Spatial distribution of food retailers in amsterdam understanding community nutrition environment in the citys context m a

Spatial distribution of food retailers in Amsterdam Understanding community nutrition environment in the city’s context Nguyễn Văn Quốc Thái Thesis submitted in fulfilment of the requ

UNIVERSITEIT VAN AMSTERDAM GRADUATE SCHOOL OF SOCIAL SCIENCES

NGUYỄN VĂN QUỐC THÁI

SPATIAL DISTRIBUTION OF FOOD RETAILERS

IN AMSTERDAM

Supervised by Asst Prof Aslan Zorlu

Amsterdam – 2016

SPATIAL DISTRIBUTION OF FOOD

RETAILERS IN AMSTERDAM

Nguyễn Văn Quốc Thái

Master thesis MSc Urban and Regional Planning Graduate School of Social Sciences

Spatial distribution of food retailers in Amsterdam

Understanding community nutrition environment

in the city’s context

Nguyễn Văn Quốc Thái

Thesis submitted in fulfilment of the requirements for the degree of

Master of Science in Urban and Regional Planning, University of Amsterdam

Supervisor: Dr Aslan Zorlu Second reader: Dr Wouter van Gent

June 2016, Amsterdam

A CKNOWLEDGEMENTS

Individual success is a myth No one succeeds all by herself – Pat Summitt

I would like to express my genuine gratitude to my supervisor Dr Aslan Zorlu, whose nonchalant and encouraging supervision throughout the stressful course of this thesis has contributed greatly to its successful completion Being trapped at the Roeterseiland Campus has turned out to be less depressing than anticipated, thanks to his consistent (literal) presence and guidance

My gratitude also goes to officials from Gemeente Amterdam, whose support has provided

me with efficient materials for the thesis I would like to thank Martin van den Oever for our cheerful and helpful conversations, and Tony Dashorst, Cor Hylkema, and Steven Poppelaars for maintaining great patience and support for my continual requests for data

I am also grateful for the support I receive from a number of academic staff at my faculty: Beatriz Pineda Revilla, for intriguing an interest in urban food system, Rowan Arundel, for living up to the expectation as a “GIS guy”, and Karin Pfeffer, for an inspiring problem-solving session

This amazing year filled with eye-opening experiences and ridiculously extravagant spending

on eating out would not have been possible without support from my family and friends The thesis is the best way I can possibly pay back their unconditional love and trust As a tribute

to my home country’s amazing cuisine: Có thực mới vực được đạo!

C ONTENTS

1 INTRODUCTION 1

1.1 THE RESEARCH 1

1.2 OUTLINE 2

2 THEORETICAL FRAMEWORK 4

2.1 FOOD RETAILERS AND URBAN NUTRITION ENVIRONMENT 4

2.2 FOODS RETAILERS AND URBAN ENVIRONMENT 8

2.3 THE CASE OF AMSTERDAM 15

2.4 CONCEPTUAL FRAMEWORK 18

3 METHODOLOGY AND DATA 21

3.1 APPROACHING GEOGRAPHIES OF FOOD CONSUMPTION 21

3.2 RESEARCH DESIGN 23

3.3 OPERATIONALISATION 25

3.4 MAPPING DISTRIBUTION USING GIS 32

3.5 REGRESSION ANALYSIS 34

3.6 DATA 35

4 RESULTS 39

4.1 SPATIAL DISTRIBUTION OF FOOD RETAILERS 39

4.2 FOOD RETAILERS IN RELATIONSHIP WITH URBAN ENVIRONMENT 48

5 DISCUSSION 58

5.1 KEY FINDINGS 58

5.2 IMPLICATIONS: SCENARIOS FOR FOOD JUSTICE 64

6 CONCLUSION 69

6.1 SUMMARY 69

6.2 PRACTICAL CONCERNS 71

6.3 RECOMMENDATIONS FOR FURTHER RESEARCH 74 REFERENCES

APPENDICES

L IST OF FIGURES AND TABLES

Figure 2.1 Model of nutrition environment 8

Figure 2.2 Conceptual framework 20

Figure 4.1 Type-based distribution of food retailers in Amsterdam, 2015 40

Figure 4.2 Number of food retailers per cell in Amsterdam, 2015 41

Figure 4.3 Number of chained supermarkets per cell in Amsterdam, 2015 42

Figure 4.4 Number of discount supermarkets per cell in Amsterdam, 2015 43

Figure 4.5 Number of non-chained grocers per cell in Amsterdam, 2015 44

Figure 4.6 Number of specialty food shops per cell in Amsterdam, 2015 45

Figure 4.7 Number of healthy food shops per cell in Amsterdam, 2015 46

Figure 4.8 Number of foreign food shops per cell in Amsterdam, 2015 47

Figure 5.1 Operation scheme of Buurtbuik 67

Figure 5.2 Meals at Buurtbuik 67

Figure 6.1 Number of supermarkets within 1000 metres (Schadenberg 2015) 68

Table 1 Filters applied to dataset 24

Table 2 Summary of GIS method 33

Table 3 Summary of regression analysis 35

Table 4 Descriptive statistics of chosen variables 37

Table 5 Summary of variables 38

Table 6 OLS regression models for number of retailers by type within 500 metre radius of CBS cells regressed on built and socio-economic environment characteristics 49

Table 7 Logistic regression models for one or more food retailers of each category within 500 m radius of CBS cells regressed on built and socio-economic environment characteristics 49

to grasp an overview of the food retailing landscape, which has been widely discussed as greatly influential on a just urban environment (Alkon & Agyeman 2011), in the socio-economic context of the Municipality of Amsterdam The food system, particularly food consumption

in this case, has been an important yet underexposed aspect of urban interrelated ecosystem (Johann et al 2014), which calls for illumination on food system’s role in the achievement of urban justice, a well-established body of research literature in urban studies

With the increasing importance of food retailers in shaping consumption in urban areas (Wrigley et al 2003), efforts have been made on observing the relation between food consumption and changing urban environments (Shannon 2014; Kwate 2008) However, a majority of studies have chosen to pay attention to accessibility or availability of specific types

of store, especially chained supermarkets, instead of a more holistic view that includes other type of food retailers Though chained supermarkets are indeed gaining popularity in when it comes to food shopping, they do not represent the whole foodscape of the city, and they should not indeed (Steel 2009) In line with this, the thesis aims to explore the panoramic picture of Amsterdam to gain both preliminary insights as well as discovering the differences between different types of food retailers in the context of built and socio-economic environment of Amsterdam It aims to answer the following research question:

How are different types of food retailers’ spatial distribution related to their surrounding built and socio-economic environment in the case of the Municipality of Amsterdam?

Methodologically, the research adopts a quantitative methodology Firstly, GIS will be used both illustrate and describe distributional patterns of food retailers Secondly, regression analysis will be used to explore the relationship between these patterns and selected aspects of

Spatial Distribution of Food Retailers in Amsterdam

urban environments Following the use of quantitative analysis, these descriptive and explanatory observations will be used to explore the implications of the current foodscape in Amsterdam The contribution of the thesis is twofold; firstly, exploration of the relationship between food retailing and urban environments help make visible the role of food system in urban context, which still remains imperceptible both academically and politically until recently (Kierans & Haeney 2010); secondly, in addition to qualitative observations, development of a complementary quantitative methodology will help diversify angles from which food problems can be approached

Thus, while not exploring original concepts or innovative methods, the thesis investigates a familiar concept yet from an under-researched aspect, using a combination of established methods The social relevance is implicated in a sense that the research topic is closely connected to urbanites, who are constantly exposed to food insecurity (Sadler et al 2015; Pothukuchi & Kaufman 1999) One of the biggest challenge to come for planning in the 21st century is ‘how to feed cities in a just, sustainable and culturally appropriate manner

in the face of looming climate change, widening inequality and burgeoning hunger’ (Morgan 2015) In a broader discussion, food consumption has great implication for maintaining urban environment quality as well, for instance, influencing external resources dependency and water footprint, which is believed to be a few among many problematic features of contemporary food system (Vanham et al 2016)

1.2 OUTLINE

The thesis is structured as follows The introduction chapter provides a brief overview

of the thesis In the second chapter, the theoretical framework is set Based on concepts of food system, food retailers and urban environment, the theoretical framework is established, preceded as well as supported by discussion on the main debates in the literature This helps inform the selection of variables as well In addition, an overview of Dutch food retailing landscape is provided to address relevant features of Amsterdam food retailers

The third chapter serves to describe in details the data and methodology employed First, the research design is discussed in order to set the research scope and provide the rationale behind the design Then, quantitative operationalisations of the main concepts, namely food retailers, built environment, and socio-economic environment are provided This helps define

methods, namely GIS and regression analysis, are delineated After that, a detailed overview

of data sources, data gathering, cleaning and transforming process is presented

The fourth chapter provides the most important information of the thesis Using the results of mapping and regression analysis, I describe firstly the general patterns of distribution

of food retailers, secondly the relationships between these patterns and the current context of Amsterdam, and thirdly the relationship with the change of such context from 2004 to 2015 Certain brief discussions/explanations are provided when necessary in this chapter as well The fifth chapter goes into further explanation, as well as trying to link the findings back

to the discussed literature, as well as some information about the Dutch foodscape As food consumption is an extremely complex and multifaceted phenomenon, the explanation is inherent incomplete In some cases, due to lack of proper academic or expert opinion, explanation is subject to a certain degree of subjectivity (from the perception of a foreign student who has only spent roughly a year in Amsterdam) Still, interpretation is much needed

as it would provide a necessary pre-condition for future research, and might be useful to set the stage for a deeper understanding of the observed patterns

Therefore, in this chapter, best efforts are made in an attempt to understand how different types of food retailers are related to their surrounding and what might explain it, and more importantly, what is missing that is much needed for such an explanation Certainly not everything is explicable, and regression analysis, when applied to a cultural phenomenon like food consumption, does not guarantee perfectly comprehensible revelations promising to cure world hunger Still, both of the more comprehensive and less comprehensive all have their benefits, which would be used in the recommendation section

The final chapter consists of three parts Firstly, the research question are revisited, summarising the main findings and discussions Secondly, practical concerns associated with chosen data and methods are mentioned at the end of the chapter Finally, recommendations for further research are made

2 Theoretical framework

2 THEORETICAL FRAMEWORK

In this chapter, I will discuss different themes with two aims: (i) to construct the backbone theoretical framework, based on which research scope, hypotheses, and research questions are developed, and (ii) to provide a theoretical argument for my choices of variables,

in addition to methodological one in the subsequent chapter Three themes will be addressed Firstly, the role of food consumption in urban context is outlined with relevant studies Based

on this, a closer look is taken to examine the food retailers in relation to nutrition environments, using a conceptual framework of nutrition environment pioneered by Glanz et al (2005) Secondly, other urban environments are discussed in relation to food retailers, with attention being paid to how these can be observed Thirdly, some main features of Amsterdam foodscape literature are outlined in support for the fourth and final part, in which a conceptual framework is constructed

2.1 FOOD RETAILERS AND URBAN NUTRITION ENVIRONMENT

2.1.1 Urban food system

Food system, while has been widely recognised for its essential role in the maintenance

of urban ecosystem, has also been less visible than other systems, such as transportation, housing, employment or environment (Johann et al 2014) Similarly, while food is central to survival for all low-income urban residents, literature on urban inequality has tended to decenter food as a focus analysis (Miewald & Mccann 2014) More than mere nutrition input, the food system comprises ‘the chain of activities connecting food production, processing, distribution, consumption and waste management, as well as other associated regulatory institutions and activities’ (Journée 2011) This system holds critical implications for urban planning (Pothukuchi & Kaufman 1999; Morgan 2009), specifically food consumption (Bedore 2010; Pottinger 2013), usually regarded as a distinguished yet closely related co-existence of food production

Food consumption itself is a broad terms, encompassing a broad ray of activities ranging from ‘buying food products and services and transporting them to preparing and eating food and finally disposing of the food wastes’ (Oosterveer & Spaargaren 2012) Efforts have been made by consumption geographers (Crewe 2001), highlighting the uneven geographies of

consumption (Mansvelt 2008) Considering food consumption’s complex dynamics, studies

on the subject often opt for specific stages of consumption Quite a few authors try to explore the psychology and patterns of food buying behaviour (Kalogeras et al 2009) Others, especially from environmental background, are deeply concerned with problems such food system’s carbon footprint or recycling food waste, either from a social (Giskes et al 2006) or technological perspective (Takata et al 2012) Another thread of literature focuses on ethnography of food system as an embedded feature of culture and society, e.g the differentiated perception of food among urban residents of varied socio - economic, ethnic or generational backgrounds (Schösler 2014) In addition, due to the development of statistics

as well geography information system technology, a rising number of studies have turned their attention to the spatiality of food retailing, as a proxy for food consumption, arguing that:

Supermarkets, specialised food shops, famers’ markets and farms all constitute consumption junction as the places where providers and consumers actually meet and their respective logics are bound to connect (Oosterveer & Spaargaren 2012)

This branch of literature is of particular interest for the thesis due to its tendency to employ quantitative data on large (enough) scales, from which patterns and predictions can be extracted Based on spatial and statistical data, researchers are allowed to visualise and correlate food retailers in relation to the surrounding contexts, which in turns allows panoramic, albeit possibly limited, understanding of how food system works in cities Abstract concepts, e.g

‘food accessibility’, or ‘food affordability’, can be quantified, coded and calculated, which contributes not only to empirical evidence but also methodology To a certain extent, this fits well with urban contexts, where administration and policy are largely based on more extensive geographical level, revealing patterns or changes among an increasingly great number of urban residents While results can be contradictory at times, food retailers have been proved to have great implications for urban foodscape, which in turn is interrelated with many aspects of urban life

2.1.2 Food retailing and its influences

Retail can be narrowly defined as the economic activity wherein physical goods are being traded from retailers to consumers (Evers & Oort 2005) However, most of the literature make

no explicit use of a defined term of ‘food retailer’ The general notion of ‘food retailer’ refers

to a location or business that provides food on a commercial basis It is based on this notion

2 Theoretical framework

that each study would establish their own definition and topology in correspondence with the research goals and questions, a process that is necessary yet problematic As Guy (1998) has argued, classification of retailers would always involve a compromise between ‘simplicity and comprehensiveness’ Overwhelming categorisation can make analysis and interpretation challenging, while too few categories can obscure essential insights In addition, secondary data is not always available at the desired topology level; thus, it is unlikely to argue for a ‘best’ system of classification As observed in the literature, topology is varied among different studies, with some being more comprehensive (Lisa Powell et al 2007; Guptill & Wilkins 2002), while others more focused (Kwate et al 2009; Smoyer-Tomic et al 2008) This has some implications for the topology of food retailers employed in this thesis, which will be explained in details in the next chapter

Within the limit of this thesis, food retailers are limited to businesses where food, either

processed or raw, can be purchased as opposed to be served, i.e service places such as restaurants,

café or fast food stores1 Certainly, such rough division would be challenged This blurred line

is evident in the literature, with a number of studies setting the same limit (Ball et al 2009; Jetter & Cassady 2006; Zenk, Schulz, Israel, et al 2005), while others trying to incorporate both (Cummins & Macintyre 1999; Cummins & Macintyre 2002; Morland, Wing, Rous, et

al 2002; Smoyer-Tomic et al 2008) The choice of this conceptual scope serves to (i) investigate food as an essential part of daily life that is less contextually restrained (comparing, for instance, shopping at supermarkets for one week’s food supply with having meals at cafés), and (ii) allow a more operationalisable yet still informative foundation definition This facilitates operationalisation of ‘food retailers’

The problem is, the literature tends to not make explicit in what ways food retailers can

be connected to a broader context of nutrition environment in cities A great numbers of studies, with a pre-defined focus on food accessibility, provide no rationale regarding this This results in some problems in terms of operationalisation, e.g whether or not to include food price as a variable, or how to justify my choice of research scope and topology Hence, it

is necessary to visualise the relationship of between the tangible food sources and intangible nutrition environment

1 Similar definition has been adopted by Morland, Wing & Diez Rouz (2002), stating that “food stores are defined

as industries that retail food and beverage merchandise from fixed point-of-sale locations, including supermarkets, grocery

2.1.3 Urban nutrition environment

In a broader sense, the thesis considers food retailers as an essential constitutive of urban nutrition environment Lake and Townshend (2006) pointed out three pathways through which environment can be linked to healthy problems, including physical design, socio-cultural rules that governs the environment, and socio-economic status of the environment Somewhat similarly yet in more details, Glanz et al (2005) suggested that urbanites are influenced by nutrition environment through various paths, which can be observed in three levels: policy, environmental, and individual (Figure 2.1) Regarding the environmental level, four sub-environments are defined with corresponding variables, including: community, organisational, consumer, and information Among these, ‘community’ and ‘consumer’ environment are assigned with highest priority, since they have been less studied and can have broader effects (Glanz et al 2005)

Certainly, all environments are interlinked, considering, for instance, barriers to healthful food access may also stem from spatial patterning of food prices, rather than solely the locations of food retailers (Breyer & Voss-Andreae 2013) The choice of which environment(s) to investigate tends to depends on the specific research goals and data available

I will focus on community nutrition, specifically the type and location of food retailers This focus fits with the data available, and remains informative enough to explore the interaction between food retailers and surrounding environments, since people tend to make food choices based on the food retailers that are available in their immediate neighbourhood (Furey et al 2001) In addition, as argued by Evers et al (2012), the Dutch retail landscape is primarily the result of their locational strategies, which pay attention to the possibility to increase the stock

of retail in use and to attract the greatest number of consumers Interestingly, even studies with focus on consumer environment tend to include community variables as well, e.g the established branch of literature investigating ‘food desert’ (detailed reviews available at Beaulac

et al 2009; Walker et al 2010; Ver Ploeg et al 2015)

2 Theoretical framework

Figure 2.1 Model of nutrition environment (Glanz et al 2005, emphsis added)

In summary, this section positions the thesis’ topic in the growing literature of urban nutrition

environment The thesis explores one component of the environmental level This approach

is used to build up theoretical framework, as well as define dependent variables

2.2 FOODS RETAILERS AND URBAN ENVIRONMENT

After having defined the role of food retailers in urban nutrition system, I turn to investigate their role in other urban environment I pay attention to the built environment and socio-economic environment These are not necessarily official, well-defined terms, but rather based on rough classification of how the urban environment is approached in food literature Factors that can be linked to neighbourhood characteristics can roughly be divided into these two categories This will be explained in details in later section

As discussed, urban nutrition system is receiving increasing academic as well as policy attention People eat – there’s no denying that Thus, food is ingrained in almost every aspects

of urban life This is evident in the literature, considering the effort to link different related concepts to urban infrastructural and socio-economic context This body of literature

food-is broad with varied themes, among which I will dfood-iscuss two main themes: (i) inequalities in food consumption, and (ii) food retailers and urban environment

2.2.1 Inequalities in food consumption

Due to the rapidly changing in how food is produced, retailed and consumed, more and more attention is being paid to this previously under-researched topic This branch of

which is currently being criticised as being exclusive, non-transparent with serious consequences on environment and society (Lang et al 2009) ‘Food justice’, the conceptual product of such a concern, points to not only to environmental but also social aspect of food

While remaining ‘relatively unformed’ (Gottlieb & Joshi 2010) and varying among authors (Bedore 2010), a common thread of the food justice literature is the inclusion of not only ecological sustainability but also social justice (Alkon & Agyeman 2011), regarding:

“The vision espoused by many food justice activists goes beyond one in which wealthy consumers vote with their forks in favour of a more environmentally sustainable food system

to imagine that all communities, regardless of race and income, can have both increased access to healthy food and the power to influence a food system that prioritizes

environmental and human needs over agribusiness profits This vision clearly weaves

together justice and sustainability”

In the first decade of the 21st century, a combination of debates warning about the threat

of food and fuel scarcity in relation to the increasingly urbanised world population (Spaargaren

et al 2012) The idea, while taking a while to catch on, has caught the attention of the public, since it also relates to the unsustainable fuel-dependent economic development of the latest decennia In the so-called oil-peak debate, the relationship between energy and food scarcity has been put on the forefront of global food agenda

In its earlier stage, this remained a mainly technical concern, dealing with more tangible questions Can the current global agricultural production maintain its capacity to feed a growing population? Is there enough space available for the further expansion of production

to feed the growing cities whose connection to traditional agricultural food system has become increasingly weakened? This leads to an increasing interest in innovation in the food industry However, as food is a phenomenon inseparable from social and cultural aspects, the issues of

access to food was raised, eventually became an overriding concern for academia as well as policy

(Spaargaren et al 2012) Since then research has broadened from a relatively minor strand in public health and community development to be addressed across a range of scientific, health and social sciences disciplines (Dowler 2008)

According to Dowler (2008), inequalities in food experience, which includes having enough money for food; being able to reach shops selling the range of commodities desired in contemporary societies, and the concomitant inequalities in nutritional and obesity outcomes, are increasingly recognised in policy and in practice, not only in the UK but also in other societies Thus, the realities of food or nutritional inequalities are: those who are richer are more likely, than those who are poor, to have better health due to better nutrient intake The

2 Theoretical framework

reasons for the existence of such inequalities diverse Regarding food consumption, most studies point to out differentiated individual as well as geographical characteristics can create such a problem It is necessary to outline some of the important findings from existing literature regarding this aspect

2.2.2 Food retailers and urban environment

Due to the extensive of the literature engaging in this topic, this part only focuses on literature that explicitly explores the relationships between food retailers and urban environment The first impression when approaching this body of literature is that ‘food desert’ appears to be the most popular theme However, not every studies use the same concept There are a number of studies, rather than trying to define certain measure of food desert, opt for a more explorative approach These studies, and some of the food desert ones, help informing the choice of independent variables in the thesis In order to remain a clear structure for the thesis, as well as cohesion of the conceptual framework, the review of these studies try to categorise these independent variables into two categories: those of built environment and those of socio-economic environment While these are all interlinked, the socio-economic ones are more explicitly used in the literature, while the built ones are more implied2

Food retailers and built environment

Studies investigating the role of neighbourhood social and physical environment in the development of diet, either healthy or not, have become common in recent years (Smith et al 2010) Accessibility and availability of food services reflect an important component of the built environment that plays a role in individual food choices (Morland, Wing, Rous, et al 2002; Wrigley et al 2003; Smoyer-Tomic et al 2008) That said, the built environment is not

a popular choice when it comes to study food accessibility and availability Some studies did make an effort in trying to incorporate this dimension This include the relationship between

food retailers and a number of physical aspects of cities, including housing, transportation and

public facilities (Bernard et al 2007)

2 In contrast, a study by Rundle et al (2007) combined both demographic and infrastructural variables into a set

Housing value, in combination with income, has been used as a measurement for wealth

by Smoyer-Tomic et al (2008) Their research concluded that that lower wealth has greater exposure to fast food outlets, even in the case of better supermarket access It has not been explicitly related to food retailing among reviewed literature, yet a great number studies on gentrification and retail in general were concerned about how increasing (or one might say, depending on the context, Neoliberalising) property value is changing the retail landscape

Higher rents in Haarlem, New York, for instance, Zukin et al (2009), replaced the local retail

capital by more powerful retailers, specifically chain stores, as well as changing the customers

base of local stores

Housing tenure is a more ubiquitous choice, albeit some might consider that this belongs

to the socio-economic group of factors rather than the physical one For instance,

Vijayaraghavan et al (2011) employed the concept of housing instability (i.e owning a home),

to measure the level of food insecurity between different neighbourhood A more straightforward approach is to consider housing tenure per se, such as Smoyer-Tomic et al (2008), Wrigley et al (2003) and Kirkpatrick and Tarasuk (2003) The assumption, and conclusion, made from these studies might vary, but in general, households residing in public/social housing tends to be on the less advantaged side when it comes to access to healthy food supply, as well as more access to fast food outlets (Kwate et al 2009) In some cases, e.g as reported by Kirkpatrick and Tarasuk (2003), among low-income households, the purchase of healthy food are significantly lower for households that had to pay rent or mortgages than for those without housing payments

In addition to the confined residential space, mobility is also an important aspect that

has attracted attention from a variety of research People tend to shops for grocery within a certain distance from their homes and proximity to stores may be especially important for low-income consumers who do not own automobiles (Eisenhauer 2001) However, this relationship proves to be relatively complicated to study, since defining pattern of mobility on its own is already challenging enough Combining mobility and food shopping patterns would equal a double challenge Studies on this relationship has been limited Nonetheless, a study

by Coveney and O’Dwyer (2009) reported that living in a food desert did not, by itself, impose difficulties on accessing food retailers Far more important was access to either independent

or public transportation

2 Theoretical framework

Clifton (2004) focused on the mobility strategies and food shopping for low-income families in an American context For these families, income, mobility, and time constraints pose the greatest limit to the range of destination within reach, and thus, the choices available

to them as consumers In this case of Austin, Texas, public transit is ill suited for most household provisioning activities, and walking is an option for only those living in neighbourhoods with good local access and pedestrian infrastructure In a context equipped with better public transportation, people tend to have more expansive access to retailers, as reported by Larsen and Gilliland (2008)

An aspect that remains under-researched is the relationship between public facilities and

food retailers It might be the case that these facilities, such as parks and pedestrian street, indeed does not have great influence on the spatiality of these retailers While certain overlapping exists between this aspect and the mobility one, when mentioned, it is often associated with anti-obesity policy (Shannon 2014), as an attempt to alleviate the consequence

of growing fast food outlets density (Kwate 2008) Thus, it does not constitute the focus of this thesis

Food retailers and socio-economic environment

The aforementioned aspects are certainly cannot be separated from the socio-economic aspect of urban life In addition, food consumption is a deeply social and cultural phenomenon, and thus, it is not surprising that most relevant studies make use of neighbourhood demographic and socio-economic characteristics (Charreire et al 2010) In general, research has found that food consumption patterns vary between neighbourhoods and that living in and disadvantaged neighbourhoods is independently associated with a poor diet (Smith et al 2010) Neighbourhoods with higher income levels and higher proportions of white residents tend to have greater access to supermarkets or large chain food stores, while poorer neighbourhoods and those with higher proportions of black or Hispanic residents might have relatively high access to small grocery stores (Bader et al 2010) However, such a conclusion might be challenged by occasionally different observations from different context In the United States, it is been reported that there is poor supermarket access and availability among low-income, inner-city, and predominantly African American or Hispanic neighbourhoods, especially compared to more affluent suburban and primarily white residential areas (Smoyer-Tomic et al 2008)

In the UK, the social determinants of inequalities in food and diet include finance, access, choice, and anxiety (Dowler 2008) Here, unsupportive local food environments in socially deprived areas were found, although in the UK these have tended to encompass suburban local authority housing estates rather than occurring primarily in inner cities as in the US However, more deprived areas in urban context tended to have better access to fruit and vegetables supply than the more developed ones (Smith et al 2010) Canadian research is more positive, with literature showing relatively good access to supermarkets even for residents’ low socio-economic status well as inner-city neighbourhoods

In addition, the number of socio-economic factors employed in these studies are great

in both number and variation, thus, it is impossible to present a comprehensive review here Detailed reviews can be found at Beulac et al (2009), Walker et al (2010) and Charreire et al (2010) For the purpose of this thesis, review of these studies will focus on a number of main

factors, including: ethnicity, income, employment, poverty and crime

“Explicit engagement with the concept of race and its diverse theoretical foundations is important because it allows scholars to make arguments about how racism, shapes food system, to understand how race changes through food, and to consider how food might enable different theorizations of race”

Understandably, residents of non-Western minority are generally assumed to be prone

to lower access to healthy or affordable food supply, which might result from a number of social and economic barriers, and in some cases, zoning policy and gentrification Similar to house value or house tenure, ethnicity is also associated with gentrification process However, the use of ethnicity is not always straightforward, since the racial segregation is not always

extreme or follows a clear pattern In addition, ‘ethnicity’ per se can be perceived in different

ways; thus, income or unemployment rate are sometimes used as instead, depending on specific

2 Theoretical framework

designs of the research Still, it remains an important factor when attempting to investigate the spatiality of food retailers

Income and employment

Closely related to ethnicity is income, with a great number of studies employing this as their main variable In neighbourhood with low average income, access to healthy food supply

is lower, while that to fast food is higher (Block et al 2004) Zenk et al (2005) reported that supermarkets have newly opened or have remained open in or nearby middle-income, yet racially transitioning from predominantly White to a more mixed composition, neighbourhoods Perceptions of the market may also cause more convenience stores and fewer grocery stores to locate in low-income neighbourhoods, while neighbourhoods preferences may block both from locating in high-income neighbourhoods (Margheim 2007) Thus, the use of income is dependent on the specific context

Relatedly, employment rate is also used, e.g by Kirkpatrick (2003) The use of income

or employment can be tricky, as the measurement of these variables can be varied among studies, with different implications Hence, there is no single ideal way to use this as a variable However, the use of this is still important, in a sense that it informs how financial characteristic

of neighbourhood can be linked to retailers’ distribution

Poverty and crime

Scholars have also paid attention to a number of factors that can be roughly grouped as social deprivation of areas The notion of deprivation, to certain extent, is actually a mere extreme form of, say, income and employment, segregation in the city Thus, various indices have been established to measure this notion Among these, two factors stand out from the literature, i.e poverty and crime

Poverty is closely associated with income The focus of discussion, instead, turns to poverty employed as an important component of deprivation index, as used by a number of studies The results tend to be a mixed one, depending on both the contexts and the measurements employed Cushon et al (2013)’s study found out that food proximity tend to improve as social deprivation increased, while food (dietary) balance worsened in the same condition On the other hand, Zenk et al (2005) reported that the most impoverished neighbourhoods in which African-American resided had fewer supermarkets and lower potential accessibility in comparison to the most impoverished White neighbourhoods

In relation to poverty, crime is also hypothesised as an important barrier to food retailers (Zenk, Schulz, Israel, et al 2005) Higher crime rates near supermarket is likely to lead to vicious cycle in which crime deters customers and reduces sales revenues, making it more difficult to stock perishable foods such as fresh produce (Bader et al 2010) This, in turn, could further reduce pedestrian traffic neat the store, diminishing informal social controls that might otherwise curb street crime (Sampson et al 1997) These factors would influence the spatiality of supermarkets, and possibly other type of food retailers as well Consequentially,

in more disadvantaged neighbourhoods, e.g with higher proportion of non-Western immigrants, spatial analysis might reveal a relatively high crim rate, in addition to deteriorating buildings as well as neglected public facilities (Kwate 2008)

In summary, a number of major relevant findings in the literature have been defined

This, once again, puts the thesis’ topic in the context of existing literature, while simultaneously helps delineate the ‘originality’ of this research In addition, it helps define and justify the independent variables In the limit of this thesis, a number of varieties are chosen, including population, income, employment and crime Further explanation on the use of these

variables are provided in the methodology chapter

2.3 THE CASE OF AMSTERDAM

Certainly, the Netherlands are not the most common name among studies of food inequalities The general perception of the Dutch context as a less segregated and unequal one might prompt one to question how the concept of food inequality can apply to the case of Amsterdam A number of authors have questioned the existence of ‘food deserts’ – commonly used as measurement of inequality in food consumption – in the European context (Beaulac

et al 2009) Still, considering the fact that 39.7% Amsterdam residents suffer from obesity and related conditions (Gemeente Amsterdam 2013), the concept of food inequality, in addition to potential framework such as the one introduced by Glanz et al (2005), might be useful in understanding such an ‘unhealthy’ figure For example, the thesis does not aim as addressing inequality, yet it can reveal how community nutrition environment is related to neighbourhood characteristics in Amsterdam; and thus, informs how consumer nutrition environment can be investigated next

In addition, literature has discussed the rise of neoliberalisation in Amsterdam, represented by e.g gentrification, and how such a change has started to “wither the ‘undivided

2 Theoretical framework

city’” (Uitermark & Bosker 2014) Findings suggest that although this process has been, for a while, mitigated, the city is increasingly subject to deepening division between the privileged and the those who are not Thus, it is safe to assume that as the development of neoliberalisation continues, further change in the retailing structure would take place, and food retailing might be prone to current problems of disturbing modern food system as well It is based on such an assumption that the research is designed

The Netherlands are a highly industrialised country whose food supply is heavily dependent on imported sources In 2009, more than 7 million tons of food were imported, of which the majority originated from Southern America Via a centralised distribution system, most of this food amount ends up in supermarkets (Vermeulen et al 2010) The traditional Dutch food shopping behaviours are generally regarded as minimalist and economical (Schösler 2014) This is in line with finding of Food Truck Monitor 2014 – 2015 reports, claiming that consumers continue to cut back on eating out this year, with 63% prefer home cooking (Food Service Instituut 2015) Thus, the role of food retailers is of great importance

in maintaining food supply to urban residents The market is dominated by chained supermarkets, among which top two performers are Albert Heijn and Jumbo, accounting for

58 percent of the market In addition, a rising market can be observed in case of discount supermarket, with that of Aldi and Lidl combined equals 17.3 percent (Pinckaers 2015) Studies have shown that people shopping at supermarkets eat and buy healthier food than those shopping at discount markets (Blok et al 2015) Based on this, the thesis seeks to investigate what type of neighbourhood characteristics can be associated with the distribution

of these stores

The organic food sector keeps growing during the last decade (Pinckaers 2015; Kalogeras et al 2009) As people are becoming more and more aware of ethical consumption, Dutch food retailers have increasingly sourced food products which are either produced sustainably or obtained in a sustainable manner (Kalogeras et al 2004) Despite this, the notion of ‘sustainability’ is often challenged, being criticised for promoting environmental elitism since they either cater to a privileged consumer group or, when needing to maintain competitive price, does not fully stay true to their ecological and ethical commitment (Hall 2012; Gottlieb & Joshi 2010; Brisman 2009) Whether or not this claim is valid in a Dutch context would add to the findings of the thesis

Independent grocers, as well as specialty stores, such as butchers, fish mongers, bread shops and cheese shops, are facing increasingly harsh competition from chained competitors Independent food retailers are increasingly leaving the scene, due to shrinking margins and on-going consolidation in the retail market The ones remaining deal with this through the use of extra quality service, sales of high-quality added value products and advantage of niche markets Traditional ‘ordinary’ shops such as butcher or fruit shops are becoming specialty shops, catering to high-end market or upper-class customers, while traditional bakery shops, previously focusing on serving daily-mean bread, are moving into pastry or offering extra products of artisanal specialty breads (Pinckaers 2015) Considering the changing customer base, it would mean that while the number of these shops are of great proportion of food retailers in Amsterdam, they actually only cater to a specified group of higher-class consumers Drawing less attention from research literature is foreign food shops, or what is termed

“exotic food” by van Otterloo et al (2007)3 From the end of the 1970s onwards, many small shops and stalls were established to supply the migrants from Surinam, the Antilles and Africa Later on, indigenous Dutch people started to be attracted to these shops as well Interestingly, their clientele appears to be varied depending on ethnic composition of the districts in which they are situated For instance, for than half of the customers in Bos en Lommer (66 per cent foreign families) are immigrants, while in Oud-West (39 per cent foreign families), 70 percent are native Dutch people (van Otterloo et al 2007) Diverse customer base is present in this case; thus, there is no assumption made for this case; interpretation will be based on the results analysis

Another characteristic of Amsterdam food retailing landscape worth notice is the rising trend of ‘blurring’ between food retailing and food servicing sector (Bureau Stedelijke Planning

& Lexence 2015) Chained retailers keeps producing more categories in order to make their range of products more comprehensive, including also takeaways or pre-cooked meals (Pinckaers 2015) as well as healthier/organic choices (Oosterveer & Spaargaren 2012) This will have multiple effects on spatial planning as well as legal policies, which in turn will have deep impacts on consumer base, as well as the meaning of food shopping in not only Amsterdam but also the Netherlands (Bureau Stedelijke Planning & Lexence 2015) Thus,

3 Definition in this case is, of course, fraught with problems van der Veen (2003) defined exotic foods to be food that are unusual or desirable because of their foreign origin This definition doesn’t help build a concrete assumption on what might be the the potential customer base of this group It is, however, helpful in conceptualising this category

2 Theoretical framework

while the thesis does not focus on food service places, in certain cases, such as Albert Heijn, such blurring effect is inevitable However, since it does not affect the ability to interpret analysis results, it is not acknowledged as thesis’ limit

Existing literature on food in Amsterdam, as well the Netherlands that is relevant to the thesis is scarce, if not almost non-existent Most of the reports from government and food research institutes focus on the population’s nutrition intake, i.e taking on a more public health perspective than geographic one A study in Eindhoven concluded, based on this case, that in the Netherlands, an individual’s socio-economic position has more influence on their diet than the socio-economic position has more influence on their diet than the socio-economic characteristics of the area in which they live (Giskes et al 2006) This is an interesting observation, since it differs from findings in the USA, the UK and Finland (Giskes

et al 2006) The results’ analysis may reveal whether or not similar interpretation can be drawn

In addition, a number of outdoors market and initiatives, such as community kitchen and food banks, are also gaining popularity in the city However, these are not the focus of the thesis and only discussed as concluding notes

Main research question

How are different types of food retailers’ spatial distribution related to their surrounding built and socio-economic environment in the case of the Municipality of Amsterdam?

Since the research question is relatively straight forward, there is no need for constructing sub-questions Answering this involves addressing three points: (i) how different types of food retailers are distributed spatially, (ii) how these distribution patterns can be linked

to the surrounding environment, and (iii) what implications these findings might have for the foodscape of Amsterdam

2.4.2 Conceptual framework

The following scheme represents the conceptual framework of the specific research problem (Figure 2.2) In choosing to focus on community nutrition environment, the study uses food retailers as an indicator, whose main characteristics of focus will be location and type While presented at separate, the two characteristics are intertwined in reality When presented

in map, both the type and location can be observed Regarding regression analysis, the aggregate of different types as well as the use of buffering cells also incorporate both type and location information Due to the thesis limit, the consumer nutrition environment is not included However, certain features of that environment can still be incorporated in building

up topology of food retailers, albeit incomprehensively The framework is relatively straightforward and self-explanatory, thus will help delineate research design and operationalisation of concepts

2 Theoretical framework

Figure 2.2 Conceptual framework (Source: Author’s illustration) (Asterisk indicates variables with separate regression models)

3 METHODOLOGY AND DATA

This chapter describes the research methodology, from the operationalization of the research question to data collecting and analysing methodology Firstly, an overview of approaches is provided Secondly, the research design is outlined, explaining and justifying the main approach to the research question Thirdly, concepts central to the research question are operationalised, based on which the research scope is also defined The fourth and fifth part describes GIS and regression methodology respectively Finally, a detailed data description is provided, with rationale for choices of dependent and independent variables

As typical of any urban problems, food consumption has been approached through different points of view Since food is closely associated with health, for a while the topic has been popular among ecological and public health scholars They hypothesised that neighbourhood food environments to which individuals are exposed can exert an independent effect on diet and diet-related chronic disease through influencing food purchase and consumption patterns This is particularly evident in the American context, considering the rising food desert body of research previously discussed Understandably, scholars in this field rely heavily on the use of quantitative data and method, e.g statistical analysis with national survey results (e.g Smith et al 2010; Zenk, Schulz, Hollis-Neely, et al 2005; Burdette & Whitaker 2004) The category does not necessarily always pay attention to spatiality of food consumption, but rather how food, as a health concerns, varies among different social groups

‘food desert’, these works investigate how spatial distribution of food retailers interact with socio-economic characteristics of urban residents With this, new concepts are applied, such

as walking distance area for food shopping (Smith et al 2010) These works help reveal certain patterns of distribution in relation to a number of factors, such as ethnicity, education, income

3 Methodology and data

and crime rate However, occasionally these observations remain inexplicable, since, after all, predicting human behaviours is much less straightforward and more difficult than, for instance, chemical reactions

Food consumption is a daily, embedded, largely unconscious practice, imbued with different meanings and acting as a marker for exchange and worth, as well as of difference (Caplan 1997; Lupton 1996) Hence, a number of authors instead opted for a more qualitative approach, which in turn can be roughly categorised into anthropological-cultural ones and social-economic ones The former category involves researches whose foci lie in the embedded meaning of food consumption in daily life Rather than using defined definitions which re more common in quantitative works, a number of these studies pay more attention to how perceptions about certain aspects of food, such as food shopping, food waste, or ethical diet is perceived among different social groups (Everts & Jackson 2009; Schösler 2014)

Closely related to this field is the investigation of social-economic impacts of certain food-related phenomenon using qualitative methods While still considering food as a deeply cultural phenomenon, these studies try to link food system to other social and economic concerns For instance, Anguelosvki’s ethnographic account on healthy food stores and food gentrification is an attempt to link the rising alternative food movement, of which organic food is a representative to gentrification, a widespread urban phenomenon with great consequences (2016) Miewald and McCann, based on in-depth interviews and participant observation, argues that food should be a more central focus of critical geographical research into urban poverty (2014) Wekerle’s political approach illustrates how the originally community concerns of food has over years transitioned into a claim for a more democratic and just society, and most recently, a framing of food justice movements with great relevance for planning and urban theory (2004)

With such a growing reservoir available, more and more researches have tried to apply a mixed methodology, making use of both quantitative and qualitative methods Though not abundant, these researches stress the important of supplementary methods, usually combines statistics analyses and interviews (Morland & Evenson 2009) or focus group discussion (Cummins et al 2008) This thesis aims to gain preliminary insights into the system of Amsterdam’s food retailers; thus, a quantitative approach has been chosen, making use of both GIS and statistical analysis This is necessary for further studies, as well as provide empirical evidence should there is demand for a more microscopic, qualitative research into a certain

aspects of food retailing GIS method helps visualise the current food retailing landscape, while regression method is expected to provide complementary and explanatory accounts Due to its multifaceted nature, a multitude of approaches have been adopted to investigate the relationship between food retailers and surrounding environments The ‘ideal’ study of access to food outlets would appear to be one that associates all dimensions related to accessibility: proximity, diversity (the types of food outlets), availability (food supply at the food outlets), affordability (cost) and perception Moore et al (2008) reported that the perceived availability of healthy foods can be different from that measured by GIS application Methodically, this would mean combining both quantitative and qualitative methods to incorporate fully the spatial and social dimension of food consumption

Within the limit of the thesis, two methods are adopted: GIS and regression analysis This is suitable for the thesis’ goals, which is to have a preliminary look at the food retail landscape of Amsterdam Based on the use of large-scale data, a number of general patterns can be observed, and different categories can be put in comparison for a more comprehensive assessment Based on this, further research can employ a more qualitative angle into, e.g., either a specific type or location

3.2 RESEARCH DESIGN

It is necessary to draw the line between my research and the well-known food accessibility/ food desert literature, since this influences the research design The existence of food deserts in the European context, in contrast to the American one, is highly questioned

by researchers (Beaulac et al 2009) This is in line with Schadenberg’s findings (2015), claiming that there is no food deserts in Amsterdam However, considering the fact that 39.7 percent of Amsterdam residents are overweight (Gemeente Amsterdam 2013), it is unlikely that food environment has a less important role here It might be the case that while food accessibility, or lack of such, is not as evident as the American or British context yet, and that other components of urban environment play a significant part in explaining geographic differences In addition, building up a topology is necessary in understanding food retailing,

as it helps delineate a complicated subject, yet still allows enough zooming-out ability to grasp

a panoramic view Most of the food accessibility literature tends to focus only on supermarkets, while this study attempts to understand the differentiated ‘story’ each type of food retailers would have to offer

3 Methodology and data

While dependence does not guarantee causality, as typical of cross-sectional study, the results are still expected to offer a panoramic picture of how the nutrition environments interact with other environments The study has a quantitative character in which two

complementary methods are employed Firstly, GIS will be used to extract information from

geocoded data, with the aim to visualise the food retailing landscape in a systematic way, at which there has been hardly no attempt in the case of Amsterdam In addition, this enhances visual juxtaposition between a major part of food supply and corresponding socio-economic

context Secondly, regression analysis is employed to provide statistical evidence for

(non-)relationship between the dependent and independent variables

The unit of observation is food retailers of different types Next, the unit of analysis is the

statistical 100x100 metres cell provided by CBS100mvierkant, which would then be buffered

to 500 metres, which assumed to be reasonable distance for people in a geographical residential unit to travel to access food4 (Cushon et al 2013; Gould et al 2012; Wrigley et al 2003; Shaw 2012b; Shaw 2012a) In order to maintain a consistent in the data among the chosen variables, two filters are applied (Table 1) Firstly, those cells with no resident in 2014 will be excluded, since the research question focus on the ones that are inhabited currently Secondly, cells with missing values are excluded as well, since they only account for a small proportion, thus allows reliable regression results In the end, 5764 cells are used

Table 1 Filters applied to dataset

The main motivation for the choice of unit of analysis depends on the availability of data

on population characteristics (Leete et al 2012) Initially, the buurten (neighbourhood) is

considered for unit of analysis, since it is official administrative geographical unit However, standards administrative areas, while common in urban management to visualise geographical

4 This does not fully justify the choice of buffering, since food accessibility literature also suggests a variety of other distances, such as 1 kilometre, 2 kilometres or 1 mile However, since the aim of the thesis not to define

patterns and compare places, are not always appropriate or well-justified (Rees 1997; Martin 2004), specifically:

(i) There is no evidence showing that people shop in a defined neighbourhood (ii) A great number of shops are located on the roads that are border between two

3.3 OPERATIONALISATION

3.3.1 Food retailers

Based on the aforementioned definition of food retailers, a topology is necessary for analytical purposes Based on reviewing various studies and data available, a custom topology

is built up This includes 6 categories, namely:

(i) Chained supermarkets,

(ii) Discount supermarkets,

(iii) Non-chained grocers,

(iv) Specialty food shops,

(v) Healthy food shops, and

(vi) Foreign food shops

The rationale behind this topology is, unfortunately, not backed by any Dutch-context policies or studies A suggested topology by Guptill and Wilkins (2002) appears to be detailed

3 Methodology and data

and well-based, relying mainly on type of food and annual revenue, and partially on number

of store on national level as well as number of employees However, this topology is exhaustive,

in a sense that there is no or little data on revenue or employment publicly accessible In addition, it would be unnecessarily detailed for the research question as well, as well as diluting the results and discussion Still, it acts as guidance for the topology

The definition of each type is also not backed by official government or academic documents, except for a few, e.g chained supermarket (Schadenberg 2015) Thus, I have resorted to website of the retailers and various news report to define which to include in category Consequentially, comprehensiveness might be prone to compromising Such problem is not unforeseeable though, considering the originality of my research Thus, I made

my best effort to maintain a comprehensive topology Each of these categories serves to tell a distinct ‘story’ – an aspect of the overall food retailing landscape

Certainly, questions might be casted over whether these categories might overlap, and certainly, in some cases, they do However, as discussed, it is implausible to create a comprehensive topology, especially when existing supportive Dutch literature is scarce Thus,

in interpreting the results, certain priorities have to made, based on the amount of empirical evidences available The detailed definition of each type is as follows:

Chained supermarkets: Supermarkets of Albert Heijn (excluding Albert Heijn to-go

shops), Attent, Boni, C1000, Coop, Deen, Deka, Jumbo, and Spar

(Chained) Discount supermarkets: Supermarkets of Aldi, Dirk van den Broek, Lidl, and

Vomar

Non – chained grocers: non-chained shops not specialising in one specific type

Specialty food shops5: shops specialising in one type of food only that contributes to basic daily meals, including: fishmongers, cheese stores, butchers, poultry stores, vegetables and fruit stores, bread stores These products are selected based on components of a basic Dutch meal (Vanham et al 2016)

Healthy food shops 6 : shops specialising in healthy/organic food, excluding supplements/functional food stores

Foreign food shops: shops specialising in foreign food

5 Similar definition can be found at Morland et al (2002)and Johann et al (2014)

6 It should be noted that Albert Heijn has also introduced their own range of organic (‘biologische’) products

Each of these category is expected to offer a distinctive insight into the food retail landscape Chained supermarkets are currently the most influential one on the Dutch scene The market share of chain supermarkets in the total expenditure on food and drinks are steadily increasing in the Netherlands; similarly, of every euro spent on food, more than a half

is spent at chain supermarkets (FoodService Instituut 2014, as cited by Schadenberg 2015) These are argued to have wider range of product, more competitive range, and in some case more range of healthy food (Zenk, Schulz, Israel, et al 2005; Morland, Wing, Rous, et al 2002) This might not totally the case in Amsterdam though, with chained supermarkets are generally perceived as middle-ranged priced Still, their distribution patterns can reveal what type of residential areas they are most associated with

Also of interest is discount supermarkets Comparing to data from 2004, the number of this type in Amsterdam has increased from 39 to 44 shops, proving their rising popularity These supermarkets are perceived to be of more competitive price, with less attractive representation and narrower selection, pared-down presentation, yet the quality is, to a certain extent, uncompromised Thus, it serves to reveal whether the assumption that discount stores caters mostly to less privileged residential areas are true In contrast, healthy food shops might reveal a more privileged target group (Hall 2012; Brisman 2009), which can then undermine the alternative food movement’s claim to create a more sustainable foodscape (Goodman & DuPuis 2002) Local specialty food shops7 and non-chained grocers are often associated with gentrification (Gonzalez & Waley 2013), as authors argued that the process has a tendency to introduce corporate players into the local retailing scene, leaving the local food shops in a less competitive position (Zukin et al 2009), as well as threatening the diversity of local foodscape (Steel 2009)

Foreign food shops represent a tricky category, since it can represent either a highly mixed foodscape, or a less diverse one with more ‘exotic’ foreign shops instead These initial observations and assumptions help establish underlying hypotheses about the interaction between food retailers and surrounding environments, which in turn help interpret the analysis results in later stage of research

3 Methodology and data

3.3.2 Built environment

As discussed, the built environment can be linked to food retailers through various ways, considering how diverse the literature on this topic is Considering the research question, it is necessary to choose variables representing built environment yet still carrying certain social and economic meaning Furthermore, it should help illustrate the change in the built environment through time as well Thus, built environment is expressed through two variables, the first is house value, and the second is house ownership (tenure) The division between built and socio-economic is, again, solely for the purpose of structuring the thesis’ ‘story’, since by

no means these are in reality separate On the other hand, since changes in the built environment, such as house value, are much more sensitive to market changes, chances are that consideration of these values will help diversify the observations

House value

Housing value has been a common indicator among gentrification literature, since it clearly represents the changes in the built environment of an area Certainly, one can also argue that this can fall under the category of economic variables (Clark 1987; Badcock 1989; Hammel 1999) Indeed, as a relatively straightforward indicator, house price incorporates both the upgrading of the built environment as well as demographic shift towards higher socio-economic status groups Still, since the relationship between housing price and social composition is not entirely straightforward (Teernstra & Van Gent 2012), and it is in close association with physical features of the cities, this remains in the built environment category

In addition, using housing value allows a more straightforward results interpretation, which is necessary in for the study, since it is exploring an under-researched topic Price per square meter is selected, as it is the most refined data available, and would also incorporate other housing features, such as size and land use

market has come under neo-liberal pressure (Kadi & Musterd 2014) Still, nowhere else in Europe does social housing dominate the housing market as it does in the Netherlands (Elsinga & Wassenberg 2007) The inhabitants of social housing tenants, compared to average household, typically are those who (Elsinga & Wassenberg 2007):

- are older

- have lower incomes

- are less likely to be in employment and more likely to be on benefit, and

- are more likely to be of non-Dutch origin

In Amsterdam as a whole, but also in gentrifying neighbourhoods, the owner occupied sector has substantively grown over time (Hochstenbach et al 2015) It was found that tenure from social housing to private housing conversions in gentrifying neighbourhoods in Amsterdam contribute to the inflow of higher income residents (Boterman & Van Gent 2014) Thus, house ownership is selected as a built environment variable to investigate how the spatiality of food retailers are associated with the housing market of Amsterdam This has been previously used by, e.g Smoyer-Tomic et al (2008), Kirkpatrick and Tarasuk (2003), though not in a Dutch context

However, considering that this variable has high collinearity level with income level as well as ethnic composition, this variable is not included in the main regression Instead, a separate model is created for this variable and will be included in appendices B

For the built environment, two variables are used It is supposed to reflect the relationship between community nutrition environment and built environment of Amsterdam,

in the context of neoliberalising house price and association between social sector and less privileged groups

3.3.3 Socio – economic environment

When it comes to socio-economic environment, literature has provided a profound list

of urban social and economic characteristics that could be associated with food consumption Among these studies, as have discussed in the previous chapter, a number of common variables stand out Among these, based on available data and time limit, five variables are selected with one additional control variables

3 Methodology and data

Detailed rationale is provided as follows:

Population

The use of population density is advocated by, e.g., Zenk et al (2005), or Wang et al (2014) However, there is no explicit assumption that can be drawn out of these studies With regards to food access, Margheim (2007) argued that stores locate where there are enough people to make a retail operation profitable The thesis still maintains the use of this variables,

as an explorative proxy to investigate if there is any association between food retailers and population density

Ethnicity

As discussed, ethnicity remains one of the most common variables among food studies Almost all of the reviewed articles mention certain extent of racial discrimination when it comes to food consumption It is common knowledge that Amsterdam has a diverse ethnical setting Thus, the use of this variable help reveal which ethnic group is more, or less spatially associated with certain type of food retailers This might reflect low purchasing power or raced-based discrimination (Bader et al 2010) While this doesn’t necessarily apply to the Dutch cases, it proves that ethnicity composition is worth consideration

Income

Another popular choice is income There is a variety of how income can be measured

A majority of reviewed literature makes use of median household income, arguing that it reflects the household dynamic of food spending The thesis follows this practice, choosing disposable household income This help reveal what type of food retailers can be associated with lower or higher rate In another scenario, considering the generally perceived economical spending of the Dutch, there might be no significance relationship at all There is also evidence that access in supermarkets in low-income urban areas has declined over time (Larsen & Gilliland 2008)

However, since this variable is prone to collinearity with ethnicity composition, two separate models are created for each variable In addition, racial segregation in Amsterdam appears to be not as extreme as the case in American or British cities (Musterd 2005) Thus, income, while incorporate certain ethnic dynamic of the population, is expected to provide more helpful insights, and will be included in the main result section

On a side note, Morland et al (2002) has argued that income and house value are related, yet with different meanings In their study, they use house value to measure neighbourhood

wealth, while using income to indicate measurement of individual wealth This, while not relevant to the conceptual framework of the thesis, is necessary to acknowledge, and also help

to delineate between the two variables

Employment

In some other studies, employment rate of residents in the geographical unit is used, e.g Inagami et al (2006) or Smoyer-Tomic et al (2008) In this study, the data is available in a different form The number of working people (werkzame persosen), represent the number of people working in the area, regardless of them being a resident there or not The use of this have several reasons Firstly, this is the data available at neighbourhood level Secondly, the use of income has, to a certain extent, incorporated the dimension of local employment rate

As suggested by an interview with van den Oever8, the unemployed in the Netherlands have relatively good access to social benefits, thus, the unemployment rate might be less meaningful when it comes to describe food retailers distribution Thus, the thesis opts for the number of working people, with full awareness of the different implications it might have

Crime

Since the relationship between food retailers and crime rate or neighbourhood safety can

be interpreted in a number of ways, the use of crime rate as a variable can a tricky one Concerns about safety could prompt shoppers to take a less direct route or seek out more distant stores (Bader et al 2010) Thus, it would help reveal more about shopping behaviours rather than food retailers’ spatiality On the other hand, high crime rate might affect strategic location of some food retailers in avoidance of deterring customers and reducing sales revenues Thus, the use of this variable is justified

Control variable

As typical of any social geography studies, there are a number of other intervening variables that need to be considered It is common knowledge that Amsterdam has become a well-known tourist destination Thus, it is safe to assume that the pattern of retail development, not only in the food sector but also other ones, is, to certain extent, in correspondence with this fact While the effect of tourism on spatial distribution of food retailers is not within the

thesis, focus, it is necessary to include tourism as a control variable, since a number of retailers

cater not only to local resident but also tourists The data available allows calculation of

8 Interviewee later introduced in chapter 5

3 Methodology and data

percentage of tourism businesses in the neighbourhood, and thus will be included in the study

to isolate the results from the effect of tourism on the spatial distribution

In addition, public transportation also contributes greatly to the food shopping practice, allowing greater mobility Even though the relationship between mobility and food consumption is extremely complicated and not the focus on this thesis, it is necessary to include this as a control variable for more precise results from the explanatory variables Another control variable to consider is universities and their campuses Although Amsterdam has the largest amount of students, Groningen has most students percentagewise (CBS 2012) Still, it

is likely that even if the great flow of students (into and out of) the city everyday can also have impact on the spatiality of food retailers; e.g most of students buy their lunch from nearby Albert Heijn, in addition to the campuses’ canteen Thus, university is also employed as an control variable

The socio-economic environment is investigated through the use of five independent variables, namely population, ethnicity, income, employment, and crime A set of control variables is used, including tourism, public transportation and university campuses

As medium of analysis, GIS is used to build maps of Amsterdam showing data described above Geocoded data is assigned with new topology, then plotted to buffered cells to count the sum of total shops, as well as number of each type Based on this spatial concentration of food retailers is mapped As presented by Pfeffer et al (2012), within a specified search radius, event data, for example existence of a chained supermarket, are aggregated to produce a continuous surface of distribution As explained in the previous chapter, buffered cells are meant to capture the food shopping environment of residents in a geographical proximity Technically, network analysis can be employed to determine the distance preferred by residents

to access food, advocated by, e.g., Cushon et al 2013; Gould et al 2012; Ball et al 2009; and Morland & Evenson 2009 However, this is very data-intensive, so instead the buffering distance are chosen based on literature, with application for different intention

The 500-metre buffering is meant to capture the ‘immediate’ food shopping environment, which would be applied to aggregate of all food retailers, as well as aggregate of each category Since the study does not include investigation of transport system, and residents

previous study has explicitly stated which distance is reasonable in the Dutch context, a second buffer of 1km is applied in order to either complement or reveal a different revelation from those of 500-meter buffering

In reality, the existence of certain physical features might undermine this assumption, among which train railroad tracks and water surface are the most noticeable ones Regarding the train railroad tracks, a number of them are elevated, and the networks are much less dense than other public transport systems Thus, this should not have heavy influence over the analysis results, especially when the research goal is not to measure accessibility

Regarding water surface, Amsterdam is well-known for its canal system Yet, with more than 1900 bridges (Gemeente Amsterdam n.d.), canals are unlikely to pose substantial obstacles for food shoppers While the IJ clearly separates Northern Amsterdam from the other side, when using 500-meter buffering, there is no situation in which one buffered cell

on one side overlaps with boundary of the other Thus, analysis results are not to be compromised

Next, since the data is available at different scale, GIS is used to modify, calculate, extract and combine information from different sources into one concrete data set The risk of doing this is data accuracy might be compromised Still, answering the research question requires certain level of unity in unit of analysis, which can be achieved through the use of cells (Shaw 2012a; Shaw 2012b) Therefore, when not available at cell level, data will be gathered or calculated at neighbourhood levels and assigned to cells whose centroids fall inside the neighbourhood’s boundary The process is summarised as follows:

Table 2 Summary of GIS method

1 Refine food retailers list ARRA food retailers

list Refined list with new topology (1)

shapefile New shapefiles with buffered cells (2)

3 Plot food retailers location to

buffered zones (1) & (2) Shapefiles with aggregate number of

stores attached to corresponding cells (3)

general foodscape and each type’s distribution

5 Calculate and assign information

to cells (3) & Administrative map at

neighbourhood level

Excel file contains data for all variables

3 Methodology and data

3.5 REGRESSION ANALYSIS

The next step is to study whether there is any statistically significant relationship between the aggregate of food retailers and built and socio-economic variables The choice of analytical method in this stage depends on the nature of variables used in the study, which is evident through a wide range of techniques from literature While some employ calculation of correlation coefficients (e.g Lu & Qiu 2015), regression analysis is a common choice (e.g Giskes et al 2009; Lisa Powell et al 2007; Moore & Diez Roux 2006; Zenk, Schulz, Hollis-Neely, et al 2005) for researches dealing with continuous independent and dependent

variables The process of this study consists of two models, investigate first the presence and then the quantity of food retailers, using STATA 13.1

Firstly, since the independent variables are of a continuous nature and there are more than one dependent variables, OLS regression is chosen This has been advocated by a number

of authors investigating the relationships between different types of food retailers and economic variables, such as Wang et al (2014) and Powell et al (2007) The thesis follows this common practice, using OLS regression, considering the nature of both dependent and independent variables In addition, OLS regression allows a more explorative interpretation that the thesis aims for Another model to use is logistic regression, in which dependent variables are encoded into dummy variables, with 0 stands for non-presence of stores and 1 stands for presence of 1 store or more, which has also been applied previously by Smoyer-Tomic et al (2008), using dummy variables to represent food outlet exposure To do this, More detailed explanation is provided in section 4.2

A number of variables with skewed distribution are log-transformed All the dependent variables are logged Logged independent variables include disposable household income and

number of working people Relationships are then considered with a p ≤ 0.05 significant Due

to multicollinearity between house ownership, ethnicity and income, three sets of described regression models are run separately for each of these independent variables Chapter 4 is reported mainly based on the income-included ones The other two results, while being reported in chapter 4 as well, are presented in Appendices B

The method and results are summarized as follows: