Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications

Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications Dominik Wiedenhofer a,n, Manfred Lenzen b, Julia K. Steinberger c a

Institute of Social Ecology (SEC), IFF-Vienna, Alpen-Adria University Klagenfurt, Schottenfeldgasse 29, 1070 Vienna, Austria Institute for Integrated Sustainability Analysis (ISA), School of Physics, University of Sydney, Australia c Sustainability Research Institute (SRI), School of Earth and Environment, University of Leeds, UK b

H I G H L I G H T S








We statistically analyze the energy requirements of consumption in Australia. Contrasting urban/suburban/rural consumption patterns and spatial inequality. Energy requirements are influenced by urban form, income and demographics. Urban households require less direct energy, but their total consumption is higher. Significant rebound effects can be expected when direct energy use is decreased.

art ic l e i nf o

a b s t r a c t

Article history: Received 6 June 2012 Accepted 5 July 2013

Household consumption requires energy to be used at all stages of the economic process, thereby directly and indirectly leading to environmental impacts across the entire production chain. The levels, structure and determinants of energy requirements of household consumption therefore constitute an important avenue of research. Incorporating the full upstream requirements into the analysis helps to avoid simplistic conclusions which would actually only imply shifts between consumption categories without taking the economy wide effects into account. This paper presents the investigation of the direct and indirect primary energy requirements of Australian households, contrasting urban, suburban and rural consumption patterns as well as inter- and intra-regional levels of inequality in energy requirements. Furthermore the spatial and socio-economic drivers of energy consumption for different categories of energy requirements are identified and quantified. Conclusions regarding the relationships between energy requirements, household characteristics, urban form and urbanization processes are drawn and the respective policy implications are explored. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Household consumption Energy requirements Urban form Energy footprint Rebound effect

1. Introduction The high levels of consumption and large dependence on fossil fuels in industrialized countries is one of the principal challenges for global sustainability. The changes which have been induced by the expansion of the fossil fueled economic system at all scales and in all the regions of the world are unparalleled in human history (McNeill, 2000). Serious concerns have furthermore been raised about the global production of conventional oil peaking sometime between now and in the next 10–30 years (Aleklett et al., 2009; Sorrell et al., 2010a, 2010b; Hirsch, 2008). Furthermore climate change can ultimately be expected to have direct socio-economic and ecological

n

Corresponding author. Tel.: +43 1 552 4000 413. E-mail address: [email protected] (D. Wiedenhofer).

consequences if the long term trend of increasing fossil fuel use does not change dramatically (Lynas, 2008). Some even argue that these outcomes cannot be avoided anymore, because of the inertia of political systems, individual consumer psychology and identity and strong time lags between cause and effect (Hamilton, 2010). Anyways it is desirable to achieve a thorough understanding of the structure, patterns and drivers of energy consumption, since they can indicate possibilities and barriers for change (Hertwich, 2005a). In this work, two different strands of research are being brought together to shed some light on these issues. From a consumptionbased accounting perspective all economic activities and the related energy use at all the stages of the economic process can be understood as being ultimately aimed at final consumption (Lenzen et al., 2008). This fruitfully expands the notion of energy use from the conceptually straightforward usage of fuel or electricity, towards an understanding that all goods and services required energy to be used

0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2013.07.035

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

2

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

during the stages of the economic process. This understanding is highly relevant when green consumerism advocates some sort of ‘shopping our way out of environmental problems’ – which does not live up to its promise if dealt with in a perspective formally incorporating the indirect or embodied energy requirements of consumption (Alfredsson, 2004). Rather more this perspective on the complexity and interdependencies of the modern production processes can contribute to a more substantial understanding of the challenges and possibilities for change towards a sustainable lifestyle transition (Lenzen et al., 2008). The second line of research on which this study draws has focused on structural and spatial determinants of energy use. The influence of spatial configurations of settlements and cities on individual mobility behavior and the subsequent transportation energy use has been under investigation for quite a while (Newman and Kenworthy, 1991; Kennedy et al., 2009). This rekindled the debate around the environmental impacts of urban and rural living (Shammin et al., 2010). The quality and quantity of the housing stock has also been identified as an important influence on the individual residential energy use for heating and cooling (Santin et al., 2009; Haas et al., 1998). These strands connect to the wider debate around the sustainability of ongoing urbanization around the globe, as well as possibilities for strategic interventions and economies of scale provided by cities (Jenks et al., 2000; Weisz and Steinberger, 2010). For this study Australia is divided into three broad human settlement categories – urban, suburban and rural – as a basis for examining the differences and similarities in direct and indirect energy requirements of the average resident of these regions, from a consumer lifestyle approach (Bin and Dowlatabadi, 2005). Secondly, the area-based inequalities in energy requirements and income on a national level, and within the urban, suburban and rural classifications are quantified, using a novel method proposed by Druckman and Jackson (2008a) and Steinberger et al. (2010). Finally, multivariate regression analyses are conducted, using socio-economic, spatial and climatic variables to model and identify the dominant drivers for total, direct and indirect energy requirements, as well as direct private transport energy use, public transport energy requirements, residential energy and food-related energy requirements. A discussion of these results yields interesting policy implications, especially for countries with similar settlement patterns and urban forms as found in Australia as well as for ongoing urbanization trends around the globe.

which interestingly is quite similar to USA and Norway of the 1970s (Herendeen, 1978; Herendeen et al., 1981). Total per capita primary energy requirements of consumption range from 283 GJ for the USA (in 2002), to 138 GJ for the UK (in 1996), to 12 GJ in India (in 1993–95) (Hertwich, 2011), 112 GJ for the Netherlands, 135 GJ for the UK, 123 GJ for Sweden and 130 GJ for Norway (Moll et al., 2005). Household energy (residential requirements), vehicle fuel and other mobility (transportation requirements) and food related requirement generally compromise the largest fractions of the total energy requirements of households, while also being the most energy intense1 ones (Hertwich, 2011). Overall these three categories are responsible for 70% of the environmental impacts of final consumption in the EU (measured as energy use, CO2 equivalents, resource use, land use, acidification and smog formation), while only representing 55% of the expenditure (Tukker and Jansen, 2006). Furthermore it has been found that with rising income, direct energy requirements only increase weakly, while the indirect requirements increase strongly (Lenzen, 1998a; Reinders et al., 2003; Lenzen et al., 2004; Moll et al., 2005). This is due to the fact “[…] that the commodities which are purchased by high income households but not by low income households are less energy intensive than the commodities purchased by both types of household. In other words, necessities are on average more energy intensive than luxuries, and the decrease of energy intensity with income is due to a saturation of necessities” (Lenzen, 1998a). Differences in energy requirements of urban and rural live are mainly due to differential expenditure patterns and inequalities in income. Suburban and rural live are about 10% more energy intense than urban live (Herendeen et al., 1981; Lenzen, 1998a; Lenzen et al., 2004; Shammin et al., 2010). This is due to the fact “[…] that the average person in a rural household spends their money on more energy intensive commodities than a person living in a city” (Lenzen, 1998a), namely residential energy use and mobility requirements (Munksgaard et al., 2005). At the same time urban households show consistently higher levels of total energy requirements than suburban or rural households, largely because of their overall higher incomes (Lenzen, 1998a; Wier et al., 2001; Lenzen et al., 2004).

2.1. Past studies on the drivers of energy requirements of household consumption 2. The literature on energy requirements of consumption and existing insights on the relationship between consumption and urban form From a consumption based accounting perspective, the primary energy supply of an economy, as well as the related emissions caused and resources used, can be differentiated into direct and indirect requirements. Households, governments and businesses consume energy carriers directly in the form of heating and cooking fuels, electricity as well as petrol through driving a vehicle. The indirect requirements of consumption include the industrial energy use throughout the whole production process which were required to produce all goods and services (Herendeen, 1978; Lenzen, 2001) – which are also commonly called embodied energy (Peters and Hertwich, 2008; Liu et al., 2010). In industrialized countries the fraction of indirect requirements is usually on a par with or even greater than the direct energy requirements (Lenzen, 1998a; Hertwich, 2005a; Moll et al., 2005; Jackson and Papathanasopoulou, 2008). For those developing countries which have been investigated yet, indirect requirements are found to be on par or slightly below direct energy use (Pachauri and Spreng, 2002; Pachauri, 2004; Cohen et al., 2005; Park and Heo, 2007),

Income/expenditure have been identified as the main determinants of total energy consumption (Herendeen, 1978; Reinders et al., 2003; Moll et al., 2005; Lenzen et al., 2006). Expenditure is usually preferred to income as a predictor variable, because it corresponds more closely to what households actually consume (Wier et al., 2001). Expenditure also includes social benefit transfers and various non-consumption expenses are already deducted, for example savings, taxes, donations and fines. Data on income levels on the other hand is much more readily available, for example from census data or international studies. This allows easier comparisons to other studies. Generally for income/expenditure much stronger correlations are found for indirect than for direct energy requirements (Reinders et al., 2003; Lenzen et al., 2004). Household composition and size has been found to have a significant influence, with more persons and especially more children per household leading to lower per capita consumption, even under comparable per capita incomes (Lenzen, 1998a; Wier et al., 2001; Lenzen et al., 2006). This effect is mainly due to increased sharing of commodities, living space and utilities, rather 1

Measured as total energy requirements per dollar spent.

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

than a significantly different consumption pattern of various household types (Lenzen, 1998a). For population density a weak negative influence on total household energy requirements has been found in several countries, even when expenditure levels are being controlled for (Lenzen et al., 2006). This link has been supported by various urban energy studies, which have highlighted the importance of high population density as a factor in reducing private transport energy requirements (Brown et al., 2008; Newman and Kenworthy, 1991; Kennedy et al., 2009). Also lower residential energy requirements can be expected in the more densely settled areas because of more shared walls, smaller living space per capita and potentially also more efficient heating technology such as district heating or natural gas (Rickwood, 2009). Furthermore a close relationship between climatic influences and residential energy requirements for thermal comfort has been documented (Kennedy et al., 2009; Wang et al., 2010). In their study on energy requirements of households in the European Union, Reinders et al. (2003) correct for climatic influences, but they then only apply a univariate analysis with expenditure as explanatory variable, therefore not statistically capturing the independent effect of climate. Another recent study even failed to find a significant effect associated with climate related variables, but as the author notes, this could also be the case of a spatially restricted sample (Sydney only) which could lead to climatic effects misleadingly being attributed to other variables or simply not showing up because of a lack of variation in the data (Rickwood, 2009). From this overview of the existing literature it becomes quite clear that some findings, like the relationship between direct and indirect energy requirements or the importance of income/expenditure, have been established quite well in the field. Others are still under debate, for example nationally differing influences of formal education and also the exact role of population density. While income and expenditure elasticities of consumption have been a wide and fruitful topic of research, driving factors besides income/ expenditure have only been investigated in some cases. Most studies only apply univariate methods, thereby possibly missing out on other influential variables. Investigating these impacts requires a multivariate regression methodology, as applied in this study. Furthermore rarely are more disaggregated consumption categories besides total, direct and indirect requirements being interrogated regarding their driving factors, which sheds some light on different specific possibilities and barriers for change. 2.2. Urban form(s), energy requirements and sustainability One of the widely discussed aspects of cities in the context of sustainability and climate policy relates to their physical structure, or urban form (various contributions in Jenks et al., 2000; Buxton and Scheurer, 2005; Gray and Gleeson, 2007; Grazi and van den Bergh, 2008). The term urban form and structure “[…] covers such aspects as density, geometric shape, use of land (residential, industrial) and infrastructure (road, rail, waterway), with implications for indicators such as density, fragmentation and accessibility” (Grazi et al., 2008). Furthermore it “[…] refers to the arrangement of the larger functional units of a city, reflecting both the historical development of the city and its more recent planning history […]” (Rose, 1967). Urban form can be measured as population density (Newman and Kenworthy, 1991; Grazi et al., 2008; Kennedy et al., 2009). Generally more complex land-use indicators would be desirable, but are very complicated to construct, mostly only feasible for intra-city research and can rarely be used in comparative studies (Rickwood and Glazebrook, 2009). For these purposes population density can serve as a useful proxy for more complex indicators of urban form (Rickwood and Glazebrook, 2009). The two consumption areas directly connected to urban form and structure are mobility and housing as well as their respective

3

energy and resource requirements (Rickwood et al., 2008). Incidentally these have also been identified as the most energy intense and environmentally problematic consumption categories (Moll et al., 2005; Tukker and Jansen, 2006). Furthermore the “physical infrastructure of a particular neighborhood could be one key determinant of lifestyle-related emissions that could also act as a barrier to lifestyle change. Such potential infrastructure bottlenecks to emission reductions are still relatively little understood and are one important avenue of research […]” (Baiocchi et al., 2010). On the other hand Rickwood et al. (2008) suggest that there might be a trade-off, where increased densification could lead to disproportionately large increases in embodied energy for infrastructure and dwelling constructions. The broader debate on the ideal urban form dates back to the early 19th century and most of the protagonists can be categorized as either ‘decentrist’ (dispersed and decentralized living) or ‘centrist’ (high density living) (Breheny, 1998). Breheny (1998) traces the origins of the debate in concerns about the effects of industrialization on cities in the 19th and early 20th century and documents the renewed interest in large scale planning interventions with the birth of ‘sustainable development’ in the 1980s. Based on a review of the whole debate (up to the publication of the book) he furthermore suggests a middle ground, incorporating the merits of all positions: “From the centrist case it can adopt continued, indeed tougher, containment, urban regeneration strategies, and a whole range of new intra-urban environmental initiatives. There will be environmental gains, but not at the expense of quality of life. From the decentrist case it can allow for the controlled direction of inevitable decentralization – to suburbs and towns able to support a full range of facilities and public transport, and to sites that cause the least environmental damage. It takes account of the grain of the market, without being subservient to it. It might allow for some development in the form of environmentallyconscious new settlements” (Breheny, 1998). The whole debate on sustainable cities is fraught with interests and values on what a city should be (Bulkeley and Betsill, 2003) and that strong ideological positions towards urban living strongly influences the various positions taken in the debate (Breheny, 1998). Recent work has also argued for an understanding that there does not exist one ideal form, but many, depending on the local context, existing urban structure and political possibilities (Guy and Marvin, 2000). Forster (2006), based on a review of existing planning strategies throughout Australia, concludes that the current ‘official’ vision of future urban structure in 20–30 years is one of “limited suburban expansion, a strong multi-nuclear structure with high density housing around centers and transport corridors, and infill and densification throughout the current inner and middle suburbs. Residents will live closer their work in largely self-contained suburban labor sheds, and will inhabit smaller, more energy-efficient and water-efficient houses. The percentage of trips using public transport, walking or cycling will have doubled. Regeneration programs will have broken up large concentrations of disadvantage, and […] low-income households will be able to find affordable dwellings […] within consolidation developments” (Forster, 2006). These planning visions have been critiqued heavily for their overly narrow focus, based on the increasing geographical complexity of urban life in Australia (Forster, 2006). For further contributions on the implications and prospects for urban consolidation in the Australian context, see (Buxton and Scheurer, 2005; Forster, 2006; Randolph, 2006; Dodson and Sipe, 2008; Gray and Gleeson, 2007).

3. Methods and data sources In this study, input–output analysis coupled with spatially resolved household expenditure data is used to calculate the primary energy requirements of household consumption in Australia divided into 85

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

4

statistical districts. The strength of the input–output approach lies in a complete coverage of the primary energy used during the whole production process. Regional production approaches on the other hand focus on the energy consumed in a certain geographically defined area and yield insights on regional economic structures and sector specific energy use patterns – but does not include embodied/ indirect energy requirements outside the specific geographical boundary. For a discussion of these two approaches applied to the same case study, their specific data needs, uncertainties and potential policy implications, see (Baynes et al., 2011). Environmentally extended input–output analysis is frequently applied to estimate the biophysical requirements of final demand in an economy, for example for energy, greenhouse gas emissions, pollutant emissions, nitrogen flows, water or ecological footprints (Leontief and Ford, 1970; Carter and Petri, 1989; Duchin, 1992; Dixon, 1996; Forssell and Polenske, 1998). This method was pioneered for energy in the 1970s (Bullard and Herendeen, 1975; Herendeen and Sebald, 1975; Herendeen and Tanaka, 1976) and has since been applied to many countries, for example Australia (Lenzen, 2001; Lenzen et al., 2008), Japan (Aoyagi et al., 1995), the Netherlands (Vringer and Blok, 1995; Biesiot and Noorman, 1999; Weber and Perrels, 2000), Brazil (Cohen et al., 2005), Denmark (Munksgaard et al., 2000; Wier et al., 2001), the USA (Herendeen et al., 1981; Shammin et al., 2010) and India (Pachauri and Spreng, 2002). A complete household expenditure input–output map for energy, water and ecological footprints in Australia has recently been published (Centre for Integrated Sustainability Analysis, Australian Conservation Foundation, 2007; Dey et al., 2007; Lenzen and Peters, 2010) and can be viewed on-line in the Australian Environmental Atlas, 〈http://www.acfonline.org.au〉. Further explanations of the combination of input–output methods and tables (ABS, 2009), energy statistics (ABS, 2003) and household expenditure data for the understanding of urban energy metabolisms can be found in Lenzen et al. (2008). Given this wealth of prior work, the standardization of the methodology and the limitations of a paper, the description of the methodology will be kept short and the interested reader is referred to the literature, for example (Lenzen, 2001; Kok et al., 2006; Suh, 2010). In short, the national Australian input–output tables T (ABS, 2009) and Australian energy statistics Q (ABARE, 2008) were combined in a generalized input–output analysis, national electricity data and petrol prices were replaced with region-specific values, which allows the calculations of energy multipliers (Lenzen, 2001). In this formulation x^ holds gross economic output, and I is the identity matrix. 1

m ¼ Q x^

^ 1 1

ð IT x

Þ

ð1Þ

These energy multipliers were then applied to spatially disaggregated household expenditure data y from the Australian Household Expenditure Survey (HES, ABS, 2009), to yield indirect energy requirements. Eind ¼ my

ð2Þ

Summing direct Edir and indirect energy requirements Eind yields total energy requirements Etot. Etot ¼ Edir þ Eind

ð3Þ

The energy requirements for different categories of household expenditure, and for each spatial region of Australia, were then available for analysis. 3.1. Correlation and regression analysis The household expenditure survey also contains a range of socio-economic-demographic variables s, which were submitted to a correlation analysis in order to control for multi-collinearity

(Table 3). These socio-economic, demographic and spatial variables were then used as explanatory variables in stepwise multiple regression analyses, in order to identify the most relevant variables in the explanations of levels of energy requirements and to eliminate less significant influences which could potentially confuse the regression estimation. This is meant by minimal model. For this purpose a cut-off point of t ¼2.2 (∼95% significance) has been chosen. To avoid the omitted variable bias, which occurs when relevant regressors are not included in the model, or issues of overfitting, which means that too many correlated variables are used (Verbeek, 2008), extensive testing has been conducted. Different variable combinations have been tried against relevant diagnostic statistics (students' t tests, F tests between models, not using highly collinear variables simultaneously), theoretical expectations from the literature and the clustering of variables to find the most stable models and therefore relevant drivers. The multivariate regression models that were used had the following general form: lnðEÞ ¼ ∑βi lnðsi Þ 2E ¼ ∏sβi

i

ð4Þ

Given that i

i

∂∏sβi βi sβi ∂E E β ¼ ¼ βi si i1 ¼ ¼ ∂si si ∂si si

ð5Þ

and in line with previous studies (for example Wier et al., 2001), we interpret the βi coefficients as consumption-elasticities of energy requirements in relation to their socio-economic drivers si as βi ¼

∂E ∂si = E si

ð6Þ

This makes the interpretation of these elasticities quite straightforward: a 1% increase in the explanatory variable si will result in a βi% change in the respective energy requirement. Furthermore if βi ¼1, the relationship is exactly proportional, if βi o1, the relationship is inelastic, if βi 41, the relationship is elastic (if βi o0, the same terms hold for – βi and the inverse of the socio-economic variable). The value of the student t test furthermore conveys information about the statistical significance of the interaction; higher values indicate a stronger relationship. The goodness-of-fit statistic R2 indicates what share of the variation found in the sample can be ‘explained’ by the regression model. 3.2. Data sources Household Expenditure data for the year 1999 has been taken from the Australian Bureau of Statistics, which regularly publishes them in the form of average annual household expenditure per statistical district (ABS, 2000). For this analysis, an expenditure dataset disaggregated into 85 districts covering the whole of Australia has been used, where about half the districts cover the major urban centers (Perth, Melbourne, Adelaide, Sydney and Brisbane) and the other half the rest of Australia. This dataset also contains a range of socio-economic and demographic variables which have been utilized as explanatory variables in the regression analysis. Heating degree days (HDD) were used as a proxy for weather and climatic conditions for the same year as the expenditure data. This indicator is frequently used for building energy demand management and approximate the heating needs of buildings in relation to a specified base temperature (Day, 2006; Kennedy et al., 2009; Hillman and Ramaswami, 2010). Average annual HDD values were extrapolated for all 85 districts based on weather data collected by the Australian Bureau of Meteorology from their 782 weather stations. In comparison, HDD values in the same year (1999) were estimated for Vienna at 1596, London 1053, New York 1238,

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Barcelona 368 and Tokyo 589 (Heating & Cooling Degree Days – Free Worldwide Data Calculation, 2010). The Australian climate can be divided into three climatic zones: tropical in the north-east, arid in the center and western areas and temperate in the south-east (Peel et al., 2007). The spatial distribution of steadily increasing HDD values from north to south(east) correspond quite well with this overall climatic pattern. 3.3. Methodological and data limitations Spatial datasets pose special methodological challenges (Páez and Scott, 2005; Getis, 2007). Firstly, the Modifiable Area Unit Problem relates to issues with the definition of boundaries for spatial districts and the subsequent process of aggregation (Openshaw, 1984; Atkinson, 2005). All studies based on spatially defined datasets (such as census data), are influenced by the underlying system of zoning and aggregation, which introduces additional uncertainties. The suggestions from the literature (using several different aggregation schemes, rastering of zonal data, individual level data) are not always possible and are still being researched. Secondly, geo-referenced variables often exhibit spatial autocorrelation, which means that not all observations in a spatial dataset can be assumed to be independent, violating the basic assumptions of any statistical analysis (Cliff and Ord, 1970; Páez and Scott, 2005; Getis, 2007). Moran's I index is commonly estimated in order to address this issue, where the residuals of a regression analysis are tested to check if the regression model properly captures all spatial effects (Getis, 2007). Moran's I ranges from 1 to +1, where values around 0 indicate no spatial autocorrelation, while 1/+1 means that there is perfect negative/ positive autocorrelation left in the residuals and that the model therefore does not properly capture the structure of the data. The methodology of input–output analysis is based on two important assumptions on the constancy of prices and the proportional relationship between demand and production. The constancy of prices is a generally applied and accepted part of the methodology for studies on environmental impacts and resource use, an explicit treatment can be found in Lenzen (2001). We do note that input–output data for electricity consumption and gasoline prices were adjusted by regional tariffs, and between final and intermediate consumers (Lenzen, 2001). Additionally, input–output analysis makes a so-called proportionality assumption, asserting that if final demand grows so does total industrial output. This of course assumes the absence of any slack capacity and the unrestricted availability of primary inputs. Therefore, most input–output-based impact studies are assumed to operate on marginal changes, assuming that current consumption and production patterns are altered only by small amounts. Regarding the data on household expenditure for 1999, which is used in this study a comparison with more recent surveys shows that overall household incomes increased since then, but the relative shares in the expenditure stayed rather similar – which supports the conclusions drawn from the earlier data. Only major difference is that “current housing costs” moved from 4th to 2nd place in relative importance (ABS, 2000, 5; ABS, 2011, 5). For a more thorough treatment of the uncertainties and limitations of input–output models and expenditure survey data, the interested reader is referred to the literature (Lenzen, 1998b; Lenzen et al., 2004; Kok et al., 2006; Girod and de Haan, 2010). In order to overcome some of the methodological and data limitations of this study, more detailed datasets, especially at an individual household level, are required. The limitations of averaged household data (Firebaugh, 2001) allows us to only draw general conclusions, in particular when it comes to potentials for intervention and changes at the individual level. More spatially disaggregated data would also yield more detailed insights into the relationship between urban form and different energy

5

requirement categories and would almost definitely refine the results. For Australia, due to confidenciality reasons, no such datasets are currently available. 3.4. Australian urban form, population density and the international context In 2001 about 11.2 million people, or 64% of the Australian population, live in major urban centers (ABS, 2003). The three metropolitan areas of Melbourne, Sydney and Brisbane alone are home to about 9 million people. The metropolitan area of Sydney has a population density of 3291 persons/km2, Melbourne has 1218 persons/km2, Brisbane 3483 persons/km2, Perth 2488 persons/km2 and Adelaide 5871 persons/km2. In comparison, for Tokyo a density of 13,092 persons/km2 has been reported, for Los Angeles 3036 persons/km2, for Paris 20,238 persons/km2, for Barcelona 1405 persons/km2 and for Mumbai 25,567 persons/km2 (UN DESA, 2007). It has to be noted, that such a comparison is always problematic, because the practices of measuring population density differ greatly between countries – the figures for Australia are based on official census districts, while the international data is based on UN sources. Even if considerable differences exist, it is quite clear that Australian settlement patterns are comparable to the USA, with high rise buildings in the central business district, surrounding areas with flats and apartment buildings and widely spread out suburbs with semi-detached and single houses. Approaches utilizing remote sensing would seem most appropriate for proper calculations (Schneider and Woodcock, 2008), but entail extensive data needs and are difficult to construct. For the sake of data compatibility between household expenditure surveys, census data and GIS maps (see data Section 3.2.) we have to stick to the official administrative boundaries and definitions for our further analysis. In Australia population density follows a steep gradient from few high density areas in inner city districts to already very low densities found in suburban and especially rural areas (Fig. 1). 3.5. Definition of urban, suburban and rural areas For this study, urban areas were defined as districts with a population density of more than 1000 persons/km2, suburban areas with 1000–100 persons/km2 and rural areas with less than 100 persons/km2. These ranges were defined based on two criteria: districts with very low densities are clustered in regards to population density and that districts connected with the CBDs of all five major cities are included in the ‘urban’ category. The remaining districts of these five major cities were defined as suburban. Based on these definitions, 24% of the Australian population lives in urban, 40% in suburban and 36% in rural areas (Table 1). To ensure that small modifications in the definition of the urban/suburban/rural boundaries do not lead to significant changes in the results, sensitivity analyses have been conducted. This is necessary to ensure that there are no districts close to the boundaries which strongly influence the results (for example with a population density of 999 persons/km2 but a strongly different consumption pattern than all other suburban districts). The results discussed below are robust under this sensitivity analysis, and therefore we are confident in the validity of our interpretations.

4. Results: the energy requirements of urban, suburban and rural households The average annual per capita income is highest in urban areas, significantly less in the suburbs and lowest in rural areas: a 26% decrease from urban to rural areas. For household income, the

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

6

same trend is visible, namely 23% lower household incomes in rural areas, compared to the average urban households (Table 1). The differences in total energy requirements are about half the income differences: a 12% decrease from urban to rural districts. Moreover, indirect energy requirements are the dominant share in total consumption in all districts, representing over two-thirds of the total. Levels of direct consumption only increase slightly from urban to rural, principally because of changes in transport energy requirements, with the more sparsely settled (and public transport poor) suburban and rural areas more dependent on private transportation. Residential energy consumption is roughly the same for all three regional groupings. Food-related (indirect) energy requirements are highest for the average urban inhabitant, with slightly lower levels for suburban and rural areas.

The demographics of the urban/suburban/rural categories (Fig. 2) show that the shares of under 18-year-olds are significantly higher in suburban and rural areas than in urban areas. The strong population density gradient between urban and rural areas (Fig. 2) is as well evident in the composition of housing types, with significantly higher shares of flats and semi-detached houses in urban areas. Variables relating to personal mobility show that the inhabitants of urban areas own significantly fewer cars and use them less to get to work, compared to people living in suburban and rural areas. 4.1. Regional inequalities in income and energy requirements To further investigate urban/suburban/rural differences, arearesource (AR) Gini coefficients are calculated in order to quantify levels of inequality between districts (Druckman and Jackson, 2008a; Steinberger et al., 2010). The AR-Gini employs geographical distribution rather than income cohorts to measure inequality, and can be used for energy or resource use as well as income. The Ginicoefficient ranges from 0 to 1, where 0 denotes perfect equality and 1 absolute inequality. The AR-Gini is formally defined as: n n 1 ARGini ¼ ð7Þ ∑ ∑ yi  yj 2 2n η i ¼ 1 j ¼ 1 where yi and yj hold average household consumption in the ith and jth district, n is the number of districts and η is the average household consumption across all districts. The Lorenz curve is then drawn by ranking the average household consumption per district and plotting it according to the cumulative mean household consumption against the cumulative number of districts. The AR-Gini coefficient then measures the area under the Lorenz curve. Because of the novelty of an AR-Gini application, no international comparisons can be made at this point. On the national level inequality in energy requirements between districts is closely related to the respective income distribution Car ownership to Work by Car urban

18 -64 years

suburban

under 18s

rural

flats semi-det. houses sep. houses 0% Fig. 1. Population density of Australia's south-east.

20%

40%

60%

80%

100%

Fig. 2. Spatial variation of socio-demographic variables.

Table 1 Average per capita energy requirements (GJ/cap) and income for the urban/suburban/rural clusters. n¼85

Urban (n¼ 18)

Suburban (n¼ 30)

Rural (n¼ 37)

Annual per capita income Annual household income Household size Total energy Indirect energy Direct energy Private transport (direct) Public transport (direct) Residential energy (direct) Food-related (indirect) Population (%)

A$21,003 (4798) A$51,572 (8245) 2.5 (0.3) 243 (34.7) 180 (30.9) 63 (7.3) 25 (3.8) 2 (0.9) 38 (4.9) 21 (3) 24%

A$17,729 (2994) A$48,449 (9696) 2.74 (0.35) 218 (23.1) 152 (21.4) 65 (7.8) 27 (3.9) 0.9 (0.6) 37 (5.9) 19 (2.1) 40%

A$15,456 (2766) A$39,978 (8517) 2.58 (0.21) 213 (20.5) 143 (18.2) 69 (8.4) 30 (5.5) 0.5 (0.5) 38 (6.2) 18 (2) 36%

100% 74% 26% 10% 1% 16% 9%

100% 70% 30% 13% 0.4% 17% 9%

100% 67% 33% 14% 0.2% 18% 8%

Note: Figures in brackets are standard deviations; sample size (n) for urban, suburban and rural denotes the number of statistical districts (and therefore average consumption) in the subgroup.

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ Table 2 AR-Gini coefficients for income and energy requirements.

Income per capita Total energy Indirect energy Direct energy Private transport Public transport Residential energy Food-related energy

National

Urban

Suburban

Rural

0.32 0.30 0.32 0.28 0.30 0.46 0.27 0.31

0.20 0.20 0.18 0.23 0.24 0.24 0.22 0.20

0.31 0.30 0.32 0.25 0.27 0.43 0.25 0.30

0.33 0.33 0.34 0.31 0.34 0.51 0.29 0.34

Note: 0 ¼perfect equality, 1¼ total inequality.

(0.32), with the significant exception of public transport, were a much higher AR-Gini coefficient of 0.46 is found (Table 2). Interestingly for indirect requirements the AR-Ginis are very closely related to income inequality, while for direct energy requirements this relationship, while obviously present, does not seem to be as strong. Within more homogeneous areas inequalities are expected to be lower than on the nationwide level. For this purpose the same analysis is performed on the subsets of urban, suburban and rural districts, in order to obtain the levels of intra-urban, intra-suburban and intra-rural inequalities (Table 2). Indeed, for the urban areas AR-Ginis are much lower in all the categories than on the national level, again closely following income inequality. Only the AR-Ginis for direct, private and public transportation energy requirements are more elevated than income inequality alone would suggest. For the suburban areas AR-Ginis are overall only slightly below national levels, with direct, private and residential energy requirements being significantly lower. In rural areas AR-Ginis are all slightly higher than nationwide results. These elevated intra-rural inequalities are an indication that this regional category is much more heterogeneous than both urban and suburban, which is to be expected: it encompasses rich coastal settled areas to the southeast as well as low income outback regions2. Overall these results also suggest that intra-regional inequality in total and indirect energy requirements are quite closely related to income inequality, while this relationship does not seem to be as strong for direct (especially private and public) energy requirements. This suggests that besides income the general spatial structure of settlements does have a significant influence. 4.2. Cross-correlations of socio-economic and spatial variables By isolating the specific effects of socio-economic, demographic and spatial explanatory factors, closer insights into the determinants of energy requirements of the average households in urban, suburban and rural districts can be gained. For this purpose an analysis of the cross-correlations between the explanatory variables and the different categories of energy requirements has been conducted (Table 3). In the next step a multivariate regression methodology can then be applied. These interpretations could also furthermore be used to derive a typology of “typical” urban/suburban/rural average households. The cross-correlations between variables suggest specific relationships, where the following interpretations are based on strong correlations and statistical significance. The first clustering is found around the variables semidetached houses and flats, which are predominantly found in inner urban areas (Fig. 2): in these districts income tends to be highest, population density is relatively large, the majority of the 2 The attentive reader may have noticed that, upon first sight, this is in contradiction with the higher standard deviations for income found for urban areas (Table 1); this can be attributed to lower sample sizes for urban (n ¼18) than for suburban (n¼ 30) and rural (n¼ 37) areas, rather than inconsistencies in results.

7

population is in their working age, households tend to be smaller, with also more single households and car use to get to work as well as car ownership rates tend to be lower than in the rest of Australia (see also Fig. 2). Such an affluent working age urban lifestyle correlates significantly with relatively high total, indirect, public transport and food-related energy requirements. A second clustering of cross-correlations can be identified along the variable separate houses, which predominantly are found in suburban and rural districts (see Fig. 2). In these areas income per capita tends to be lower than in urban areas (see also Table 1), population density is also lower, while households are generally larger and the share of under 18 year olds is significantly higher. Furthermore car use and ownership rates are generally higher (see also Fig. 2). According to the cross-correlations the total, indirect and food-related energy requirements are relatively lower in these areas, but direct and private transport energy requirements are significantly elevated. Overall the results suggest two consumption patterns: of affluent urban dwellers vs cardependent suburban and rural families. Furthermore, as expected from the literature review, population density correlates positively with total, indirect and public transport energy requirements and negatively with direct and private transport requirements. Heating Degree Days (HDD) also exhibit the expected positive relationship with direct and residential energy requirements. 4.3. Isolating the influence of different drivers of energy requirements Applying a multivariate regression approach allows for the isolation of each specific effect: as expected, income per capita exhibits a strong influence on total (β¼0.41) and indirect (β¼0.51) energy requirements, but significantly weaker impact on direct (β¼0.20) energy requirements (Table 4). Because the β regression coefficients can be interpreted like consumption-elasticities (see Section 3), this means that changes in income levels can be expected to have the strongest influence on indirect energy requirements, and the least on direct energy consumption. Furthermore household size (β¼  0.19) has a significant negative influence on total energy requirements, possibly due to economies of scale achieved by increased sharing of living space and appliances among household members and/or because of lower requirements of children and elderly household members. Overall larger households have lower per capita expenditure on the goods and services affected by such household economies of scale. But at the same time the shares of expenditures on various commodities generally does not change with household size and the energy intensity of consumption is similarly unchanged (Lenzen, 1998a). It is quite surprising that household size is only significant for total energy requirements, since other studies also found such an influence for direct and indirect requirements, although the strength of the effect is usually much smaller than the one identified here (Wier et al., 2001; Lenzen et al., 2004; Cohen et al., 2005; Druckman and Jackson, 2008b). However, since a different set of variables was used than in the previous studies, it is possible that the influence of household size seen in other studies appears in this work through other variables, or vice versa. For instance the variable ‘travel to work by car’ is positively correlated with household size (Table 3) and part of its negative influence on indirect energy requirements could be seen by other studies with different variable sets as a household size effect. The effects of “travel to work by car” (β¼  0.28) suggest that independent of income, higher shares of commuting by car are associated with lower indirect energy requirements (Table 4). Other studies usually do not include such a variable, usually only car ownership or related proxies and different housing type

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

8

Table 3 Cross-correlations of socio-demographic, spatial and climatic variables and energy requirements. n ¼85

1

Income per capita Separate dwellings Attached dwellings Apartments Under 18-year-olds 18–64 years olds Household size Population density To work by car Car ownership Heating degree days Total energy Indirect energy Direct energy Private transport Public transport Residential energy Food-related energy

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

 0.56 0.50 0.52  0.57 0.67  0.18 0.47  0.53  0.27  0.05 0.80 0.82 0.05  0.17 0.45 0.29 0.72

2

3

4

5

6

7

8

9

10

11

 0.56

0.50  0.66

0.52  0.76 0.65

 0.57 0.80  0.51  0.65

0.67  0.57 0.47 0.45  0.61

 0.18 0.46  0.24  0.38 0.60  0.08

0.47  0.40 0.63 0.51  0.42 0.36  0.03

 0.53 0.83  0.52  0.64 0.58  0.46 0.33  0.22

 0.27 0.41  0.35  0.40 0.17  0.35  0.13  0.20 0.62

 0.05 0.17  0.16  0.24  0.02  0.15  0.01 0.02 0.22 0.35

 0.66  0.76 0.80  0.57 0.46  0.40 0.83 0.41 0.17  0.51  0.58 0.21 0.29  0.43  0.03 -0.55

0.65  0.51 0.47  0.24 0.63  0.52  0.35  0.16 0.37 0.50  0.32  0.43 0.62  0.03 0.46

 0.65 0.45  0.38 0.51  0.64  0.40  0.24 0.48 0.59  0.31  0.41 0.52  0.01 0.50

 0.61 0.60  0.42 0.58 0.17  0.02  0.57  0.58  0.02 0.13  0.32  0.21  0.63

 0.08 0.36  0.46  0.35  0.15 0.47 0.53  0.06  0.09 0.32 0.07 0.54

 0.03 0.33  0.13  0.01  0.32  0.27  0.16 0.00  0.11  0.24  0.23

 0.22  0.20 0.02 0.31 0.47  0.42  0.46 0.72  0.13 0.33

0.62 0.22  0.53  0.59 0.18 0.28  0.39  0.05  0.51

0.35  0.16  0.27 0.29 0.31  0.38 0.09  0.23

0.09  0.05 0.42 0.18 0.01 0.45  0.06

Note: Bold values indicate 99% statistical significance (p 40.99), italic values indicate 95% and non-bold values are not significant.

Table 4 Minimum multivariate requirements. n ¼85

Income per capita Household size Population density To work by car Heating degree days R2 Adj. R2 F stat Moran's I

regression

models

for

total,

indirect

and

direct

Total energy requirements

Indirect energy requirements

Direct energy requirements

β

|t|

β

|t|

β

|t|

0.41  0.19 – – – 0.67 0.66 81.9 0.02

11.8 2.9 – – –

0.51 – –  0.28 – 0.70 0.69 96.0  0.01

9.8 – – 3.1 –

0.20 –  0.02 – 0.03 0.45 0.43 60.9 0.04

3.8 – 6.4 – 5.4

Note: Blank fields indicate that the variable was omitted in this model for reasons of colinearity or lack of significance. The R2, F stat, t-statistic and Moran's I are all described in Section 3.

indices are used (Lenzen et al., 2006; Baiocchi et al., 2010; Shammin et al., 2010). But “travel to work by car” seems like an useful proxy for behavior, capturing the way people deal with their mobility requirements in the light of overall urban forms and their perceived options. Furthermore this strong effect can also be understood in the wider context of car dependency in sparsely settled suburbs and rural areas with poor public transport options, combined with oil price vulnerability and high shares of mortgage financed housing – which in combination strongly constrain household spending on not housing or mobility related goods and services (Dodson and Sipe, 2007, 2008). As expected, heating degree days (β¼0.03) and population density (β ¼  0.02) have highly significant (t ¼5.4 and 6.4) but relatively weak influences on direct energy requirements. This is consistent with previous findings as discussed below.

4.4. A further disaggregation of energy requirements and their drivers To shed further light on these relationships and as an extension on previous work, a further disaggregation of direct (into private and

Table 5 Minimum multivariate regression results for private and public transport, residential and food-related energy requirements. n¼ 85

Income per capita Under 18-year-olds Population density Car ownership Heating degree days R2 Adj. R2 F stat Moran's I

Private transport

Public transport

Residential

Foodrelated

β

|t|

β

|t|

β

|t|

β

|t|

– –  0.02 0.38 – 0.26 0.25 14.6 0.15

– – 4.3 2.4 –

– – 0.2  1.96 – 0.57 0.56 54.5  0.01

– – 9.0 3.3 –

0.35 –  0.02 – 0.04 0.41 0.38 18.5 0.25

5.0 3.8 – 5.6

0.31  0.15 – – – 0.60 0.59 60.9  0.01

6.3 3.9 – – –

Note: Blank fields indicate that the variable was omitted in this model for reasons of colinearity or lack of significance. The R2, F stat, t-statistic and Moran's I are all described in Section 3.

public transport as well as residential energy requirements) and indirect (food-related) consumption should yield additional insights. Surprisingly our results suggest, that income does not play a statistically significant role for the average levels of public and private transport energy requirements (Table 5). Using the variables at our disposal, private transport energy requirements are best fitted by population density (β¼  0.02) and car ownership (β¼0.38), although these only explain 25% of the variation in the data. The remaining spatial autocorrelation indicates that there are spatial effects which, with the variables at hand, cannot be captured properly (see Moran's I, Table 5 and Section 3). The same variables, however with much stronger relationships and opposite signs (β¼0.20 and  1.96), cover 55% of the variation in public transport energy requirements across districts, with no spatial autocorrelation remaining in the residuals. The connection between population density and levels of individual mobility energy requirements has been well documented (Newman and Kenworthy, 1991; Camagni et al., 2002; Grazi et al., 2008). In more densely settled areas average trip lengths are shorter and above a certain threshold public transport systems can also supply enough coverage to attract sufficient riders (Rickwood and Glazebrook, 2009). Besides well documented correlations there is still some debate about specific causalities (Cameron et al., 2003) of demographic influences (young

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

affluenct professionals in inner city districts versus generally families in suburbs) (Rickwood and Glazebrook, 2009), effects of residental self-selection in location choice and direct impacts of urban form on travel mode choice. In a meta-review across 38 studies, Cao et al. (2009) found strong evidence that the built environment does indeed influence travel behavior, even when self-selection processes are controlled for. In regards to residential energy requirements income (β¼0.35), population density (β¼  0.02) and heating degree days (β¼ 0.04) were found to be statistically significant, although only providing a moderate fit (adj. R2 ¼ 38%) and not completely removing the spatial autocorrelation from the data (Moran's I ¼0.25). The positive relationship of heating degree days and residential energy requirements is consistent with other findings (Kennedy et al., 2009), where also a large variability of heating energy consumption at similar levels of heating degree days have been found (Balaras et al., 2005). Parts of this variation is captured by the variable income, which is related to appliance ownership rates, floor sizes as well as individual heating behavior. The negative influence of population density is due to less per capita floor space in higher density residential areas, where flats and semi-detached houses are dominant, which also tend to be more energy-efficient than separate houses (Rickwood, 2009). The variable ‘under 18-year-olds’ interestingly exhibits a negative influence on food-related energy requirements, even when controlling for household income. The findings of Dodson and Sipe (2007, 2008) seem especially relevant here: in Australia young families with children often live in car-dependent, mortgage-financed separate houses in suburban areas, which significantly impacts their budgetary options. Further research with household-level data and complementary information would be needed to explore this interpretation.

5. Conclusions and policy implications Overall this work has revealed significant differences in per capita energy requirements of household consumption for the average urban, suburban and rural households in Australia. The results are consistent with earlier findings for Sydney (Lenzen et al., 2004): in the more urban and wealthier districts, indirect and total energy requirements are highest, which translates into higher consumption of all kinds of goods and services. At the same time direct energy use, for example petrol or electricity, is lowest for urban households, compared to average suburban and rural households. For public transport energy requirements a steady decrease when moving from urban to rural districts has been found, while private transport energy use, i.e. petrol, is generally higher in suburban and rural areas. Even at equal incomes, districts with high levels of car dependency have also lower levels of indirect energy requirements, meaning that more of their household budgets is spent on petrol and relatively less can be used to consume all other goods and services which are not energy carriers. Whilst we have cast our analysis in terms of energy requirements only, our findings are also applicable to greenhouse gas emissions, at least that part caused by fossil fuel combustion. However due to limitations in journal space, we report on energy requirements only. Interestingly rather similar results have also been found for greenhouse gas emissions due to household consumption in Finland (Heinonen and Junnila, 2011a, 2011b). These results indicate that rising fuel costs are going to impact car-dependent and on average less affluent districts, such as most of the suburbs and rural Australia, much more than already wealthy urban areas, where there are also more possibilities and resources for adaptation, for instance, increased public transport use (Dodson and Sipe, 2008). In particular, suburban areas in their current form are locked into an energy intensive, automobile-dependent lifestyle built

9

on uncontrolled low-density urban sprawl (Buxton and Scheurer, 2005; Gleeson, 2006; Frost and O'Hanlon, 2009; Spearritt, 2009). Interestingly the results for private and public transport requirements both point to the same phenomenon: that mobility behavior depends on availability and access to public transport infrastructure as well as broader issues related to urban form (such as mixed vs single use areas), spatial distribution of workplaces, facilities and shops, as well as individual preferences regarding automobile ownership and use (Camagni et al., 2002; Buxton and Scheurer, 2005; Grazi et al., 2008). Dense urban form is also a prerequisite for an efficient and attractive public transportation system, as well as greater possibilities for walking and cycling because of shorter distances (Grazi et al., 2008). Furthermore, as our analysis shows, lowering private transportation energy requirements for environmental reasons may not be as effective as expected. In order to make these changes politically and socially viable, they would probably need to be cost saving in comparison to private car use, which in turn can be expected to lead to increases in indirect energy consumption due to re-spending of saved costs. Although expenditure on indirect energy requirements is likely to be less energy intensive than car use (Lenzen, 1998a), shifts in expenditure patterns from automobile use to other goods and services will lead to significantly lower reductions in total energy requirements than expected from the decrease in car use alone. This is an important finding for policy making, as it suggests that focusing on restricted categories of residential energy use or private transportation alone is probably missing out on significant rebound effects, thereby undermining the intentions of the respective policy because of shifts in expenditure patterns. Under future climate change the findings on residential energy requirements are also relevant, as large parts of the existing housing stock will still be in use over the coming decades, thereby exhibiting strong pressures on local energy systems (Wang et al., 2010). Residential energy requirements generally are a function of climatic effects, the specific building envelope construction, heating system typology, thermal efficiency, controls, annual hours of use and highly important, occupant behavior (Balaras et al., 2005). As technical efficiency increases through building standards and renovation activities, behavioral aspects become ever more important (Haas et al., 1998; Santin et al., 2009). Furthermore while a warming climate can be expected to lead to decreases in annual heating degree days, cooling energy requirements especially in the sub-tropical regions of Australia are probably going to rise (Wang et al., 2011). Mitigation strategies therefore need to take high efficiency standards for new buildings, improvements of the existing housing stock and future temperature levels as well as increased variability into account, so as not to overestimate potential gains (Wang et al., 2011). More to the point, renovation of the existing building stock should probably be even prioritized, because the construction of new buildings, even highly efficient ones, primarily causes additional energy use and greenhouse gas emissions during production of materials and construction activities, while the benefits of lower heating/cooling demand are only reaped in the mid- to long term (Säynäjoki et al., 2012). Overall, while the structural aspects of urban form allow for lower direct energy use and therefore lower energy intensities of consumption, these inherently positive aspects are usually negated by generally more affluent lifestyles in urban areas (Lenzen et al., 2008), which go hand in hand with higher availability of a multitude of goods and services in these locations. The discussion of different drivers of energy requirements also suggests possibilities for change leading to savings in total energy requirements, for example due to lower private transportation requirements or intra-household sharing. But it is important to keep in mind that these changes do not automatically translate to proportionally lower total energy requirements, because of the rebound effect, also known as Jevon′s paradox

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

10

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

(Hertwich, 2005b). Furthermore Abrahamse and Steg (2009) conclude that “contextual variables such as income shape households' opportunities for energy consumption, whereas reductions in energy use require conscious efforts to change behaviors/adopt energy-saving measures” (Abrahamse and Steg, 2009). Following from all this, a new form of urbanism and human settlement patterns are necessary to enable households to actually decrease their total energy requirements. A sustainable lifestyle transition would need to occur (Lenzen et al., 2008). However, current developments in Australian urban and transport policy seem to keep focusing on large infrastructure projects as well as unchecked and hardly discussed (sub)urban sprawl (Dodson, 2009; Spearritt, 2009). Furthermore current urban consolidation efforts in Australia simplistically focus on solely increasing density, without taking into consideration the interplay between land-use planning, transport and infrastructure planning and the role of prices on total energy requirements of consumption (Gray and Gleeson, 2007). Additionally purely technical solutions, like a rapid transformation of the energy system, which in Australia is highly dependent on coal and oil, to renewable electricity can be expected to yield greenhouse gas reductions probably not until the 2nd half of the 21st century (Myhrvold and Caldeira, 2012). This is because new infrastructure primarily requires energy for its production and these CO2 emissions have long atmospheric lifetimes, while savings accrue only slowly over time. Well designed combinations of planning and non-planning policies will therefore be necessary to counteract rebound effects and simple shifts in consumption patterns, taking the specific situation of a city or region and the households into account (Kyrö et al., 2012; Jones and Kammen, 2011). Overall, Australia has one of the highest levels of per capita energy use, greenhouse gas emissions and materials use (Wood et al., 2009). The energy system is mostly dependent on coal for electricity, except in Tasmania where hydropower delivers an important share. Exports earnings of extractive and energy intensive industries play a major role for the economy, while climate change threatens the agricultural system and extreme events like floods or heat waves are already common. Overall, it is clear that the reductions required for serious climate change mitigation will be only possible with much more systemic and fundamental changes of production and consumption patterns than indicated here – a challenge faced by all developed and developing societies.

Acknowledgments Some of these results appeared in a shortened and earlier version as a book chapter in “Urban Consumption” (Wiedenhofer et al., 2011), edited by Prof. Peter Newton and published by CSIRO Publishing. Many thanks go to Christoph Plutzar for his GIS support and to Shonali Pachauri for helpful comments on an earlier draft. Alpen-Adria University Klagenfurt supported Dominik Wiedenhofer with a travel grant for his research visit at the University of Sydney. We furthermore thank the reviewers for constructive and insightful feedback.

References Aleklett, K., Hook, M., Jakobsson, K., Lardelli, M., Snowden, S., Soderbergh, B., 2009. The peak of the oil age – analyzing the world oil production: reference scenario in world energy outlook 2008. Energy Policy 38 (3), 1398–1414. ABARE, 2008. Australian Energy Statistics: Australian Energy Update 2008 [WWW Document]. URL 〈http://www.abare.gov.au/interactive/energyUPDATE08/index. html〉. Abrahamse, W., Steg, L., 2009. How do socio-demographic and psychological factors relate to households' direct and indirect energy use and savings? Journal of Economic Psychology 30, 711–720. ABS, 2000. Household Expenditure Survey: Summary of Results 1998–99. Australian Bureau of Statistics, Canberra, Australia.

ABS, 2003. 2001 Census Household Sample File (CURF). Australian Bureau of Statistics, Canberra, Australia. ABS, 2009. 1998–99 Household Expenditure Survey: Detailed Expenditure Items. Australian Bureau of Statistics, Canberra, Australia. ABS, 2011. Household Expenditure Survey: Summary of Results 2009–10. Australian Bureau of Statistics, Canberra, Australia. Alfredsson, E.C., 2004. “Green” consumption – no solution for climate change. Energy 29, 513–524. Aoyagi, M., Moriguchi, Y., Kondo, Y., Shimizu, H., 1995. Characteristics of household's energy consumption. Energy and Resources, 59–67. Atkinson, P.M., 2005. Problems of spatial scale, In: Encyclopedia of Social Measurement. Elsevier, New York, pp. 639–648. ISBN: 978-0-12-369398-3. Baiocchi, G., Minx, J., Hubacek, K., 2010. The impact of social factors and consumer behavior on carbon dioxide emissions in the United Kingdom. Journal of Industrial Ecology 14, 50–72. Balaras, C.A., Droutsa, K., Dascalaki, E., Kontoyiannidis, S., 2005. Heating energy consumption and resulting environmental impact of European apartment buildings. Energy and Buildings 37, 429–442. Baynes, T., Lenzen, M., Steinberger, J.K., Bai, X., 2011. Comparison of household consumption and regional production approaches to assess urban energy use and implications for policy. Energy Policy 39, 7298–7309. Biesiot, W., Noorman, K.J., 1999. Energy requirements of household consumption: a case study of The Netherlands. Ecological Economics 28, 367–383. Bin, S., Dowlatabadi, H., 2005. Consumer lifestyle approach to US energy use and the related CO2 emissions. Energy Policy 33, 197–208. Breheny, M., 1998. Centrists, decentrists and compromisers: views on the future of urban form. In: Jenks, M., Burton, E., Williams, K. (Eds.), The Compact City: A Sustainable Urban Form?. E & FN Spon, London, New York. Brown, M.A., Southworth, F., Sarzynski, A., 2008. Shrinking the Carbon Footprint of Metropolitation America, Metropolitatn Policy Program at Brookings. Brookings Institution, Washington, DC. Bulkeley, H., Betsill, M., 2003. Cities and Climate Change: Urban Sustainability and Global Environmental Governance. Routledge, New York. Bullard, C.W., Herendeen, R.A., 1975. The energy cost of goods and services. Energy Policy 3, 268–278. Buxton, M., Scheurer, J., 2005. The Expanding Urban Frontier: Urban Form in Melbourne's Growth Areas. RMIT Publishing, Melbourne, Australia. Camagni, R., Gibelli, M.C., Rigamonti, P., 2002. Urban mobility and urban form: the social and environmental costs of different patterns of urban expansion. Ecological Economics 40, 199–216. Cameron, I., Kenworthy, J.R., Lyons, T.J., 2003. Understanding and predicting private motorised urban mobility. Transportation Research Part D: Transport and Environment 8, 267–283. Cao, X., Mokhtarian, P.L., Handy, S.L., 2009. Examining the impacts of residential self-selection on travel behaviour: a focus on empirical findings. Transport Reviews: A Transnational Transdisciplinary Journal 29, 359–395. Carter, A.P., Petri, P.A., 1989. Leontief's contribution to economics. Journal of Policy Modeling 11, 7–30. Centre for Integrated Sustainability Analysis, Australian Conservation Foundation, 2007. ACF Consumption Atlas [WWW Document]. URL 〈http://www.acfonline. org.au/consumptionatlas/〉. Cliff, A.D., Ord, K., 1970. Spatial autocorrelation: a review of existing and new measures with applications. Economic Geography 46, 269–292. Cohen, C., Lenzen, M., Schaeffer, R., 2005. Energy requirements of households in Brazil. Energy Policy 33, 555–562. Day, T., 2006. Degree Days: Theory and Application (No. TM41). Chartered Institution of Building Services Engineers, London, UK. Dey, C., Berger, C., Foran, B., Foran, M., Joske, R., Lenzen, M., Wood, R., 2007. Household environmental pressure from consumption: an Australian environmental atlas. In: Birch, G. (Ed.), Water Wind Art and Debate: How Environmental Concerns Impact on Disciplinary Research. University Press, Sydney, Sydney. Dixon, R., 1996. Inter-industry transactions and input–output analysis. Australian Economic Review 29, 327–336. Dodson, J., 2009. The “infrastructure turn” in Australian Metropolitan Spatial Planning. International Planning Studies 14, 109–123. Dodson, J., Sipe, N., 2007. Oil vulnerability in the Australian City: assessing socioeconomic risks from higher urban fuel prices. Urban Studies 44, 37–62. Dodson, J., Sipe, N., 2008. Shocking the suburbs: urban location, homeownership and oil vulnerability in the Australian City. Housing Studies 23, 377–401. Druckman, A., Jackson, T., 2008a. Measuring resource inequalities: the concepts and methodology for an area-based Gini coefficient. Ecological Economics 65, 242–252. Druckman, A., Jackson, T., 2008b. Household energy consumption in the UK: a highly geographically and socio-economically disaggregated model. Energy Policy 36, 3177–3192. Duchin, F., 1992. Industrial input–output analysis: implications for industrial ecology. Proceedings of the National Academy of Sciences of the United States of America 89, 851–855. Firebaugh, G., 2001. Statistics of ecological fallacy, In: International Encyclopedia of the Social & Behavioral Sciences. Pergamon, Oxford, pp. 4023–4026. Forssell, O., Polenske, K.R., 1998. Introduction: input–output and the environment. Economic Systems Research 10, 91–97. Forster, C., 2006. The challenge of change: Australian cities and urban planning in the new millennium. Geographical Research 44, 173–182. Frost, L., O'Hanlon, S., 2009. Urban history and the future of Australian cities. Australian Economic History Review 49, 1–18.

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ Getis, A., 2007. Reflections on spatial autocorrelation. Regional Science and Urban Economics 37, 491–496. Girod, B., de Haan, P., 2010. GHG reduction potential of changes in consumption patterns and higher quality levels: evidence from Swiss household consumption survey. Energy Policy 37, 5650–5661. Gleeson, B., 2006. Australian Heartlands Making Space for Hope in the Suburbs. Allen & Unwin, [S.l.]. Gray, R., Gleeson, B., 2007. Energy demands of urban living: what role for planning? In: Presented at the State of Australian Cities Conference 2007. Adelaide, SA, Australia. Grazi, F., van den Bergh, J.C.J.M., 2008. Spatial organization, transport, and climate change: comparing instruments of spatial planning and policy. Ecological Economics 67, 630–639. Grazi, F., van den Bergh, J.C.J.M., van Ommeren, J.N., 2008. An empirical analysis of urban form, transport, and global warming. Energy Journal 29, 97–122. Guy, S., Marvin, S., 2000. Models and pathways: the diversity of sustainable urban futures, In: Achieving Sustainable Urban Form. E & FN Spon, London, New York. Haas, R., Auer, H., Biermayr, P., 1998. The impact of consumer behavior on residential energy demand for space heating. Energy and Buildings 27, 195–205. Hamilton, C., 2010. Requiem for a Species: why we Resist the Truth About Climate Change. Earthscan, London; Washington, DC. Heating & Cooling Degree Days – Free Worldwide Data Calculation [WWW Document], 2010. BizEE Degree Days. URL 〈http://www.degreedays.net/〉. Heinonen, J., Junnila, S., 2011a. A carbon consumption comparison of rural and urban lifestyles. Sustainability 3, 1234–1249. Heinonen, J., Junnila, S., 2011b. Implications of urban structure on carbon consumption in metropolitan areas. Environmental Research Letters 6 (1), 014018. Herendeen, R., 1978. Total energy cost of household consumption in Norway, 1973. Energy 3, 615–630. Herendeen, R., Sebald, A., 1975. Energy, employment and dollar impacts of certain consumer options. In: Williams, R.H. (Ed.), The Energy Conservation Papers. Ballinger, Cambridge, Mass. (pp. 131–170). Herendeen, R., Tanaka, J., 1976. Energy cost of living. Energy 1, 165–178. Herendeen, R.A., Ford, C., Hannon, B., 1981. Energy cost of living, 1972–1973. Energy 6, 1433–1450. Hertwich, E., 2011. The life cycle environmental impacts of consumption. Economic Systems Research 23, 27–47. Hertwich, E.G., 2005a. Life cycle approaches to sustainable consumption: a critical review. Environmental Science & Technology 39, 4673–4684. Hertwich, E.G., 2005b. Consumption and the rebound effect: an industrial ecology perspective. Journal of Industrial Ecology 9, 85–98. Hillman, T., Ramaswami, A., 2010. Greenhouse gas emission footprints and energy use benchmarks for eight U.S. cities. Environmental Science & Technology 44, 1902–1910. Hirsch, R.L., 2008. Mitigation of maximum world oil production: shortage scenarios. Energy Policy 36, 881–889. Jackson, T., Papathanasopoulou, E., 2008. Luxury or “lock-in”? An exploration of unsustainable consumption in the UK: 1968–2000. Ecological Economics 68, 80–95. Jenks, M., Williams, K., Burton, E., 2000. Achieving Sustainable Urban Form. E & FN Spon, London, New York. Jones, C., Kammen, D., 2011. Quantifying carbon footprint reduction opportunities for U.S. households and communities. Environmental Science & Technology 45, 4088–4095. Kennedy, C., Steinberger, J., Gasson, B., Hansen, Y., Hillman, T., Havránek, M., Pataki, D., Phdungsilp, A., Ramaswami, A., Mendez, G.V., 2009. Greenhouse gas emissions from global cities. Environmental Science & Technology 43, 7297–7302. Kok, R., Benders, R.M.J., Moll, H.C., 2006. Measuring the environmental load of household consumption using some methods based on input–output energy analysis: a comparison of methods and a discussion of results. Energy Policy 34, 2744–2761. Kyrö, R., Heinonen, J., Säynäjoki, A., Junnila, S., 2012. Assessing the potential of climate change mitigation actions in three different city types in Finland. Sustainability 4, 1510–1524. Lenzen, M., 1998a. Energy and greenhouse gas cost of living for Australia during 1993/94. Energy 23, 497–516. Lenzen, M., 1998b. Primary energy and greenhouse gases embodied in Australian final consumption: an input–output analysis. Energy Policy 26, 495–506. Lenzen, M., 2001. A generalized input–output multiplier calculus for Australia. Economic Systems Research 13, 65–92. Lenzen, M., Dey, C., Foran, B., 2004. Energy requirements of Sydney households. Ecological Economics 49, 375–399. Lenzen, M., Peters, G.M., 2010. How city dwellers affect their resource Hinterland. Journal of Industrial Ecology 14, 73–90. Lenzen, M., Wier, M., Cohen, C., Hayami, H., Pachauri, S., Schaeffer, R., 2006. A comparative multivariate analysis of household energy requirements in Australia, Brazil, Denmark, India and Japan. Energy 31, 181–207. Lenzen, M., Wood, R., Foran, B., 2008. Direct versus embodied energy – the need for urban lifestyle transitions, in Urban Energy Transition. Elsevier, Amsterdam. (pp. 91–120). Leontief, W., Ford, D., 1970. Environmental repercussions and the economic structure: an input–output approach. Review of Economics and Statistics 52, 262–271.

11

Liu, H., Xi, Y., Guo, J., Li, X., 2010. Energy embodied in the international trade of China: an energy input–output analysis. Energy Policy 38, 3957–3964. Lynas, M., 2008. Six Degrees: Our Future on a Hotter Planet. National Geographic, Washington, DC. McNeill, J., 2000. Something new Under the sun: An Environmental History of the World in the Twentieth-Century. Allen Lane, London. Moll, H.C., Noorman, K.J., Kok, R., Engström, R., Throne-Holst, H., Clark, C., 2005. Pursuing more sustainable consumption by analyzing household metabolism in European countries and cities. Journal of Industrial Ecology 9, 259–275. Munksgaard, J., Pedersen, K.A., Wien, M., 2000. Impact of household consumption on CO2 emissions. Energy Economics 22, 423–440. Munksgaard, J., Wier, M., Lenzen, M., Dey, C., 2005. Using input–output analysis to measure the environmental pressure of consumption at different spatial levels. Journal of Industrial Ecology 9, 169–185. Myhrvold, N.P., Caldeira, K., 2012. Greenhouse gases, climate change and the transition from coal to low-carbon electricity. Environmental Research Letters 7, 014019. Newman, P.W.G., Kenworthy, J.R., 1991. Transport and urban form in thirty-two of the world′s principal cities. Transport Reviews: A Transnational Transdisciplinary Journal 11, 249–272. Openshaw, S., 1984. Ecological fallacies and the analysis of areal census data. Environment and Planning A 16, 17–31. Pachauri, S., 2004. An analysis of cross-sectional variations in total household energy requirements in India using micro survey data. Energy Policy 32, 1723–1735. Pachauri, S., Spreng, D., 2002. Direct and indirect energy requirements of households in India. Energy Policy 30, 511–523. Páez, A., Scott, D., 2005. Spatial statistics for urban analysis: a review of techniques with examples. GeoJournal 61, 53–67. Park, H.-C., Heo, E., 2007. The direct and indirect household energy requirements in the Republic of Korea from 1980 to 2000 – an input–output analysis. Energy Policy 35, 2839–2851. Peel, M.C., Finlayson, B.L., McMahon, T.A., 2007. Updated World Map of the KöppenGeiger Climate Classification. Peters, G.P., Hertwich, E.G., 2008. CO2 embodied in international trade with implications for global climate policy. Environmental Science & Technology 42, 1401–1407. Randolph, B., 2006. Delivering the compact city in Australia: current trends and future implications. Urban Policy and Research 24, 473–490. Reinders, A.H.M.E., Vringer, K., Blok, K., 2003. The direct and indirect energy requirement of households in the European Union. Energy Policy 31, 139–153. Rickwood, P., 2009. Residential operational energy use. Urban Policy and Research 27 (19), 137–155. Rickwood, P., Glazebrook, G., 2009. Urban structure and commuting in Australian cities. Urban Policy and Research 27, 171–188. Rickwood, P., Glazebrook, G., Searle, G., 2008. Urban structure and energy – a review. Urban Policy and Research 26, 57–81. Rose, A.J., 1967. Patterns of Cities, Nelson's Australasian Paperbacks. Nelson, Melbourne, Australia. Säynäjoki, A., Heinonen, J., Junnila, S., 2012. A scenario analysis of the life cycle greenhouse gas emissions of a new residential area. Environmental Research Letters 7, 034037. Santin, O.G., Itard, L., Visscher, H., 2009. The effect of occupancy and building characteristics on energy use for water heating in Dutch residential stock. Energy and Buildings 41, 1223–1232. Schneider, A., Woodcock, C., 2008. Compact, dispersed, fragmented, extensive? A comparison of urban growth in twenty-five global cities using remotely sensed data, pattern metrics and census information. Urban Studies 45 (3), 659–692. Shammin, M.R., Herendeen, R.A., Hanson, M.J., Wilson, E.J.H., 2010. A multivariate analysis of the energy intensity of sprawl versus compact living in the U.S. for 2003. Ecological Economics 69, 2363–2373. Sorrell, S., Miller, R., Bentley, R., Speirs, J., 2010a. Oil futures: a comparison of global supply forecasts. Energy Policy 38, 4990–5003. Sorrell, S., Speirs, J., Bentley, R., Brandt, A., Miller, R., 2010b. Global oil depletion: a review of the evidence. Energy Policy 38, 5290–5295. Spearritt, P., 2009. The 200 km city: Brisbane, the gold coast and the sunshine coast. Australian Economic History Review 49, 87–106. Steinberger, J.K., Krausmann, F., Eisenmenger, N., 2010. Global patterns of materials use: a socioeconomic and geophysical analysis. Ecological Economics 69, 1148–1158. Suh, S., 2010. Handbook on Input–Output Economics for Industrial Ecology. Springer, Dordrecht; London. Tukker, A., Jansen, B., 2006. Environmental impacts of products: a detailed review of studies. Journal of Industrial Ecology 10, 159–182. UN DESA Department of Economic and Social Affairs, 2007. 2005 Demographic Yearbook, Fifty-sixth issue. United Nations Publication, New York. Verbeek, M., 2008. A Guide to Modern Econometrics. John Wiley & Sons, Chichester, England; Hoboken, NJ. Vringer, K., Blok, K., 1995. The direct and indirect energy requirements of households in the Netherlands. Energy Policy 23, 893–910. Wang, X., Chen, D., Ren, Z., 2010. Assessment of climate change impact on residential building heating and cooling energy requirement in Australia. Building and Environment 45, 1663–1682. Wang, X., Chen, D., Ren, Z., 2011. Global warming and its implication to emission reduction strategies for residential buildings. Building and Environment 46, 871–883.

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i

12

D. Wiedenhofer et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Weber, C., Perrels, A., 2000. Modelling lifestyle effects on energy demand and related emissions. Energy Policy 28, 549–566. Weisz, H., Steinberger, J.K., 2010. Reducing energy and material flows in cities. Current Opinion in Environmental Sustainability 2, 185–192. Wiedenhofer, D., Lenzen, M., Steinberger, J.K., 2011. Spatial and socio economic drivers of direct and indirect household energy consumption in Australia. In: Urban Consumption. CSIRO Publishing, Collingwood, Vic.

Wier, M., Lenzen, M., Munksgaard, J., Smed, S., 2001. Effects of household consumption patterns on CO2 requirements. Economic Systems Research 13, 259–274. Wood, R., Lenzen, M., Foran, B., 2009. A material history of Australia. Journal of Industrial Ecology 13, 847–862.

Please cite this article as: Wiedenhofer, D., et al., Energy requirements of consumption: Urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.07.035i