Small is green? Urban form and sustainable consumption in selected OECD metropolitan areas

Land Use Policy 54 (2016) 212–220

Contents lists available at ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

Small is green? Urban form and sustainable consumption in selected OECD metropolitan areas Alex Y. Lo ∗ University of Hong Kong, Hong Kong, China

a r t i c l e

i n f o

Article history: Received 23 June 2015 Received in revised form 15 January 2016 Accepted 11 February 2016 Keywords: Urban form Urban density Sustainable consumption Compact city Environmental behaviour Greenspace

a b s t r a c t This article examines the relationship between urban form and sustainable consumption practices across selected OECD metropolitan areas. Using resources available from the OECD database and the International Social Survey Programme, the research identifies factors associated with the stated frequency of making efforts to adopt a sustainable lifestyle across the selected areas. Results confirm that everyday sustainability practice is a function of personal factors, i.e. the socio-economic characteristics of the individuals and environmental concern, whereas the impacts of contextual factors related to urban form are modest. Residents living in compact developments are not more likely to drive less or reduce energy consumption as a means to mitigate their environmental impacts. The article discusses the implications for spatial planning and policy-making in terms of the primary effects of the built environment on sustainable consumption and the tension between urban density and greenspace provision. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Urban form, defined as the spatial configuration of fixed elements within a metropolitan area, has significant implications for enhancing environmental sustainability (Anderson et al., 1996). In a review of these implications, Jabareen (2006) proposes a typology of sustainable urban form involving multiple spatial dimensions, such as density, compactness, and mixed land use. Concentration of urban dwellers and their activities in the built-up area is widely regarded as a key factor determining cities’ sustainability performance (Jenks et al., 1996; Williams et al., 2000). Compact forms of urban development, however, create both opportunities and challenges for achieving environmental sustainability (Jabareen, 2006; Ravetz, 2000; Satterthwaite, 1997). The sustainability implications of low urban densities are manifold and complex and, in some circumstances, come into conflict with each other, undermining the case for increasing urban concentration. Some authors, such as Williams and Dair (2007), have called for more attention on the primary effects of the built environment on people’s behaviour. Apart from assessing physical or ecological outcomes, there is a need to explore the behavioural dimensions

∗ Correspondence address: T.T. Tsui Building, University of Hong Kong, Pokfulam Road, Hong Kong, China. E-mail address: [email protected] http://dx.doi.org/10.1016/j.landusepol.2016.02.014 0264-8377/© 2016 Elsevier Ltd. All rights reserved.

of sustainability, which refer to the sustainable actions of those individuals living, working and enjoying their leisure time within a particular spatial environment (Williams and Dair, 2007; Williams et al., 2010). Similarly, Holden’s (2004, p. 91) study is premised upon the assumption that physical surroundings influence human behaviour. The agency of individuals is given greater emphasis in these analyses. Following Holden (2004) and Williams and Dair (2007), this article addresses the question about urban form and people’s sustainable consumption practices, with an aim to examine the systematic relationship between them. It makes use of an opensource international dataset and involves an exploratory study of the cross-national variations among 24 metropolitan areas in the member countries of the Organisation for Economic Co-operation and Development (OECD). Spatial attributes, such as population density and polycentricity, are used in a multi-level regression analysis to predict the stated frequency of everyday sustainability practices. This knowledge can help understand the effects of the built environment on sustainable consumption and are valuable for specific spatial planning projects that seek to achieve environmental sustainability at the primary level, i.e. households and communities. It is one of the few studies that introduce city-level spatial attributes in the analysis of sustainable lifestyles and adopt a wider range of behavioural indicators. The next section offers a brief review of studies on compact developments and sustainable consumption. This is followed by

A.Y. Lo / Land Use Policy 54 (2016) 212–220

a description of data sources and research methods. Statistical results are then presented. Conclusions provide a discussion of these results and their policy implications.

2. Sustainable consumption in compact developments Reducing fuel consumption is an important component of sustainable lifestyle. Scholarly and practitioners’ debates are often revolved around the question that whether high-density developments can raise prospects for energy conservation. For example, there is widespread belief that the frequencies of automobile use and travel decline as urban density increases. Evidence has demonstrated that compact urban form encourages the use of sustainable modes for commuting. A US research sponsored by the Federal Transit Administration has indicated that residents of higher-density residential areas are more likely than those in lower-density areas to commute by transit, walking, bicycling, or their combinations, and less likely to drive (Transportation Research Board of the National Academy, 1996). Holden and Norland (2005) have also shown that, in Oslo, longer distance between home and city centre increases the amount of energy used for everyday travel. In the Netherlands, higher residential densities and proximity to employment centres have reduced the frequency of solo driving (Buliung and Kanaroglou, 2006; Schwanen et al., 2004). These studies corroborate the conclusion of an earlier report that higher urban population densities are negatively associated with gasoline consumption (Kenworthy and Newman, 1990), which is a main source of urban air pollution and climate change. Furthermore, firm and household concentration in a compact city has wider implications for reducing energy use in other forms of energy-consuming activities. High-density and vertical development can contribute to environmental sustainability by promoting housing forms with lower energy consumption, such as multistorey apartment blocks, semi-detached houses, and multi-family housing. These residential spaces are likely to have lower energy requirements for space conditioning and lighting that those in a more dispersed city, and their clustered pattern are likely to be more efficiently served by district heating (Anderson et al., 1996; Holden and Norland, 2005; Høyer and Holden, 2003). The structure of the built environment also determines the capacity for collecting polar energy. Jabareen (2006, p. 42) has suggested that “potentially more solar radiation can be collected on a built urban site than on a flat, open terrain”, because “built urban sites have larger areas of exposed surfaces per unit area of ground cover”. This implies that high-density urban areas can create a range of sustainable energy options for urban planners. However, the view that compact urban form is environmentally sustainable is far from a consensus. Higher densities may compromise the liveability of city (Neuman, 2005). The inverse relationship between liveability and sustainability implies that the demand for better quality of urban life among the denser population can undermine or offset the environmental benefits of the compact city. Gordon and Richardson (1997) challenge the argument for compactness and doubt that high-density development has contributed to reduction in vehicle miles travelled. Leisure-time travel is a function of lifestyle that is subject to spatial constrains in the city. According to Holden and Norland (2005), for example, access to a private garden is related to lower energy consumption for leisure-time travel by car and plane. Since areas of higher-density housing typically offer limited supply of private gardens, or at a lower quality, residents who have to share public gardens and can afford longer trips might tend to seek cross-border recreational options. Also, higher-density development leaves little space for urban greening (Jim, 2004; Lin and

213

Yang, 2006; Lo and Jim, 2012, 2015a,b), which is important for improving urban air quality and biodiversity (Beer et al., 2003; Swanwick et al., 2003). Finally, mixed land use in these areas places residential sites in close proximity to sources of pollution or environmental nuisances, such as noise from vehicles and refuse and hazardous materials on the roads, which have been a major public concern in Asian cities (Rabianski et al., 2009; Zhang, 2000). Complexities in the ways in which urban form affects sustainable consumption have contributed to these pieces of conflicting evidence. Neuman (2005) has argued that the key issue to address is whether the processes of building cities and the processes of living, consuming, and producing in cities are sustainable, rather than the form of the city per se (Neuman, 2005, p. 22). People’s lifestyles are therefore one critical issue for scientific investigation. Yet, Anderson et al. (1996) contend that attempts to assess the relationship between urban form and environmental performance encounter difficulties in accounting for the differences in individual behaviour or preference. Williams and Dair (2007) further elaborate on this argument and stress that cities and neighbourhoods not only need to support technical sustainability, but also behavioural sustainability, which involves behaviours by individuals or groups that reduce resource consumption and pollution. A related argument has been raised by Scoffham and Vale (1996, p. 70): “it is not the physical environment that is the bar to using less resources and causing less environmental pollution”, but an attitude toward how many resources are needed to sustain a decent urban living with “access to all possible information and activities”. The behaviour or attitude discourse focuses on the agency of individuals in responding to spatial constraints and enablers on sustainable consumption practices. This perspective is particularly important for understanding the tensions between the different implications of high-density development for sustainability, such as reducing driving to work but increasing demand for leisure-time travel. While the conditions for sustainable consumption are spatially bound, the extent to which these spatial elements determine the actual or intended practice of people remains unclear. There has been a large body of knowledge about the causal linkage between urban form and environmental impacts, but less on people’s decision or action per se. As Williams and Dair (2007, p. 162, original emphasis) put it, “little attention has focused on the validity of arguments relating the physical environment to behavioural sustainability”. This drives home the analytical focus on people’s lifestyles, which are manifest as the primary effects of spatial conditions, instead of the consequences of their lifestyles. Williams and Dair (2007) have provided a list of sustainable behaviours and their spatial or physical determinants. Apart from energy or fuel consumption, the framework they proposed includes water conservation, waste recycling, supporting ecologically important local habitats, and using local services that contribute to environmental or social sustainability, etc. These practices are subject to a wide range of spatial or physical constraints, such as proximity to recycling facilities, access to greenspace, and the diversity of service providers or food producers (e.g. community farms). Small and high-density residential areas offer varying opportunities for these practices. For example, while centralized recycling arrangements are likely to be more cost-effective, the availability of public greenspace is often limited in space-poor areas. In contrast, lower urban densities make land available for creating private or public outdoor space and provide conditions for public contribution to the protection of local flora and fauna, but people take longer distance to travel on a daily basis. Sound knowledge about the relationship between the built environment and sustainable behaviour is crucial for understanding the environmental implications of compact urban form. The views

214

A.Y. Lo / Land Use Policy 54 (2016) 212–220

of Scoffham and Vale (1996), Williams and Dair (2007), and others can be boiled down to a general research question: does city dwellers’ engagement in sustainable consumption vary across different urban forms? 3. Methods 3.1. Data The main hypothesis of this study is that the sustainable behaviours of individuals are a function of the spatial attributes of the metropolitan area where they live in. Measures of behaviours are available from the openly available dataset offered by the International Social Survey Programme (ISSP) (ISSP Research Group, 2012). The ISSP is a continuing annual programme of cross-national collaboration on surveys covering topics important for social science research. The Environment Module of the ISSP maintains a collection of national surveys conducted during 2009–2011 (mainly 2010) and involved 36 countries worldwide. Participating countries are represented in the program by independent academic research units or commercial research service providers that are appointed to collect individual data in their countries. Most participating countries solicited data through face-to-face interviews, but mixed survey formats (telephone, mail or web surveys) were employed in some cases. Stratified random sampling methods were adopted to obtain demographically representative national samples. These surveys included a battery of questions to record the frequencies of making an effort to adopt a sustainable lifestyle at the household level. These questions pertain to waste recycling, water conservation, reducing driving and energy consumption, buying green products or avoid buying products with adverse environmental consequences (see Section 4 for exact wording). Respondents were requested to state how often they make these adjustments for environmental reasons, with four options ranging from ‘Never’ to ‘Always’. Responses were used as dependent variables in the regression analysis. The entire ISSP data file contains more than 50,000 observations. Since the present research mainly focuses on metropolitan areas, the first level of data selection was based on the respondents’ place of living. The questionnaire provided five response options on place of living: three of them pertained to the city (i.e. a big city, suburbs of a big city, or a small city), whereas the other two referred to ‘country village’ or ‘farm or home in the country’. The results reported here involved the survey responses collected from the city only, which account for 84% of the entire sample. The second level of selection was defined by the availability of relevant data about the study areas included in the ISSP. Only some of the ISSP national surveys, such as UK, Mexico, South Korea, recorded information about the cities or metropolitan areas the respondents currently living in, and not all of these surveys specified every city or metropolitan area in the respective country. Reliable data on urban form are available from the OECD database for about half of these local areas. The OECD database (stats.oecd. org) provides a list of urban form indicators for a number of metropolitan areas in OECD countries, such as London, Mexico City, and Seoul etc. These estimates include population density, size of population, polycentricity, concentration of people in core areas, and per capita area of greenspace. Only those metropolitan areas that can be identified and extracted from the ISSP dataset and are included in the OECD database were selected for the present study. The selection of study areas, therefore, was driven by availability of the data required. These procedures selected 24 study areas, from 14 countries, for analysis (Table 1), with a total sample size of 3418. Although the sample does not demographically represent the

entire OECD population, it includes representatives from different urban forms, such as Seoul (very high density, bi-centric, concentrated, fewer greenspaces, Asian), Gothenburg (relatively lower density, mono-centric, dispersed, more greenspaces, European), Paris (medium-high density, polycentric, concentrated, fewer greenspaces, European), Querétaro (medium density, monocentric, less concentrated, more greenspaces, Latin American), and Santiago (high density, mono-centric, concentrated, fewer greenspaces, Latin American) (relative assessment based on the OECD data). While being far from the full spectrum of urban form, this sample is able to provide an adequate amount of variations for the regression analysis primarily aimed at identifying predictors of the behavioural variables. Table 2 contains the descriptive information for these spatial indicators, which are used as explanatory variables. A few other variables were included in the analysis in order to control for confounding effects: GDP per capita, unemployment rate, CO2 emissions per capita, and estimates on exposure to air pollution. The first two indicate the economic conditions of the metropolitan areas, while the latter two represent environmental quality, which may influence the tendencies to take environmental action and therefore should be taken into consideration. All of these spatial, economic, and environmental indicators were extracted from a single official source (i.e. OECD Statistics) to ensure consistency and reliability, and measured at the time point when the ISSP surveys were conducted, i.e. 2010 (Table 2). The variables for GDP, density, total population, greenspace area were log-transformed to correct skewness. In addition, the analysis included five socio-economic variables available from the ISSP data file: gender, education, age, household income, and environmental concern. Dummy variables were created for gender and education. To make comparison across countries possible, household income was standardized as a z-score and created individually for each country. Environmental concern was introduced to control for individuals’ general environmental attitude, which is likely to be a determinant of environmental action.

3.2. Multilevel regression analysis Using multilevel regression techniques for analyzing crossnational data has become increasingly popular (Franzen and Meyer, 2010; Gelissen, 2007; Marquart-Pyatt, 2012). As Marquart-Pyatt (2012) points out, previous research usually either aggregates individual-level data to country (or city) level, or disaggregates country-level data to the individual-level. Both treatments are methodologically problematic. In aggregating individual-level data, variation at the individual-level is lost, whereas in disaggregating data, the individual cases are correlated, violating the assumption of independence of measures. Either of these treatments can lead to incorrect results. Multilevel regression modelling is a more advisable approach because it can be applied to situations where individuals may be classified by higher level groupings, e.g. within different cities. Multilevel models incorporate individual (e.g. personal attributes) and contextual (e.g. city attributes) factors and examine which part of the variance in a dependent variable can be attributed to these factors, i.e. with-group and between-group variances. The cross-sectional analysis employed random intercept model, in which groups differ with regard to the mean value of the dependent variable so that the random intercept is the only random group effect and individual-level coefficients are not allowed to vary (Gelissen, 2007). Following Marquart-Pyatt (2012), the levelone (individual) model of the multi-level model for sustainable behaviour is specified as:

A.Y. Lo / Land Use Policy 54 (2016) 212–220

215

Table 1 Descriptive statistics for 24 OECD metropolitan areas. Country

Metropolitan area

GDP (US$)

Unemploy (%)

CO2 (t/person)

PM2.5 (␮g/m3 )

Population (persons)

Polycentricity Concentration Density (%) (persons per km2 )

Greenspace (m2 per person)

Austria Belgium Belgium Switzerland Chile Chile Czech Germany Spain France UK S. Korea S. Korea S. Korea Mexico Mexico Mexico Mexico Mexico Sweden Sweden Slovakia Netherlands Netherlands

Vienna Brussels Antwerp Zürich Santiago Valparaiso Prague Berlin Madrid Paris London Seoul Incheon Busan Daegu Mexico City Chihuahua Veracruz Querétaro San Luis Potosí Stockholm Gothenburg Bratislava Utrecht Amsterdam

40,107 45,749 37,327 48,128 16,558 11,220 41,543 29,971 34,735 49,498 46,532 26,243 19,829 17,478 16,061 21,902 14,972 18,912 15,952 48,364 32,344 45,414 42,060 39,596

6.0 11.7 5.1 4.2 8.2 9.7 4.2 12.4 16.2 9.0 8.2 4.4 3.2 3.4 6.9 6.8 2.8 7.2 4.2 7.2 8.5 7.0 4.4 4.4

14.23 9.48 15.13 6.70 3.42 2.16 13.04 12.10 7.14 5.80 7.87 5.74 3.68 5.38 3.42 4.74 1.56 2.11 1.24 6.17 6.56 9.97 6.39 15.06

20.84 23.13 23.02 18.95 16.00 5.62 24.17 20.32 14.49 14.44 19.67 27.10 19.42 17.64 25.76 11.57 17.46 22.98 16.30 7.80 10.01 20.44 19.85 22.57

2,683,251 2,485,480 1,053,725 1,206,312 6,393,831 954,333 1,829,843 4,374,708 6,507,502 11,693,218 11,793,530 22,451,402 3,447,182 2,639,116 19,255,925 830,231 782,301 1,119,642 1,185,716 1,964,829 877,150 715,456 716,648 2,360,958

1 1 1 1 1 1 2 1 2 5 6 2 2 1 1 1 1 1 1 3 1 1 1 4

231.12 297.10 314.53 277.60 18.26 3.60 275.68 207.17 26.36 91.55 36.97 5.94 9.09 33.51 27.95 2.32 558.67 342.61 31.50 117.77 299.30 453.05 234.53 215.21

63.81 44.40 46.38 30.56 94.34 91.56 71.21 79.64 77.34 79.35 79.87 91.28 95.13 94.28 95.46 98.71 88.23 71.62 87.75 74.44 56.26 60.20 43.45 69.25

295.09 761.09 901.19 1025.63 1402.98 720.88 465.72 708.30 564.02 967.23 1704.05 4804.36 4756.27 2370.26 3774.43 93.69 740.36 456.15 226.05 276.47 227.82 275.11 1150.92 837.27

Source: OECD.StatExtracts ‘Metropolitan areas’ dataset (stats.oecd.org), 2010 estimates.

Table 2 Description of model variables. Individual-level variablesa FEMALE DEGREE AGE INCOME CONCERN

1 = Female, 0 = otherwise 1 = University degree completed, 0 = otherwise Continuous Standardized household income (z-transformed) 1 = Not at all concerned, 5 = very concerned in environmental issues (five-point scale)

City-level variablesb GDP UNEMPLOY CO2 PM2.5 POPULATION POLYCENTRICITY CONCENTRATION DENSITY GREENSPACE

GDP per capita of the metropolitan area (2005 constant US$ and constant PPPs) Number of unemployed divided by labour force (%) CO2 emissions in metropolitan areas divided by population (tonnes per person) Estimated population exposure to air pollution PM2.5 (␮g/m3 ) Total population of the metropolitan area (persons) Number of non-contiguous core areas by metro area Share of population living in the core areas over the total metropolitan population (%) Ratio between total population and total land area of the metropolitan area (persons per km2 ) Land in the metropolitan area covered by vegetation, forest and parks divided by the population (m2 per person)

a b

Data extracted from 2010 Environment Module of the International Social Survey Program. Data extracted from OECD.StatExtracts ‘Metropolitan areas’ dataset (stats.oecd.org), 2010 estimates.

Y ij = ˇ0j + ˇ1j FEMALE + ˇ2j DEGREE + ˇ3j AGE + ˇ4j INCOME + ˇ5j CONCERN + r ij where Yij is the behaviour score for person i in city j. Behaviour Yij depends on the individual’s characteristics as denoted by the above equation. Contextual factors are incorporated by varying the intercept ˇ0j which depends on the metropolitan-level variables. Specifications of the level-two (metropolitan) model are: ˇ0j = γ 00 GDP + γ 01 UNEMPLOY + γ 02 CO2 + γ 03 PM2.5 + γ 04 POPULATION + γ 05 POLYCENTRICITY + γ 06 CONCENTRATION + γ 07 DENSITY + γ 08 GREENSPACE + u0j

where i = 1,2,3,. . .Nij (Nij = 3,418) and j = 1,2,3, . . . J(J = 24). Following Heck et al.’s (2010) suggestion, the Restricted Maximum Likelihood is used as the estimation method as it is a better choice with small number of groups. Proportional reduction in the estimated variance components between the one-way ANOVA

(null) model and the estimated model (with predictors), was used to calculate a reduction in variance estimate, or R2 , as follow (for each level): (M1 2 − M2 2 ) M1 2 where  2 M1 refers to the one-way ANOVA level-one or level-two variance components and  2 M2 refers to the variance components of the model with predictors. The proportion of total variance that exists between cities was estimated by calculating the Intraclass Correlation Coefficient (ICC): b 2 b 2 + w 2 where  2 b and  2 w refer to between-group and within-group variance in the dependent variable, respectively.

216

A.Y. Lo / Land Use Policy 54 (2016) 212–220

Table 3 Measures for sustainable consumption. How often do you. . .. . .

N

Recycle/reuse a. Make a special effort to sort glass or tins or plastic or newspapers and so on for recycling b. Choose to save or re-use water for environmental reasons?

Mean

S.D.

Factor loading

Cronbach’s alpha

3310 3378

3.13 2.31

1.01 1.02

0.49 0.72

– –

Save energy Cut back on driving a car for environmental reasons? c Reduce the energy or fuel you use at home for environmental reasons? d.

2465 3360

2.08 2.50

0.94 0.97

0.70 0.77

– –

Buy green products e. Make a special effort to buy fruit and vegetables grown without pesticides or chemicals? f. Avoid buying certain products for environmental reasons?

3210 3355

2.19 2.18

0.95 0.94

0.64 0.76

– –

Composite scale Sustainable consumption (a)–(f) Adjusted sustainable consumption, excluding (c) and (d)

2281 3090

14.66 9.86

3.85 2.74

– –

0.77 0.66

Notes: 1. Statements extracted from the 2010 Environment Module of the International Social Survey Program (ISSP), Question 20. 2. Only urban residents in selected OECD metropolitan areas are included (N = 3418). 3. (a)–(f) were measured by a four-point scale (reverse coded): 1 = Never, 2 = Sometimes, 3 = Often, 4 = Always. 4. All items were coded so that higher scores indicate a pro-environment position. 5. Factor extraction method: principal component analysis with varimax rotation.

4. Results Table 3 provides the descriptive statistics for the dependent variables. These survey items were reverse-coded so that higher scores indicate a pro-environment position. A relatively high score of 3.13 was recorded for waste recycling (a), suggesting that it is a common sustainable practice in the 24 metropolitan areas. All other items fall within the category of ‘sometimes’. In particular, cutting back on driving (c) is less popular; however it should be interpreted with care because the survey question is only relevant to car owners or drivers and this explains the large number of non-responses. Since the need to drive and the costs of driving are central to the debates on development density and substitution of private transport by transit, excluding this item from analysis would be problematic. One solution is to isolate it from the main analysis and put it into a separate regression model, which is presented below. For corroboration purposes, this treatment was also applied to another energy-related item (d), which allowed nondrivers to respond. All of the six items loaded on one factor and yielded a Cronbach’s alpha of 0.77, indicating scale reliability. A composite variable was created by combining them, labelled as ‘sustainable consumption’, with a maximum value of 24. The mean scores for each metropolitan area, ranging from 17.68 (Berlin) to 11.04 (Valparaiso), are displayed in Fig. 1. The figure also shows the adjusted scores for sustainable consumption, created by excluding the energy-related items. The composite scale for the adjusted sustainable consumption managed to achieve a satisfactory alpha of 0.66. The mean scores of the two energy-related items are displayed in Fig. 2. A full descriptive Table is included in Appendix A. The four lifestyle variables described in Figs. 1 and 2 were individually regressed on the abovementioned explanatory variables. As shown in Table 4, sustainable ways of living were closely related to individual-level variables. Gender, age, and environmental concern created significant positive impacts, suggesting that female and older respondents indicating a higher level of environmental concern are more likely than others to adopt a sustainable lifestyle. At the city level, only greenspace area achieved statistical significance. People living in cities with a generous supply of greenspaces more frequently made consumption choices that could contribute to environmental sustainability. The ICC for sustainable consumption indicates that only 4.5% of the total variance can be attributed to between-group differences. Polycentricity, concentration of people in the urban cores, population and population density were

Table 4 Multilevel regression analysis for sustainable consumption. Sustainable consumption (a–f)

Adjusted sustainable consumption, excluding (c) and (d)

Individual-level variables

Coefficient

Std. error

Coefficient

Std. Error

Intercept FEMALE DEGREE AGE

−22.670 0.723*** 0.376 0.038***

13.504 0.169 0.197 0.006

−22.513 0.537*** 0.252* 0.022***

8.992 0.103 0.125 0.003

INCOME CONCERN GDP (log) UNEMPLOY CO2 PM2.5 POPULATION (log) POLYCENTRICITY CONCENTRATION DENSITY (log) GREENSPACE (log)

−0.085 1.087*** 1.177 −5.165 0.102 −0.100 0.944 −0.322 −0.003 0.726 0.497*

0.076 0.085 1.226 8.435 0.075 0.067 0.435 0.241 0.027 0.413 0.199

0.064 0.681*** 1.386 −4.330 0.030 −0.050 0.640 −0.266 0.002 0.619* 0.424**

0.047 0.050 0.840 6.016 0.052 0.046 0.313 0.168 0.019 0.275 0.133

Variance component Individual-level variance City-level variance

11.076 0.523

5.403 0.303

0.173 0.779

0.155 0.793

Reduction in variance Individual level City level

Note: ***p < 0.001; **p < 0.01; *p < 0.05.

not key predictors. The results remain largely unchanged when the energy-related items were removed, except that the contributions of population density became significant. Table 5 shows that the decision to cut back driving for environmental reasons mainly depends on individual-level factors. The stated frequency was higher among those respondents who are older, more concerned about the environment, have a university degree, and lower incomes. The size of the metropolitan population had a marginal positive effect, implying that driving less in order to save energy or reduce pollution is more common among residents of bigger cities than the smaller ones. Gender effects on the frequency of reducing energy or fuel consumption were evident. Age and income continued to play a key role, but education and environmental concern have lost significance. Female, older and lower-income individuals more frequently engaged in energy or fuel conservation. None of the city-level variables demon-

A.Y. Lo / Land Use Policy 54 (2016) 212–220

217

Fig. 1. Sustainable consumption by metropolitan area. Source: 2010 Environment Module of the International Social Survey Program (ISSP). Range: 6–24 (sustainable consumption), 4–16 (adjusted sustainable consumption)

Fig. 2. Sustainable consumption (reduce driving and reduce energy use) by metropolitan area. Source: 2010 Environment Module of the International Social Survey Program (ISSP). Range: 1–4

strated significant impacts. Whether individuals often make efforts on energy or fuel conservation for environmental reasons is not associated with the spatial attributes investigated. 5. Discussion Decades of research have demonstrated that everyday environmental actions are strongly influenced by personal characteristics, such as age and gender (Gilg et al., 2005; Hines et al., 1987). The national context, most often represented by the country’s GDP and indicators of environmental conditions, is also found to be a key determinant (Dunlap and Mertig, 1995; Franzen and Meyer, 2010; Lo and Chow, 2015). Yet, empirical investigations at the city or metropolitan level are rare. Those research that have examined the impacts of spatial attributes, such as population density, on envi-

ronmental sustainability have put emphasis on a particular form of consumption choice or multiple cities in a particular country. For instance, Høyer and Holden (2003) and Holden and Norland (2005) have shown that household consumption is shaped by urban structures, but their primary focus is travel behaviour. Lin and Yang’s (2006) case study includes a number of Taiwanese cities, but has only investigated the relationship between spatial attributes and a limited set of environmental performance indicators, such as air quality and fuel consumption. Missing from the literature are a cross-national inquiry and the use of a suite of broadly defined sustainable consumption measures. The present research addressed the structural differences between metropolitan areas across countries and adopted a broader view about sustainable consumption (Jenks et al., 1996; Satterthwaite, 1997; Williams et al., 2000). Contrary to the usual

218

A.Y. Lo / Land Use Policy 54 (2016) 212–220

Table 5 Multilevel regression analysis for reducing driving and energy consumption. Reduce driving (c)

Reduce energy use (d)

Individual-level variables

Coefficient

Std. Error

Coefficient

Std. Error

Intercept FEMALE DEGREE AGE INCOME CONCERN GDP (log) UNEMPLOY CO2 PM2.5 POPULATION (log) POLYCENTRICITY CONCENTRATION DENSITY (log) GREENSPACE (log)

4.800 0.067 0.126* 0.007*** −0.062** 0.188*** −0.629 −2.626 −0.003 −0.003 0.277* 0.059 −0.013 −0.114 0.103

3.742 0.043 0.050 0.001 0.019 0.021 0.346 2.441 0.021 0.018 0.126 0.068 0.008 0.113 0.055

−2.143 0.092* −0.008 0.005*** −0.051** 0.203 0.431 0.569 −0.008 0.022 −0.098 −0.023 0.008 −0.035 −0.026

3.140 0.039 0.048 0.001 0.018 0.019 0.294 2.127 0.018 0.016 0.110 0.059 0.007 0.096 0.046

Variance component Individual-level variance City-level variance

0.758 0.049

0.831 0.036

Reduction in variance Individual level City level

0.067 0.430

0.088 −0.263a

Note: ***p < 0.001; **p < 0.01; *p < 0.05. a The negative value of R2 should be interpreted with caution. It is statistically possible for this reduction-in-variance type of R2 estimates to produce negative values because the variance components may be less accurately estimated when there are no predictors in the model (i.e. null model). See Heck et al. (2010, p. 86) for a full explanation.

expectation, it offers little evidence for the view that residents in compact cities tend to adopt low-energy ways of living, e.g. driving less or reducing fuel consumption, for environmental reasons. A possible explanation is that individuals who adhere to sustainability principles opt to live in a place that already has lower energy and fuel requirements for urban living (e.g. metropolis with a more extensive public transport network), i.e. a ‘self-selection bias’ (Holden and Norland, 2005), and there is not much room to curb energy consumption any further. Another possibility is that they reduce the use of energy because of financial rather than environmental reasons (with unintended environmental benefits) and therefore found the options provided by the questionnaire, which specify ‘for environmental reasons’, irrelevant. Either case, as Holden and Norland (2005) argue, urban form still matters and contributes to environmental sustainability. The lack of significant association between urban form and energy conservation may be due to the specific framing and wording of the ISSP survey questions. However, some of the spatial attributes investigated are statistically related to sustainable lifestyles. The size of metropolitan population size has modest positive impacts on the tendency to drive less. This is probably because a large number of commuters can significantly reduce the marginal costs of public transport and make it cheaper to use, thus encouraging car owners to switch to the more sustainable commuting option. Higher population densities increase the frequency of practicing other forms of sustainable consumption, such as recycling waste and buying green products. Similarly, this may be attributed to the economies of scale: more options for sustainable consumption are available in densely populated areas, as the marginal costs (or benefits) of setting up sustainable infrastructure or businesses, such as recycling facilities (e.g. recycling bins) and selling organic food in the markets, are likely to be lower, whereas such opportunities may be limited or costlier in smaller towns with fewer people. A possible role of the compact urban form is to give residents greater access to sustain-

able consumption options that would otherwise come at higher costs or require greater efforts to reach. It is difficult to determine causality in examining the relationship between the built environment and people’s behaviour (Williams and Dair, 2007). The present study shows a positive linkage between sustainable practice and per capita area of greenspaces, but causality remains unclear. Individuals and households who have already adopted a sustainable lifestyle may choose to live in a place with better environmental quality, such as having more greenspaces in the city and closer to natural areas. Yet, the quality of the living environment can also influence lifestyle choice; a generous supply of greenspaces can potentially strengthen the motivation of people to appreciate the idea of environmental sustainability by offering a sense of nature and enhancing life satisfaction (Lo and Jim, 2010). The possible causal linkage between the availability of greenspace and the choice of sustainable lifestyles requires further evidence to validate. The findings open up some directions for future research. While other authors have shown that household consumption of energy is related to urban form (Holden and Norland, 2005), the present study offers no such evidence. Holden and Norland’s (2005) research is based on actual energy consumption reported by respondents, rather than stated frequency of reducing energy use. Energy bills provide a measure of the outcome of households’ consumption practices during a particular period of time, regardless of their attitude or preference toward environmental sustainability, and are therefore a passive indicator of sustainable behaviour. In contrast, the present study makes use of an active indicator that assumes a conscious attempt, i.e. a pair of survey questions that probed the frequency of saving energy for environmental reasons. Although the environmental outcomes of the households’ consumption decision are independent on their intention, the different observations seem to suggest that the compact urban form has a passive effect driven by physical-structural conditions, rather than an active one driven by awareness. Questions then arise as to how much consideration should be given to the public’s explicit support to the intrinsic sustainability arguments for increasing the city’s compactness, if all that matter is the outcomes. Moreover, the observation that greenspace and sustainable consumption are inter-related, regardless of causality, reinforces the tension between increasing densities and greening within the builtup area. The debate on the optimal level of population density and its threats to environmental conditions is contentious (Williams, 2009). High-density developments in the city often result in fewer greenspaces (Lin and Yang, 2006; Jim, 2004; Jabareen, 2006), although it depends on how density is defined (Anderson et al., 1996). Few greenspaces, as the present study has shown, are associated with lower tendencies for adopting a sustainable lifestyle. However, this contradicts another observation that these tendencies increase with population density. This leaves open the question as to how increasing densities and greening come into play, given that both of these two factors positively contribute to sustainable lifestyle. The tension between urban density and greenspace provision may be over-stated to the extent in which high density is defined in terms of the concentration of people within the developed districts of the city (provided that ample open spaces exist between districts) (Anderson et al., 1996). This definition issue is related to the ability of local urban planners and designers to influence decisions about development density. Nonetheless, the concentration factor did not indicate significance in the regression models presented in the last section, leaving the alternative explanation unproven. It is then worthwhile to explore another aspect of the compact urban form, i.e. the concentration of greenspaces at the peripheries of developed districts, which might help us understand how sustainable lifestyle is related to high population densities and

A.Y. Lo / Land Use Policy 54 (2016) 212–220

adequate supply of greenspace. The agency and influence of local stakeholders, planning professionals and residents in defining and re-defining the spatial norms by which greenspaces are inserted and distributed also warrant further work.

Mexico Czech Chile Mexico Mexico Chile

Veracruz Prague Santiago San Luis Potosí Chihuahua Valparaiso

Total

6. Conclusions This study examined the key factors that explain the variations in sustainable behaviours across 24 selected OECD cities or metropolitan areas, mostly from Europe. The regression analysis focused on attributes of urban environment and controlled for confounding factors that might account for these variations, such as income, environmental concern, GDP, and air quality. In general, results show that everyday sustainability practice is a function of personal factors, i.e. the socio-economic traits of the individuals and the level of environmental concern, whereas the effects of contextual factors related to urban form are not clear. In particular, residents in compact developments are not more likely to drive less or reduce energy consumption in order to mitigate their environmental impacts. The claim that there is no clear linkage between the built environment and sustainable consumption should be qualified, as sustainable consumption is measured in terms of self-reported, conscious environmental behaviours (Williams and Dair, 2007). An implication for spatial planning and policy-making is that compact urban developments might not lead to a conscious attempt by individuals to change everyday household practice toward sustainability, or those individuals who live in a compact urban environment are not more likely than others to voluntarily adopt a sustainable lifestyle. Any behavioural change driven by a sustainable urban design is likely to be passive, indirect and societal, rather than active, direct and individual. In addition, while the tendencies for reducing energy consumption are not found to be related to urban form, other forms of sustainable consumption are linked to population density and the area of greenspace in the city. Sustainable consumption practices are more common in high-density developments with large areas of greenspace. An implication, or challenge, for planning and policy-making is that high densities increase competition between different land uses and often result in few spaces for ground-level greening. Nonetheless the tension could be alleviated by increasing the ‘loading’ of human activity onto developed areas separated by ample greenspaces (Anderson et al., 1996). Further clarifications are needed to understand how these factors play out in practice. Appendix A. Sustainable consumption by metropolitan area Country

Metropolitan area

Sustainable consumption (a–f)ˆ

Adjusted sustainable consumption, excluding (c) and (d)#

Reduce driving Reduce energy (c) use (d)

N Germany France Mexico Switzerland Belgium UK Austria Mexico Belgium S. Korea

Berlin1 Paris Querétaro Zürich Brussels London Vienna Mexico City Antwerp Daegu

25 60 93 73 27 70 141 187 94 56

Mean

N

Mean

N

Mean

17.68 16.78 16.65 16.60 16.59 15.93 15.86 15.61 15.55 14.91

44 76 103 101 30 85 203 238 117 68

11.07 11.39 11.17 11.42 11.47 10.91 10.85 10.47 10.74 10.50

27 66 99 77 27 71 146 209 103 57

2.67 2.53 2.65 2.42 2.37 2.23 2.37 2.46 2.12 1.86

44 78 105 99 30 87 200 263 128 69

2.77 2.64 2.79 2.54 2.77 2.51 2.32 2.49 2.76 2.54

S. Korea S. Korea Netherlands Netherlands Sweden Sweden Slovakia Spain

Seoul Incheon2 Busan Amsterdam Utrecht Stockholm Gothenburg Bratislava Madrid

323 87 70 50 176 87 68 151

14.62 14.49 14.47 14.40 14.30 14.25 14.22 13.99

367 108 85 59 216 106 88 208

10.27 10.15 9.78 9.75 9.76 9.85 9.58 9.81

326 89 78 51 182 92 85 164

1.96 1.99 2.19 2.14 1.97 1.97 1.76 1.62

371 110 96 60 224 113 111 223

2.41 2.60 2.59 2.53 2.47 2.48 2.54 2.56

N

Mean

219 53 121 192 23 28 26

13.38 13.21 12.80 11.96 11.89 11.04

111 159 364 39 45 70

9.32 9.19 8.05 7.92 7.38 7.00

67 125 222 27 39 36

2.12 1.70 2.10 1.85 1.69 1.58

123 167 459 45 59 96

2.46 2.29 2.44 2.44 3.14 2.07

2281

14.66

3090

9.86

2465

2.08

3360

2.50

Source: 2010 Environment Module of the International Social Survey Program (ISSP), Question 20. ˆ six items, range 6–24, ˛ = 0.77; # four items, range 4–16, ˛ = 0.66. Ranked by sustainable consumption (a–f). Notes: 1. Created by merging the ISSP entries for Berlin East and Berlin West. 2. Created by merging the ISSP entries for Seoul and Incheon.

References Anderson, W.P., Kanaroglou, P.S., Miller, E.J., 1996. Urban form, energy and the environment: a review of issues, evidence and policy. Urban Stud. 33, 7–35. Beer, A.R., Delshammar, T., Schildwacht, P., 2003. A changing understanding of the role of greenspace in high-density housing. Built Environ. 29, 132–143. Buliung, R.N., Kanaroglou, P.S., 2006. Urban form and household activity-travel behavior. Growth Change 37, 172–199. Dunlap, R.E., Mertig, A.G., 1995. Global concern for the environment: is affluence a prerequisite? J. Soc. Issues 51, 121–137. Franzen, A., Meyer, R., 2010. Environmental attitudes in cross-national perspective: a multilevel analysis of the ISSP 1993 and 2000. Eur. Soc. Rev. 26, 219–234. Gelissen, J., 2007. Explaining popular support for environmental protection: a multilevel analysis of 50 nations. Environ. Behav. 39, 392–415. Gilg, A., Barr, S., Ford, N., 2005. Green consumption or sustainable lifestyles? Identifying the sustainable consumer. Futures 37, 481–504. Gordon, P., Richardson, H.W., 1997. Are compact cities a desirable planning goal? J. Am. Plan. Assoc. 63, 95–106. Heck, R.H., Thomas, S.L., Tabata, L.N., 2010. Multilevel and longitudinal modeling with IBM SPSS. Routledge, New York. Hines, J.M., Hungerford, H.R., Tomera, A.N., 1987. Analysis and synthesis of research on responsible environmental behavior: a meta-analysis. J. Environ. Educ. 18, 1–8. Holden, E., 2004. Ecological footprints and sustainable urban form. J. Hous. Built Environ. 19, 91–109. Holden, E., Norland, I.T., 2005. Three challenges for the compact city as a sustainable urban form: household consumption of energy and transport in eight residential areas in the Greater Oslo Region. Urban Stud. 42, 2145–2166. Høyer, K., Holden, E., 2003. Household consumption and ecological footprints in Norway—does urban form matter? J. Consum. Policy 26, 327–349. I.S.S.P. Research Group, 2012. Environment III—ISSP 2010 (ZA5500 Data file Version 2.0.0 10.4232/1.11418), in: GESIS Data Archive (Ed.), Cologne 2012. Jabareen, Y.R., 2006. Sustainable urban forms: their typologies, models, and concepts. J. Plan. Educ. Res. 26, 38–52. Jenks, M., Burton, E., Williams, K., 1996. The Compact City: A Sustainable Urban Form? E & FN Spon, London, pp. 350. Jim, C.Y., 2004. Green-space preservation and allocation for sustainable greening of compact cities. Cities 21, 311–320. Kenworthy, J.R., Newman, P.W., 1990. Cities and transport energy: lessons from a global survey. Ekistics 34, 258–268. Lin, J.J., Yang, A.T., 2006. Does the compact-city paradigm foster sustainability? An empirical study in Taiwan. Environ. Plan. B: Plan. Des. 33, 365–380. Lo, A.Y., Chow, A.Y., 2015. The relationship between climate change concern and national wealth. Clim. Change 131 (2), 335–348. Lo, A.Y., Jim, C.Y., 2010. Differential community effects on perception and use of urban greenspaces. Cities 27, 430–442. Lo, A.Y., Jim, C.Y., 2012. Citizen attitude and expectation towards greenspace provision in compact urban milieu. Land Use Policy 29, 577–586. Lo, A.Y., Jim, C.Y., 2015a. Community attachment and resident attitude toward old masonry walls and associated trees in urban Hong Kong. Cities 42, 130–141. Lo, A.Y., Jim, C.Y., 2015b. Protest response and willingness to pay for culturally significant urban trees: Implications for Contingent Valuation Method. Ecol. Econ. 114, 58–66. Marquart-Pyatt, S.T., 2012. Contextual influences on environmental concerns cross-nationally: a multilevel investigation. Soc. Sci. Res. 41, 1085–1099. Neuman, M., 2005. The compact city fallacy. J. Plan. Educ. Res. 25, 11–26. Rabianski, J., Gibler, K., Tidwell, O.A., Clement III, J.S., 2009. Mixed-use development: a call for research. J. Real Estate Lit. 17, 205–230. Ravetz, J., 2000. Urban form and the sustainability of urban systems: theory and practice in a northern conurbation. In: Williams, K., Burton, E., Jenks, M. (Eds.), Achieving Sustainable Urban Form. E & FN Spon, London, pp. 215–228. Satterthwaite, D., 1997. Sustainable cities or cities that contribute to sustainable development? Urban Stud. 34, 1667–1691. Schwanen, T., Dieleman, F.M., Dijst, M., 2004. The impact of metropolitan structure on commute behavior in the Netherlands: a multilevel approach. Growth Change 35, 304–333. Scoffham, E., Vale, B., 1996. How compact is sustainable—How sustainable is compact? In: Jenks, M., Burton, E., Williams, K. (Eds.), The Compact City: A Sustainable Urban Form? E & FN Spon, London, p. 66.

220

A.Y. Lo / Land Use Policy 54 (2016) 212–220

Swanwick, C., Dunnett, N., Woolley, H., 2003. Nature, role and values of green space in towns and cities: an overview. Built Environ. 29, 94–106. Transportation Research Board of the National Academy, 1996. National Academy Press, Washington, DC. Williams, K., 2009. Space per person in the UK: a review of densities, trends, experiences and optimum levels. Land Use Policy 26 (Suppl. 1), S83–S92. Williams, K., Burton, E., Jenks, M., 2000. Achieving Sustainable Urban Form. E & FN Spon, London, pp. 388.

Williams, K., Dair, C., 2007. A framework of sustainable behaviours that can be enabled through the design of neighbourhood-scale developments. Sustain. Dev. 15, 160–173. Williams, K., Dair, C., Lindsay, M., 2010. Neighbourhood design and sustainable lifestyles. In: Jenks, M., Jones, C. (Eds.), Dimensions of the Sustainable City. Springer, Netherlands, pp. 183–214. Zhang, X.Q., 2000. High-rise and high-density compact urban form: the development of Hong Kong. In: Jenks, M., Burgess, R. (Eds.), Compact Cities: Sustainable Urban Forms for Developing Countries. Spon Press, New York.