The determinants of domestic energy consumption in France: Energy modes, habitat, households and life cycles

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser

The determinants of domestic energy consumption in France: Energy modes, habitat, households and life cycles Jean-Pierre Lévya, Fateh Belaïdb, a b

⁎

CNRS, UMR, LATTS, Ecole des Ponts ParisTech, 6 & 8 avenue Blaise Pascal, Cité Descartes, F-77455 Marne-la-Vallée cedex, France CSTB/ Department of Economic Studies. Paris Est University, 84 Avenue Jean-Jaurès, Champs-sur-Marne, 77447 Marne-la-Vallée cedex 2, France

A R T I C L E I N F O

A BS T RAC T

Keywords: Energy consumption Housing practices Life cycles Household energy use Socio-economic effects

This article explores the question of domestic energy consumption. It concentrates on the example of France and using micro-data from the 2002 and 2006 National Housing Surveys conducted by INSEE (National Institute of Statistics and Economic Studies). The empirical analysis is divided into three stages: the first verifies the correspondences between modes of energy consumption, household profiles and housing profiles, in order to classify consumer mode types; the second looks at the stabilities and variations over time of each of these types; the last seeks to identify the causes of these changes. The findings reveal sharp divergences between the factors affecting global consumption, consumption per m2 and per person. These variations can be explained by the impact of the demographic characteristics of households, residential mobility and life cycles. Therefore, these findings demonstrate the flexibility over time of domestic energy consumption, which is still too often approached as a static variable solely associated with building characteristics. They need to be taken further through a longitudinal and multidisciplinary approach to energy consumption patterns.

1. Introduction

consumption of buildings have changed little since the 1970s, despite the genuine advances that have taken place in construction techniques. Whatever the (economic or engineering) methods employed [5] the models relating to energy consumption and demand draw on a set of variables relating – amongst other things – to the properties of buildings, the price of energy, household incomes, the number of household electrical appliances, or climate (indoor temperature). For their part, certain economists stand out for their interest in more macro factors, such as inflation [6]. The great weakness of these models, however, is that they ignore consumer lifestyles, and therefore the energy use patterns of households, even when they take into account their socio-demographic characteristics [7]. In this respect, they overlook the leeway people have in their specific relations to energy consumption, and therefore treat the consumption patterns of users as governed purely by supply. While this perspective is undoubtedly relevant, it is also partial in the sense that it reduces the individual to a mere energy consumer subject to the economic rationales of

According to the International Energy Agency's 2013 reference scenario, by the year 2040, 14% of worldwide energy consumption will come from households, an increase of 57% compared with the 2010 rate. The residential sector is thus responsible for a large proportion of energy consumption. In France, for example, it is the second largest source of final consumption, at 46 million tonnes of oil equivalent (Mtoe) in 2012, just behind the transport sector (49 Mtoe, approximately 30% of final energy consumption).1 Far from being governed by the energy efficiency of buildings alone, residential consumption largely reflects the domestic energy practices of households. This fact explains the increasing and interrelated interest of researchers and governments in understanding the determinants of domestic energy use in order to develop measures to rationalise consumption [1–4]. However, studies (and regulations) on this issue remain very techno-centric and the thematic frameworks of research into the energy

⁎

Corresponding author. E-mail addresses: [email protected] (J.-P. Lévy), [email protected], [email protected] (F. Belaïd). 1 MEDDE (Ministry of Ecology, Sustainable Development and Energy), National Sustainable Development Indicators 2010–2013. Energy consumption by the residential and tertiary sectors. http://www.statistiques.developpementdurable.gouv.fr/indicateursindices/f/1932/0/consommation-denergie-secteurs-residentiel-tertiaire.html.

http://dx.doi.org/10.1016/j.rser.2017.06.022 Received 5 July 2016; Received in revised form 16 February 2017; Accepted 9 June 2017 1364-0321/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Lévy, J.-P., Renewable and Sustainable Energy Reviews (2017), http://dx.doi.org/10.1016/j.rser.2017.06.022

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

2. Data and methods

producers (of buildings and energy). However, attempts to include socio-demographic variables reveal results in which the age of individuals and household size emerge as decisive variables in understanding domestic consumption processes [8].2 Over the past few years, empirical studies on residential energy consumption have received considerable attention. A growing literature has focused on exploring the determinants of residential energy consumption using different bottom-up statistical approach [7,9–12]. Valnezuela et al. [4] used a quantile regression model to examine predictors of household energy consumption among single-family residences in a Texas urban area. Their results indicate potential opportunities to lower consumption among the highest energy-consumption households including those with pools, with no central cooling, with people working from home, those built on pier/post foundation, and those that are renter-occupied. They found that: households in larger housing units consumed less energy per square meter than smaller housing units, married households consumed less energy per square meter than not married households, as the number of adults increases energy per square meter also increases, etc. Recently Belaïd [13] used a bottom-up statistical approach to tease out the impacts of various factors on the domestic energy consumption in France across different population groups using data from the most recent National Housing Survey. The multivariate analysis has led to identify four main consumption typologies. Results revealed that energy prices were the most important factors determining domestic energy consumption. In addition, the study showed that occupant characteristics significantly affect domestic energy use. Thus, results call for combine all efforts, multiple strategies and smart policies, to incorporate household and consumption behaviours in managing domestic energy consumption. Therefore, by shifting the focus to the energy behaviour of individuals or groups, sociological, anthropological and geographical approaches reveal the importance of residential practices and lifestyles in better understanding energy consumption processes [15–20]. Going beyond the analysis of individual behaviours fashionable in the 1980s and 1990s, the most current research tackles energy consumption as a set of practices that can only be identified through crosscutting approaches [21]. From this perspective, energy demand becomes a social construction structured by norms and conventions [22–25]. Beyond building-related factors, occupant behaviour plays a not insignificant role in the intensity of energy consumption in residential buildings. It remains the case that these two dimensions are rarely brought together in research on energy consumption. That is the essential goal of this article, which seeks to study the effect of household characteristics on the modes and intensities of French domestic energy consumption. Using a national approach, we will endeavour to understand if there are links between the types of habitat occupied (houses, apartment buildings, residential sectors), the characteristics of residents and the energy combinations used. More specifically, the aim is to understand whether socio-demographic factors affect the typological profiles of domestic energy consumers, and whether the latter have a direct impact on the intensity of and trends in consumption. The findings reveal the decisive role of the socio-demographic characteristics of households, of their residential mobility and of their life cycle on domestic consumption. They thus demonstrate the flexibility over time of domestic consumption, which is still too often approached as a static variable solely associated with the characteristics of buildings.

The approach taken in this article is to link – from a sociodemographic point of view – the technical performances of buildings and the residential behaviours of inhabitants. This approach has been applied using methods of quantitative analysis based on the secondary processing of existing national surveys: the surveys on the housing conditions of French households (ENL) conducted by INSEE in 2002 (32,000 households) and in 2006 (31,000 households). These contain very detailed information on the characteristics of dwellings (size, amenities, number of rooms, renovation, etc.) and of buildings, as well as of housing occupants. However, they also have the advantage of including a large section on “dwelling amenities and energy used”. Part of this information was initially used to construct summary indicators for the type of domestic energy used by households, combined with the intensity of consumption (from high to low). These indicators were then applied to the characteristics of the households, the buildings and the dwellings, on the assumption that the types of energy used at home reflect particular household and dwelling structures. The empirical analysis is divided into three stages: the first seeks to verify the correspondences between the combined use of domestic energy, the household profiles and the dwelling profiles, in order to establish consumer mode typologies. The second looks at the stabilities and variations over time of consumption in each of these types, endeavouring to distinguish between the role of household characteristics (consumption per head) and of dwellings (consumption per m2) in these intensities. The final stage seeks to identify the causes of these changes, notably by introducing a life-cycle based analysis. In the first step, in order to obtain a typology of consumers based on the combinations of energy types used by French households, a multiple correspondence analysis (MCA) and a hierarchical cluster analysis (HCA) were conducted. MCA is a generalisation of factorial correspondence analysis to multivariate cases [26]. The distances calculated between the different households within the factorial axis space is used to classify the closest individuals and merge them, at each successive stage, using a proximity criterion called Ward's minimumvariance method. In the second phase, in order to analyse the morphological, social and demographic determinants of domestic energy consumption, we employed a logistic regression in which the variable to be explained is low energy consumption per person and per m2, and the explanatory variables are the socio-demographic characteristics of the households and of their habitat. Finally, in the third and last stage, for the analysis of household consumption on a life-cycle basis, the age of the reference person was combined with household size. In this way, we identify the major life cycles of households, assuming linear family development (without separation, death or divorce). We then obtain four snapshots of life-cycle stages, meaning that we observe these stages from a transversal perspective on a given date, rather than from a longitudinal perspective. 3. Results 3.1. First stage: a typology of domestic energy consumers in France In order to construct a typology of energy consumers, we began with the combinations of energy types used by French households in 2002, in order to be able to conduct a trend analysis by comparison with the housing survey conducted in 2006. Independently of their specific uses, we chose to consider the combinations of domestic uses of six types of energy: electricity, gas, oil, propane gas, wood, coal and district heating. The intensities and uses of these different forms of energy can be very variable: for example, oil is generally a fuel used for heating, mains gas for heating and cooking, and propane solely for cooking. Combining them can nevertheless make sense and distinguish users socially, for example when a household only has electricity in the

2 The final report of the Energy Transports Habitat Environment Localisations (ETHEL) project thus notes that “one of the major obstacles to a steady reduction in the consumption of energy for heating lies in the expansion of the areas heated”. For example, living space per person in France increased from 31m2 in 1984 to 37m2 in 2002, because of the rise of detached houses as a proportion of new housing and a fall in the average size of households. Once children have left home, many older households occupy dwellings with a high occupancy cost (areas to be heated), which no longer correspond to their need for living space [14].

2

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 1. Multiple Components Analysis (MCA) of the types of domestic energy sources in France in terms of intensity of consumption.

based on the MCA, we therefore isolated 10 types of French energy consumer (Fig. 2). We then endeavoured to find out if there were dominant household and dwelling profiles corresponding to each of the 10 types of consumer (Table 1),3 identified in terms of the mode and intensity of the energy forms used. At first sight, this table would suggest that the mode of energy used reflects the particular characteristics of households and dwellings, and vice versa. As regards habitat, the type and construction date of the building, the living space and number of rooms in the dwelling, and the spatial context, would seem to be major differential factors. With regard to the households, the presence or absence of children, the age of the reference person and social status, are very significant factors. Table 1 also highlights the fact that in 2002, there were three predominant types of consumer in the population, accounting for 3/5 of households (types 3, 5 and 7). The combinations of the energy modes used, and the dominant profiles of households, buildings and dwellings, give us a glimpse of the leeway and constraints on energy use in the households belonging to these three types. Thus, type 3 represents consumers who combine high electricity use with the use of propane and wood (26.3%), typically characterised by average income families aged over 40 and under 60, who are homeowners in a rural or suburban area, and can therefore reduce their electricity consumption by means of an open fire (very common in houses, especially in rural and periurban areas) for heating and by the use of propane for cooking; type 7 represents high electricity consumers without access to other energy forms (22.2%), typically young people living in small rented apartments built after 1975, where the morphology generally precludes the combination of energy sources and imposes intense consumption; and type 5 who are low consumers of electricity and gas (13.2%), typically childless, working-class and sometimes precarious households, living in small rented social housing where appliances can generally be individually regulated to reduce consumption and avoid excessively high bills. Two types appear marginal in the total population: type 6, with a profile similar to the previous one, consisting of households in which high electricity consumption com-

home, and uses propane for cooking for economic reasons, or when a household has oil for heating, but uses an open fire to reduce consumption. Moreover, it is well known that the type of energy used is not totally dependent on household choices, but also reflects the equipment, infrastructures and networks available in buildings and neighbourhoods. In light of this, we made the assumption that the modes of energy used are partly dependent on the type of dwelling occupied. Household energy expenditure was used to establish a uniform value for the intensity of consumption in kWh. We excluded from the analysis households declaring no electricity consumption. For the other energy sources, we considered non-consuming households as a group in its own right. On this basis, using the median of intensities specific to each type, we identified two groups (high/low). Ultimately, therefore, for each type of energy except electricity, we obtained three categories of household: no consumption, high consumption, low consumption. In this way, we arrive at a typology of consumers: – – – – – –

of electricity (high [E+] or low [E-]). of gas (no consumption [G0], high [G+] or low [G-]). of fuel oil (no consumption [F0], high [F+] or low [F-]). of propane (no consumption [P0], high [P+] or low [P-]). of wood (no consumption [W0], high [W+] or low [W-]). of district heating and coal (no consumption [DH/C0], high [DH/C +] or low [DH/C-]).

These categories were studied using a multiple correspondence analysis (MCA, Fig. 1). The factor analysis map (43.7% of the information) reveals the differences between high consumers (upper part of the drawing) and low consumers (lower part of the drawing). It should be noted that this difference is reversed for district heating and coal (DH/C) consumers. This means that high consumers of district heating and coal are low users of other types of energy, and vice versa. Another noteworthy observation from the MCA is the contrast between users of wood and electricity and users of district heating and coal, on the one hand; and between users of oil and propane and those who use gas, on the other hand (cross axes). These oppositions reflect the existence of types of consumers whose combined intensity of use of the different energy sources depends on their financial capacity, supply constraints and their need for comfort. Using a hierarchical cluster analysis (HCA)

3 These profiles were calculated using the Maximum Percentage Difference (MPD), which makes it possible to estimate the force of attraction between two modes (the closer MPD is to 100%, the greater the attraction between modes) combined with the weight of the characteristic as a proportion of the mode (type of consumers studied) [27].

3

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 2. Dendrogram of the hierarchical cluster analysis of the 10 consumer types. Note. The figures in boxes represent the types of habitat, the figures in brackets represent the number of people in the database at each row and node of the hierarchical classification. Table 1 Dominant profiles of households and dwellings in the 10 types of French consumers in 2002. Sources: secondary processing of the 2002 national housing survey (INSEE). Example of interpretation: the consumption modes not in brackets represent a minimum of 60% of the type. The consumption modes in brackets are in a minority and are represented at least 1.5 times more within the type than in the total population. Types of consumers

%

Energy types

Dominant profiles

Type 1

4.6

E- F- P0 (DH/C+)

Type 2

8.5

E- F0 P0 (W0)

Type 3

26.3

E+ P0 W0 (DH/C-)

Type 4

1.5

E+ F+ P0 W0

Type 5

13.2

E- G- (DH/C+)

Type 6

1.1

E- G- F-

Type 7

22.2

E+

Type 8

7.2

E- P0 (DH/C0 F+)

Type 9

9.7

E- G+

Type 10

5.5

E+ G+ (W-)

Childless couples or young or older single people belonging to the working classes with medium or low incomes; tenants of an apartment with less than forums and smaller than 100 m2 in a building constructed between 1950 and 1974 in a rural or urban area Childless couples or single retired people belonging to the working classes; owner-occupiers of a house larger than 100 m2 built before 1975 in a rural or periurban area Middle-income families aged 40–59; owner occupiers of a house larger than 70 m2 built before 1948 or after 1975 in a rural or periurban area Families aged over 40, upper or middle management levels (including retired); owner occupiers of an old house larger than 100 m2 in a rural or periurban area Young working-class middle-income or low income childless couples or single people (sometimes precarious); tenants in social or private housing with 4 rooms or less built between 1949 and 1974 in an urban area Childless couples or single retired people, management or working class category; tenants in social or private housing with 4 rooms or more built between 1949 and 1974 in an urban area Childless young couples or single people; tenants in social or private housing with one or two rooms, smaller than 70 m2, built after 1975 in an urban area Low income, working class childless couples or retired single people; tenants in a house with 4 rooms or more and smaller than 70 m2, built before 1975 in a rural or periurban area Management level or working class childless couples or retired single people; tenants in social or private housing with 3 or 4 rooms, between 70 and 100 m2 built between 1949 and 1974 in an urban area High income upper or middle management families aged 40–59 ans; owner occupiers of a house larger than 100 m2 built before 1975 in an urban area

bines with low gas and oil consumption (1.1%); and type 4 consisting of households in which high electricity and oil consumption is combined with the use of wood and propane (1.5%), largely families who own large houses in a rural or periurban area and do little to regulate their energy consumption. This therefore raises the question of the stability of the proportions of each of the ten consumer types within the French population, insofar as they are simultaneously dependent on energy supply policies, on the structure of the real estate stock and on population structure (Table 2). Without lingering on these different factors, we would note that the results of the 2002 and 2006 housing surveys show that, under the impact of public policy incentives to adopt less energy intensive practices and of building refurbishments to improve insulation (stricter energy standards), French energy consumption fell significantly (by 18.2%) over the period. Nonetheless, this reduction varied sharply between the energy types: it applied less to electricity and gas than the other modes. Logically, it is the proportions of consumer types that exclusively

Table 2 Changes in French energy consumer types between 2002 and 2006. Sources: National Housing Survey 2002 and 2006, (INSEE). Example of interpretation: the consumption modes not in brackets represent a minimum of 60% of the type. The consumption modes in brackets are in a minority and are represented a minimum of 1.5 times more within the type than in the total population.

4

Type of consumer

2002

2006

Variation

Type 1 (E- F- P0 [DH/C+]) Type 2 (E- F0 P0 [W]) Type 3 (E+ P0 W0 [DH/C0]) Type 4 (E+ F+ P0 W0) Type 5 (E- G- [ DH/C +]) Type 6 (E- G- F-) Type 7 (E+) Type 8 (E- P [DH/C0 F+]) Type 9 (E- G+) Type 10 (E+ G+ [W-]) Total

4.8 8.5 26.3 1.5 13.2 1.1 22.2 7.2 9.7 5.5 100

2.6 7.1 23.1 1.3 17.2 0.2 22.5 6.5 9 10.6 100

− 2.2 − 1.4 − 3.2 − 0.2 4 − 0.9 0.3 − 0.7 − 0.7 5.1 100

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

under the age of 40 (37%), whereas the highest are families (42%) aged over 40 (55%) or single people and childless couples in the higher age bracket (42%). Surprisingly, on the other hand, socio-professional categories (with the exception of farmers and shopkeepers) and income, by contrast with the consumption levels specific to each type (Table 1), have little impact on intensities, probably because they are tackled globally here. Indeed, Fig. 3 also shows that there is no automatic correspondence between the level of global consumption intensity per m2 and per person in each of the types. Beyond the types of energy used, it is therefore probable that the characteristics of the dwelling and the household have a different impact on consumption intensities, which could explain their contribution to the construction of the ten types of consumer. Does this mean that the factors that influence consumption by people and by buildings do not reflect the same processes? In order to understand these factors better, we conducted a logistic regression in which the variable to be explained is low energy consumption per person and per m2.4 Table 3 sets out the significant explanatory factors that impact differently on consumption per person and per m2. What we see is a sharp contrast in household characteristics depending on whether consumption is measured per person or per m2. As regards the socio-economic determinants, we see that – all else being equal – a foreign reference person is 1.2 times less likely than a French person to be a low consumer per m2, but 1.6 times more likely to be a low consumer per person (probably because they are a member of a large family, living in a small dwelling, in buildings of mediocre quality). Similarly, a private sector tenant is as likely as a social housing tenant to be a low consumer per m2, but 2.7 times more likely to be a low consumer per person (probably because of household size). This difference between the factors affecting consumption per m2 and per person is confirmed for household size or income. As regards dwelling characteristics, we find the opposite trends, associated with dwelling size or number of rooms. Thus, a household occupying a dwelling of less than 70 m2 is – all else being equal – 2.5 times less likely to be a low consumer per m2 than a household living in a dwelling measuring 100 m2 or more, and 1.2 times more likely to be a low consumer per person. In the light of this, the fact that consumption intensity per person increases with household size and diminishes with dwelling size might seem counterintuitive, especially if one begins with the assumption that big households occupy large dwellings, and that logically, total consumption should be distributed between family members and across housing space. It becomes less so if one considers that there is no automatic match between household size and dwelling size, and that small households can also live in large dwellings, for example when the children of a large family occupying a large dwelling leave the family home, while the parents stay.

and predominantly use gas and/or electricity, even at low intensity, which are on the rise within the national population. These are notably low consumers of electricity and gas (type 5), i.e. childless workingclass households living in small rented accommodation in urban areas; and large electricity and gas consumers (type 10), corresponding to well-off home-owning families with large houses in urban areas. It should be noted that this increase is not apparent in households that regulate their electricity consumption and not their gas consumption (type 9), probably because they are retired people living alone or in couples, occupying a 3 or 4 rooms apartment in an urban area, which may be assumed to be equipped with central heating over which residents have little or no control. It can also be noted that households that use only electricity represent a relatively stable proportion of the population (type 7), probably because they live in small rented apartments, new or recent, fitted with the electrical convection heaters found in numerous buildings constructed between 2002 and 2006. The proportions of all the other types fell. Ultimately, it may be observed that – against a background of global reduction in domestic energy consumption – changes in the distribution of consumer types relate more to the mode of energy used than to the intensity of consumption. This then raises the question of the connection between changes in consumer types and changes in buildings and households, insofar as – as we have seen – the former are indicative of the latter and, above all, insofar as the structure of households and of the housing stock changed little between 2002 and 2006, whereas that of consumer types changed significantly. 3.2. Stage two: the morphological, social and demographic determinants of domestic energy consumption In order to try to answer this question, we initially compared the average annual consumption intensities of the ten types in 2002, in a threefold approach: a global approach to consumption, another per m2 of housing and a third per person in households, during the 12 months preceding the survey (Fig. 3). Fig. 3 clearly shows that global annual average consumption varies greatly from one type of consumer to another. These variations do not directly reflect the intensity of use of the energy used: for example, type 7, which covers the highest exclusive energy users are among the lowest overall consumers, which is not much of a surprise in that they are young single people or young childless couples in small rented accommodation. Global consumption intensities therefore rather reflect the correspondences between the energy mode used, the characteristics of the dwellings occupied and the characteristics of the consuming households. Thus, as regards the energy mode used by the four types of biggest energy consumers (types 3, 10, 2, 4), 75% are big consumers of electricity, 79% are users of propane (which they combine with other modes) and they are heavily represented amongst the big oil users. Conversely, the types of consumers who use the least energy (types 5, 7, 8) do not use gas, district heating (and coal) and oil, recognised to be energy modes associated with high consumption. These types of energy mode are associated with forms of habitation. It would seem that the biggest energy consumers are owner-occupiers (79%) living in houses (89%) with multiple rooms (56% in dwellings with 5 rooms or more) and large living space (56% in dwellings in excess of 100 m2), generally located in periurban (27%) or rural (30%) areas. For their part, the least energy intensive types of consumer are those living in rented collective housing, either private (20%) or social (27%), with less than 4 rooms (77%), 40% of them with living space of less than 70 m2, located in an urban municipality with a population in excess of 200,000 (79%). There is therefore a sharp contrast in the housing conditions of high and low consumers. These same contrasts are found in the characteristics of the households corresponding to each of the consumer types. The lowest energy consumers are people living alone or childless couples (60%)

3.3. Stage three: analysis of consumption in relation to life cycle How can this inverse effect be explained? To try to understand it, we shifted the focus to the role of life cycle on consumption, with the idea that it provides a better explanation than household or dwelling size in isolation. In order to identify the stages in the family life cycle, we linked the family situation of households to the age of the reference individual. We therefore marked out the major stages in the life-cycle snapshots of households in 2006: young childless couples or young single people (reference person aged under 30), couples with child (ren) where the age of the reference person is between 30 and 39, couples with child (ren) where the age of the reference person is between 40 and 49, couples with child (ren) where the reference person is aged between 50 and 59, childless couples and single 4 Identified on the basis of the median of average annual consumption per person and per m2.

5

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 3. Average French energy consumption in the 12 months preceding the survey, by household, per person and per m2. Presentation of consumer types from the lowest average annual household consumption to the highest. Example of interpretation: the consumption modes not in brackets represent a minimum of 60% of the type. The consumption modes in brackets are in a minority and are represented a minimum of 1.5 times more within the type than in the total population.

In order to study these effects, we began with the hypothesis that a household owning a large house in an urban area, and a household renting an apartment with 4 rooms or fewer in an urban area, would show the same pattern of accommodation throughout their life cycle. For owner-occupier households, we based ourselves on the energy consumer from type 10, 89% consisting of high electricity consuming households, 100% of high gas consumers, sometimes complemented by the use of an open fire (19%); these consumers are predominantly families aged 40–59, executives or middle managers with high incomes. For households in rented accommodation, we based ourselves on the type 7 energy consumer, 61% of whom come from households that are high electricity consumers (without using other energy sources), who are predominantly childless young couples or single people. Fig. 4 provides an example of the average annual consumption per m2 and per person of type 10 consumers, i.e. owners of large houses located in urban areas. Their consumption is presented according to their position in the household life cycle (childless couples aged under 30 are not included, because they constitute too small a proportion in this type of housing). Although the data are transversal, we opted to present them through curves, rather than histograms, in order to emphasise the sense of continuity of consumption. The tables under the graphs show the average surface area of the dwellings and the average household size of the type at each position in the life cycle. The left-hand figure distinguishes between consumption by working-class categories (office and manual workers, a minority in the type) and consumption by higher class categories (managerial and professional positions). It shows that average annual consumption per person is the same for both categories, whatever their stage in the life cycle. On the other hand, it is very clear that working-class households consume more energy per m2 than higher class households, since for the same position in the life cycle, household size is greater (an average of 1.7–

Table 3 Logistic regression on the factors that have different impacts on consumption intensities by household and by dwellings per person and per m2 in France. Factor to explain

Low consumption kW h/m2

Low consumption kW h/pers

Factor

Signif

Odd-ratio

Signif

Odd-ratio

1 or 2 people 3 or 4 people 5 people or more Foreigners French Low income per CU Middle-income per CU High income per CU 1 or 2 room(s) 3 or 4 room 5 rooms and + Less than 70 m2 71–99 m2 100 m2 and + Owners Social housing tenant Private sector tenant Free accommodation Farmers

Rf *** *** *** Rf **

1 0.687 0.676 0.843 1 0.932 0.968 1 1.194 1.181 1 0.398 0.57 1 0.575 1 0.936 0.753 0.446

Rf *** *** *** Rf *** *** Rf ***

1 2.306 5.804 1.579 1 1.49 1.261 1 0.859 0.954 1 1.224 1.256 1 0.494 1 2.662 1.495 0.837

Rf *** *** Rf *** *** Rf *** Rf *** ***

Rf *** *** Rf *** Rf *** ***

(Rf) Reference group.

people where the reference person is aged over 60 (where children are assumed to have left home). Of course, nothing tells us that the young couples of 2006 will behave in the same way as their elders when they are 60. Nonetheless, a transversal analysis provides information on the impact of position in the life cycle on the average intensities of energy consumption in 2006. 6

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 4. Average annual consumption per m2 and per person based on a life-cycle snapshot for type 10 in France. Note. Table interpretation example: Working-class households consisting of a couple with children, where the reference person is aged between 30 and 39, on average occupy a dwelling of 125 m2 and consist on average of 4.5 people. A higher class household in the same place in the life cycle occupies on average a dwelling of 136.3 m2 and consists on average of 4.2 people.

Fig. 5. Average annual consumption per m2 and per person based on a life-cycle snapshot for type 7 in France. Note. Table interpretation example: High electricity consuming households in private rented accommodation of less than 70 m2, consisting of a single person or a childless couple, where the reference person is aged 30–39, have an average dwelling size of 34.9 m2 and an average household size of 1.3. Households in the same position in the life cycle living in rented social housing with living space of less than 70 m2, have an average dwelling size of 43.6 m2 and an average household size of 1.3 people.

7

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 6. Average annual consumption per m² according to a life-cycle snapshot of owner-occupiers of houses of more than 70 m² in France in 2006.

for consumption per person shows that childless households (in the first and last life-cycle sequence) consume more than families, but with smaller differences than those observed for owner-occupied houses. There is also a very clear reduction in consumption per m2 at the end of the residential life cycle. As the tables in Fig. 5 show, this is explained by the fact that – unlike the occupants of owner-occupied houses – tenants move when the children leave home, and reduce the size of their dwelling (between the last two life-cycle sequences, dwelling size decreases by 10.7 m2 for private sector tenants and by 15.3 m2 for social housing tenants). Tenants therefore seem to adapt their dwelling size to changes in household size. In this case, by comparison with owner-occupied housing, we see the very clear energy impact of this adjustment. The reduction in dwelling size automatically causes a fall in energy consumption per m2, whether in social or private sector rented housing. These observations are not limited to consumer types 7 and 10, but apply to all owners of houses larger than 70 m2 (Figs. 6 and 7)6 and tenants of apartments (Figs. 8 and 9).7 Here again, we see the impact of the mode of energy used, which combines with the life-cycle positions in determining consumption intensities that vary according to the types of housing, the types of households occupying it and the residential environment.

4.5 people in the working class category, compared with an average of 1.8–4.3 in higher class households) and that they occupy smaller dwellings (an average of 116–125 m2 for working-class households, compared with 136–141 m2 for the higher class categories). It may also be that their houses are less energy-efficient than those of the higher class categories. The right-hand figure presents consumption in dwellings built before and after 1975. The table under the right-hand figure shows that, for the same place in the life cycle, the more recent houses are bigger than the older houses, without necessarily being occupied by larger households. Nonetheless, the graphs very clearly show that the older buildings consume more energy per m2, and that for the same place in the life cycle, the occupants of the older dwellings consume more per person than those in the more recent dwellings, probably because the latter are better insulated and more energy-efficient. Independently of the social characteristics of the occupants and the construction date of the dwellings, the consumption curves above all illustrate the importance of the life-cycle position to average annual household consumption per person. For equivalent social status and construction date, consumption per person is relatively stable for families, whatever the age of the reference person. However, consumption explodes when the household reference person is aged over 60, and when the household consists of just one person or a childless couple. This observation can be related to the relative invariance of dwelling size in the last two stages in the life cycle (see tables in Fig. 4) combined with the stability of energy consumption per m2 in those stages (see graphs in Fig. 4). This would seem to suggest that households remain in their dwellings after the children have left home (stability of dwelling size and logical reduction in household size), while maintaining their energy practices (increase in consumption per person and maintenance of consumption per m2). It can therefore be argued that, for owner-occupier families in a large house, the departure of the children does not result in energy practices being adapted to the new family situation. What is the situation in the rented sector? When we look at average annual consumption in the collective rented sector through the example of type 7 (high exclusive consumers of electricity), distinguishing between those living in rented social housing and in private rented housing, Fig. 5 shows that private tenants consume more energy per m2 than social housing tenants (probably better insulation).5 Moreover, the flattened U-shaped curve

4. Results discussion The role of life cycle in energy consumption is not really a new discovery. The effects of family structures and their changes were demonstrated in 1977 by a longitudinal study conducted over 5 years 6 Type 6 does not appear in the figures, because it contains only 18 individuals. 9 sequences are represented by fewer than 60 individuals: couples with children aged 30– 39 in types 1 (38), 4 (33) and 5 (58); couples with children aged 40–49 in types 1 (48), 4 (58) and 5 (58); childless couples aged 50–59 in types 1 (21), 4 (27), 5 (37) and 8 (52) The number of individuals represented in the other sequences varies from 75 (type 5 couples with children, aged 50–59) to 1099 (type 3 single people or childless couples aged 60 and above). 7 Types 1 (12), 2 (8), 4 (1) and 6 (4), were not included, since they amounted to less than 20 people. Other types are not shown, because certain sequences contain fewer than 20 individuals: this is the case for couples with children, aged 40–49, in types 8 (11), 9 (15) and 10 (15); couples with children, aged 50–59 in types 8 (6), 9 (6) and 10 (10); type 10 single people or childless couples aged 60 and above (17). Among types 5 and 7 presented in the Fig. 5 sequences contain fewer than 60 individuals: type 5 couples with children, aged 30–39 (51); type 5 couples with children, aged 40–49 (37); couples with children, aged 50–59, in types 5 (25) and 7 (35). The number of individuals in the other sequences varies from 51 (type 5 couples with children, aged 30–39) to 899 (type 7 single people or childless couples aged under 30).

5 This finding does not corroborate the results of the logistic regression, where consumption per m2 by social housing and private sector tenants appears – all else being equal – globally the same (see Table 3).

8

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 7. Average annual consumption per person according to a life-cycle snapshot of owner-occupiers of houses of more than 70 m² in France in 2006.

Fig. 8. Average annual consumption per m² according to a life-cycle snapshot of apartment tenants in France in 2006.

focusing on life-cycle sequences and more recent work focusing more on demographic characteristics alone. Indeed, our research was primarily based on the hypothesis that there are links connecting demographic and housing characteristics with the combinations of domestic energy modes used and their intensities. It was only in the course of our research that we identified the importance of life-cycle sequences in consumption processes. In this sense, and since life-cycle sequences were not one of the input variables in our study, their role in energy consumption would seem to be a strong finding from our research. The influence of life-cycle sequences shows that domestic energy consumption is highly dependent on adaptations of household size to dwelling size, and therefore, on household residential mobility. Insofar as household composition and size change throughout the household life cycle, our results illustrate the existence of flexibility over time in the processes of domestic energy consumption, with variations depending on whether these adaptations are made or not. Our findings therefore show that this adaptation seems more common in households in rented collective housing, because of their residential mobility, than in owner occupier households, which are known to be more stable in

with 217 households in the Lansing metropolitan area (Michigan) by Morrison and Gladhart [28]. The 1980s saw the publication of numerous studies linking life cycle with energy consumption. For example, following in the lineage of marketing typologies [29,30]. Fritzsche [31] studied household energy consumption by reference to stages in their life cycles, as did Frey and LaBay [32]. These different works drew on different samples and methods, and their conclusions did not converge. Since then, with the exception of a few cases [33,34], studies on household energy practices have abandoned the life-cycle approach. Nevertheless, though without necessarily employing this concept, the most recent studies on energy consumption ascribe great importance to the demographic characteristics of households [4,7,35]. Although other studies have endeavoured to combine the demographic characteristics of households and of dwellings within a single approach to energy consumption [7]. To the best of our knowledge there has up to now been no research showing the links between the combination of domestic energy modes used, the demographic characteristics of users, their location and the type of housing occupied. Our approach therefore stands at the interface between pioneering work 9

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

Fig. 9. Average annual consumption per per person according to a life-cycle snapshot of apartment tenants in France in 2006.

the household's life cycle (before the arrival of children). However, Fritzsche did not use consumption per person nor consumption per m2, but only global consumption, which may explain the differences between his conclusions and ours. Our results can also be compared with more recent work by Brounen et al. [7] based on a survey of more than 300,000 Dutch households. Using a multiple regression method, it stressed the benefit of including demographic components in the analysis of domestic consumption of heating gas and electricity. Like us, Brounen et al. [7] used the intensity of consumption per person (but not per m2). Their study brings out results that differ between modes of energy consumption. According to them, families consume more gas than older childless households, but the result is reversed when it comes to consumption per person. For electricity, older childless households consume less than families with children, both globally and per person. If our results are compared with these two studies, our conclusions differ greatly from those of Fritzsche [31], but are similar to those of Brounen et al. [7] with regard to gas consumption per person. Nevertheless, it is difficult to compare these studies, which were conducted by different methods, at different periods and in different conditions. As Fritzsche points out [31] (page 232), it is difficult to reach robust conclusions in the absence of studies based on comparable samples and methods.

their residential patterns. While these conclusions accord with several hypotheses about the variables that affect residential energy use, which have already been verified in the literature, they nevertheless do not necessarily converge with other studies based partially or exclusively either on the demographic characteristics of users, or on housing characteristics, or more rarely on both. Our analyses thus show divergences between the factors affecting global consumption per m2 and per person, depending on the different types of consumer typology constructed. Indeed, by neutralising the factors of dwelling type, location and energy modes, it becomes possible to fully explore the concept of efficiency and to evaluate the households most likely to regulate their consumption on the basis of their social and demographic characteristics [4]. Moreover, the classification of consumers into homogeneous groups made it possible to conduct a close analysis of energy consumption, by differentiating the profiles of user households and the dwellings occupied. The findings thus show the need to combine measurements of consumption per m2 (highly dependent on dwelling size) with measurements of consumption per person (highly dependent on family structures). Incentive policies seeking to cut domestic energy consumption do not generally take account of the first of these measurements. These findings argue for the construction of complex consumption models that incorporate consumption per m2 and per person in a dynamic approach that includes the changing patterns of household behaviour over the life cycle depending on residential stability or mobility. Indirectly, and independently of the consumption dynamics over time, the importance of the life-cycle sequence shows the impact of the age of heads of households and of family size on intensity of consumption. In terms of consumption per person, intensity decreases with the presence of child (ren) within the household, whatever the age of the reference person. Across all energy types, domestic energy consumption per person therefore becomes higher with the birth of children and after they leave home. Measured at each life-cycle sequence, they thus take the form of a U-shaped curve (Figs. 6–9). These findings should first be compared with those of Fritzsche [31], who employed a variance analysis (ANOVA) to study energy consumption in the different stages of the household life cycle in the USA. Unlike our own, his findings show that global energy consumption intensity takes the form of an inverted U in which the right side is higher than the left side. According to Fritzsche, when children leave home, consumption falls to a level lower than that in the first stage of

5. Conclusion and policy implications In the outcome to this study, we have been able to show the essential role of the combination of the energy mode used by households in the intensity of domestic energy consumption; these combinations are markers for specific population profiles and housing characteristics. Another strong result is to have shown that – for identical housing types and locations – consumption intensity per person varies according to households’ position in the life cycle, whereas consumption per m2 remains relatively stable. This finding reflects the importance of residential mobility in the adjustment of energy consumption to changes in household characteristics over their life cycle. It means that the intensity of consumption is associated with the differing levels of constraint imposed on households in their residential choices. And finally, it shows the flexibility of domestic energy consumption, which is still too often treated as a static variable. This study seeks to make a contribution to the understanding of the 10

Renewable and Sustainable Energy Reviews xxx (xxxx) xxx–xxx

J.-P. Lévy, F. Belaïd

determinants of domestic energy consumption in France. Its results are a vehicle of strategic information that decision-makers can use in the development of public policies to reduce and manage energy consumption in the housing stock. These initial results nevertheless need to be extended through a longitudinal approach, the only method that can further elucidate the means whereby households adapt (or fail to adapt) their energy consumption to their family and residential situations.

[16] [17]

[18] [19]

[20]

References [21]

[1] Moore JR. U.S. energy situation: exploring consumption and supply issues. Pet Account Financ Manag J 2010;29(1):1. [2] Musti S, Kortum K, Kockelman KM. Household energy use and travel: opportunities for behavioral change. Transp Res Part D: Transp Environ 2011;16(1):49–56. http://dx.doi.org/10.1016/j.trd.2010.07.005. [3] Weber C, Perrels A. Modelling lifestyle effects on energy demand and related emissions. Energy Policy 2000;28(8):549–66. [4] Valenzuela C, Valencia A, White S, Jordan JA, Cano S, Keating J, et al. An analysis of monthly household energy consumption among single-family residences in Texas, 2010. Energy Policy 2014;69:263–72. [5] Richard GN, William A. Carbon mitigation costs for the commercial building sector: discrete-continuous choice analysis of multifuel demand. Resour Energy Econ 2008;30:527–39. [6] Summerfield AJ, Pathan A, Lowe RJ, Oreszczyn T. Changes in energy demand from low energy homes building. Res Inf 2012;38(1):42–9. [7] Brounen D, Kok N, Quigley JM. Residential energy use and conservation: economics and demographics. Eur Econ Rev 2012;56:931–45. [8] Dureau F, Lévy J-P. Morphologie urbaine et consommations énergétiques: un éclairage à partir de la recherche française. In: Coutard O, Lévy J-P, editors. Ecologie urbaine, Paris, Économica-Anthropos; 2010. p. 3–14. [9] Yoo S-H, Lee JS, Kwak S-J. Estimation of residential electricity demand function in Seoul by correction for sample selection bias. Energy Policy 2007;35:5702–7. [10] Ewing R, Rong F. The impact of urban form on US residential energy use. Hous Policy Debate 2008;19(1):1–30. [11] Kaza N. Understanding the spectrum of residential energy consumption: a quantile regression approach. Energy Policy 2010;38(11):6574–85. [12] Sanquist TF, Heather O, Shui B, Bittner AC. Lifestyle factors in U.S. residential electricity consumption. Energy Policy 2012;42:354–64. [13] Belaïd F. Understanding the spectrum of domestic energy consumption: empirical evidence from France. Energy Policy 2016;92:220–33. [14] Raux C, Traisnel J-P, Pochet P, Maizia M, Croissant Y, Bagard V, et al. Analyse et modélisation des comportements transports-habitat-localisations, ETHEL, Rapport R3. Lyon: CNRS – Ministère de la Recherche, LET,; 2006. [15] Thomas M-P, Pattaroni L. Choix résidentiels et différenciation des modes de vie des

[22] [23]

[24]

[25] [26] [27] [28] [29] [30] [31] [32]

[33] [34] [35]

11

familles de classes moyennes en Suisse. Espaces Et Sociétés 2012;1(148– 149):111–27. Desjeux D. La consommation. Paris, PUF, coll. Que sais-je?; 2006. Zelem M-C. Politique de maîtrise de la demande d'énergie et résistances au changement. Une approche socio-anthropologique, Paris, l'Harmattan, coll. Logiques sociales; 2010. Devalière I. De l’inconfort thermique à la précarité énergétique. Inf Soc 2009;155(5):90–8. Maresca B, Dujin A, Picard R. La consommation d’énergie dans l’habitat entre recherche de confort et impératif écologique. Paris, CREDOC, Cahier de recherche no. 264; 2009. Belaid F, Garcia T. Understanding the spectrum of residential energy-saving behaviours: French evidence using disaggregated data. Energy Econ 2016;57:204–14. Lutzenhizer L, Hill Gossard M. Lifestyle, Status and Energy Consumption. In: Proceedings, American council for an energy efficient economy. Washington, DC: ACEEE Press 8; 2000. p. 207–22. Shove E, Pantzar M, Watson M. The dynamics of social practice: everyday life and how it changes. London: Sage; 2012. Wilhite HL. A socio-cultural analysis of changing household electricity consumption in India. In: Spreng D, editor. . TaCkling Long-term Global Energy Problems: the Contributions of Social Science, 6. Kapittel: Springer; 2012. p. 97–115. Subremon H. Pour une intelligence énergétique: ou comment se libérer de l’emprise de la technique sur les usages du logement. Métropolitiques 2012, [URL 〈http://www.metropolitiques.eu/Pour-une-intelligence-energetique.html〉]. Chappels H, Shove E. Debating the future of comfort: environmental sustainability, energy consumption and the indoor environment. Build Res Inf 2005;33(1):32–40. Benzécri J-P. L’analyse des données, Tome 2: L’analyse des correspondances. Paris: Dunod; 1973. Cibois P. Le PEM, pourcentage de l’écart maximum: un indice de liaison entre modalités d’un tableau de contingence. Bull Méthodologie Sociol 1993;40:43–63. Morrison BM, Gladhart PM. Energy and families: the crisis and the response. J Home Econ 1976;68:15–8. Wells WC, Gubar G. Life cycle concept in marketing research. J Mark Res 1966;3:355–63. Murphy PE, William AS. A modernized family life cycle. J Consum Res 1979;6:12–22. Fritzsche DJ. An analysis of energy consumption patterns by stage of family life cycle. J Mark Res 1981;18(2):227–32. Frey CJ, LaBay DG. A comparative study of energy consumption and conservation across family life cycle, In: Bagozzi RP, Tybout AN, Abor A, editors. Advances in consumer research, Vol. 10; 1983. p. 641–6. Garabuau-Moussaoui I. Behaviours, transmissions, generations: why is energy efficiency not enough? In: ECEEE summer study proceedings; 2009. p. 33–43. Lévy J-P, Roudil N, Belaïd F, Flamand A. Les déterminants de la consommation énergétique domestique. Flux 2014;96:40–54. O’Neill BC, Chen BS. Demographic determinants of household energy use in the United States. Popul Dev Rev 2002;28:53–88.