000
similar to occupant age, it is empirically shown that as the size of a house increases, so does its energy consumption [16,40]. The result in this paper suggests that larger homes, which consume the most energy, were not associated with any specific barrier, and therefore cannot easily be targeted with a barrier-specific incentive. Surprisingly, household income was not correlated with any barrier, remarkably not even Cost. This lack of association indicates that the upfront cost of energy efficiency measures may not be as important as often expected for low income families, or re-phrased, that other barriers are as equally important as cost. From a demand perspective, this result was in line with other studies reviewed by Guerin et al. [40] which found no consensus linking income and energy consumption. As such, this result implies that price incentives or subsidies that address the cost of energy efficiency measures may be equally effective for high income families as for low income families. The following sub-sections discuss each of the remaining significant variables in detail. 5.1. [D1] Sex The mosaic plot in Fig. 2 shows that females were strongly associated (odds ratio greater than 2) with the barriers [B1] and [B3] (Beliefs/information and Family/partner/housemate), while males were more strongly associated with [B4] and [B5] (Institutional and Landlord-tenant/housing associations). Females were also weakly associated with barrier [B8] (Property itself). It is not immediately clear why the barriers were split along these gender lines, but it was an interesting result which should encourage further research by gender behaviour specialists. As a hypothesis, perhaps the woman interviewed were more concerned with ‘internal’ barriers such as beliefs and family, while the men were more concerned with ‘external’ barriers such as institutions and landlords. In the context of energy efficiency, Burgess and Nye [42] also found unexpected gender differences when
0
Fig. 1. Barriers towards the adoption of energy efficiency measures in the home.
60
70
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
353 Marital.status
[B 1] Be lie fs
Male
5.4. [D6] Dwelling As expected, Fig. 5 shows that residents living in apartments or flats were strongly associated with barrier [B5] (Landlord-tenant/
High school / trade
ie
fs
Education Degree / more
ts [B
2]
C
Fig. 3 shows that marital status was strongly associated with barriers [B5] and [B8] (Landlord-tenant/housing associations, and Property itself). Marital status was also weakly associated with [B1] and [B3] (Beliefs/information and Family/partner/housemate). Specifically, married/common-law interviewees were strongly associated with the Property barrier, while currently single interviewees were strongly associated with the Landlord/tenant/ housing associations barrier. Upon closer examination of the data, this result was not surprising since married/common-law interviewees in the sample were correlated to property ownership, and correspondingly single interviewees with tenancy (c2 ¼ 16.69, df ¼ 1, p < 0.001). The importance of this result is that interventions which specifically target barriers [B5] and [B8] should be designed to consider marital status as an important factor.
os
5.2. [D4] Marital status
It was also interesting to compare the results between education and income. Why is the level of education correlated with barriers while income is not? This result indicates that using education level as a factor to address energy efficiency barriers is more effective than income. Having a degree or more was also strongly associated with both barriers [B7] and [B8], implying that any intervention which specifically targets the landlord-tenant split incentive or the physical property should specifically consider education level as an important variable.
]B el
measuring the impact of energy monitoring equipment in the home.
Fig. 3. Mosaic plot for marital status with the left-hand side shading for ‘single’ (total number of responses 108 from 77 participants) and the right-hand side shading for ‘married/common-law’ (total number of responses 144 from 96 participants) representing the strength of the odds-ratio.
[B 1
Fig. 2. Mosaic plot for sex with the left-hand side shading for ‘females’ (total number of responses 135 from 88 participants) and the right-hand side shading for ‘males’ (total number of responses 117 from 85 participants) representing the strength of the odds-ratio.
[B [B [B4 [B [B [B ] 7] 5] 8] 6] Pe La Ins 3] F Pr N tit op am nd rs o u on ne t er lo rd iona ily ty al l
Barriers
on al
7] Pe rs ro pe r ty
[B
8] P
[B
Married/commonlaw
[B 2] C
2] C [B [B 4] [B 5] La Ins 3] F tit am nd on u t lo e rd iona ily l [B
6] N
[B
Barriers
Single
os ts
os ts
[B
1] B
el ie fs
Sex Female
l na
itu
tio
[B 3]
rd
st
lo
In nd
4]
La
[B
8]
Pr
op
er
ty
[B 7]
Pe r
so n
al
[B 6]
N
on
e
[B
5]
[B
With respect to education level, Fig. 4 shows that interviewees with a single university degree level qualification or more were strongly associated with barriers [B5], [B7], and [B8] (Landlord/ tenant/housing associations, Personal behaviour, and Property itself), and interviewees with up-to a high school diploma or trade qualification were only weakly associated with barriers [B3], [B4], and [B6] (Family/partner/housemate, Institutional, and None). This result indicates that individuals with a degree or more of education saw their personal behaviour/lifestyle as a barrier (i.e. they may be less willing to compromise comfort or time). Investigating this association is an area for further research, for example, does education lead to an increase in self-awareness and therefore a better ability to acknowledge one’s own lifestyle as a barrier? Or, are the educated simply more selfish and less willing to adapt and compromise their lifestyles?
Barriers
Fa m
ily
5.3. [D5] Education level
Fig. 4. Mosaic plot for education level with the left-hand side shading for ‘degree/ more’ (total number of responses 143 from 91 participants) and the right-hand side shading for ‘high school/trade’ (total number of responses 97 from 71 participants) representing the strength of the odds-ratio.
354
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357 Number.of.occupants 1−2
fs
Semi/detached home
3 or greater
lie
Terrace
tio
na
l
Fa m itu
rd
st
lo
In 4]
nd
[B
La
5]
Barriers
rd lo
[B
3]
l na
tio
on
e
[B 6]
N
N
on
e
[B
Pe r
so
so
na
l
na
l
[B
6] [B
7]
Pe r
Pr
op
op
er
er
ty
ty
[B
7]
8] [B
The location demographic variable is of particular interest, because it highlighted city level barrier correlations. Specifically, Residence Rent/live with family/friends
tio
La
nd
4]
lo
In
rd
st
itu
[B
3]
Fa m
ily
[B
2]
C
os
ts
[B
1]
Be
lie
fs
Own
5]
[B
e
7] [B
8]
Pr
op
er
ty
[B
The results shown in Fig. 7 for variable [D9] Residence were consistent with the above insights which suggested that owner occupied homes were strongly associated with barrier [B8] (Property itself). Correspondingly, Fig. 7 also showed a strong relationship between tenants and barrier [B5] (Landlord-tenant/housing associations). This indicates that interventions which address the property barrier should be targeting owner occupied households and interventions which address the landlord-tenant split incentive barrier
Pe r
so n
al
[B
5.6. [D9] Residence
6]
N
on
Not surprisingly, Fig. 6 shows a strong association observed between households with 3 or more occupants and barrier [B3] (Family/partner/housemate). This result is of interest because it highlighted the importance of inter-family/occupant relationships. Often the household is treated as a single unit by economic models; however, this result showed that the inter-occupant relationships are the most significant barrier for homes with more than 3 occupants. This was consistent with other studies which found that inter-occupant relationships were an important factor affecting the adoption of energy efficiency measures [21e23]. This result implies that interventions to promote energy efficiency in homes with 3 or more occupants should be designed to specifically address the family/partner/ housemate barrier.
5.7. [D10] Location
Barriers
5.5. [D8] Number of occupants
should be targeting tenanted households (not surprisingly). These results were expected a priori, but it was reassuring to observe them empirically. Owner occupied homes were also weakly associated with barriers [B2] and [B6] (Cost and None/no barriers), and tenanted properties were weakly associated with [B4] and [B7] (Institutional and Personal behaviour). The underlying reason for these weak associations was not immediately apparent, but may be an area for future research.
[B
housing associations), which was not a surprising result given that apartment and flat residents were more likely to be tenants (c2 ¼ 28.79, df ¼ 2, p < 0.001). However, it is interesting to note that terraced homes (end or mid) and semi/detached homes were not strongly associated with barrier [B8] (Property itself). This result implies that for terrace and semi/detached homes, the other barriers were of equal importance. The one exception is barrier [B1] (Beliefs/information) which was strongly associated with semi/detached homes. The reason for this association was not immediately apparent, but may be an area for future research.
Fig. 6. Mosaic plot for number of occupants with the left-hand side shading for ‘1e20 (total number of responses 120 from 90 participants) and the right-hand side shading for ‘3 or greater’ (total number of responses 132 from 83 participants) representing the strength of the odds-ratio.
l
[B Pr 8] [B
Fig. 5. Mosaic plot for dwelling with the left-hand side shading for ‘flat’ (total number of responses 58 from 41 participants), the middle shading for ‘terrace’ (total number of responses 61 from 44 participants), and the right-hand side shading for ‘semi/detached home’ (total number of responses 131 from 87 participants) representing the strength of the odds-ratio.
na
[B
itu st
In 4]
nd
[B
La
5]
Barriers
3]
Fa m
ily
ily
[B
[B
2]
2]
C
C
os
ts
os
ts
[B
1]
[B
1]
Be
Be
lie
fs
Dwelling Flat
Fig. 7. Mosaic plot for residence with the left-hand side shading for ‘own’ (total number of responses 140 from 100 participants) and the right-hand side shading for ‘rent/live with family/friends’ (total number of responses 112 from 73 participants) representing the strength of the odds-ratio.
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
Location Cardiff
[B ]I ns 3] F tit ut am dl io or na i l y d l an
[B 4
]L [B 8
]P
ro p
er
ty
[B 7
]P er so
na
l
[B 6
]N on
e
[B 5
Barriers
[B 2
]C os
ts
[B 1
]B
el
ie
fs
Manchester
Fig. 8. Mosaic plot for location with the left-hand side shading for ‘Manchester’ (total number of responses 119 from 75 participants) and the right-hand side shading for ‘Cardiff’ (total number of responses 133 from 98 participants) representing the strength of the odds-ratio.
Fig. 8 showed that Manchester was strongly associated with [B1], [B2], [B4], and [B7] (Beliefs/information, Cost, Institutional, and Personal behaviour) and weakly associated with [B3] (Family/ partner/housemate); while Cardiff was only strongly associated with [B6] and [B8] (None/no barriers, and Property itself). This difference should be of interest for the local governments in each city, since the results allow local councils to design interventions to specifically address strongly associated barriers. These results can also guide further research at the regional level. For example, future research may investigate why Beliefs/ information and Institutional barriers were strongly associated with Manchester over Cardiff. Was there a greater level of mistrust towards the Manchester city council versus the Cardiff city council? If so, what lessons from Cardiff can be applied in Manchester? Similarly, was there a price difference for energy efficiency measures between the two cities, or was the Manchester and Cost barrier association due to resident perception differences? 6. Limitations The results and discussion highlighted strong and weak associations between demographics and barriers; however, as with all complex surveys, there are caveats that require addressing. The underlying limitation of this analysis was the possibility of sample bias, since the street interviews were only conducted over 3 days in each city. However, previous research [6] attempted to reduce any
355
possible bias by varying interview locations, and choosing to conduct the interviews over an entire bank holiday weekend, when the majority of people were off work (i.e. able to participate). Nevertheless, caution should be taken before generalising the results across all UK cities. A second limitation with this type of study is the temporal nature of opinions. Fundamentally, the interviews transcribed subjective barriers, i.e., barriers which were expressed as modifiable opinions from the interviewees rather than stable observed truths. As such, those opinions are liable to change with the passing of time. However, while the recorded barriers may shift, the measured demographic variables are stable and unlikely to change significantly. The consequence is that future research can easily use the same demographic variables to repeat this study, and measure any change in the resulting association with identified barriers. Finally, only individual demographic variables were used in the analysis; however, it is possible for future research to investigate the association of two or more demographic variables with each barrier. The results of any future multi-demographic analysis can be improved by increasing the total number of participants. 7. Conclusions The study presented in this paper linked demographic variables with empirically identified barriers affecting the adoption of energy efficiency measures in the home. The results revealed that strong associations exist with certain demographic variables (Fij > 2). Table 5 summarises these associations, as well as highlights where to target interventions. Specific guidance is presented to help city policy makers, local councils, and members of the construction/ retrofit industry increase the adoption of domestic energy efficiency measures. 7.1. Guidance 7.1.1. Awareness/information campaigns The results imply that awareness/information campaigns promoting energy efficiency measures should be targeting woman and residents living in semi/detached dwellings. This insight can be tested with a large scale awareness/information trial specifically targeting woman. The results of such a trial can be compared against a control group to empirically determine the effect of targeting a specific sex. Similarly, another awareness/information trial can be designed to test the effect of specifically targeting residents of semi/detached homes. 7.1.2. Upfront cost The upfront cost of energy efficiency measures was not a barrier strongly associated with any demographic variables, apart from Location (Manchester). In order to better understand why Cost is associated with only Manchester and not Cardiff, local policy
Table 5 Summary table of demographic variables versus barriers. Barriers
Demographic variables Sex
Beliefs Cost Family Institutional Landlord-tenant Behaviour Property itself
Marital
Education
Woman Woman Men Men
Dwellings
Occupants
Residence
Semi/detached
Location Manchester Manchester
3þ Manchester Single Married
Degreeþ Degreeþ Degreeþ
Flats/terraces
Tenants Owner
Manchester Cardiff
356
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357
makers should compare the two regions in greater detail for potential insights. Another consequence of this result is that the upfront cost barrier cannot be easily addressed by simply targeting a certain demographic segment. This insight raises questions about the conventional practice of subsidising energy efficiency measures for the elderly. If the upfront cost was not consistently the main barrier for the elderly, then it may be more effective to target this age demographic with other types of incentives. 7.1.3. Inter-occupant relationships Inter-occupant relationships are an important factor affecting the adoption of energy efficiency measures [21e23]. As such, since this factor was associated with households with 3 or more occupants, future energy efficiency campaigns or interventions should address all of the occupants/family in a household, rather than just individuals. The uptake of energy efficiency measures may benefit by considering the dynamics of inter-occupant relationships in the initial stages of design. 7.1.4. Institutional barriers Since the Institutional barrier (government/energy companies) was strongly associated with Manchester, local policy makers and utilities within Manchester may benefit by comparing their institutions with those in Cardiff. Such city comparisons may lead to the identification of policy differences and/or practices which may help Manchester improve the uptake of energy efficiency measures. 7.1.5. Landlord-tenant split incentive In order to address the landlord-tenant split incentive to adopting energy efficiency measures in the home, the UK government introduced the Green Deal; a policy framework enabling firms to offer consumers energy efficiency improvements to their homes at no upfront cost, and recoup payments through a charge of instalments on their energy bill [43]. To support the Green Deal and overcome the landlord-tenant split incentive barrier, the results of this study imply that interventions should target: men; individuals who are currently single; individuals with a degree or more of educations; flats and terraced homes; and tenants. 7.1.6. Personal behaviour This was a complex barrier which conflated behaviour and attitudes, and was only strongly associated with individuals with a degree or more of higher education. This association does not provide clear guidance for designing future interventions, but does provide a sign-post to help guide interventions which aim to change behaviour as a barrier. 7.1.7. Physical property Apart from the non sequitur barrier of ‘None’, this was the only other barrier which was strongly associated with Cardiff; hence, possible insights may emerge by comparing the Cardiff built environment with Manchester in more detail. These insights would benefit the local construction industry, as well as local policy makers. The results also imply that interventions which focus on the physical form of the house and/or planning issues should specifically target: individuals who are married/common-law; individuals with a degree or more of higher education; and owner-occupied homes. Using the guidance above, and the associations between demographic variables and energy efficiency barriers presented in this paper, this research aimed to assist policy makers, local councils, and members of the construction/retrofit industry. The final aim was to improve the energy efficiency of the built environment and contribute to the body of knowledge surrounding the Energy Efficiency Gap.
Role of the funding source The site visits and sundries were funded as part of the EPSRC funded Retrofit 2050 project [8]. All other aspects of the research (study design, collection, analysis, interpretation of data, and writing of the report) were conducted fully by the authors. Acknowledgements This research was conducted at the Department of Engineering, Centre for Sustainable Development, University of Cambridge, and is part of the EPSRC funded research project: Re-Engineering the City 2020-2050 (RETROFIT 2050) Urban Foresight and Transition Management. Finally, the authors would like to thank all of the friendly interviewees in Manchester and Cardiff, and the anonymous paper reviewers for their invaluable feedback. References [1] Lior N. Sustainable energy development: the present (2009) situation and possible paths to the future. Energy 2010;35(10):3976e94. [2] Kelly MJ. Retrofitting the existing UK e building stock. Building Research & Information 2009;37(2):196e200. [3] Allcott H, Mullainathan S. Behavior and energy policy. Science 2010; 327(5970):1204e5. [4] Jaffe AB, Stavins RN. The energy-efficiency gap what does it mean? Energy Policy 1994;22(10):804e10. [5] Weber L. Some reflections on barriers to the efficient use of energy. Energy Policy 1997;25(10):833e5. [6] Pelenur M, Cruickshank H. The social barriers towards adopting energy efficiency measures and behaviours in the home: a manchester and cardiff case study. In: proceedings of the RETROFIT 2012 Conference. Salford: UK; 2012, Jan, P. 24e26. [7] European Spatial Planing Observation Network [Internet]. British urban pattern: population data [cited 2011 Oct 10]. Available from: http://www.espon.eu/ export/sites/default/Documents/Projects/ESPON2006Projects/StudiesScientific SupportProjects/UrbanFunctions/fr-1.4.3_April2007-final.pdf#page¼119; 2007. [8] RETROFIT 2050 [Internet]. Re-engineering the city 2020e2050 urban foresight and transition management [cited 2012 Mar 28]. Available from: http://www. retrofit2050.org.uk/; 2012. [9] Committee on Climate Change [Internet]. Climate change act. Tech. Rep.; Committee on Climate Change; 2011. [cited 2011 Jun 17]. Available from: http:// www.theccc.org.uk/about-the-ccc/climate-change-act. [10] The Climate Group [Internet]. UK government’s commitment on long term emissions cuts is step in right direction [cited 2011 6 Jun]. Available from: http://www.theclimategroup.org/what-we-do/news-and-blogs/; 2011. [11] Committee on Climate Change [Internet]. Homes [cited 2011 Jun 6]. Available from: http://www.theccc.org.uk/sectors/buildings-a-industry/homes; 2011. [12] Institution of Mechanical Engineers [Internet]. UK 2050 energy plan making our commitment a reality. Tech. Rep.; IMechE; 2009. [cited 2011 Apr 3]. Available from: http://www.imeche.org/knowledge/themes/environment/climate-change/ copenhagen-conference/uk-2050-energy-plan. [13] Stern N. Stern review on the economics of climate change. Tech. Rep.; Stern Review on the Economics of Climate Change, UK Treasury; 2006. [14] Upham P, Whitmarsh L, Purdam K, Darnton A, McLachlan C, Devine-Wright, P. Public attitudes to environmental change: a selective review of theory and practice. Tech. Rep.; Research Councils UK; 2009. [15] Darnton A. GSR behaviour change knowledge review. Tech. Rep.; Practical guide: an overview of behaviour change models and their uses. Report for government social research unit. London, Centre for Sustainable Development, University of Westminster; 2008. [16] Abrahamse W, Steg L. How do socio-demographic and psychological factors relate to households direct and indirect energy use and savings? Journal of Economic Psychology 2009;30(5):711e20. [17] Faiers A, Cook M, Neame C. Towards a contemporary approach for understanding consumer behaviour in the context of domestic energy use. Energy Policy 2007;35(8):4381e90. [18] Madlener R, Alcott B. Energy rebound and economic growth: a review of the main issues and research needs. Energy 2009;34(3):370e6. [19] Lutzenhiser L. A cultural model of household energy consumption. Energy 1992;17(1):47e60. [20] Stephenson J, Barton B, Carrington G, Gnoth D, Lawson R, Thorsnes P. Energy cultures: a framework for understanding energy behaviours. Energy Policy 2010;38(10):6120e9. [21] Hargreaves T, Nye M, Burgess J. Making energy visible: a qualitative field study of how householders interact with feedback from smart energy monitors. Energy Policy 2010;38(10):6111e9. [22] Darby S. Smart metering: what potential for householder engagement? Building Research & Information 2010;38(5):442e57.
M.J. Pelenur, H.J. Cruickshank / Energy 47 (2012) 348e357 [23] Shove E, Lutzenhiser L, Guy S, Hackett B, Wilhite H. Energy and social systems. Human Choice and Climate Change 1998;2:291e325. [24] Pelenur M, Cruickshank H. The subjective view of energy in the urban built environment what are the social factors that affect our interaction with energy? In: Proceedings of the 6th UNESCO sustainable development of energy water and environment systems conference. Dubrovnik, Croatia: University of Dubrovnik; 2011, Sept 25e29. SDEWES11e0422. [25] Boyatzis RE. Transforming qualitative information: thematic analysis and code development. SAGE, ISBN 9780761909613; 1998. [26] Fereday J, Muir-Cochrane E. Demonstrating rigor using thematic analysis: a hybrid approach of inductive and deductive coding and theme development. International Journal of Qualitative Methods 2008;5(1):80e92. [27] Marks D, Yardley L. Research methods for clinical and health psychology. SAGE, ISBN 9780761971917; 2004. [28] Yates D, Moore DS, McCabe GP. The practice of statistics AP: TI-83 graphing calculator enhanced. 1st ed. W. H. Freeman, ISBN 0716733706; 1999. [29] Bilder CR, Loughin TM. Testing for marginal independence between two categorical variables with multiple responses. Biometrics 2004;60(1):241e8. [30] Thomas DR, Decady YJ. Testing for association using multiple response survey data: approximate procedures based on the Rao-Scott approach. International Journal of Testing 2004;4(1):43e59. [31] Loughin TM, Scherer PN. Testing for association in contingency tables with multiple column responses. Biometrics 1998;54(2):630e7. [32] Decady YJ, Thomas DR. A simple test of association for contingency tables with multiple column responses. Biometrics 2000;56(3):893e6. [33] Rao JNK, Scott AJ. The analysis of categorical data from complex sample surveys: chi-squared tests for goodness of fit and independence in Two-
[34] [35] [36]
[37]
[38] [39] [40]
[41]
[42] [43]
357
Way tables. Journal of the American Statistical Association 1981;76(374): 221e30. Rao JNK. On chi-squared tests for multiway contingency tables with cell proportions estimated from survey data. The Annals of Statistics 1984;12(1):46e60. Agresti A, Liu I. Modeling a categorical variable allowing arbitrarily many category choices. Biometrics 1999;55(3):936e43. Thomas DR, Decady YJ. Analyzing categorical data with multiple responses per subject. In: Statistics society of canada proceedings of the survey methods section; 2000. p. 121e30. Bilder CR, Loughin TM, Nettleton D. Multiple marginal independence testing for pick any/C variables. Communications in Statistics e Simulation and Computation 2000;29(4):1285e316. Baker KJ, Rylatt RM. Improving the prediction of UK domestic energy-demand using annual consumption-data. Applied Energy 2008;85(6):475e82. Friendly M. Mosaic displays for multi-way contingency tables. Journal of the American Statistical Association 1994;89(425):190e200. Guerin DA, Yust BL, Coopet JG. Occupant predictors of household energy behavior and consumption change as found in energy studies since 1975. Family and Consumer Sciences Research Journal 2000;29(1):48e80. Lenzen M, Wier M, Cohen C, Hayami H, Pachauri S, Schaeffer R. A comparative multivariate analysis of household energy requirements in Australia, Brazil, Denmark, India and Japan. Energy 2006;31(23):181e207. Burgess J, Nye M. Re-materialising energy use through transparent monitoring systems. Energy Policy 2008;36(12):4454e9. Department of Energy & Climate Change [Internet]. The green deal: a summary of the government’s proposals [cited 2011 Jun 17]. Available from: http:// www.decc.gov.uk/en/content/cms/tackling/green_deal/green_deal.aspx; 2010.