Understanding household food metabolism: relating micro-level material flow analysis to consumption practices

Understanding household food metabolism: relating micro-level material flow analysis to consumption practices

Journal of Cleaner Production 125 (2016) 44e55 Contents lists available at ScienceDirect Journal of Cleaner Production journal homepage: www.elsevie...

2MB Sizes 0 Downloads 36 Views

Journal of Cleaner Production 125 (2016) 44e55

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Understanding household food metabolism: relating micro-level material flow analysis to consumption practices Loïc Leray*, Marlyne Sahakian, Suren Erkman Faculty of Geosciences and Environment, Institute of Earth Surface Dynamics (IESD), University of Lausanne, CH-1015 Lausanne, Switzerland

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 July 2015 Received in revised form 22 March 2016 Accepted 23 March 2016 Available online 7 April 2016

Public and private food consumption is responsible for significant environmental impacts, resulting in numerous studies that highlight the problem and reveal its magnitude at global and national scales. Drawing on a high level of data aggregation and focussing on individual choices and attitudes, current accounts stop short of grappling with the underlying complexity of the phenomenon. In this paper, we explore the conceptual value and methodological feasibility of linking Material Flow Analysis (MFA) and Social Practice Theory (SPT) to apprehend household food consumption dynamics. We develop and pilot a “Practice-extended MFA” framework among selected households in Bangalore, India. While MFA modelling serves to describe and quantify all food consumption processes and related flows at the microlevel, SPT is applied to investigate how individual, technological and sociological aspects of consumption practices converge towards household food “metabolic profiles”. The results revealed a complex system of interactions between food provisioning, storage and management practices, as well as socio-cultural norms. The paper concludes by emphasizing the contribution of a reflective stance between household metabolisms and consumption practices revealing not only what and how much food is consumed and wasted, but why and in what way. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Household consumption Material flow analysis Food Social practices Metabolism Conceptual framework

1. Introduction The need to reduce energy and material throughput associated with human activities is widely recognized. Food consumption is critical in this respect: 20e30% of all private and public consumption life cycle impacts are related to food (Tukker et al., 2006). Globally, an estimated 30% of edible food was wasted in 2010 (Parfitt et al., 2010; Gustavsson et al., 2011), leading to unnecessary resource depletion and GHG emissions (Kummu et al., 2012). Although all actors in the food system should be involved in tackling this issue (Gustavsson et al., 2011; Papargyropoulou et al., 2014), households play a significant role by staging the majority of transformation processes of food into waste (Timmer et al., 2009). Developing an inter-disciplinary framework to understand household-level food consumption patterns and dynamics would be an important first step towards informing and designing effective policy intervention regarding food waste. Numerous studies

* Corresponding author. Faculty of Geosciences, Institute of Earth Surface Dynamics (IESD), University of Lausanne, Geopolis Building, UNIL-Mouline, CH-1015 Lausanne, Switzerland. Tel.: þ41 216923562. E-mail address: [email protected] (L. Leray). http://dx.doi.org/10.1016/j.jclepro.2016.03.055 0959-6526/© 2016 Elsevier Ltd. All rights reserved.

have been conducted to quantify and qualify household food consumption and related waste. Some rely on secondary datasets such as EUROSTAT or national statistics adapted to a specific country. In EU-27, Monier et al. (2011) showed that 25% of food purchased by European households is discarded. The situation is similar in Switzerland with households responsible for 45% of losses within the whole food supply chain (Beretta et al., 2013). A second type of quantitative study draws on primary datasets collected at the national level using different tools and methods, such as waste composition analysis (Holding, 2010; Quested et al., 2013a) or food diaries and interviews (Lyndhurst et al., 2007) combined with n et al., 2006). These findings have been statistical methods (Sibria complemented by research focussing on the causes of this phenomenon, linking the amount of food waste to socio-demographic variables (Friedl et al., 2006; Koivupuro et al., 2012; Quested and Luzecka, 2014); evaluating or quantifying food waste in relation to individual beliefs, attitude and behaviours (Lyndhurst, 2011; Graham-Rowe et al., 2014; Parizeau et al., 2015); focussing on specific elements of food products such as packaging (Williams et al., 2012); or the question of food provisioning linked to mobility (Sonesson et al., 2005). While these results are of critical importance to highlight the problem and reveal its magnitude, they

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

do not take into account the underlying complexity of the phenomenon (Evans, 2012, 2014; Quested et al., 2013b). On the one hand, the high level of data aggregation on which these studies draw does not provide specific insights on food consumption and wastage patterns; on the other hand, focussing on individual choices and attitudes fails to address the diversity of household food consumption practices formed by overlapping behavioural, technological and institutional interactions among Food Supply Chains (FSCs) stakeholders and consumers within specific cultural contexts (Evans et al., 2013). The aim of this paper is to contribute to addressing this gap by exploring the value of an explicit connection between Material Flow Analysis (MFA) and Social Practice Theory (SPT). While MFA can describe and quantify all food consumption processes and related flows at the household level, SPT can be applied to understanding how individual, technological and sociocultural aspects of food practices relate to household food metabolism. We propose to address a methodological question: How can a metabolism study be complemented through a qualitative study of social practices in everyday life? To illustrate this, we present the results of a micro-study that took place among households in the growing mega-city of Bangalore, India. 1.1. Socially “extended” material flow analysis Material Flow Analysis (MFA) is a family of tools aimed at describing and quantifying the metabolism of human activities, or the underlying set of biophysical flows and stocks linking society and natural environment (Ayres, 1994; Fischer-Kowalski, 1998). Formally an MFA is a systematic assessment of flows and stock of material e energy or substance e within a system defined in space and time (Brunner and Rechberger, 2004). Although the family of tools called “MFA” includes a diversity of methods and specific standards (Finnveden and Moberg, 2005), the core objective is to track every stage of the material or substance flows (the sources, transfer, accumulation and fate) through and within a system, based on the mass balance principle (Huang et al., 2012). This implies that MFA can be applied to any organisational level (individuals, households, firms) as well as various spatial and temporal scales. Defining the MFA system and its components ultimately depends on the object of study, what Brunner and Rechberger (2004) refer to as an activity. An activity comprises all relevant flows, stocks and processes of the material and energy necessary to fulfil a particular human need. For example and in the context of this paper, “to nourish” constitutes the activity on which we focus. In relation to existing MFAs focussing on this activity (Beretta et al., 2013), one limitation to apprehending a national food system is that households remain a “black box” of complexity: little is known about what happens to food from the moment it enters the households, to when it leaves it as “waste”. Households are often represented as a single process, transforming food input into waste output. Getting into that black box to understand how food is consumed and eventually wasted at the household level requires increasing the system's level of resolution: to make an analogy with a screen, the more pixels, the sharper and more detailed the image. Towards defining MFA subprocesses and disaggregating household activities, a finer reading requires engaging with the following questions: what type and how much food is being purchased? Where does it come from? How is it stored and eaten? And more specifically, how does this vary among different households and how, in turn, does this relate to the amount and types of food being consumed and wasted? Answering these questions could help identify all steps necessary for fulfilling the activity of food consumption and define the MFA system. However, MFA remains a descriptive tool: MFA reveals what and how much material (and

45

energy) is consumed, but stops short of providing insights on why and in what way food is consumed and wasted, leading to a better understanding of local patterns and environmental impacts (Burger Chakraborty et al., in press). Researchers in the field of industrial ecology and ecological economics recognize that material and energy flows are embedded in a social context (Binder, 2007a,b; Boons and Howard-Grenville, 2009). When it comes to explaining what cause physical flows to “be and become” in particular ways (Green and Foster, 2005: 664), MFA can be usefully completed with sociological analysis (Hoffman, 2003; Lifset, 2008; Schiller, 2009). Studies have been conducted to “socially” extend the MFA framework. Elaborating on a Structural Agent Analysis (SAA), Binder (2007a,b) investigates how social institutions and actor's decisions affect wood flows in a Swiss region; Hobbes et al. (2007) used an Action-in-Context approach to identify actors' actions and motivations driving the unsustainable extraction of timber in a Vietnamese village. These approaches proved to be useful in understanding how actor's decision-making processes affect concrete material flows. However, they rely on behavioural and microeconomic models, built around the assumption of rational choice, an approach that has been criticized to understanding choices in sustainable consumption studies (Cohen and Murphy, 2001; Shove, 2009). 1.2. Overcoming individualistic perspectives in consumption studies In recent years, and building on the seminal work of Bourdieu (1979) and Giddens (1984), researchers have engaged with social practices theories to understand consumption as part of everyday life (Schatzki, 1996; Reckwitz, 2002), and in relation to natural resources and environmental concerns (Randles and Warde, 2006; Shove, 2010; Halkier and Jensen, 2011; Spaargaren, 2011). Rather than relying on overly simplistic visions of consumers as driven by individual need or greed, social practices fix the analytical lens on people and their doings and sayings, in relation to activities that are often routinized and habitual. Indeed, much of our everyday life involves mundane actions, such as turning on lights or running a shower, which consume resources but are very much inconspicuous (Shove and Warde, 1998). Although social practice theories are subject to different interpretations, some authors (Wilhite, 2013; Sahakian and Wilhite, 2014) suggest that social practices are a “block” of routinized actions, (re)produced and ordered in space and time, through the distributed agency across three “elements”: people (their competences, understanding and engagement); the physical world (objects, infrastructures and technologies); and social context (culture, norms and institutions). The concept of “practice” implies that consumption activities are not conceived as the result of isolated individual and rational decision-making processes, but as the emergence, persistence and disappearance of practices formed by interactions among the “elements” of which practices are composed (Shove, 2015). It is the performative dimension of social practices (Schatzki, 1996), enacted in specific moments and places (Shove, 2009), which shapes the quality and intensity of resources used to fulfil an activity. One interesting aspect of social practices is that they can change over time (Spurling and McMeekin, 2015), no doubt when one or more of the elements holding together a practice shift (Sahakian and Wilhite, 2014). Given our focus on methodological considerations in this paper, the social practices approach enhances the ability to detect and inform the elements influencing the practical performances of food provisioning, storing, cooking and eating, rather than individual choices and characteristics regarding shopping, food and tastes (Halkier and Jensen, 2011). Consequently, we are primarily interested in piloting our method and testing its ability to uncover processes shaping food consumption patterns in the home. In the

46

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

following section, we present the methods we used to design and quantify the MFA system, informed by social practices, as well as the overall approach we developed to link sets of data, both qualitative and quantitative. In Section 3, we present the results of this research. Finally, the last section discusses opportunities and limits of the proposed framework. 2. Research methods 2.1. Population and sample This research draws from a pilot survey that took place in the city of Bangalore in April 2014, including one week of sampling among six households whose socioeconomic status was representative of the new Indian urban middle classes.1 We assumed that one week corresponds to the smallest relevant cycle where food consumption practices can be observed, including potential changes between week days and weekend (Williams et al., 2012). Participants were asked to choose a week corresponding to a “normal” situation (without major social events, family visits or other special occasions). To recruit households, emails were sent to acquaintances in Bangalore, with attached leaflets describing the objectives of the survey, and tasks expected from the participants. The heterogeneity of demographics, household composition, housing structures and locations in the city ensured a reasonable diversity. Two households were composed of young adults in their 30s, including a single woman self-employed (HH-2) and three people sharing a flat as roommates (HH-4). The roommates are all graduated and active people working full time. Three households were made up of families: one was composed of a mother and her 20 year-old daughter (HH-5) and the second was composed of three young adults (20se30s) and their parents (HH-1). One family with two young children (HH-6) was composed of 4 people. Both parents were working full time for companies. Finally, one couple was made up of “empty nesters” in their late 50's (HH-3). 2.2. System design and data collection According to Brunner and Rechberger (2004), the steps to be fulfilled to apply an MFA are i) to clearly define the scope and objective; ii) to set the system's spatial and temporal boundaries; iii) to select the relevant flows and processes; iv) to calculate the flows, stocks and to consider uncertainties; and v) to interpret and present the results in an appropriate way. Flows and stocks of food are defined as all edible products e excluding drinks and pet food e entering the spatial boundaries of household and are expressed in kilocalories. A first phase of interviews and observations within households and food provisioning infrastructures allowed us to design the MFA system corresponding to the activity “to nourish” in the specific context of this study. The model presented in Fig. 1 maps the disaggregated flows and processes related to the activity. In order to analyse how each part of the system responds to specific practices, we broke-up the activity into five interlocking practices: food provisioning, storing, consuming, managing leftovers and disposing. The first practice considered is food “provisioning”. Preliminary interviews and observations revealed that respondents access food from three different retailing categories. To represent this diversity, the food provisioning system has been added to the model as a stand-alone process. Local stores include specialized stores and are known as Kirana stores (Sengupta, 2008). While a supermarket

1 Household incomes, housing standards and education degree were used as indicators to determined the affiliation of respondents to the targeted group.

generally refers to local supermarket chain, Kirana stores are typically independent and family-owned. Local supermarkets are distinguished from hypermarkets and international chains, which are relatively new in India and where none of the respondents reported buying food, albeit solely during the one-week sampling. The second category includes all wet markets and pushcart vendors. While the former typically deal with the entire range of fresh food products (Sood and Mishra, 2013), pushcart vendors are usually specialized, selling seasonal fruits and vegetables. Finally, the third category includes all food products home-delivered, whether raw or cooked. The food inputs were disaggregated into three flows (Fp1eFp3) to account for this diversity. The second practice is “storing” (Fs1eFs3) and represents the allocation of food inputs to cupboard, refrigerator and freezer, the three storage options available to all participants and the only processes including stocks. During preliminary phase, we noticed that measuring initial stocks would considerably increase the required effort from participants while providing minimal information, the main objective of MFA being to observe food flowing within households. Initial stocks are therefore not considered in the model. The third practice is “consuming” and includes: all food consumed from stock (Fc2, Fc4); food consumed without being stored (Fc1) or directly after being home-delivered (Fc5). Cooking processes include peeling raw food ingredients, heating and preparing the meals, as well as heating up leftovers. We distinguished “direct” food consumption in two flows (Fc1, Fc5): when food is home-delivered it is either ready to eat (meals and snacks) or raw; when food is purchased from retailing it can be cooked and consumed on the same day. Making this distinction was important to observe how “consuming” and “provisioning” food practices are potentially interacting. The fourth practice is related to leftovers flows (Fl1eFl3). Here again, preliminary interviews pointed us toward a process that would remain hidden otherwise: in Bangalore it is common for households to share their leftovers with domestic help; this is labelled as the “given away” process. However, as the consumption of food per se is occurring out of the system boundaries,2 the process is represented as a stand-alone process in the model. Finally “disposing” is the fifth practice and includes all food that has been disposed either raw or cooked (Fw2) from stocks (Fw1) or directly after eating (Fl2). The model mass balance is ensured by the digestion flow (Fd).

2.2.1. Food diary and consumption measurement The data have been collected at the level of food ingredients, based on self-reporting approach (Fig. 2). All families were provided with a sampling “tool kit”, including three bins, one cooking scale and a food diary. The food diary was divided up into four distinct phases in order to collect disaggregated data related to each flow. Respondents completed the food diary on a daily basis during one week. In phase 1, respondents reported all quantities of food entering the home, the source of food and storage option (including food directly consumed). In phase 2, quantities of food consumed from the stock were reported, including indication regarding preceding actions: when food ingredients were combined to cook a meal, the name of the meal and its composition was reported. The third and fourth phases were intended to measure the amount of i) nonavoidable food waste such as peelings, bones and eggshells and ii) avoidable waste including all edible food discarded during or after storage and consumption phases. Three plastic bins were provided to respondents to distinguish between avoidable wastes discarded from the stock (pre or post consumption as leftovers);

2

All respondent's domestic helpers were leaving outside the household.

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

47

Fig. 1. MFA system for the activity “to nourish” and the set of five related practices.

waste discarded directly after consumption; or waste related to food processing (peelings, egg shell, bones, etc.). All bins were weighted and emptied once a day. If leftovers occurred, respondents were asked to weight them and report the quantity with the corresponding meal name and the associated action (stored back, disposed, given away). Finally, it should be noted that the level of involvement required to fill the diary might result in respondents skipping to report data. Furthermore, measuring food consumption requires engaging all household members in the survey since food can be consumed apart from regular meals and at different times. As a consequence, chances to introduce bias in data collection or to be confronted with incomplete results are not negligible.

2.2.2. Semi-structured interviews Following the sampling week, each respondent was contacted to participate in a semi-structured interview. Building on Wilhite (2013), Sahakian and Wilhite (2014), we designed the interview

guide as to explore how each practices “elements” e individual understanding, competences, engagement; objects, infrastructures and technologies; culture, norms and institutions e are articulated and perceived by respondents. A list of open-ended questions was developed for each of the five consumption practices considered in the model. As the conceptual distinction between food consumption practices and their elements serves an analytical purpose, the interview guide was only used to ensure that all aspects of food consumption practices were covered during discussions. For example, it might sound difficult for respondents to talk about food provisioning practice without simultaneously talking about the quality and the price of the products on different markets, the way food provisioning is embedded in daily life schedule and routines, the family preferences for some meals and his/her cooking competences. In order to identify and structure the practice's elements that emerged throughout the conversation, interviews were recorded, transcribed and coded using NVivo software. Some of the coding also emerged through an analysis of the transcripts.

2.3. Data processing and analysis

Fig. 2. Schematic representation of data collection method.

Data modelling and analysis have been conducted in three steps. First, food diaries data were processed to calculate the flows and stocks of each household. Second, the calculated flows and stocks have been used to build metabolic profiles corresponding to each of the five practices considered in the model. Finally, practiceperformance reported during interviews served as a heuristic framework to iteratively investigate how and why metabolic profiles converge or diverge among households.

48

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

2.3.1. Flow and stock calculations The food diaries data were converted in kilocalories to avoid misinterpretation when assessing food categories contribution to each flow and stock due to the large water contain of certain products. As underlined by Lipinski et al. (2013), a kilogram of wheat flour contains on average 12 percent water and 3643 kcal whereas a kilogram of apples contains on average 81 percent water and 1704 kcal. Therefore, measuring by weight does not adequately represent the energy consumed or wasted by people. Since the conversion of raw data implied several assumptions and data sources, detailed explanations and examples of calculation are provided in supplement information referenced as “SI”. The raw data reported as grams of ingredients have been multiplied by their average kilocalories content (SI 1.1e1.3). Avoidable wastes are either ingredients or meal's leftovers. When ingredient is wasted, its mass is multiplied by its average caloric content. The details provided in diaries about meals composition were used to derive the proportion of each ingredient in leftovers. These proportions were then used to calculate the caloric content of leftovers (SI 1.4). Non-avoidable wastes were converted using empirical data and assumptions on the percentage of mass removed when fruits/ vegetables, eggs, chicken and fish are consumed (SI 1.5). The theoretical mass obtained was crosschecked with empirical measurements of non-avoidable “cooking bin” (Fig. 2, Section 2.2.1) and range respectively within 1.5% overestimate and underestimate. The mass was then converted in kilocalories. Finally, the results have been aggregated in 6 categories, namely fruits and vegetables; milk and dairy; eggs; meat; cereals; and others (including snacks such as biscuits, chips and sweets). 2.3.2. Metabolic profiles To analyse and condense the dataset into a clear message, MFA's results are usually presented and summarized using visual Sankeytype diagrams where the width of a flow is proportional to its value. While this approach is convenient to communicate the results of a single MFA over a region or an industry, we found that it does not fit the need to clearly compare consumption patterns and associated practice performances over multiple MFAs. Sankey diagrams tend to reduce the dynamic aspects of food consumption pattern: all flows and stocks are expressed in absolute values, using the selected system's unit and aggregated over the system's temporal boundaries. In our case, a Sankey diagram would represent absolute values expressed in kilocalories per week, neglecting any variation in diet, i.e. flows and stocks composition. Moreover, comparing absolute values (e.g. kilocalories consumed per household) is not relevant to identifying similar or contrasting consumption patterns over multiple households, in which demographic profiles and absolute level of food consumption vary widely. Finally, analysing flows and stocks as kilocalories per week leads to a “snapshot” effect preventing the consideration of daily variations within and between households, and relating these variations to the associated practices (e.g. food provisioning). In order to circumvent these limitations, we propose a method that term “metabolic profiles”. The metabolic profiles can be defined as a set of histograms, builds on flows and stocks' normalized values presented over different descriptive domains. A descriptive domain is the quantitative or qualitative dimensional spaces within which the dataset is presented and linked to each of the five practices involved in the activity “to nourish”. For example, when looking at food provisioning, one can compare food provisioning patterns using the sum of input flows (Fp1eFp3) to normalize the input values of each household and then calculate the allocation of food inputs among the provisioning processes, i.e. which shares of each food inputs are purchased in each of the retailing system categories and then analyse how this differs or converges from one household to another.

Because the two descriptive domains e food and retailing categories e are both related to the provisioning practice, it is also possible to qualitatively investigate why we observe convergence or divergence. Conversely, the same approach can be applied to analyse the temporal distribution of food inputs e the provisioning intensities and frequencies e by deriving the daily proportion of food purchased by each household from the total purchased during the week. Both are metabolic profiles related to provisioning practice, however they are expressed through different descriptive domains, either product specific or temporally oriented. It is precisely because household metabolisms are analysed and compared using multiple descriptive domains that they can be explicitly related to practice performances. 2.3.3. Linking metabolic profiles and practice-performances through reflexivity We assumed the biophysical flows of the activity “to nourish” as being embedded in and resulting from social practices. The method we developed draws on the distinction between practice-as-entity, a “block” or “pattern” of elements that can be spoken about and readily identified as shared practices (e.g. provisioning and cooking) and practice-as-performance, the actual enactment of the practice and its successive moment of reproduction (Shove, 2012). The practice-as-entity has been used as a conceptual framework to develop the interview guide sections and to classify the data at two levels: i) the type of practice involved and ii) the nature of its elements (individual, material and technological, normative and institutional). This classification allowed us to analyse how each of the five practices' elements converge or diverge amongst respondents, i.e. how practices are performed and why they are performed as they are. In order to detect causal links between the measured metabolic profiles and the observed practiceperformances, we engaged in a process of analytical reflexivity (Mauthner and Doucet, 2003), where each piece of data are iteratively visited and revisited as new insights emerged, leading to progressively refine focus and understanding (Srivastava and Hopwood, 2009). As an example, a first step of analysis leaded us to identify “food freshness” as a normative element involved in multiple practices (provisioning, storing and cooking) and common to all respondents. We then draw several metabolic profiles for provisioning practices (e.g. caloric intensities and temporal frequencies, origins and categories of food), which in turn, revealed both convergence and divergence among respondents. Going back to the qualitative data new elements emerged (e.g. routines and individual engagement) and provided support to sharpening our analysis of why we observed such convergence and divergence in metabolic profiles. This iterative process has been applied at both classification levels and to each practices considered in the model (Fig. 3). Although this stance has been debated as a methodological predicament in social sciences (Breuer, 2003), we argue that reflexivity can contribute to improve the analytical process of interdisciplinary consumption studies, including metabolism studies. It is through this process that causal links between the biophysical and sociological dimensions of food consumption can be elucidated e as differences and similarities in practiceperformances affect the metabolic profiles, while metabolic profiles reflect specific performances. 2.4. Limitations As mentioned above, the proposed framework requires a significant level of engagement from both researchers and respondents. This raises question about the method scalability, especially regarding the number of case studies required to achieve

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

49

Fig. 3. Practice-extended MFA framework linking sociological and biophysical dimensions of household consumption through analytical reflexivity.

results generalization. This question has been debated in social sciences realm and is not specific to our pilot study (Mason, 2010). Combining quantitative and qualitative methodologies necessarily involves a trade off between different epistemic foundations. The conceptual distinction between statistical and analytical generalizations, proposed by Yin (2009), offers a valuable support to this end. Statistical generalization is intended to infer knowledge about a specific population based on data, which confidence is determined regarding the size and internal variations within population and sample. However, analytical generalization relies on a different logic where the previously developed theory is used as a template to confront with the empirical results of the case studies, not being considered as “sampling units” but as multiple “experiments” (Yin, 2009: 31). In that sense, the proposed method should be considered only as the first steps towards a deeper inquiry into the theoretical framework relevant for understanding the dynamic relationship between household food consumption practices and corresponding household metabolisms. 3. Results 3.1. Provisioning The investigation of food provisioning practices revealed a general tendency for food origin and provisioning frequencies to be articulated with a set of normative elements such as food freshness, perceived quality and convenience. Fig. 4 shows the distribution of food inputs using food product categories and food origin as descriptive domains. About 75% of cereals are purchased in supermarkets and Kiranas while the majority of fruits and vegetables (70%) are purchased from wet markets and pushcart vendors, even though supermarkets do also provide a large choice of fruits and vegetables. How can this preference for more traditional forms of retailing be explained in relation to specific products? Going back to data gathered during interviews, all respondents declared purchasing dry products

monthly (lentils, rice, species, etc.) in supermarkets and convenience appeared to play a major role. For example, HH-4 purchased cereals products in a supermarket because they also find other nonfood products that have similar consumption cycles, e.g. cleaning detergent and body care products. Because supermarkets offer all these items under the same roof, they are more attractive and convenient than other forms of retailing. Respondents claim, however, that they would prefer to purchase fruits and vegetables in wet markets, because of higher quality and freshness. For example, HH-6 stated that wet markets offer better quality products and allow customers to “touch” and “feel” the products, while having a large choice of stalls and offers to choose and compare from. By changing the descriptive domains, one can look at the temporal dynamic of food provisioning practices, i.e. the food inputs frequencies and intensities. Provisioning frequencies refers to the number of shopping trips carried out during the week for each household; provisioning intensity expresses the proportion of food purchased by shopping trips over the total food purchased by each household during the week. Fig. 5 shows provisioning frequencies and intensities per household. Two different patterns emerge on this figure. One group (HH-1, HH-3 and HH-4) presents a high shopping frequency but a low intensity. The second group (HH-6, HH-2 and HH-5) exhibits a lower frequency but higher intensity. In a study on consumption habits, Haden and Vittal (2008) found that Indian consumers have one of the highest food shopping frequency in the BRIC countries with an average of 6 shopping trips carried out per week. While their findings are consistent with ours (each household carried out between 5 and 7 shopping trips during the week), they rely on a high level of data aggregation and do not consider potential variation in provisioning practice performances, neither explore why and in what way differences occur. However, routines, norms and daily schedules seem to be important factors in shaping provisioning patterns. Among the group presenting the lowest shopping intensity and highest frequency, HH-1 declares to buy fruits and vegetables at the wet market everyday as part of an

50

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

Fig. 4. Food origins by food categories normalized over the total food purchased by all households (n ¼ 6).

Fig. 5. Shares of food purchased per household and shopping occasions.

early morning walk in neighbourhood and considers that this is an occasion to get the “fresher” products on the market, right after opening. For HH-3, buying fresh products everyday is a requirement for her stepmother's health and also a way to accommodate a specific caste-based diet. The same pattern emerges from HH-4 but for different reasons: the three roommates are sharing the responsibility for food provisioning and found it difficult to plan meals and grocery shopping a week ahead, consequently they tend to buy food on a daily basis according to their needs and schedules. Conversely, the single woman living in HH-2 reported to buy most of her produce once per week in order to save time for other activities. The same logic has been reported by HH-5 and is confirmed by the metabolic profile above. Finally, HH-6 reported only one shopping trip at the wet market during the week, explaining why it exhibits the highest intensity (100%) in the sampling. This particular situation will be further discussed in Section 3.1.2. Finally, patterns of consumption are also productoriented. Fig. 6 shows provisioning frequencies and intensities for milk among each household during the week. This metabolic profile contrasts with the previous findings: homogeneous patterns can be observed for all households, except HH-6 who did not report any purchase in that category during the sampling period. We found that milk provisioning is influenced by

several interacting elements among food consumption practices. Indeed, half of respondents (HH-1, HH-3 and HH-5) are still homemaking curd, ghee and paneer3 and consider that full milk in soft plastic bags e locally called Nandini packets e is of better quality than UHT milk found in supermarkets to prepare this type of food. Moreover, this product is distributed through a well-established door-to-door delivery system, supported by a large network of milkman in the city, accounting for 40% of milk origin among respondents (Fig. 4). Interestingly, even though the remaining share is UHT milk originating from supermarkets, product that could be kept and stored to a larger extends than Nandini packets, respondents are still purchasing it on a daily basis. Indeed, this finding illustrates how provisioning practices elements such as normative perception of milk quality; distribution infrastructures; and culinary traditions are interacting with personal cooking competences and shape provisioning practice. Overall, these results indicate that despite the massive expansion of modern commercial food retailing in India, which grew by 1500% between 2005 and

3 Curd and paneer are words used in Indian subcontinent to designate homemade yogurt and cheese made out of raw milk. Ghee is a clarified butter originated in India prepared by simmering butter and removing the liquid residue.

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

51

Fig. 6. Shares of milk purchased per household and shopping occasions.

2013 (Sood and Mishra, 2013), western style hypermarkets and international chains are still facing strong resistance from well established and sophisticated food markets (Bijapurkar, 2009), largely supported by Indian consumers habits and preferences, themselves embedded in and maintained by the availability of this specific retailing infrastructures. 3.1.1. Storing and direct consumption Fig. 7 presents the storage distribution of the food input by households. Variations among household food storage practices are useful to better understand systemic interactions with provisioning practices and other steps of household food metabolism. This figure shows that HH-1 and HH-4 exhibit the highest shares of direct consumption; at the same time they also have the highest shopping frequencies (Fig. 5). Conversely, HH-2 and HH-6 exhibit the lowest shopping frequencies and did not consume any food directly after purchasing it. Another interesting finding is the way freezer is used or rather not used for long term storage. Because of erratic power supply in the city, it is not possible to assert whether frozen products have remained frozen prior to arriving at the supermarket; nor whether home freezers can maintain products at set temperatures. The only respondent who reported storing food in the freezer (HH-2) was also the only one living in a gated community equipped with a generator offsetting power shortages. Moreover, storage practice strongly relies on

habits and normative understanding of what is an appropriate use of storage options: when asking about food storage, HH-2 respondent declared that freezer is a good way to prevent insects degrading wheat and flour; HH-1 respondent stated that freezer is not intended to store “food” but ice cubes and ice cream. Looking at storage metabolic profile by food categories (Fig. 8), we found that on average 75% of fruits and vegetables have been stored in cupboard, despite the high temperatures and humidity levels at the time of the sampling in Bangalore. In a study on food storage behaviour, Lyndhurst (2011) qualified as “sub-optimal” storage to keep fruits and vegetables at room temperature rather than in the fridge; loose rather than in its packaging or wrapped up. These opposing findings suggest that looking at all steps of household food metabolism could help to reframe the problem of food storage behaviour and associated waste emphasized by Lyndhurst (2011), not only in terms of individual knowledge about appropriate storage behaviour; but also in terms of food consumption practices within a specific context. In Bangalore, where both a cultural perception of food being fresh only when bought loose and cooked daily from scratch is combined with erratic power stability, respondents tends to maintain traditional storage management habits, which means buying fresh produce more regularly, even though they widely have access to modern appliances such as refrigerator and freezer.

Fig. 7. Distribution of food storage normalized over of the total food purchased by each household.

52

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

Fig. 8. Distribution of food storage by categories normalized over of the total food purchased by each household (n ¼ 6).

Fig. 9. Food consumption by food categories normalized over the total food consumed by each household.

3.1.2. Cooking and eating Among all respondents, HH-6 has the least diversified diet (Fig. 9). Recalling what we observed earlier, HH-6 exhibits also the lowest shopping frequency with only one shopping trip (Fig. 5), the rest being purchased and stored prior to the week the survey was conducted. As an opposite pattern, HH-4 has the most diversified diet and is also the household with the highest shopping frequency. Indeed, 71% of caloric intakes in HH-6 are coming from cereals and lentils, products that can be stored longer than fruits, vegetables and meat. Therefore, diet is an important factor to consider when analysing household food metabolism dynamic. In fact, variations in provisioning intensity and frequency observed above between HH-4 and HH-6 are better understood as differences in households' diets than differences in particular provisioning practice elements such as food stock management routines or habits. Even though more data would be required to assert this finding, comparing this metabolic profiles to the Indian national average suggest a shift in urban middle classes diets. India still has the smallest share of meat consumption in the world, accounting on average for 1% of caloric intakes at the national level (FAOSTAT, 2013). This value is contrasting with our observations where meat consumption occurs in two thirds of the surveyed households. Regarding milk and dairy, the same trend can be observed: in 2011, the national milk and dairy average consumption was reported to reach 8% of caloric intake (FAOSTAT, 2013).

The consumption of this food category among the respondents is almost twice as large, covering on average 15% of caloric intakes. This suggests that beside population growth, shifting in consumption patterns among middle classes with higher disposable incomes, also significantly contributes to the increase in dairy products consumption. Finally, one important element of cooking practice has to be considered to understand how respondents manage to provision and eat fresh meals cooked from scratch everyday. In Bangalore, it is rather common for the middle classes to rely on domestic helpers and here five households out of six were employing domestic help cooking either part-time or for the totality of the daily meals. The availability of domestic help therefore allows for upholding food consumption practices by mediating temporal constraints associated with traditional culinary preferences (e.g. cooking daily as part of an healthy diet) and busy “modern” professional and recreational activities. 3.1.3. Managing leftovers and disposing Fig. 10 shows the share of avoidable and non-avoidable food wastes over the total food consumed by household.4 On average (n ¼ 6), only 2% of the total food disposed was avoidable. Since

4 Total food consumed ¼ direct food consumption (Fc1, Fc5) þ storage outputs (Fc2, Fc4).

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

53

interesting to note that HH-2 and HH-4 exhibit what has been quoted the “two stage holding process” (Evans, 2012: 6) where the refrigerator is used as a buffer stock for the leftovers. In HH-4, 56% of leftovers ended up in the bin, but almost the total (51%) have been previously stored back in the perspective of being eaten at some point in the future. This underlines once again that looking simultaneously at the biophysical and sociological dimensions of the activity “to nourish” leads to the same conclusion: provisioning and leftover management are key processes in household food metabolism, but their dynamic seems to be essentially shaped by practices elements e retailing infrastructure, access to domestic helpers, norms around food freshness and healthy diet e articulated around routines and habits.

4. Discussion and conclusion Fig. 10. Avoidable and non-avoidable food waste normalized over each household's food consumption.

detailed data on household food waste are quite scarce, that no data are available in the Indian context and methodologies are differing from one study to another, it is challenging to draw comparison. However, compared to available studies conducted over UK and Swiss households, these values are very low. Quested et al. (2013a) and Holding (2010) found that 17e19% of food purchased e excluding drinks e has been wasted in UK, while Beretta et al. (2013) indicate that 23e25% of purchases are wasted in Switzerland. Metabolic profiles and interviews indicate that half of the respondents draw on a “buy today, eat today” mentality, enhanced by a normative understanding of food freshness, supported by elements such as food retailing infrastructure diversity, domestic helpers, and influenced by contextual factors such as erratic power supply. All these elements contribute to shape provisioning and storage practices, which in turn leads to minimal food wastage. These findings emphasize that the food waste issue could be more effectively (re)framed and described as a pattern emerging from provisioning and managing practices embedded in a larger socio-cultural context, rather than simply framed as part of careless behaviour, individual choices and personal attitudes. The metabolic profiles of HH-2 and HH-4 regarding food waste seem well suited to enlighten this statement. Contributing to 98% of the avoidable wastes among the sampling, these two households are both corresponding to what Quested and Luzecka (2014) identified as the “high food wasters” profile: young professionals, aged 16e34, and working full time. Interestingly, interviews revealed that this apparent similar socioeconomic situation affects practices performances in different ways. HH-2 is living in a single household and declared to shop once a week in order to cope with a busy professional schedule while saving time for other activities. Living alone, the respondent recognized that she has to throw food away because it goes bad or too much has been cooked. On the other hand, HH-4 shared a flat with other people and declared shopping everyday precisely to accommodate and manage erratic schedules. But here the lack of coordination and communication among roommates is leading to over-provisioning or stock mismanagement, ultimately turning edible food into waste. Finally, Fig. 11 presents the distribution of leftovers fates among households using management categories as a descriptive domain. Except HH-6 who did not report any leftovers during the sampling week, the larger share have been either consumed after storage or shared with domestic helpers, a finding consistent with the minimal amount of food waste reported above. However, it is

Our goal was to pilot an inter-disciplinary method to quantify the disaggregated flows and processes of the activity “to nourish” in the home and analyse how they relate to household's consumption dynamic using a social practice theory perspective. In particular, applying analytical reflexivity proved to be valuable to combine both frameworks and to enlightening the links between metabolic profiles and consumption practices. Reflexivity implied to engage in an iterative process where both sets of data e quantitative and qualitative e are collected, structured and analysed in order to inform each other: the unified set of food consumption practices (as-entity), reproduced in space and time, provide a structure toward selecting relevant e possibly hidden e flows and processes along with the model's spatial and temporal boundaries; because all flows and processes are attributed to and treated as being embedded in practices (as-performance), the building of metabolic profiles based on different descriptive domains allows for investigating which practices and elements configurations lead to variations and similarities in food consumption patterns. Our findings indicate that food-retailing infrastructures, system of provision's reliability, cultural values around food freshness combined with upholding of culinary competences, and socioeconomic conditions allowing respondents to rely on domestic helpers, notably influence food-provisioning frequencies. This in turn affects storing practices resulting in a more efficient food stock management and less food wasted than what has been accounted for in industrialized countries. Among our sample and based on a one-week consumption cycle, food waste is very low e only 2% of the food waste would be deemed avoidable. Conversely, our findings also revealed that variations in elements such as professional schedule, communication and planning, along with diet content influence provisioning patterns in a way that could increase food wastage. Finally, we would like to emphasize that our results cannot be used to statistically infer the food consumption pattern of Bangalore's middle classes. Rather they are an empirical illustration of the process through which insights about substantive and methodological issues can be revealed and concepts clarified. By opening up the household “black box” and through a micro-study of households, we showed that a systemic approach combining quantitative and qualitative perspectives yields a better heuristic device towards understanding and explaining food consumption complexity in the home. This points to the need for further research to apply and extend these conceptual and methodological developments to a larger sample in order to derive general food consumption patterns related to specific social practice typologies. Hence, scaling up the proposed method would be an important step towards the formulation of recommendation aimed at sharpening the design of sustainable consumption policies that account for sociocultural contexts and consumption dynamics.

54

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55

Fig. 11. Food leftovers normalized over the total food reported as leftovers by each household.

Acknowledgements We would like to express our gratitude to all the research respondents for their engagement and to all our colleagues in Bangalore, including Malavika Belavangala, Megha Shenoy and Sunayana Ganguly.

Appendix A. Supplementary material Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jclepro.2016.03.055.

References Ayres, R.U., 1994. Industrial metabolism: theory and policy. In: Ayres, R.U., Simonis, U.E. (Eds.), Industrial Metabolism: Restructuring for Sustainable Development. United Nations University Press, Tokyo, p. 390. Beretta, C., et al., 2013. Quantifying food losses and the potential for reduction in Switzerland. Waste Manag. 33 (3), 764e773. Bijapurkar, R., 2009. We Are like that Only: Understanding the Logic of Consumer India. Penguin Books, India. Binder, C.R., 2007a. From material flow analysis to material flow management, part I: social sciences modeling approaches coupled to MFA. J. Clean. Prod. 15 (17), 1596e1604. Binder, C.R., 2007b. From material flow analysis to material flow management, part II: the role of structural agent analysis. J. Clean. Prod. 15 (17), 1605e1617. Boons, F., Howard-Grenville, J.A., 2009. The Social Embeddedness of Industrial Ecology. Edward Elgar Publishing. Bourdieu, P., 1979. La distinction critique sociale du jugement. Les Editions de Minuit, Paris. Breuer, F., 2003. Subjectivity and Reflexivity in the Social Sciences: Epistemic Windows and Methodical Consequences. Brunner, P.H., Rechberger, H., 2004. Practical Handbook of Material Flow Analysis. Lewis Publishers, Boca Raton, FL. Burger Chakraborty, L., et al., 2016. Urban food consumption in Metro Manila: interdisciplinary approaches towards apprehending practices, patterns, and impacts. J. Ind. Ecol. (in press). Cohen, M.J., Murphy, J., 2001. Exploring Sustainable Consumption: Environmental Policy and the Social Sciences. Elsevier, Oxford. Evans, D., 2012. Beyond the throwaway society: ordinary domestic practice and a sociological approach to household food waste. Sociology 46 (1), 41e56. Sage Publications. Evans, D., 2014. Food Waste: Home Consumption, Material Culture and Everyday Life. Bloomsbury Publishing. Evans, D., et al., 2013. Waste Matters: New Perspectives of Food and Society. WileyBlackwell. FAOSTAT, 2013. What the World Eats. Retrieved April 10, 2014, from: http://www. nationalgeographic.com/what-the-world-eats/. Finnveden, G., Moberg, Å., 2005. Environmental systems analysis tools e an overview. J. Clean. Prod. 13 (12), 1165e1173. Fischer-Kowalski, M., 1998. Society's metabolism. J. Ind. Ecol. 2 (1), 61e78.

Friedl, B., et al., 2006. Socio-economic drivers of (non-) sustainable food consumption. An analysis for Austria. In: Proceedings to the Launch Conference of the Sustainable Consumption Research Exchange (SCORE), Vienna. Giddens, A., 1984. The Constitution of Society: Outline of the Theory of Structuration. University of California Press. Graham-Rowe, E., et al., 2014. Identifying motivations and barriers to minimising household food waste. Resour. Conserv. Recycl. 84, 15e23. Green, K., Foster, C., 2005. Give peas a chance: transformations in food consumption and production systems. Technol. Forecast. Soc. Change 72 (6), 663e679. Gustavsson, J., et al., 2011. Global Food Losses and Food Waste. Food and Agriculture Organization of the United Nations, Rome. Haden, P., Vittal, I., 2008. The Great Indian Bazaar: Organized Retail Comes to an Age in India. McKinsey & Company, New Delhi, p. 80. Halkier, B., Jensen, I., 2011. Methodological challenges in using practice theory in consumption research. Examples from a study on handling nutritional contestations of food consumption. J. Consum. Cult. 11 (1), 101e123. Hobbes, M., et al., 2007. Material flows in a social context: a Vietnamese case study combining the materials flow analysis and action-in-context frameworks. J. Ind. Ecol. 11 (1), 141e159. Hoffman, A.J., 2003. Linking social systems analysis to the industrial ecology framework. Organ. Environ. 16 (1), 66e86. Holding, J., 2010. Household Food and Drink Waste Linked to Food and Drink Purchases. Department for Environment, Food and Rural Affairs (DEFRA), York, p. 11. Huang, C.-L., et al., 2012. Using material/substance flow analysis to support sustainable development assessment: a literature review and outlook. Resour. Conserv. Recycl. 68 (0), 104e116. Koivupuro, H.-K., et al., 2012. Influence of socio-demographical, behavioural and attitudinal factors on the amount of avoidable food waste generated in Finnish households. Int. J. Consum. Stud. 36 (2), 183e191. Kummu, M., et al., 2012. Lost food, wasted resources: global food supply chain losses and their impacts on freshwater, cropland, and fertiliser use. Sci. Total Environ. 438, 477e489. Lifset, R., 2008. The quantitative and the qualitative in industrial ecology. J. Ind. Ecol. 12 (2), 133e135. Lipinski, B., et al., 2013. Reducing Food Loss and Waste. World Resources Institute Working Paper (June). Lyndhurst, B., 2011. Consumer Insights: Date Labels and Storage Guidance. Waste and Resources Action Program, Manchester, p. 194. Lyndhurst, B., et al., 2007. Food Behaviour Consumer Research: Quantitative Phase Manchester. Waste and Resources Action Program, p. 43. Mason, M., 2010. Sample size and saturation in PhD studies using qualitative interviews. Forum Qual. Sozialforschung/Forum Qual. Soc. Res. 11 (3). Mauthner, N.S., Doucet, A., 2003. Reflexive accounts and accounts of reflexivity in qualitative data analysis. Sociology 37 (3), 413e431. Monier, V., et al., 2011. Preparatory Study on Food Waste across EU 27. European Commission. Papargyropoulou, E., et al., 2014. The food waste hierarchy as a framework for the management of food surplus and food waste. J. Clean. Prod. 76, 106e115. Parfitt, J., et al., 2010. Food waste within food supply chains: quantification and potential for change to 2050. Philos. Trans. Roy. Soc. B Biol. Sci. 365 (1554), 3065e3081. Parizeau, K., et al., 2015. Household-level dynamics of food waste production and related beliefs, attitudes, and behaviours in Guelph, Ontario. Waste Manag. 35, 207e217. Quested, T., et al., 2013a. Household Food and Drink Waste in the United Kingdom 2012. Waste and Resources Action Programme, Manchester, UK, p. 17.

L. Leray et al. / Journal of Cleaner Production 125 (2016) 44e55 Quested, T., Luzecka, P., 2014. Household Food and Drink Waste: a People Focus Manchester. Waste and Resources Action Program, p. 131. Quested, T.E., et al., 2013b. Spaghetti soup: the complex world of food waste behaviours. Resour. Conserv. Recycl. 79, 43e51. Randles, S., Warde, A., 2006. Consumption: the View from Theories of Practice. Industrial Ecology and Spaces of Innovation. Edward Elgar Publishing, Inc., Cheltenham, UK. Reckwitz, A., 2002. Toward a theory of social practices: a development in culturalist theorizing. Eur. J. Soc. Theory 5 (2), 243e263. Sahakian, M., Wilhite, H., 2014. Making practice theory practicable: towards more sustainable forms of consumption. J. Consum. Cult. 14 (1), 25e44. Schatzki, T.R., 1996. Social Practices: a Wittgensteinian Approach to Human Activity and the Social. Cambridge University Press. Schiller, F., 2009. Linking material and energy flow analyses and social theory. Ecol. Econ. 68 (6), 1676e1686. Sengupta, A., 2008. Emergence of modern Indian retail: an historical perspective. Int. J. Retail Distrib. Manag. 36 (9), 689e700. Shove, E., 2012. The dynamic of social practices. The Dynamic of Social Practices: Everyday Life and how it Changes. Sage, London, p. 208. Shove, E., 2009. Everyday Practice and the Production and Consumption of Time. Time, Consumption and Everyday Life: Practice, Materiality and Culture. Berg Publishers, London, pp. 17e34. Shove, E., 2010. Beyond the ABC: climate change policy and theories of social change. Environ. Plan. A 42 (6), 1273. Shove, E., 2015. Linking low carbon policy and social practice. In: Strengers, Y., Maller, C. (Eds.), Social Practices, Institution and Sustainability. Routledge, Oxon, p. 201. Shove, E., Warde, A., 1998. Inconspicuous Consumption: the Sociology of Consumption and the Environment.

55

n, R., et al., 2006. Estimating Household and Institutional Food Wastage and Sibria Losses. Food and Agriculture Organization, Rome, p. 30. Sonesson, U., et al., 2005. Home transport and wastage: environmentally relevant household activities in the life cycle of food. Ambio 34 (4e5), 371e375. Sood, D., Mishra, S., 2013. India Retail Foods. D. Williams. Global Agricultural Information Network, New Delhi, p. 20. Spaargaren, G., 2011. Theories of practices: agency, technology, and culture: exploring the relevance of practice theories for the governance of sustainable consumption practices in the new world-order. Glob. Environ. Change 21 (3), 813e822. Spurling, N., McMeekin, A., 2015. Interventions in practices: sustainable mobility policies in England. In: Strengers, Y., Maller, C. (Eds.), Social Practices, Intervention and Sustainability. Routledge, Oxon, p. 201. Srivastava, P., Hopwood, N., 2009. A practical iterative framework for qualitative data analysis. Int. J. Qual. Methods 8 (1), 76e84. Timmer, V., et al., 2009. Sustainable Household Consumption: Key Considerations and Elements for a Canadian Strategy. Consumers Council of Canada, p. 68. Tukker, A., et al., 2006. In: Peter, E., Luis, D. (Eds.), Environmental Impact of Products (EIPRO) e Analysis of the Life Cycle Environmental Impacts Related to the Final Consumption of the EU-25. European Commission, p. 141. Wilhite, H., 2013. Energy consumption as cultural practice: implications for the theory and policy of sustainable energy use. In: Strauss, S., Rupp, S., Love, T. (Eds.), Culture of Energy. Left Coast Press, San Francisco, pp. 60e72. Williams, H., et al., 2012. Reasons for household food waste with special attention to packaging. J. Clean. Prod. 24 (0), 141e148. Yin, R.K., 2009. Case study research. Des. Methods 4.