© Daniel Nettle, CC BY http://dx.doi.org/10.11647/OBP.0084.03
Don’t push me ‘cos I’m close to the edge;
I’m trying not to lose my head...
This chapter deals with prosocial behaviour; that is, behaviour that helps someone else. The beneficiary can be specific, as when one aids a friend in need, or more diffuse, as when one cleans up the street, benefitting anyone who happens to use it. The behaviour is prosocial in either case. Note that a behaviour being prosocial does not mean self-interest must be absent. Many of the things people do together involve both parties benefitting, and these would still be classed as prosocial. Even where one party does not obviously benefit in the short term, she may in the longer run. For example, donating money to a community organization appears on the face of it to be completely altruistic, but the donor may obtain long-term favours or reputational benefits by doing so. We will not be concerned here with whether prosocial behaviour can always be shown to be a form of enlightened self-interest, but rather with whether and why patterns of prosocial behaviour differ across the two neighbourhoods.
There are many kinds of prosocial interactions woven into the life of any community. Indeed, community is sometimes defined in terms of social interactions beyond the scope of the household but short of the scope of formal or governmental institutions. Prosocial interactions beyond the household relate closely to the notion of social capital: resources and investments embedded or distributed in a social network. You may recall from chapter 1 that we have two broad and opposing narratives about deprivation and prosociality available to us. The Kropotkinian narrative leads us to expect greater prosociality in Neighbourhood B, and the Mountain People narrative leads us to expect greater prosociality in Neighbourhood A. The bulk of this chapter is devoted to examining our various datasets in the light of these opposing sets of expectations.
Round one: Social interactions in the streets
We should be able to learn a lot about people’s social behaviours by seeing how they associate on the streets. First, then, let us examine the patterns in Observational Dataset 1. The simplest question we can ask is: how many people are on the street? The upper plots of Figure 3.1 show the data by time of day, separately for the main and residential streets. As you can see, the two main streets follow roughly the same pattern across the day, though main street A is busier than B: people come out into the main street during commercial hours to do what they need to do in the shops and banks; by evening, the main streets have become empty of people. The more interesting difference is in the residential streets. In Neighbourhood A, there is a steady flow of pedestrians in the residential streets through the day, but after 17:00, it drops off precipitately. People have gone into their houses. In Neighbourhood B by contrast, they continued to be outside their houses in considerable numbers until I stopped recording data at 21:00. In fact, the residential streets of Neighbourhood B are busier in the evening than at any point during the day.
Part of the relative excess of people on the residential streets of B in the evening is attributable to the greater prevalence of children ‘playing out’, a phenomenon to which we return in chapter 5. But the focus of this chapter is on adult social behaviour, so let us ignore children for now. The lower panels of Figure 3.1 plot the same data, but for adults only. We see a similar pattern; after around 18:00, there are essentially no adults on the residential streets of A, whereas they continue to be on the residential streets of B in considerable numbers until the end of recording at 21:00. (Recall that these data were gathered in the summer when dusk is considerably later than 21:00.)
Not only are the adults in Neighbourhood A inside their houses in the evening; they have closed their doors. Figure 3.2 plots the number of front doors open by time of day in the residential streets of each neighbourhood. As is clear, the residents of B, as well as being more likely to be outside, are much more likely to have their front doors open. The meaning of this is hard to determine; their houses are smaller on average, for one thing. However, the door being open at least allows for the possibility of social interaction, and it does suggest that people are not secluding themselves, embattled, from the danger of encounter with their neighbours. I have had occasion to wander into houses with open doors in Neighbourhood B, in search of someone to give money to, including on occasions when the resident turned out to be elsewhere despite the door being open.
So adults are more likely to be on the street in Neighbourhood B, especially in the evenings. What are they doing there? One thing they are doing that you do not see in Neighbourhood A is street drinking. The rising cost of drinking in pubs, coupled with people’s relatively straitened means, has largely destroyed that traditional working-class institution in Neighbourhood B. However, alcoholic beverages are inexpensive in shops, where they are bought for off-the-premises consumption. People consume beer in particular in impromptu gatherings in gardens, parks, and street corners. It is not uncommon to see sofas and dining chairs dragged outside to facilitate these gatherings, and it is my impression that they typically involve residents of multiple households. In Observational Dataset 1, I observed 38 instances of street drinking in Neighbourhood B as against 1 in Neighbourhood A; these involved a mean of 3.21 people (standard deviation 1.89). Most people on the residential streets of Neighbourhood B were not involved in street drinking. Nonetheless, there was plenty of social interaction going on, as the street drinking parties demonstrate.
These evening gatherings exemplify a more general pattern in Observational Dataset 1: adults are less likely to be alone in Neighbourhood B than A. Figure 3.3 shows the mean number of adults in a social group by time of day and neighbourhood. Groups consisting wholly of children are excluded. As you can see, the mean is fairly close to 1 at all times. Adults are always moderately likely to be alone (or at least, the lone adult) as they go about their neighbourhoods. However, there is a clear neighbourhood difference; adult group sizes are consistently larger in Neighbourhood B, and this is driven by adults in B, at all times of day, being rather more likely to be with another adult.
It is very unlikely that this pattern simply reflects adults going outside with the people they live with. As you may recall from chapter 2, in Neighbourhood B, there are many more households headed by a lone adult than is true in Neighbourhood A. Thus, if people go out of the house with the people they live with, we should expect to see smaller adult group sizes in B than A. The fact that we observed exactly the opposite suggests strongly that adults in B are more heavily involved than those in A in day-to-day social interaction with other local adults who are not members of their households.
We can drill down further into what kinds of associations are driving the differences. Figure 3.4 shows the relative prevalence of men alone, women alone, male-female couples, and other types of group across the two neighbourhoods in Observational Dataset 1. The other groups of interest are multi-male groups (several men but no women), multi-female groups (several women but no men), and mixed groups (more than two adults, both sexes present). As the figure shows, adults of both sexes are less likely to be on their own in Neighbourhood B, though the neighbourhood difference is much stronger for women than men. The reduction in aloneness is made up for by a relative increase in same-sex associations for both sexes, and also in more mixed-sex groups of more than two, in Neighbourhood B. Only some of this is driven by the street drinking groups. There are also, for example, many more groups consisting of several females and their young children out and about in Neighbourhood B (93 such groups observed in B against 21 in A). Male-female couples are about equally prevalent in the two neighbourhoods, though this in itself is something of a surprise, since there are many more lone-female-headed households in Neighbourhood B than A. We might thus have expected fewer opposite-sex couples on the street, but this is not so.
Observational Dataset 2 provided similar information to Observational Dataset 1, albeit gathered three years later. Jessica and Ruth confirmed many of the same patterns I had seen (see Hill, Jobling, et al., 2014 for a detailed analysis). Adult group sizes were significantly larger in B than A. This was again driven by adults of both sexes being significantly less likely to be on their own. Indeed, in Observational Dataset 2, the odds of an adult being with another adult were twice as high in B than A. The novel addition of Observational Dataset 2 was Jessica and Ruth’s examination of ‘new’ social interactions. To recall, these were defined as instances when an individual or group in the street engaged in conversation with another individual or group they had not previously been interacting or moving with. ‘New’ interactions give us a metric of how often someone in the neighbourhood bumps into someone else they know well enough to want to say hello. Jessica and Ruth observed 62 ‘new’ interactions in A and 120 in B during their period of sampling. Correcting for the different total numbers of people observed, the odds of an adult engaging in a ‘new’ interaction in B were 2.3 times higher than in A. The difference was particularly marked in the early evening (as we approach the time of street socializing in B), when they saw 37 ‘new’ interactions in B and just 6 in A (Figure 3.5).
The patterns in Observational Datasets 1 and 2 bring to mind Kropotkin—and Young and Willmott’s (1957) depiction of working-class Bethnal Green—much more than they bring to mind the Mountain People narrative. In the deprived Neighbourhood B, it seems, adults seek out other adults and interact socially with them. They are more likely to go around the neighbourhood with someone else, especially in the case of women; their front doors are more likely to be open; there are more informal social gatherings on the street; and they stop and greet each other more as they move around. We do not know what social interactions are going on inside the houses in either neighbourhood, or being transacted via phones and email. This is a big limitation. However, the difference in life on the streets is fairly marked, and it would not be totally unreasonable to assume that it is roughly representative of the difference in social interactions of other kinds too. We have no sense from the observational findings of what the quality of social relationships is, or what kinds of prosocial services are provided through them. Nonetheless, it seems that Neighbourhood B is the more social place, where interaction between adults is much more pervasive.
Round two: Self-reported social capital
Social Survey 1 asked people six key questions about social networks and social capital (to recall, these concerned trust within the neighbourhood, knowing neighbours, having good friends living locally, people in the neighbourhood looking out for one another, social contacts, and support cliques). Given the results of the previous section, the clear expectation is that residents of B will report that they know their neighbours better and have more good friends living locally than residents of A. All those open doors and street greetings might mean than they trust their neighbours more in B than A, and feel more strongly that people look out for one another. When asked about social contacts and social support, residents of B might well produce a greater number of contacts than those of A, though this prediction is less clear than the previous ones, because these last two survey questions do not exclusively concern social interactions within the neighbourhood. Residents of A might have large social networks and engage in extensive social interaction, just geographically elsewhere. Nonetheless, the result that would be most consilient with the observational data, as well as easiest to make sense of in a Kropotkinian kind of way, would be higher social capital in B than A by all measures.
The results are exactly the opposite. There are substantial neighbourhood differences on all the social capital measures, but social capital is in all cases higher in A than B (Figure 3.6). If we take the right-hand two pairs of bars in Figure 3.6 first, we see that respondents from B have smaller numbers of people in their sympathy groups and support cliques than respondents from A. There is no necessary contradiction between this and the results we have already seen showing more spontaneous social interaction in B. Perhaps people in B concentrate on fewer, deeper relationships, whilst people in A have more numerous but less intense social contacts. The first four pairs of bars, though, present a paradox: although the observational data shows that there is demonstrably less interaction between neighbours in A, respondents there trust each other more, feel they know each other better, feel that they have better friends locally, and feel more strongly that people in their neighbourhood look out for one another.
Before turning to the question of how to reconcile these findings with those from the direct observational data, we will just check that they are reproducible in other datasets. In Social Survey 2 (which, remember, was a different and larger set of respondents), Kari also had questions about trust. Trust is a key marker of social capital, and in Social Survey 1, the trust response was highly correlated with the responses to the other social capital questions. In Social Survey 2, Kari distinguished between social trust, which is trust of people you do not know well, and personal trust, which is trust of people you do know well. Both were measured on a 10–point scale. The results were much the same as those from Social Survey 1, in that there was significantly lower trust in B than A. Interestingly, this applied to both types of trust. For social trust, the mean in A was 5.00 (standard deviation 1.86), compared to 3.53 (standard deviation 2.05) in B. For personal trust, the mean in A was 8.61 (standard deviation 1.24), compared to 7.97 (standard deviation 1.88) in B. Thus, in Neighbourhood B, the pattern of relatively low trust encompassed not just strangers, but also people the respondent knew well. This is a potentially important observation, to which I will return.
The self-report data, then, all point in the same direction. In the deprived Neighbourhood B, people trust each other less, feel they know each other insufficiently, and generally report feeling that they have less capital embedded in their social networks than people from A. These findings are consistent with the large-scale patterns detected by Haushofer (2013), and they point much more toward the Mountain People narrative and away from Kropotkin. Moreover, the neighbourhood differences are large. One way to illustrate this is to compare the data from Social Survey 1 to the World Values Survey data from other populations. The World Values Survey uses its questions about trust to compute a country-level trust index; this index is on a scale where 0 represents total distrust, 200 total trust, and 100 an equal balance of trust and distrust. With a little kneading, I can use our data to produce the same index for Neighbourhoods A and B as if they too were countries.
Figure 3.7 compares Neighbourhoods A and B to a selection of World Values Survey countries. Neighbourhood A scores a little less than 100. This puts it somewhat below the remarkably trusting Scandinavian countries, but slightly above many of the major industrial economies such as Germany and the USA. Its score is very similar to Canada’s. Neighbourhood B, by contrast, scores about 28. This puts it in the company of populations in the developing world; ahead of Kenya, about equal to Zimbabwe, but some distance below Burkina Faso and Colombia. This seems like a difference big enough to concern us—a difference that is socially and not just statistically significant.
So far, we have one set of measures lining up on the side of each of our two narratives: direct behavioural observation for Kropotkin and self-report surveys for the Mountain People. Perhaps the third type of measure, economic games, will help adjudicate between them.
The economic games for studying prosociality that we implemented in Neighbourhoods A and B were attached to Social Survey 1. The respondents had been told that they would be recompensed £10 for their trouble. At the end of the survey, they found a form asking them to specify how that £10 would be given out. They could choose some amount to be delivered in cash to their own address, with the remainder to be delivered to someone else (see below on exactly who the someone else was). We were thus implementing a version of the Dictator Game. Because of concerns about demand characteristics (people making particular decisions because they know they are taking part in a study of prosociality for which the researcher has certain expectations), we tried to make our game surreptitious. That is, the participants were not told that this decision about the disposition of the £10 payment was itself part of the study. We do not know the extent to which they figured out that it was, and whether this differed between neighbourhoods.
The Dictator Game is perhaps best thought of as a measure of the participant’s motivation to make a social investment with a monetary resource, rather than keeping it for private use. In general terms, then, we could make predictions about neighbourhood differences either way. The Kropotkinian view, the direct behavioural observation results, and previous studies on socioeconomic variation in generosity (Piff et al., 2010) clearly suggest that transfers to the other party would be higher in B than A. The Mountain People hypothesis, and the data on social capital and trust, points firmly in the opposite direction.
In fact, we implemented three subtly different versions of the Dictator Game, each with a separate group of participants. In the first version, the standard condition, the instructions allowed the respondent to specify an amount in pounds, including zero, to be delivered to a randomly-chosen name and address in their neighbourhood. The balance would be delivered to the respondent’s own house. We stressed that the respondent would remain anonymous whatever decision she made, and that she would also not know the identity of any beneficiary. Thus, this situation equates fairly closely to a laboratory Dictator Game as usually performed.
In the second version, the friend condition, we allowed the respondent to nominate the recipient, with the condition that the person must be someone in the neighbourhood. In addition, we explained that we would double any amount transferred. The motivation for this non-standard version of the Dictator Game was the hypothesis that people in Neighbourhood B might be less inclined to help someone in general, but more inclined to make a social investment in a particular individual they were close to (that is, their cooperation might be more parochial, centred on people they knew well). Allowing the respondent to nominate the recipient had the potential to reveal the existence of close prosocial ties within the neighbourhood. Moreover, the doubling of the stake made cooperation relatively attractive: the respondent could nominate a friend over the road and transfer £10, which would be turned into £20. The two could then meet up and split the money evenly and the respondent would have lost nothing. This would work fine as long as he or she had a trusted partner suitable for the endeavour living nearby. We predicted we might see greater transfers in Neighbourhood B than Neighbourhood A in this second condition, even if no differences (or the opposite difference) could be seen in the first, standard condition Dictator Game.
In the final version, the charity condition, any money transferred would again be doubled, but the recipient a locally-based charity (specified by us) that was well-known in both neighbourhoods. Thus, this condition featured doubling like the second condition, but without the respondent being able to choose the recipient.
The results are shown in Figure 3.8. In Neighbourhood A, the results for the standard and friend conditions are much the same as those observed in many studies of developed Western populations; many but not all people transfer something, and the mean transfer is of the order of 40–50% of the stake. (Rather surprisingly, transfers were no higher in the friend than the standard condition.) The charity condition produced greater generosity still.
Neighbourhood B looked quite different; transfers were dramatically lower across the board. In fact, in Neighbourhood B, almost nobody transferred anything in the standard or friend conditions (2 people out of 33, against 21 out of 45 for Neighbourhood A). The results were particularly striking for two reasons. The first is that (to our disappointment) it did not matter which version of the Dictator Game we looked at; the neighbourhood difference was always in the same direction. People in B were just less likely to make transfers regardless of which condition they were in. Second, the differences were large. To put them into context, a famous cross-cultural study including societies on different continents and at radically different stages in economic development showed what they characterized as dramatic variation in, inter alia, Dictator Game behaviour (Henrich et al., 2010). We can overlay the results from our standard condition on theirs (Figure 3.9). As you can see, the difference we observed across neighbourhoods within this single city was much more dramatic than the largest they documented by comparing people from rural Missouri and from small-scale societies in Papua New Guinea or East Africa.
There are some important technical differences between the studies. Henrich et al. (2010) adjusted the stake to local resource values. We did not do this and, ideally, the stake would have been larger in A than B to make its relative value more comparable. However, the general consensus in the behavioural economic literature is that decisions in games such as these are relatively impervious to variation in the size of the stake (Carpenter, Verhoogen, & Burks, 2005; Forsythe, Horowitz, Savin, & Sefton, 1994). On the other hand, in many ways, our study was much better controlled than that of Henrich et al. (2010), since we were able to deliver exactly the same instructions, in the same language and the same way, in our two sites. Thus, the magnitude of the difference does seem noteworthy, and its direction decidedly unfavourable to Kropotkin and reminiscent of the Mountain People.
The neighbourhood difference in Dictator Game transfers was tightly bound up with the neighbourhood difference in self-reported social capital. That is, people in B tend to have lower social capital than those in A; people with lower social capital transferred less in the Dictator Game; and this explains a sizable portion of the neighbourhood difference in Dictator Game transfers. It makes sense that people with lower social capital would transfer less. Presumably we are willing to invest resources in our social network to the extent that we feel we have a potentially fruitful social network to invest in; the poorer we feel it is, the more we will want to keep our resources to ourselves.
This section is devoted to trying to make some theoretical sense of the data presented so far. Behavioural observation shows people interacting more in the streets, socializing more, being alone less often, showing signs of knowing one another, and having their front doors open more in Neighbourhood B than A. On the other hand, people in Neighbourhood B say they feel that they don’t know their neighbours as well, they don’t trust them as much, and they don’t feel they have such supportive social networks. Moreover, when you give them, through the Dictator Game, a chance to make a social investment, they are much less keen to do it, even when they can choose the beneficiary. How can we make sense of this pattern?
I am going to assume that each type of measure has some value. That is, I am not going to dismiss one kind of measure (e.g. self-report) as just less reliable than another (e.g. direct observation), and therefore claim that one gives us the ‘true’ picture and the other, junk. Rather, I will assume that all kinds of measures are giving us real information about something, but that something might not be the same in each case. A corollary of this is that both of the Kropotkin and Mountain People narratives are probably capturing something about real life.
To make sense of the data, I would like to build on a number of propositions, some of which I can evidence later and some of which I cannot, though they are plausible. The first of these is that people in Neighbourhood B tend to be, in a sense I will explain shortly, closer to the edge than those in Neighbourhood A. What do I mean by close to the edge? There are various domains of life where, as things get increasingly unfavourable, it becomes worthwhile to do things that would, the rest of the time, be extremely unwise. This prediction is set out with particular clarity in so-called risk-sensitive foraging models in animal behaviour (Stephens, 1981). An animal with sufficient energy reserves should prefer a food patch that provides a small sure yield over one that sometimes provides a bonanza and often provides nothing. However, when that animal’s reserves get so low that it is close to the edge of starvation, only a bonanza would be enough to get it back up to safety. It should thus go to the often-bad-but-occasional-bonanza patch; it will probably get nothing and starve, but it is so close to the edge that it will certainly starve anyway if it goes to the safe patch. So, a gamble that might normally be a really bad idea becomes attractive when you are close to the edge. Importantly for our purposes, the models suggest that there is a point close to the edge of starvation where the individual’s behaviour should flip, from strongly preferring the safe option to strongly preferring the gamble. (The theoretical predictions from risk-sensitive foraging models are clear. However, it is not clear that animals’ behaviour actually conforms to these predictions, though people have tried hard to find evidence that they do [Kacelnik & Bateson, 1996; Kacelnik & El Mouden, 2013]. That however is a story for another day.)
We can generalize the idea of people’s behaviour flipping when they are close to the edge (Nettle, 2009). It is not just starvation that might cause such a flip, but financial crisis or any other form of existential threat. For example, it is normally a really bad idea to take things from your neighbours, as they will probably provide you with more benefits over the long run through friendship than the short-term resource you might get away with (and, needless to say, you will lose their friendship if you take things from them). However, as you get close to the edge of financial crisis, you begin to need money just to get to the end of the week. You no longer have the luxury of considering the benefits that might accrue over many years; you will not be able to have many years unless you can overcome your immediate shortfall. So, your behaviour might flip and you might do something impulsive like robbing your neighbours. We can think of many domains where this principle might apply: financial risk-taking, betraying friendships, breaking social norms and rules, using coercion or deceit. As you get close to the edge, there may come a point where you suddenly start to have a reason for doing these things.
I am not saying that deprivation provides uniformly greater incentives for committing desperate or risky acts. If limited finances mean that you can’t get yourself out of trouble or move away to another town, then it might be all the more important, most of the time, not to be deceitful, selfish, or risk-prone. However, deprivation might move people into a space where they are hovering so close to an edge that they oscillate stochastically between a state of necessary prudence and a state where they feel they have to do desperate things urgently just to get through their immediate crisis. This is the edge principle: if you live your life close the edge, your behaviour is likely to be more variable than if you live your life far from the edge.
Let us see how this might work. In a very deprived context, most people most of the time are hard at work just getting by and have to be very careful in doing so, since their means are limited and prospects uncertain. However, life shocks (financial, health, personal) are constantly happening. People far from the edge can absorb these shocks with no radical changes to their behaviour, since they have enough of a financial and emotional buffer to do so. For people already under great strain, a life shock might be enough to push them to the edge, and thus their behaviour will suddenly become very different from what it was last week. As if to illustrate this point, on June 26th 2014, the Newcastle Evening Chronicle reported the case of a pair of men who stormed into one of the bookmaker’s shops on the main street of Neighbourhood B and threatened staff with metal bars, making off with several hundred pounds. They were instantly caught by police, and turned out to be locals. ‘I can’t believe I did it. I go in that bookies all the time’, one of them commented, adding: ‘I did it to pay for my dad’s funeral’. Such striking temporal variability in behaviour would be unlikely to happen in a population where everyone was far from the edge to begin with.
The more variable something is, the more information you require to predict it and therefore make appropriate decisions about it (Frankenhuis & Panchanathan, 2011). If the people in your neighbourhood are—at least some of them, at least some of the time—close to the edge, then it makes sense that even if you had a lot of interaction with them, you might not feel that you had enough interaction to say you knew what they were going to do next. By contrast, if the people in your neighbourhood were always far from the edge, then even if you had only very limited and infrequent interaction with them, it might well be enough—enough to say you knew them, enough to feel you could turn to them, enough to trust them—simply because there was so little variability in their behaviour over time.
Now let us apply these principles to the interpretation of our data. First, as the behavioural observation suggests, it is likely that the residents of B do rely more on their neighbourhood social networks than the residents of A to accomplish the things they need. After all, they have less money on average and poorer access to the technologies and institutions that money and professional status can give access to. Yet at the same time as they possibly rely more on their social networks, the surveys show that they feel the inadequacy of these bonds all the more keenly. The stakes are probably higher in that they really need those bonds to work, yet the predictability of their social ties is lower because of the greater behavioural variability discussed above. The greater investment in informal socializing in Neighbourhood B might be considered an attempt to reduce uncertainty, by gathering more data on how other people are currently disposed. Even with this extra data, though, people’s subjective experience is of greater mistrust and social anxiety; the extra data people have about the behaviour of others is not enough to outweigh the extra variability in that behaviour. This view makes a novel testable prediction: interpersonal relationships in Neighbourhood B should show greater temporal variability than those in A; at times people will be close, but they will also have occasional, dramatic fallings out. By contrast, relationships in A should show a flatter line over time. I haven’t directly tested this hypothesis. In Observational Dataset 1, I coded the occurrence of clear altercations going on in the street. I saw 3 in Neighbourhood B and none in A. The numbers are too small to warrant much of a conclusion, especially given that people in A may choose to conduct their altercations in the privacy of their homes.
What I have said so far in this section appears sufficient to explain the gulf between behaviour and self-report. It also explains why trust would be lower in Neighbourhood B, even trust of people well-known to the respondent, since trust is to a very considerable extent a metric of our belief in the predictability of others’ behaviour. How though can we explain the Dictator Game results? The Dictator Game is an assay of motivation to make an avoidable social investment. (It is avoidable since by transferring zero, in our implementation of the Dictator Game at any rate, you can avoid the need for any third party to become involved in the strange business at all.) The attractiveness of making an avoidable social investment will depend a lot on your perception of the predictability and reliability of the social actors with whom that investment would be made. If predictability and reliability are low, there is a risk inherent in the social investment that you do not run if you just keep all the money to yourself. This might explain the unwillingness of respondents in B to make transfers. It is not that their transfers were small, but rather that they did not want to get another person involved at all. Interestingly, in the charity condition, where the other party was not a person in the same neighbourhood, but a well-known and presumably reliable regional organization, the proportion of Neighbourhood B respondents making a transfer dramatically increased.
We may also have explained the diametric difference between our findings and those of Piff and colleagues (Piff et al., 2010). To recall, they found that people of lower socioeconomic position were relatively more generous using assays like the Dictator Game. However, they measured prosociality in general settings, not prosociality specifically directed towards other people who are experiencing deprivation. In most of their studies, the participants were individuals who had made it as far as university. This means that the participants’ backgrounds were probably a lot less deprived than those of our Neighbourhood B participants. More importantly, it means that the targets of the participants’ social investments were not people close to the edge. They were implied to be others from the university community. Thus, the participants in those studies were facing a very different social allocation decision compared to ours: whether to invest in a generally middle-class social group, regardless of their own social background. It is perhaps not surprising that those from humbler backgrounds wanted to invest more in that social group, but this is a different question from whether the residents of Neighbourhood B want to invest in the other residents of their neighbourhood.
The return of the lost letter, and other encounters
The conclusion of the foregoing section was that people in Neighbourhood B are more uncertain about what others in their neighbourhood will do, which makes them feel anxious and negative about social relationships even in the face of abundant interaction. It also means that if they have a choice between opening up an additional, avoidable social relationship and not doing so, they more often veer towards not doing so; hence why so few wanted to involve a third party by making any transfer at all in the Dictator Game. If this view is roughly right, then it is a great lesson in the value of multiple methods. If we had just had the Dictator Game, or only the observational data, then we would have come to very different and much less nuanced generalizations about prosociality in the two neighbourhoods.
There were still other measures of prosociality that we gathered, in the form of small, naturalistic field assays: dropping lost letters, asking people for directions, asking people for change, and dropping objects like pens and keys in the street. I like these simple measures because they are more naturalistic than the economic games, and closer to the flow of behaviour on the streets than the surveys are. The results (presented in detail in Nettle et al., 2011) were quite illuminating in view of the foregoing discussion. For the lost letters, there was a huge neighbourhood difference: they pretty much all came home from neighbourhood A, with very few from Neighbourhood B. On the other hand, in the cases of asking directions, making change, or dropping an object, there were no discernible neighbourhood differences, despite considerable care taken to perform the assay in a standardized way. There could be banal explanations for this pattern, such as the fact that lost letters may disappear amongst the litter that is a depressing feature of the streets of B. Equally, though, they could relate to the patterns we have already discussed. With the lost letter, passers-by have to decide whether to become momentarily involved in the affairs of some person unknown, by picking up their letter and posting it. What if it turns out to contain something illegal, controversial, or nefarious? You are now implicated because you handled it. If you feel that others around you are close to the edge, and might therefore be sending all kinds of strange letters, it might be better to walk on by without anyone noticing. By contrast, in asking directions and making change, you can’t walk on by: you have already been accosted, and thus you have no real choice but to follow the social interaction to some kind of ending. And once you are part of the social interaction, you might as well be helpful, because whatever state your interlocutor is in, they are going to prefer helpfulness. This interpretation does not, admittedly, explain the lack of neighbourhood difference for the dropped object, which would appear to be more like the lost letter in that the participant can plausibly walk on by without initiating any social interaction. But it does suggest why the lost letter patterns with the Dictator Game, whilst asking for directions and making change do not.