Neighbourhood wealth, not urbanicity or population density, predicts prosociality towards strangers

Neighbourhood wealth, not urbanicity, predicts prosociality towards strangers. Elena Zwirner and Nichola Raihani. Proceedings of the Royal Society B: Biological Sciences, October 7 2020. https://doi.org/10.1098/rspb.2020.1359

Abstract: Urbanization is perhaps the most significant and rapid cause of demographic change in human societies, with more than half the world's population now living in cities. Urban lifestyles have been associated with increased risk for mental disorders, greater stress responses, and lower trust. However, it is not known whether a general tendency towards prosocial behaviour varies across the urban–rural gradient, or whether other factors such as neighbourhood wealth might be more predictive of variation in prosocial behaviour. Here, we present findings from three real-world experiments conducted in 37 different neighbourhoods, in 12 cities and 12 towns and villages across the UK. We measured whether people: (i) posted a lost letter; (ii) returned a dropped item; and (iii) stopped to let someone cross the road in each neighbourhood. We expected to find that people were less willing to help a stranger in more urban locations, with increased diffusion of responsibility and perceived anonymity in cities being measured as variables that might drive this effect. Our data did not support this hypothesis. There was no effect of either urbanicity or population density on people's willingness to help a stranger. Instead, the neighbourhood level of deprivation explained most of the variance in helping behaviour with help being offered less frequently in more deprived neighbourhoods. These findings highlight the importance of socio-economic factors, rather than urbanicity per se, in shaping variation in prosocial behaviour in humans.

4. Discussion

Our data do not support the idea that urbanicity is associated with reduced generalized prosociality. Instead, most of the variation in whether help was offered was explained by neighbourhood wealth, with help being more forthcoming in higher-wealth neighbourhoods. Our findings contrast with previous theories and empirical studies on the role of urbanicity in shaping prosocial behaviour (see [21,50] for reviews) but support more recent work that has implicated relative deprivation as being key to understanding variation in generalized prosociality and collective behaviour [34–36,51–54]. The effect of neighbourhood wealth on the tendency to help a stranger might help to explain why previous studies have reported decreased prosociality among urban dwellers. Although the pattern in the UK is quite mixed, there is quantitative evidence that rural areas are typically less deprived than more urban locations (a pattern than we also observe in the locations selected for this study) [42]. Previous findings that generalized cooperation is reduced in urban areas might, therefore, be masking the underlying instrumental variable, which relates to deprivation rather than to population density.

These data from a battery of real-world experiments lend more weight to the hypothesis that relative deprivation is negatively associated with generalized trust and prosociality. Several other studies have reported a similar pattern [34,52–61]. In experimental settings, exposure to harsh environments has been associated with an increased willingness to defect in a Prisoner's Dilemma game, and a reduced tendency to send money to a partner in a Dictator Game [36,62,63], and experimentally induced financial deprivation can increase the willingness to cheat for financial rewards [64]. Studies using the lost letter technique to measure prosociality have consistently reported increased return rates from higher-wealth neighbourhoods [34,35,56]. Large-scale, survey-based studies report similar findings. One recent study using more than 30 000 observations based on nationally representative samples concluded that high socio-economic status was associated with increased willingness to donate to charity, to volunteer to help, to contribute a higher proportion of income to charity, and to choose the prosocial option in an economic game [59] (see also [58]), while another large study (a total sample of greater than 60 000 and with participants from more than 30 countries) reported positive effects of household income on tendency to volunteer or to donate to charity ([60], but see [65], who performed a similar study, obtaining a null result). Finally, a study using data from more than 40 000 responses to the World Values Survey and the European Values Study, respectively, also suggests a negative link between exposure to environmental harshness and the tendency to invest in cooperative behaviour [52].

Nevertheless, other studies have reported negative effects of socio-economic status on prosocial and ethical behaviour [66–72]. We cannot account for the apparently contradictory findings in this field though we note that some of these earlier studies have been based on relatively small undergraduate samples [67,69] and several key results have failed to replicate in large-N, pre-registered replication attempts [73]. Other work has not found a main effect of social class on prosocial behaviour but has shown, instead, that the effect is moderated by third variables. For example, one study found that the effect of social class on social behaviour depends on whether the behaviour takes place in private or public [74], though we note that the effects of the interaction between class and anonymity on giving was inconsistent across different experiments in the study mentioned. Another study argued that the effect of social class on prosocial behaviour depends on whether lower-status individuals are exposed to high economic inequality [75], though two subsequent studies have failed to replicate this effect using similarly large datasets [58,60].

We note that our study differs from much of the work reported above in that we measured behaviour in the real world, rather than via online surveys or in the context of experimental games. In this true field experiment, it is possible that participants were also affected by the environment in which the behaviour was measured. For example, people are more likely to violate prosocial norms of collective behaviour where there is evidence that others do the same [76]—evidence that these norms are violated (e.g. litter and other indices of disorder) is typically higher in more deprived communities [77]. Other work has shown that prosocial behaviour is more likely when people have access to green space [78]. Importantly, access to green space is often lower in more deprived areas [77,79] or else is used less frequently by residents due to perceptions of being inaccessible or unsafe [79,80]. With our current dataset, it is impossible for us to determine whether the link between helping behaviour and deprivation pertains to the deprivation experienced by the participants or, alternatively, to the nature of the environment in which the studies were conducted, though we note that these explanations are not mutually exclusive.

Assuming that the negative association between neighbourhood levels of deprivation and the reduced tendency to help a stranger does exist, it begs the question as to how such a relationship might arise. There are several plausible routes by which deprivation might lead to reduced tendency to help a stranger. One of the most plausible routes might be through the effects that environmental harshness or unpredictability has on the tendency to invest to achieve larger rewards in the future, rather than taking immediately available, smaller pay-offs now. Investing to help a stranger has this incentive structure, where any downstream benefits of the helpful action are typically delayed and/or uncertain [38,81,82]. The effect of reduced income or lowered neighbourhood quality on the tendency to help a stranger could also be mediated by reduced social capital and generalized trust [83–86]. For example, a natural experiment in Russia found that a 10% decrease in average national income following the 2009 recession was associated with a 5% decrease in social trust [87]. Given that one of the help measures in our study (dropped item) involved a live interaction with a stranger, trust may well have been a relevant concern for people deciding whether to help or not. Another potential explanation for the link between adversity and willingness to help a stranger might be the role of material security and how this impacts the scale at which people cooperate [88–91]. In brief, this hypothesis predicts that as material security increases, people are more able to expand their social network, offering impartial help and cooperation to people beyond their core social group of known and regular interaction partners. We note that a prediction that derives from all these hypotheses (albeit one that we cannot test with our data) is that higher indices of deprivation in the UK will only affect the willingness to help a stranger, but not the willingness to help others that are part of one's existing community. Some empirical work supports the hypothesis that exposure to adversity or low socio-economic status is associated with an increased tendency to help friends or in-group members [70,71] though this hypothesis deserves further empirical attention.

We found mixed support for the hypothesis that help would be more likely when participants were directly requested to help. In the dropped item experiment, the direct request increased the likelihood of receiving help, whereas in the lost letter experiment, there was no difference in return rates across the direct request and no request conditions. These patterns may stem from the fact that the direct request in the dropped item experiment was made face-to-face, whereas in the lost letter experiment, the request was made remotely. The perceived costs of refusing to help when asked are likely to be higher in a face-to-face interaction, where the helper can be seen and potentially identified by the requester [26], and other work has shown that people are more cooperative in face-to-face interactions and/or when their name and picture are shown to others [26,30]. The difference we observe between the lost letter and the dropped item experiment helps to further quantify when and how direct requests might elicit help. Specifically, our study suggests that direct requests might increase helpful behaviour not because this reduces the diffusion of responsibility, but because a direct request increases the perceived reputation costs of refusing to help. This hypothesis could be explored in further experimental work.

We expected individuals in a group to help more than lone individuals as the presence of others would create the opportunity to accrue reputation benefits (e.g. [33,92,93]). However, in contrast with our expectations, we recorded higher helping from lone individuals than from individuals in groups in one of our experiments. One possibility is that individuals in social groups might be under greater perceptual load (e.g. in conversation with their acquaintance) which means that they do not note the helping opportunity. Another possibility is that people in groups experience greater perceived costs from pursuing an independent course of action (stopping to help an experimenter) that requires them to temporarily deviate from the group action, and also to impose the time costs associated with helping onto their acquaintance. We do not know of any study that addresses these latter possibilities empirically.

These results contribute to our understanding of the factors affecting variation in human cooperation. These data challenge the folk view that city dwellers are less cooperative than town dwellers and show that this variation may be a by-product of the association between urbanicity and deprivation. More generally, this study supports the hypothesis that deprivation reduces the willingness to extend impartial norms of cooperation towards strangers.