Version classiqueVersion mobile

Tyneside Neighbourhoods

Daniel Nettle

2. Study sites and methods

Texte intégral

I tried to get away, but I couldn’t get far,
‘Cos a man with a tow-truck repossessed my car.

1This chapter describes the main study sites, discusses how the methods were chosen and developed, and provides an overview of the datasets that were generated in the course of the Tyneside Neighbourhoods Project.

The study neighbourhoods

2I chose Neighbourhoods A and B as the main study sites after some archival research and a number of informal visits. Very extensive social and economic data are available for census areas within the UK. The most convenient size of census tract, the Lower Super Output Area, is a little small for a neighbourhood study, and in Newcastle, the neighbourhoods as recognized by local people tend to be cut in pieces by Lower Super Output Area boundaries. Thus, I decided that each of the study neighbourhoods would consist of two adjoining Lower Super Output Areas that together roughly corresponded to a neighbourhood as recognized and named by residents. All census and other third-party statistics given for the study neighbourhoods given in this book have been created by taking a summation or population-weighted average of the figures for the constituent Lower Super Output Areas. The two Lower Super Output Areas making up each study neighbourhood are pretty similar to each other on most indices.

3The search strategy was to find two neighbourhoods that were as far as possible similar in every respect except the level of deprivation. Of course, deprivation changes a lot of things: the state of the properties, the shops present, the types of food available and many other things are different in deprived and in affluent neighbourhoods. Thus, the neighbourhoods would inevitably differ in myriad ways. The search was thus not for an exact match, but rather for two studies where everything that could be similar given contrasting economic fortunes was similar, and the differences were as far as possible related to socioeconomic deprivation, rather than extraneous factors.

4I selected a site in the North East of the city for Neighbourhood A, and in the West End for Neighbourhood B. The sites selected were satisfactory in a number of ways. They were about the same distance (3km) from the city centre. Each consisted of a recognizable main street with a supermarket or two and a selection of smaller shops and food outlets. The main street ran North-South in one case, and East-West in the other. Leaving the main street in both perpendicular directions were residential streets; these led to networks of other residential streets and then to the neighbourhood boundaries. Both neighbourhoods were generally low-rise and contained a great deal of housing around or slightly more than one hundred years old, much of it terraced. Beyond this there were some differences, such as a greater number of large detached houses in Neighbourhood A, and a greater number of smaller post-1945 houses and flats in Neighbourhood B, although there were some of these in Neighbourhood A too. In neither neighbourhood were the residents mainly students. Students are a very particular population, high in individual socioeconomic status, but with no long-term ties or investment in the community, and comparing a student to a non-student population would be misleading. On the whole, it was possible to imagine without too much a stretch that if economic fortunes improved, Neighbourhood B would start to look rather like Neighbourhood A does now, and, if the economy collapsed, Neighbourhood A would start to look somewhat like B does now.

5The neighbourhood statistics backed up how good the match was (Table 1.1). The population sizes were about the same. The population structure differed in predictable ways (proportionately fewer men and more children in Neighbourhood B), but the differences were slight. The median age was only slightly higher in Neighbourhood A. You will notice that the number of households is rather larger in Neighbourhood B. This is because in Neighbourhood B, half of households with children contain only a lone adult, whereas in Neighbourhood A, the great majority of households with children (88%) contain two adults (Nettle, 2011a). This is itself an interesting pattern to which we will return.

Table 1.1 Key statistics for the two study neighbourhoods. Data are reproduced from Nettle (2011a), and the ultimate sources are the 2001 census and 2004 indices of multiple deprivation

Neighbourhood A

Neighbourhood B

Total population




1502 (48.5%)

1508 (46.8%)


708 (22.9%)

810 (25.1%)

Median age






Percentage population born in UK



Index of Multiple Deprivation



Index of Multiple Deprivation percentile



6Particularly important for the conception of the study was the fact that the ethnic composition of the two sites was very similar, with both study sites containing, at the 2001 census, 92% people born in the UK. Thus, whatever else the study was doing, it would not predominantly be comparing the behaviour of different ethnic groups, or immigrants to non-immigrants. An important caveat is in order here, which is that there has been considerable immigration into Neighbourhood B in the years since the 2001 census. The data from the 2011 census, which were not available at the beginning of the study, have now been released, and the percentage UK-born for Neighbourhood B is now closer to 70% (Hill, Jobling, Pollet, & Nettle, 2014). Thus, it remains the case that we are comparing two predominantly British-born white communities, but the difference in the number of immigrants should be noted.

7If the upper rows of Table 1.1 show statistics that are similar across the two neighbourhoods, the lower rows show the opposite. The Index of Multiple Deprivation is a composite statistic on an arbitrary scale, so values of 9 and 76 will not mean much out of context. The percentages are more informative: these show that if all of England’s 34,000-odd Lower Super Output Areas were laid out from the most to least deprived, the areas making up Neighbourhood A would be found around position 27,000. The areas making up Neighbourhood B would lie in the first few hundred. In the Tyneside context, Neighbourhood A is about as affluent as it gets whilst still remaining within the city, whilst Neighbourhood B is typical of the concentrations of deprivation in the West End and Eastern riverside of the city.

8A more dramatic way of seeing the difference comes from the UK government’s Neighbourhood Statistics website. The idea of this site is that residents or potential house buyers can enter a postcode and obtain a dashboard of useful statistics about the area around that particular street. Central to these is a handy visual aid: the meter of deprivation. As Figure 2.1 shows, the meter of deprivation looks pretty healthy for a postcode in the middle of Neighbourhood A, but it is maxed out for a postcode in Neighbourhood B. Quite what the creators of this web site were planning to do for neighbourhood regeneration, or whether they had thought about the possible adverse consequences, I am not sure.

Figure 2.1 The meter of deprivation for an address in Neighbourhood A (left) and one in Neighbourhood B (right). Image from, subject to the terms of the Open Government Licence.

9It is important to be clear about what the design of the Tyneside Neighbourhoods Project does and does not allow us to claim. It looks rather like a quasi-experimental design: two sites that are the same, except that fate has assigned one of them to socioeconomic deprivation. Thus, any differences in behaviour that we see between them tell us what the causal consequences of deprivation are, clean of other confounding variables. This seems to me an advance on traditional one-site ethnographies, in that at least there is one explicit and systematic comparison. My original conception of the Tyneside Neighbourhoods Projects was indeed as a quasi-experimental comparison along these lines. However, in truth the project does not allow such strong or such general claims.

10The reason concerns a basic principle of research design: replication. If we want to conclude something about variation at level X, we need multiple replications of entities at level X. What this means in this case is that to say something general about the consequences of neighbourhood deprivation, we would need multiple pairs of matched neighbourhoods, one in each pair deprived and the other one not deprived. Without such replication, any differences we observe between the two neighbourhoods are interesting but do not necessarily generalize to other affluent-deprived comparisons. They might not be due to deprivation at all, but to some other idiosyncratic feature of one of the sites I happen to have chosen. It might be that what we see in Neighbourhood B is very specifically to do with the uncertainty created by the threat of large-scale demolition, or the historical legacy of heavy industry, or any other specific feature of Neighbourhood B, rather than an epitome of what deprivation always does. Getting more participants in Neighbourhoods A and B to take part does nothing to alleviate this problem. That gives us more replication within our neighbourhoods, but does not increase our generalizability at the neighbourhood level at all. Partly for this reason, a critique of community studies and ethnographies in general is that they are not cumulative: we learn something about another location, but we don’t progress toward theory or knowledge of a more general character (Crow & Allan, 1994 p. 195).

11I acknowledge this limitation, but do not feel it undermines the value of the data presented here. By studying Neighbourhood B in contrast to Neighbourhood A, we will uncover patterns, generate ideas, and form hypotheses. The same patterns can subsequently be looked for, and the hypotheses tested, across a broader range of places. Some things will turn out to be similar elsewhere, and some will not, but that is an empirically tractable question. Related to this point is our commitment to developing repeatable quantitative measures that could be applied elsewhere to allow for direct comparability, a commitment I discuss in the next section. Moreover, this is not just a study of neighbourhoods, but also a study of humans living in those neighbourhoods. Quite a lot of the data to be presented in this book actually tests hypotheses at the level of the individual person rather than the neighbourhood, since the studies were concerned with understanding the individual-level processes whose emergent consequence was the neighbourhood-level difference. We had plenty of replication at the individual level, since our samples of people were typically quite large.

Development of methods: General considerations

12We now had two study sites. What were we going to do in them? There was no off-the-shelf set of standard methodologies. In this section, I discuss some of the general considerations that went into the methods; in the final section of the chapter, I discuss in more detail the individual datasets that we ended up collecting.

Qualitative and quantitative methods

13The reporting of the results of ethnographic research in anthropology does not tend to be very quantitative. Community studies in sociology are somewhat more quantitative, but patchily so. By contrast, I felt from early on that in the Tyneside Neighbourhoods Project that I wanted to quantify everything. The neglect of quantification in studies of social life is something of a curiosity, since researchers gather a large amount of data in the course of them, and there are many simple ways that they could use quantification more widely than they typically do. I can only attribute it to a widespread but unhelpful view that there are two different classes of methods, quantitative and qualitative.

14If you accept the premise that there are two classes of methods, qualitative and quantitative, then it is easy to see how you end up preferring quantitative methods if you have clearly specified a priori hypotheses to test between, but qualitative methods if you want to learn about the experiences of a particular group of people in a rich and open-ended way. However, this argument rests on a false premise, which is that quantitative and qualitative methods are mutually exclusive classes. In fact, all methods in the social and behavioural sciences are qualitative. This is because they all involve decisions about what to ask, what to record, how to represent and encode the data, and what the typology of outcomes is. These are fundamentally qualitative decisions, regardless of whether or not you go on to count the occurrences you have decided are significant. So, to the extent there is a choice of methods, it is between qualitative data that are represented numerically, and qualitative data that are excerpted or summarized verbally or visually.

15Faced with this choice, it is my view that you can get more from the data, and allow the data to speak more, if you attempt to quantify the entities or occurrences you are interested in. It is true that any particular quantification you do loses some of the information in the underlying data. However, so does any verbal summary, thematic analysis, or representative quotation. Using numbers versus not using numbers does not make any difference to this. Non-quantitative analyses are generally less transparent in terms of what encoding and editing processes have gone on than quantitative analyses are. With a well-presented quantitative analysis, filtering of the data has occurred, but at least the reader has some hope of understanding what kind of filtering it was. As for open-endedness, the important issue is not whether you quantify or not, but whether you can return to the underlying raw data and re-encode them differently if you have some new insight, idea or question. My personal experience has been that through trying to get substantial-sized samples and pull out numbers from them, I have often seen things that I did not expect and even ran counter to my phenomenological impressions. I don’t know if the same would have happened if I had not forced myself through the imperfect but self-improving discipline of quantification.

16Quantification of results, if coupled to explicitly described data-collection methods, has the potential to enhance the generalizability of social research and make the knowledge it produces more cumulative. Instead of merely noting that things are different in our study site than that of another study, we can take the same measures, compare the data, and test statistically where the similarities and differences lie. Quantification is also potentially useful for the legitimation of ethnography. Ethnographers make generalizations of record about the behaviour and experiences of particular groups of people at particular times. Sometimes, quite properly, those generalizations are contested. We have already met two cases of this in chapter 1: Young and Willmott (1957) and Turnbull (1972). Such contestations can quickly descend into one person’s anecdote against another’s, and that means we need to know the basis of each party’s claims. Quantification, in tandem with adequate description of sampling methods, can help here. Yes, you saw X happen, but how many observation bouts did you make in which X could have happened, and during how many of them did it actually happen? Quantification does not of course guarantee that all the generalizations we make will be fair or uncontroversial. I have come to appreciate through this project how every measure has multiple possible interpretations, and often raises as many questions as it answers. At least with quantification, though, we know how the generalizations were arrived at, and that can be the starting point for a dialogue between views.

17Thus, I decided early on that though the Tyneside Neighbourhoods Project was ineluctably a qualitative endeavour, nothing was to be admissible as evidence unless a number could be put on it. It was an essay in quantitative ethnography, and the methods for quantitative ethnography (or as I would prefer it, the quantitative methods for ethnography), are still fragmentary (Efferson, Takezawa, & McElreath, 2007). Thus, we were going to have to invent, borrow and adapt a range of assays, sampling strategies and coding schemes, to which I will turn shortly.

Direct observation and self-report

18Alongside the commitment to quantification sat a commitment to multiple methods. In particular, a prominent role would be assigned to direct observation of behaviour (also known as systematic social observation). Direct, systematic observation of behaviour is strangely little used in contemporary social science, although there was more done back in the 1950s than there is now. As Payne and Payne (2004 p. 157) drily put it: ‘People coming new to social research are often surprised to learn that social scientists […] do not use observation very much as a method’. This is all the more puzzling since anthropology, for example, prides itself on its detailed characterization of human life at relatively small and fine scales. Direct behavioural observation avoids recruitment bias (people who choose to participate being atypical of the population from which they are drawn), avoids demand characteristics (people behaving differently because they know they are taking part in a study), is faithful to life because it is part of life, and allows for almost limitless detail in what is recorded. Ethnographers do, of course, do a great deal of informal social observation. They just don’t tend to systemize it or turn it into structured quantitative data. The neglect of direct behavioural observation is not restricted to anthropology; it is underused in psychology too (Baumeister, Vohs, & Funder, 2007; Furr, 2009), and there is been a suggestion that sociologists should use it more to understand neighbourhood effects (Sampson et al., 2002).

19There are some weak reasons for the under-use of direct behavioural observation, and some more defensible ones. The weakest reasons are pragmatic: direct observation is time-intensive — for data capture, transcription and processing — and often just harder work than other methods, whereas researchers want usable data quickly and cheaply. The strong conformity-preserving features of the academic system mean that once direct observation is rare, it tends to remain rare. Researchers want their studies to look like the other studies they read in the literature, and peer reviewers enforce this. This conformity has good aspects, such as increasing the possibility of replication and cumulation, but bad aspects, such as imposing a rare-type disadvantage on methodological innovations that would be genuinely useful.

20The most defensible reason for the neglect of direct behavioural observation is that observation tells us little or nothing about the cognition or goals of the people involved. Max Weber gave the example of seeing a man chopping wood: from observing him, we cannot necessarily tell if he is doing wage labour, stocking his hearth, or just enjoying some exercise. To gain insight into this, you need to ask him; hence the centrality of questionnaires and interviews in social science. This is a good point but incomplete as it stands. To understand a man’s chopping of wood, it may not be sufficient to observe that he chops it, but it is necessary. He might not actually chop wood, even when he claims he does. There can be important and systematic gaps between what people do and what they say they do. Thus, even if our research interest is in the subjective meaning of the wood-chopping, we should use multiple methods, including a naturalistic observation component. Without this component, our picture of people’s lives is incomplete in a fundamental way. In a prefatory passage, Young and Willmott (1957 p. xxxi) warn the reader that ‘for the most part, we can only report what people say they do, which is not necessarily the same as what they actually do’. Yet what people actually do would seem to be a fairly central aspect of the enquiry. Thus, I determined early on that we would gather as many different kinds of data as we could, but that recording what people actually do in Neighbourhoods A and B would form a central plank of the research. Some of the most interesting insights came at the moments where the self-report and behavioural data pointed in opposite directions, as we shall see.

21My desire to give direct behavioural observation a central role was no doubt influenced by the fact that I also do research in animal behaviour, where for obvious reasons direct observation looms much larger. However, there was also a social science influence, coming from the urban studies of people like William H. Whyte. (William H. Whyte is not to be confused with his unrelated contemporary William Foote Whyte, also a social researcher. William Foote wrote a famous ethnography called Street Corner Society, which would have been a fine title for William H.’s book City: Rediscovering the Center. However, William Foote’s methodological approach was actually quite different.) William H. Whyte spent much of his career doing direct behavioural observation of how people actually used city environments, such as sidewalks and plazas, particularly in New York. Whyte’s work stands today, despite his — for my taste — insufficient quantification of the very interesting data he and his collaborators gathered. In City: Rediscovering the Center, Whyte notes that there were already many studies on how people use cities prior to his. He says:

Some of the studies were illuminating. Taken together they suffered one deficiency: the research was vicarious; it was once or twice removed from the ultimate reality being studied. That reality was people in everyday situations. That is what we studied. (Whyte, 2009 p. 5).

22I feel the same about the contemporary use of questionnaires, surveys, and interviews as social science methods of choice. To the extent that these are taken as assays of people’s behaviour, they are vicarious, once or twice removed from the reality they purport to represent. Thus, it is essential to also study directly the ultimate reality of people behaving in everyday situations.

Economic games

23Economic games were also part of the plan from early on. Economic games came into prominence particularly in the 1990s and 2000s. They are, in essence, standardized financial decisions. They are always incentivized with real money. Economic games gave economists a way to find out how people actually allocate a scarce resource in a social context, rather than relying on a priori assumptions about how an economic agent ought to do so. For example, in the classic Ultimatum Game, one player can propose a split of the stake (say £10) between himself and another player. If the second player accepts the proposed division, they both go away with their share, whereas if the second player rejects it, they both leave with nothing. The players are anonymous from one another and will not interact again. In a world of actors motivated only to maximize their financial gain from the interaction, the first player ought to offer the second player the minimum possible amount from the £10: 1 penny if the money can be broken down that small. The second player ought to accept this, as 1 penny is more than nothing. Thus, we should observe splits favouring the first player as extreme as £9.99 to £0.01.

24In fact, this is not the result we typically observe at all. Second players reject low offers, even though they hurt themselves as well as the other player by doing so. First players don’t make offers as low as they should — they tend to offer quite a lot. This may be in part due to their understanding that low offers may well get rejected, and they will lose everything. However, they tend to offer more than is necessary to maximize their expected return given how second players actually behave. In other words, rather than maximizing the monetary return from the interaction like economic automata, the players behave like people actually do in social situations, with pride, spite, conscience, and so on. (I will not make a cheap aside about this being a major discovery for economists.)

25The first wave of findings from economic games tended to be interpreted in pan-human terms. The game results were held up as showing how humans behave, which was often at variance with the way simple economic models suggested they ought to behave. This was of course rather a naïve conclusion, since usually each study was done on a single population of university undergraduates in a Western country; whether the patterns of behaviour would generalize to other people was not known. The second wave of economic game findings led to more nuance; they revealed that how people behave in economic games depends. Macro studies revealed differences in average behaviour between different populations (Henrich et al., 2005; Herrmann et al., 2008), whilst micro studies revealed that small changes to the ways the games were set up led to quite different outcomes (Bardsley, 2007; Cronk & Wasielewski, 2008; Haley & Fessler, 2005). Economic games thus clearly had potential as assays of prosocial and antisocial motivations in our project, both for investigating neighbourhood differences and probing the decision-making processes that might underlie such differences. We ended up carrying out two different rounds of economic games, the first led by myself and Agathe Colléony, and the second by Kari Britt Schroeder. In each case, the questions we wanted to ask led us to develop new variant games, as well as new protocols for administering them that would allow people to play them with their neighbours and in their homes.

26A major issue with economic games concerns their external validity. This has led to increasing discussion in the last few years (Bardsley, 2007; Burton-Chellew & West, 2013; Winking & Mizer, 2013). Do people really understand exactly what the contingencies of the game are? Would they behave in the same way if they did not know they were taking part in an experiment? Does a person’s behaviour within the game bear any relationship to how they behave in everyday life? Given that we were going to use multiple methods, we would have the possibility of relating patterns of game play to other social measures, both at the individual and neighbourhood level. Both correspondences and non-correspondences would be interesting.

Correlational and experimental approaches

27Social science is traditionally a correlational business, though people are increasingly interested in the potential for doing experiments. Correlational research means that you measure the pre-existing variation in something, say income or neighbourhood deprivation, and you see how it relates to variation in something else, say health. The Tyneside Neighbourhoods Project was basically a correlational study, with neighbourhood as the initial predictor variable, and a whole host of other variables as secondary or mediating predictors. As I hinted in chapter 1, though, one of the key potential advantages of working at the meso-scale is that it raises the possibility of doing at least some experimental work.

28It is important to be clear what experimental means here. Economic games are sometimes described as experimental, but this is a misnomer, since there is not necessarily experimental manipulation of any independent variable. The most famous results showing different economic game results in different cultural groups, for example, are purely correlational. When I say experimental work, I mean randomly assigning participants to a treatment that might produce a difference in a measured outcome — what in medicine and public policy is called a randomised control trial. The experiment in this sense is science’s blue-chip method for providing insights about causality rather than just association. We did make a little progress in developing experimental approaches of this kind. We carried out two large-scale experiments and a few smaller ones. The design of each of these experiments was quite complex, and will make sense only once I have presented the bulk of the correlational data. I will therefore say nothing more about these experimental studies until it is time to look at their results, in chapter 4, and in chapter 6.

Ethical considerations

29All empirical research raises ethical considerations, and the Tyneside Neighbourhoods Project was particularly ethically challenging. There were really two clusters of ethical issues: the first concerned the ethics of participation, and the second, the ethics of representation. The ethics of participation encompasses questions about what it was like to take part in the research: could any of the things we asked people to do cause them distress or harm, lead them to regret their participation, or make them feel aggrieved? These were, by and large, the easier issues to address. All of our research was scrutinized and approved by the Faculty of Medical Sciences ethics committee at Newcastle University. For observational studies, we only observed people in public settings where they could reasonably expect their behaviour to be observable anyway, and we never made any attempt to identify individuals. We were ready to be completely open about what we were doing if challenged, which we never were. For other kinds of studies, participants gave their written informed consent to participate, were fully debriefed, and of course were represented anonymously in all of our data.

30We aimed to not deceive people. For example, when gave a financial dilemma about how to divide £10 between self and an anonymous third party, the money was always real, and we really find an anonymous third party and give her whatever was due. We did admittedly flirt with the boundaries of deceit at times. Dropping your keys on purpose to see if someone will pick them up for you is a kind of misrepresentation. We satisfied ourselves that the interaction was fleeting and the burden very small. In Kari Britt Schroeder’s experiment described in chapter 4, we wanted to manipulate the social information that people had about what their neighbours thought, but we did not want to do this by making up false information. We found a solution whereby we gave them real information from real neighbours, but manipulated which neighbours they saw the information from. This achieved our experimental requirements whilst stopping short of deceit, though it could rather fairly be described as being economical with the truth.

31The issues of ethics of representation concern the possible negative consequences for participants and their communities of the way we portrayed them in our publications and presentations. If the issues around the ethics of participation were soluble, those around the ethics of representation were much more of a concern to me, and remain so. Rather than discussing these any further at this point, I will return to them at the end of chapter 7, once we have seen the data. It is partly a concern with the ethics of representation that has led me to anonymize Neighbourhoods A and B. Anyone prepared to do some digging would be able to work out more or less where they are, but their specific identities are not put in the centre of the spotlight.

32In the interests of full disclosure I should also say something about funding. I began the project with no funding beyond the fact that Newcastle University pays my salary. Students and other helpers mainly gave their labour voluntarily or as part of their programme of study. North Tyneside Council provided benefits in kind in the conduct of the School Survey described below. At later points, we were able to draw on resources from Newcastle University’s funding of Gillian Pepper’s PhD, and especially from a fellowship grant from the US National Science Foundation to Kari Britt Schroeder. The funders played no role in decisions about what to do or how to publish the results.

Description of datasets

  • 1 An archived version of the dataset, preserving the data in the form it was at the time publication (...)

33Subsequent chapters of the book are organized thematically; each chapter draws on several different datasets, and each dataset is referred to in several different chapters. In the remainder of this chapter, then, I briefly describe each of the main datasets that we produced (as mentioned, leaving aside for now the two experimental datasets to be discussed in chapters 4 and 6). I will not repeat information such as sample size and method of collection when presenting results in future chapters; to find such information, the reader should return to this section. The citations shown as footnotes to the headings introducing each dataset refer to the published sources where the methods for that dataset are fully described. The raw data are freely available via an Open Science Framework project (Nettle, 2015) that can be found at​ys7g6.1

Observational Dataset 12

  • 2 Nettle, Colléony, & Cockerill, 2011; Nettle, 2011a, 2011c.

34The first and largest dataset involved direct behavioural observation on the streets of each neighbourhood. It was gathered, by me, between April 19th and July 8th 2010. The idea was to record everything happening on the streets for one whole spring/summer day, 9am to 9pm. Clearly, it would be impossible and undesirable to get all of the data on a single day. Instead, I divided the day into 24 half-hour segments (9:00–9:30, 9:30–10:00, etc.). I then chose a time segment pseudo-randomly (i.e. starting neither at the beginning of the day and working forwards, nor at the end and working backwards), went to one neighbourhood and recorded for that time segment. The next weekday I went to the other neighbourhood and recorded at the same time. Sometimes a weekend intervened, but the median delay between observing a time segment in the first neighbourhood and in the second was one day, and the maximum was four days. The order of the two neighbourhoods was counter-balanced across time segments. This process continued until all time segments had been filled in. All five weekdays were represented at least three times in both neighbourhoods. The observation period was outside the school holidays.

35The first ten minutes of each segment were spent walking a transect up one side of the main street and down the other, the starting end varying randomly. The remaining twenty minutes were spent taking a random walk around the residential streets. I wore an unobtrusive digital voice recorder to capture the data to be transcribed later. With practice I developed an efficient numerical code for doing this; there was not a single instance where I felt that anyone noticed my sotto voce muttering.

36My focus was on people — what they were doing and what kind of social groups they were doing it in — though I did capture some other information. For each social group that came into my sight and was not inside a building, I recorded the number of men, women, children, and babies. I recorded whether any of the individuals were smoking, drinking an alcoholic drink, or running. Social groups were defined on the basis of interaction or common movement; this was rarely difficult. A person on their own was classified as a social group of one. Individuals re-encountered within the same time segment were not recorded again. I additionally noted instances of some miscellaneous categories: babies crying, open front doors, people with walking sticks or wheelchairs, the dropping of litter, police patrols, and things on fire. The transcribed dataset contained data on 4123 social groups in Neighbourhood A, and 3773 in Neighbourhood B. In terms of people-observations, this equated to 5884 in Neighbourhood A and 6757 in Neighbourhood B.

Observational Dataset 23

  • 3 Hill, Jobling, et al., 2014.

37Observational Dataset 2 was gathered by Jessica Hill with the assistance of Ruth Jobling in the summer of 2013. Rather than sampling every minute of the day as in Observational Dataset 1, they sampled the first 30 minutes of every hour between 9am and 6pm inclusive (that is, 9–9:30, 10–10:30, etc.). They were able to complete this in two weeks, on some occasions gathering the data from the two neighbourhoods simultaneously on the same day, on others completing the time segment in the second neighbourhood the day after the first. Again, their sampling occurred during weekdays outside the school holidays.

38Jessica and Ruth largely used the same protocol as in Observational Dataset 1. They observed a total of 3665 people across the two neighbourhoods. Their innovation was to code ‘new’ interactions. A ‘new’ interaction was defined as when an individual or group engaged in conversation with another individual or group they had not previously been interacting with. This could be anything from a fleeting greeting to stopping in the street to talk to someone. Jessica was interested in social ties, and this behavioural measure would give us some insight into the extent to which people in each neighbourhood knew others nearby well enough to talk to them. I am not sure whether Jessica was aware of this at the time, but her focus on new interactions relates closely to a dataset in Young and Willmott (1957 p. 84) where a resident recorded how many people that she knew she bumped into on the streets of Bethnal Green as she went about her daily business. Young and Willmott too saw this as an index of the density of social ties in the neighbourhood beyond the household. The findings from Observational Dataset 2 will be discussed in chapter 3 in particular.

Street Ages Dataset4

  • 4 Nettle, Coyne, & Colléony, 2012

39As a result of gathering Observational Dataset 1, we became interested in knowing which adults were using the streets, and in particular how this related to their age. To investigate this, Rebecca Coyne (in July and August 2010) and later Agathe Colléony (in April 2011) walked transects through the main streets of each neighbourhood, recording the sex and estimating the age of every person they passed. Unlike Observational Datasets 1 and 2, the Street Ages Dataset was gathered during the school holidays. Rebecca and Agathe each made six visits to each neighbourhood, alternating neighbourhoods and balancing approximate times of day. Rebecca coded for 15 minutes each time and Agathe for 30 minutes, giving a combined sample size of 2533 age-ratings. There were no significant differences in the distributions of ages in Rebecca and Agathe’s data, and they were pooled for analysis.

40Rebecca and Agathe’s data only tell us about their judgements of age; we do not know the actual ages of the people that they saw. Previous research has indicated that visual judgements of age are fairly accurate (George & Hole, 1995). We knew from the 2001 census data what the ‘true’ distribution of chronological ages had looked like just a few years earlier, and so the null hypothesis was that their judged distribution of ages would be the same. Any systematic departure from the census distribution in one or both neighbourhoods would be telling. It might suggest neighbourhood-specific biases among those who, of all the people living there, used the streets.

Social Survey 15

  • 5 Nettle et al., 2011.

41Social Survey 1 was the first major self-report assay we used in Neighbourhoods A and B. It also provided the vehicle for the first economic game, which will be described below. Its focus was on social relationships and prosocial behaviour. Social Survey 1 was largely implemented by Agathe Colléony. During the spring of 2011, we chose names and addresses in each neighbourhood from the electoral register, and sent a survey with a cover letter asking the addressed recipient if they would be so kind as to fill the survey out. A stamped return envelope was enclosed, and the respondent would also receive £10 for their trouble. 124 people returned surveys; 74 from Neighbourhood A and 50 from Neighbourhood B. This was after we had boosted the sampling in Neighbourhood B, since return rates were only 21.7% in Neighbourhood B as compared to 43.5% in Neighbourhood A.

42The survey contained a number of demographic and attitudinal questions. Central to it was a measure of neighbourhood social capital. The concept of social capital is used somewhat variably in the literature (Halpern, 2005). Originally, it referred to resources that are embedded in the individual’s social network; that is, having other people who know you, are well disposed to you, and are able to help you get things done. It is often used rather more loosely to denote positive feelings about social relationships or high trust within some group.

43Our six key items to measure neighbourhood social capital included four questions on 7-point ratings scales: how much did the respondent trust other people in the neighbourhood, how much did they feel people in the neighbourhood looked out for one another, how well did they know their neighbours, and to what extent did they feel they had good friends living locally. The remaining two items were crude social network measures. First, we asked respondents to list all of the people they had contacted for social reasons within the last two weeks. They could use anonymizing initials for this, since we weren’t interested in the names, but rather in how many of them there were. Second, we asked respondents to list all the people they could turn to if they had a problem. Human social networks are hierarchically nested; every person is at the centre of several concentric circles, with the total number of people increasing as you move from the centre outwards, and emotional closeness and interaction intensity decreasing (Hill & Dunbar, 2003). Our two network measures were designed respectively to provide a rough measure of the size of the sympathy group, as the inner circle of frequently interacting friends and relatives is usually known, and the support clique, the smaller core of people one would rely on for the most serious needs. Previous studies have found that sympathy groups tend to number 10–12 individuals, and support cliques around 5 (Dunbar & Spoors, 1995). The means in our Social Survey 1 were correspondingly 11.65 and 6.23. Both measures had a right-skewed distribution, with some sympathy groups as large as 35 and support cliques as large as 32, but most much smaller. This is typical of previous studies. To a greater extent than some previous studies, we found women tended to have larger sympathy groups and support cliques than men (means for sympathy groups 13.90 vs. 9.01; for support cliques 6.84 vs. 5.16). Our six social capital items were rather heterogeneous in meaning, but, usefully, they were all substantially correlated with one another. This allowed us to combine them into a single index of social capital. The index had a Cronbach’s α statistic of 0.81, which means that all six items correlated well with each other. Higher scores on this index equate to having a larger sympathy group and support clique, and to giving higher ratings of trust, indicating that people look out for each other, that the respondent knows neighbours, and that she has good friends living locally.

44The first economic game was an integral part of Social Survey 1, but we did not want to draw attention to it as a measure of the study. This was because of the concerns that have been repeatedly raised about the effects of knowing that one is participating in a cooperation experiment on economic game behaviour (Bardsley, 2007). We therefore made our game relatively surreptitious. The respondent was told at the end of the questionnaire that she had finished and had earned £10 as a thank-you. A separate payment form then asked the respondent to provide the name and address to which she would like the money delivered. However, it also gave her the option of transferring some or all of the £10 to someone else rather than receiving it herself. This was in effect a version of the Dictator Game, the well-studied economic game in which an actor decides how a sum of money should be divided between himself and another individual, with that other individual having no say in the matter and no comeback. The Dictator Game is the simplest possible index of prosocial motivation: how much money will you keep for yourself, and how much will you transfer to the other party? We will return to exactly what the versions of the Dictator Game we used were, and the motivation for them, in chapter 3. Once payment forms were received in the post, we hand-delivered the money, in cash, to the relevant addresses within one week.

Social Survey 26

  • 6 Schroeder, Pepper, & Nettle, 2014.

45Social Survey 2 was envisaged and designed by Kari Britt Schroeder. She implemented it with assistance from Gillian Pepper and, to a much lesser extent, myself during 2012 and early 2013. Again, names and addresses were randomly chosen from the electoral register and the survey was delivered by mail with a return envelope. Care was taken not to resample individuals from Social Survey 1 and (for reasons to do with the associated economic game) not to choose immediate neighbours. It proved challenging to get a sufficient sample in Neighbourhood B given the constraints, and we extended the boundary of Neighbourhood B a few streets beyond its original limit. The extension area was similar to the original area on the relevant measures, such as the Index of Multiple Deprivation. The final sample size was 260, well balanced across the two neighbourhoods (A: 133; B: 127). Another 168 participants took part in an experimental add-on, to be described in chapter 4.

46Whereas the focus of Social Survey 1 was on positive social resources and prosociality, the focus of Social Survey 2 was on the negative side of social interactions: breaking rules and harming others. The survey asked about a number of things, again including trust. This time we distinguished between trusting people in general and trusting people you know well. This had the potential to differentiate parochial social resources (‘there is a small group of people I know well on whom I can rely’) from more generalized trust (‘I know that when I need something, anyone will help me out’). These could pattern differently across the two neighbourhoods. In addition, Kari investigated perceived norms of social behaviour. We will examine in more detail exactly how she did this in chapter 4. Kari’s particular interest in perceived norms was in how they would relate to decisions in the economic game associated with Social Survey 2, to which I now turn.

47The economic game associated with Social Survey 2 was a lot more complex than a Dictator Game. It thus could not be surreptitious, and a substantial fraction of the survey consisted of the questions involved in it. Given the focus on antisocial behaviour and its determinants, the game allowed some people the opportunity to behave antisocially, and others to the opportunity to administer justice by punishing that antisocial behaviour if they wished. The game had three roles, which will be explained in chapter 4. One third of the surveys sent out assigned their readers the role of player 1, the next third that of player 2, and the rest player 3. Specific triads were formed at random from surveys returned at around the same time from each neighbourhood; no respondent knew who the other members of their triad were. As before, all monies arising were delivered in cash to the respondents’ addresses within one week.

Police Crime Dataset7

  • 7 Nettle et al., 2011.

48As we were interested in the topic of crime and antisocial behaviour, Agathe obtained data on all incidents reported to Northumbria Police between December 2010 and March 2011 falling within the study neighbourhoods (this information can be freely found via As well as tabulating the number of incidents overall, she was able to break them down into different categories of crime or antisocial behaviour, as we will see in chapter 4.

Assorted assays of prosociality8

  • 8 Nettle et al., 2011.

49During 2011, we also undertook a number of other field assays of prosocial behaviour. The first of these used the lost letter paradigm, a classical unobtrusive measure of passer-by willingness to do an anonymous kindness. A stamped, addressed letter is left on the pavement in the general vicinity of a postal box. The measure is simply the proportion of letters that ever find their way to their recipient. The probability of return has been shown to vary, and to depend on a number of micro factors, such as the implied characteristics of the recipient, as well as macro factors, such as the community where the letter is left (Levine, Martinez, Brase, & Sorenson, 1994; Milgram, Mann, & Harter, 1965). We left 22 letters in each neighbourhood on rain-free mornings over the course of several months.

50In addition to the lost letter, we employed three other helping assays from classic social psychological literature (Levine et al., 1994): dropped object, asking directions, and making change. In dropped object, 24 volunteer field assistants walked along the street and, seemingly inadvertently, dropped a small object such as keys, a glove or a pen around 10m in front of a pedestrian approaching in the other direction. We scored whether or not the target person helped by drawing attention to or picking up the object. The same volunteer field assistant then dropped the same object in the other neighbourhood. There were a total of 60 drops in each neighbourhood. Characteristics of both dropper and target, such as age and sex, were recorded. Asking directions and making change were very similar paradigms, except that the volunteer field assistants (the same ones as for dropped object) asked either for directions to a nearby hospital (there was a suitable one close to both neighbourhoods) or to make change from a large-denomination coin. Trials were scored as minimal or no help versus substantial help. There were 30 trials in each neighbourhood for each assay.

51Field assays of social behavioural such as these have been extensively used in the past. Lost letter and making change are becoming obsolete due to changes in technology (what do you need a 10p coin for anymore?). Even asking for directions seems a little unnatural when most young people carry smartphones. Nonetheless, we were able to obtain datasets using all of these, and they will be discussed in chapter 3.

School Survey9

  • 9 Nettle & Cockerill, 2010.

52The final dataset is the only one that was not gathered in Neighbourhoods A and B, and also the only one where we heard from young people. In 2009, Maria Cockerill and I surveyed 1046 school students aged 9–15 from various parts of Tyneside other than Neighbourhoods A and B. They were from eight different schools, and their residential addresses could be assigned to eight different neighbourhoods, for each of which we obtained an Index of Multiple Deprivation. The School Survey proved useful in a number of ways. It provided some data from very deprived neighbourhoods of Tyneside other than Neighbourhood B (along the Eastern riverside); it provided some data from an affluent area of Tyneside other than Neighbourhood A (in the East of the conurbation); and it provided data from some intermediate neighbourhoods too, rather than just having the two extremes as in Social Surveys 1 and 2. Thus, it helped to generalize beyond the main two study neighbourhoods and across the spectrum of deprivation. Moreover, by surveying children at different ages, we could get some ideas about how the characteristic attitudes and feelings of deprived neighbourhoods develop through childhood. I say some ideas rather than any stronger inference, because this was not a truly longitudinal study; rather, we surveyed cross-sections of Tyneside children at several different ages. Nonetheless, it did allow us to identify which patterns were apparent in which age groups.

53The School Survey was a comprehensive questionnaire whose original aims were to do with the psychological consequences of deprivation and the factors related to early childbearing. It contained questions about the respondents’ ideal age of parenthood and anticipated life expectancy. It also contained measures of the neighbourhood’s perceived safety and the respondent’s perceived support from their family. Finally, it asked about trust, linking it the social capital work in Social Surveys 1 and 2.

54I have now described all of the main datasets that made up the Tyneside Neighbourhoods Project. With this preliminary work done, we can turn to the substantive findings. These are laid out in chapters 3–6. Chapters 3 and 4 respectively examine prosocial and antisocial behaviour amongst adults. Chapter 5 explores how neighbourhood differences vary with age. Chapter 6 focuses on questions of psychological mechanism, asking how variation in experience might lead to variation in attitudes and hence to variation in social behaviour. Chapter 7 concludes with my overall reflections on what we learned.


1 An archived version of the dataset, preserving the data in the form it was at the time publication, is available at

2 Nettle, Colléony, & Cockerill, 2011; Nettle, 2011a, 2011c.

3 Hill, Jobling, et al., 2014.

4 Nettle, Coyne, & Colléony, 2012

5 Nettle et al., 2011.

6 Schroeder, Pepper, & Nettle, 2014.

7 Nettle et al., 2011.

8 Nettle et al., 2011.

9 Nettle & Cockerill, 2010.

Table des illustrations

Légende Figure 2.1 The meter of deprivation for an address in Neighbourhood A (left) and one in Neighbourhood B (right). Image from, subject to the terms of the Open Government Licence.
Fichier image/jpeg, 42k


Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search