This lab experiment tests whether adjustments to the social environment affect an individual’s propensity to give to an unrelated cause. The quality of interpersonal relationships is complex and hard to evaluate (Sobel 2002; Guiso et al. 2010; Quibria 2003), but if we find that the aggregate effect of people being drawn together in a positive way translates directly into an individual’s decision to give, then giving behaviours may offer us a tangible way of monitoring the qualities of those relationships.

Giving is a prosocial behaviour form in that a person chooses to allocate her resources in a pro-collaborative manner. The drivers of giving may include other-centred values and inequality aversion, but giving also depends on what other people are doing and on the state of the wider social environment: the motivation of reciprocal returns from giving, concerns about reputation if a prosocial response is not forthcoming, the expectation that others are ‘doing their bit’ as well, and the cooperative norms signalled by others (Kolm and Ythier 2006; Konow 2010; Gui and Sugden 2010). All these social influences are important not just to giving, but are indicative of social norms and trust levels that are helpful to any collaborative effort. Thus, ‘giving’ may be representative of a much wider range of prosocial behaviours. These prosocial behaviours facilitate collaboration, whilst antisocial behaviours hinder it (Adhikari and Goldey 2010).

Prosocial behaviours not only signal trustworthiness then, they are also dependent on trust in others. There may be an interaction going on between these two: a closer, more trustworthy social environment stimulating individuals to put more weight on other people’s interests or communal interests over their own private interests, and then these prosocial responses feeding back to sustain and improve the quality of the wider social environment. The prosocial character of this interaction between the individual and the wider social environment may be possible to evaluate by the existence of giving behaviours.

The idea of an interaction between individuals and their social environment being involved in the way that social structures change over time is not new; see for example Berger and Luckman (1966) or Giddens (1984). Dasgupta (2009) also describes trust formation in this manner, highlighting how individual actions affect as well as respond to wider social norms. Krishna and Uphoff (2002) distinguish between ‘structural’ and ‘cognitive’ social capital, and discuss how visible relational structures interact with individual preferences and perceptions to determine the development of future relational structures. This paper adds to the literature by deliberately setting out to test the link between a closer relational environment and individual prosocial preferences (manifest in giving) in a lab experiment.

Previous lab experiments suggest that giving does respond to the immediate social context (e.g. Hoffman et al. 1996; Hornstein et al. 1975; Holloway et al. 1977; Ross and Ward 1996; DeScioli and Krishna 2013). And over time, Kosse et al. (2016) found that offering children one-on-one time for one year had a significant and persistent impact on the prosocial manner in which those children then went on to treat others. Our experiment complements these findings and is novel in several ways. Firstly, instead of evoking one social context or another in the minds of the participants by the wording of the instructions, we created two different social environments by manipulating the way people were allowed to interact with their partners, and then checked whether this had an immediate impact on giving. Secondly, giving was not framed as the main decision of the game. Instead it added in at the end of other tasks, and was directed towards a third party. And thirdly, we tested for social and materialistic drivers of giving in the same experiment, which allows us extra insight into what motivates giving. The impact of material incentives on giving was tested by treating half of the subjects in each relational environment to an unexpected doubling in pay. Thus the experiment comprised four groups, each containing 10 participants, who were treated as follows:

After this treatment we measured the effect that these differences had on mood, on desire to meet one’s partner again, and on levels of giving to charity at point of payment and exit from the experiment.

The experiment helps us to address three main issues: Firstly, it tests the effect of the relational environment on an individual’s mood and on the way that individual chooses to allocate his or her own resources. If a change in the relational environment changes that individual’s decision to give, we may conclude that prosocial preferences (conducive to social cohesion) may be adjusted; social preferences are not a static endowment for which no policy can be relevant.

Secondly, we want to see whether these giving behaviours are sensitive enough to the relational environment to be used as its proxy. Should we find that giving behaviours respond significantly to changes in a relational environment, then it would suggest that giving provides us with useful information about the prosocial, cohesive qualities of civic sector relationships. In other words, we might evaluate how prosocial the civic sector is by the resources that its members are allocating to one another, quantifying a complex social stock by easy-to-measure giving patterns. Such a finding might offer decision makers a tool by which to evaluate the effect of their interventions on prosocial inclination and, ultimately, on social cohesion.

Thirdly, the inclusion of an endowment differential allowed us to test whether the association between giving and the social environment is spurious. For example, people could be responsive to any kind of feel-good factor, irrelevant of whether the feel-good factor arises from relational considerations or material considerations. Certainly this is a point of contention in the literature. Giving and wealth tend to go together (The Center on Philanthropy at Indiana University in the USA 2017; Bauer et al. 2012; Holland et al. 2012; Lindsey 2012) and even in lab experiments a higher show-up fee was linked to higher levels of giving (e.g. Chowdhury and Jeon 2014). However, a closer look reveals that richer people are not automatically more generous (Mayo and Tinsley 2009; Auten and Rudney 1990; Breeze 2006). Our experiment allows us to test whether giving requires both a resource endowment advantage and the relational motivation. The results will add to our knowledge of where attention must be focussed for the increase of prosocial/collaborative behaviours.

In addressing these three issues, the experiment contributes to our understanding of civic sector relationships. We gain insight into whether it is possible to influence prosocial inclination. We see whether giving offers a way of evaluating the prosocial qualities of civic sector relationships. And the inclusion of an endowment differential confirms whether it is really relational factors, not just any mood-altering boost to welfare, which fosters prosocial behaviours.

Figure 1 summarises the order of events during the experiment. The participants were divided into a closer or else into a more distanced relational environment. In these environments, they completed a demographics questionnaire and then carried out a series of tasks partially in pairs. Afterwards, in an identical, non-interactive social environment, half the people in each relational environment, unbeknown to the other half, were confidentially informed of a windfall pay bonus. After these treatments (relational and material) they completed private mood surveys, a question about how much they would want to see their partner again, and were provided with the option to give to charity at pay and exit. Mood, feelings towards one’s partner and charitable giving constitute the measurable ‘outcomes’ of the treatments.

To produce the differences in the relational environment, we manipulated conditions according to the five relational parameters affecting relational proximity identified by Schluter and Lee (2009). Table 1 outlines these relational parameters and how treatment differed between the two groups.

Schluter and Lee identified these parameters primarily for the use of managers and executives in the state and market sector, but they correspond to wider research into factors affecting the proximity relationships between people: For directness and multiplexity and their association with closer, strong relationships, see Hess et al. (2007), Boyd (2014), Tillema et al. (2010) and Grootaert and Van Bastelaer (2002). For how the consideration of common interests is vital to collaborative relationships, see Bardsley (2000), Sen (2009), Kolm and Ythier (2006) and Durlauf and Fafchamps (2004). Without parity and a fair distribution of resources, cooperation and social cohesion are compromised (Adhikari and Goldey 2010; Vajja and White 2008; Du Toit 2004; Wilkinson and Pickett 2009). And finally, continuity in a relationships makes it more worthwhile to maintain a positive collaboration with others (Kolm and Ythier 2006; Durlauf and Fafchamps 2004; Schneider and Weber 2013).

We may perceive that all these elements overlap to determine one factor: relational proximity. It is not usual practice to change many things at once in an experiment, but here we altered the social parameters in a unified direction to create one single factor of comparison: a ‘close’ relational environment in which the cohesive aspects of the interaction were emphasised compared to a more ‘distant’ relational environment in which barriers between people were emphasised.

The hypothesis is that the close or ‘cohesive’ relationships will be reflected in how much one party factors the other into their decision-making process; a process evident in the way an individual allocates his or her own resources into a common activity or for the benefit of other people. Thus in this experiment, we specifically test whether a change in relational proximity affects the ‘other-centred’ way in which people handle their resources. Money was given to a third party (to charity, not to other participants in the experiment) so the stimuli can be monitored in terms of other-centeredness in general; it is not just an in-group strategic manoeuvre. The test reveals then whether giving behaviours are a sensitive barometer of the changes in relational parameters.

Experiment participants were drawn from undergraduates in their first year, who were offered £5 for their participation. Two similar classrooms in the same building were chosen for the experiment, one in which to create the close relational environment, and one for the more distant relational environment. The seating and tables were prearranged such that every person was seated next to one other person. Each individual’s place was numbered. Most of the papers for use in the experiment were already on the tables in both rooms, but face down (or enveloped) with ‘Do not turn over these papers or open any envelopes until told to do so’ printed on the top.

On presentation in the foyer outside the classrooms, student consent forms were collected. The students were then divided into alternate rooms, men in order of arrival, and women in order of arrival, with 20 altogether into each room. This was to produce a split of minimal bias. Bias in the composition of the groups was also minimised by requesting people to sit male-female where possible and female-female only where necessary. We also let people choose their own seats, without their knowing that every other pair of seats in each room was pre-determined to receive a bonus payment. This randomised the allocation of windfall benefits. Electronic networking during the experiment was banned to avoid external influences. We ensured a lack of a pre-existent relationship between participants by selecting only undergraduates in their first week of university and by telling them to sit next to persons they did not already know. This, plus similarities in their stage of life, eliminated many possible biases between the groups.

There were two invigilators in each classroom as students entered, one to speak, and the other a timekeeper. The timekeeper was to ensure that the lengths of the exercises were exactly the same in both classes, making the groups comparable. Table 2 documents the parallel progression of events in each room, with the relational differences in treatment clearly indicated. Details of the contents of each paper and the exact wording of invigilator instructions may be found in the online appendix (Zischka 2016).

Session effects are potentially a problem, although the following four precautions had been put in place to ensure that the students did not influence each other in their decision to give: (1) No talking. Once the treatments had been completed, students had been working individually and in silence for several minutes before each getting their money and having to decide whether to give to charity; (2) The payment envelope was A5 size – large enough for people to choose coins without being seen by their partner; (3) Everyone had to fill in slips from the envelope whether they donated or not; and (4) all the envelopes had to remain behind, with only the money not being donated being handled by the students. These measures made it easy for students to donate or not to donate without anyone else being able to observe their choice.

We could also check that these measures were sufficient by observing who gave and where they were sitting, especially in terms of whether former partners were influencing one another. We found that in nine occasions neither partner gave. In eight occasions one person out of the partnership gave. And only on three occasions both persons in a partnership gave. This provides assurance that the decision to give had not been biased by people having somehow seen what their neighbour was doing and just doing the same; our observations on how giving responded to the relational environment appeared to be independent.

This is a 2x2 experiment, the key factors being differences in relational proximity and endowment. Care was taken to avoid other differences in treatment between the groups, with the tasks being of the same length and cognitive demand so as to maintain a fair basis for comparison. Competitive dynamics were also avoided, since introducing competition (market norms) in itself suppresses cooperation and giving (Ariely 2008; Kolm and Ythier, 2006). Even with the windfall payment, the idea of ‘lucky seats’ was evoked to keep the spirit away from competition. In everything to do with money, the treatments were identical in both groups. We could check there was no bias in the composition of the groups by considering the spread of demographics such as age, race, gender, religious practice, financial pressures, life-experience, test-scores and so on between groups.

First we consider how differences in the relational environment and monetary endowment affect giving, and then how they affect mood. The choice of charities students could opt to give to were taken from various websites citing popular charities in Britain, and selecting nine charities that represented the widest possible range of interests (Table 3).

Giving is a prosocial behaviour form in that it involves one party in a positive interaction with another party. Since giving is to a third party and not to one’s partner, it offers more powerful evidence of other-centred motivations in play; there is no direct, reciprocal motivation behind the decision to give. We are measuring then whether relational proximity has an impact on prosocial motivations affecting persons outside of the experimental environment.

All results were subjected to tests of statistical significance, using a chi-squared test for proportional differences and a Mann-Whitney test for unrelated samples. OLS was used to determine the statistical significance of unrelated samples with more than two groups. Figures 2 to 5 display histograms of the choice people made about giving to charity by treatment group, and sections 3.1 to 3.4 discuss these results in more detail.

As expected, relational proximity positively influenced giving. We were also expecting that more money would stimulate more giving and would have a positive effect on mood, but neither of these expectations were met. As is described in the following sections, it turned out that these expectations were negated by the largeness of the impact of relational proximity as it interacted with these factors.

This experiment set out to discover if changes to relational parameters drove other-centred giving patterns. Three linked issues were addressed, to which this experiment gave the following answers. Firstly, we wanted to see how the relational environment and individual prosocial preferences interact, and it was found that the relational variables strongly influence an individual’s decision to give. This bears out the results of DeScioli and Krishna (2013) and also of Attanasi et al. (2013), showing that a prosocial attitude is not just a static, pre-existent, integral emotion, but can also be modified by momentary conditions.

Secondly, we wanted to see whether giving is a useful indicator of the cohesive qualities of the wider social environment, and found that giving is indeed a sensitive barometer of these qualities. Giving is indicative of individual prosocial inclination, and this prosocial inclination interacts with the nature of the wider social environment. Since giving in this experiment was to a third party, we have evidence that changes in the social environment affect prosocial preferences generally. In other words, how people are treated affected how they went on to treat others.

And thirdly, we wanted to confirm that it really was the relational component that motivated giving, and not just a mood-altering improvement to welfare that could be achieved in other ways (a by-product of being better off). Here it was found that a welcome monetary windfall was impotent to stimulate giving by itself; it was relational proximity upon which giving behaviours pivoted, and the desire to maintain that proximity.

Although we note an interaction between individuals and their social environment, this was not a repeated experiment, so the continuity of the feedback loops are not proven. However we do have a social treatment, a social response, and an indicator of a wish for further social engagement. Moreover, the donations to charity that emerge as a positive social response do not go out of the social network; they must change the social parameters experienced by the third party who receives the money, so now their social parameters have changed. How the treatment that third party receives goes on to affect their activities is beyond the scope of the experiment. What we can affirm, however, is that the treatments each group and its invigilators were giving to one another systematically impacted the individual’s decision to consider the interests of a third party, bringing them to give their own money to others. This may be represented as a positive interaction between an individual and their social environment having further positive knock-on consequences to the welfare of society as a whole.

As outlined in the introduction, this interaction between individual prosocial inclinations and the wider social environment (response and counter-response) could be dictating whether the overall cohesion of these relationships is increasing or decreasing. Prosocial inclination (affecting the way a person allocates their resources to the benefit of others) can be modified by changing certain parameters of that interaction. Furthermore, the prosocial qualities of this modification can be quantified by changes in giving flows. There are a couple of far-reaching implications to this research.

Firstly, if relational parameters interact with individual attitudes to have such a big knock-on effect on the way social cohesion develops over time, the implication is that these parameters should be carefully taken into consideration by decision makers and development agents in their social interventions. Schluter and Lee’s mix of directness, parity, commonality, multiplexity and continuity are all malleable and may be useful points of departure to consider.

Secondly, we know that relationships between people with their highly complex structural and cognitive elements are difficult to evaluate, and yet the giving behaviours that flow from the mix are easier to trace. Giving behaviours provide a useful proxy for the prosocial nature of civic sector relations in that giving is ultra-sensitive to differences in the social environment and far more sensitive than subjective responses to questions along the lines of, ‘do you want to meet again.’ Giving reflects a positive interaction between individuals and their social environment; interactions which add to social cohesion and which help to redress factors that might disturb that social cohesion. Those involved with community evaluation might therefore help themselves by asking questions about giving, and monitoring the way giving patterns change over time. In monitoring giving norms, decision makers can evaluate the effect of various interventions on relational health, and can make adjustments that are appropriate to the cause of social cohesion.