When the two ends meet: an experiment on cooperation and social capital

Abstract

We study the behaviour of individuals with different geographic origins interacting in a same public good game. We exploit the peculiar composition of the experimental sample to compare the performance of groups where individuals have mixed origins to homogeneous groups. We find that, despite the absence of any geographic framing, mixed groups exhibit significantly lower contributions. We also find that cooperation levels differ significantly across geographic origins, in line with the existing literature. This is explained by a different impact of coordination opportunities, such as communication, as we show by manipulating them. Our results point towards integration as a crucial aspect for the economic development of intercultural societies. They also confirm that, rather than being explained just by the differences in institutions and economic opportunities, the Italian North–South divide embeds elements of distrust, prejudice and a consequent path dependence in the level of social capital

Appendices

Appendix A: Algorithm for the creation of groups

The following algorithm was implemented to subdivide participants of each session in four groups. Importantly, in each session, each school was represented by a maximum of three students.

• 1. Create three empty lists: ${\mathcal {S}}$(outh) with six slots, ${\mathcal {N}}$(orth) with six slots, ${\mathcal {M}}$(ixed) with 12 slots. A slot is occupied whenever a student is appended to a list.

• 2. If the session has strictly less participants from the North (South), remove one slot to the ${\mathcal {N}}$ (${\mathcal {S}}$) list, respectively.

• 3. Let I be the school with the most students among schools still not processed.

• 4. Let ${\mathcal {L}}$ be the list ${\mathcal {S}}$ if the school is from the South, ${\mathcal {N}}$ otherwise.

• 5. If ${\mathcal {L}}$ has a free slot, append a randomly selected student from I to it.

• 6. If there are still students to be placed from I, append them to ${\mathcal {M}}$.

• 7. If there are still schools to be processed, go back to point 3.

• 8. Create two lists ${\mathcal {M}}1$ and ${\mathcal {M}}2$ from elements of ${\mathcal {M}}$ in odd and even positions, respectively.

The rationale for ordering schools by size was to guarantee that no two students from the same school would end up in the same group (i.e. that schools with more students, and hence more difficult to place, would get their students assigned first).

For completeness, Table 2 presents a brief overview of the schools involved.

Table 2 Subdivision of participants across schools and cities

Appendix B: Instructions

Participants in each session spent the entire time of the experimental session sitting in circle in the same room. In order to limit communication to the phases designed for this (see Sect. 2.1), subjects were, since the beginning, instructed not to talk among them about the experiment, and to direct any question to the experimenter in charge of explaining the activity (the experimenter was always the same across all sessions). For the same reason, instructions and control questions were not provided in written form, and were instead verbally issued (in Italian).

• Initially, subjects entered the room together. The experimenter and a helper (the same across all sessions) were present. A number of chairs, one for each participant, were distributed in circle around the room, except for a portion of one wall, where a table was placed, holding the material used for the experiment. Such material included a computer connected to a projector (where no image was projected initially), the playing cards, and an empty box with a small opening in its top.

  “On each of these chairs you will find a note with a name written on it: please find your name and sit on the corresponding chair. Please do not discuss this game with other participants until its end. If you have any questions, please ask them to me. You can do so at any time.”

• After all participants had sat down, the playing cards used for communicating their decision were distributed, according to a predefined random mapping. The helper went distributing the cards in circle, verifying the names of participants against the list he had.

  “My colleague will now assign four cards, covered, to each of you. We will explain you in few moments what use you will make of these cards. It is important that they remain secret: one basic rule of this game is that nobody, at no time, must see your cards.”

• After all playing cards had been distributed, the functioning of the activity was explained.

  “Please now look at your cards, while keeping their face hidden from the view of other participants. You will notice that you have two cards with a red suit, and two with a black suit. The red cards are worth one point, the black cards are worth zero points. You will also notice that the two cards of a black suit come from different decks: one has a blue back, one has a red back. The same applies to the cards of a red suit.”

  “The game will be composed of several rounds. At each round, we will collect two cards, covered, from each of you. You all have the possibility, and can freely decide, to give zero, one or two cards from a red suit, that is, to contribute zero, one or two points. The points you will not contribute will remain to you. Please give one card with a blue back and one with a red back.” “Although their composition is, at this point, unknown to you, participants in this room have been subdivided in four groups of approximately the same size. The points you decide to contribute when we collect the cards will flow into a common fund—one for each group. The content of this fund will be multiplied by two and subdivided in equal shares between members of the group. So, at the end of each round, each of you will obtain the points he or she decided to keep, plus the total number of points that your team members decided to contribute, multiplied by two, and divided by the number of team members.”

• Subjects were then proposed the following three test questions in order to ensure they had correctly understood. For each question, after one or more participants had answered correctly, the experimenter would explain the answer to all participants, and the steps (calculations) required to obtain it.

  • 1. “Let us assume that you are in a group of four people and that, at a given round, each of you decides to keep both points: how many points does each of you get for that round?”

  • 2. “What if, in this same group, each of you decides to contribute both points?”

  • 3. “What if in this same group you decide to keep both points, but the three other group members decide each to contribute both points?”

• Finally, information was given about the rounds and payoffs.

  “You now know the basic functioning of the game. The points you make in each round will be added up and, at the end of the game, the three people with most points will win a prize. You don’t know now the number of rounds which you will play, but I will inform you before playing the last round. Moreover, I will provide you with other pieces of information in the following rounds, but details will be provided at that time only.”

• The experiment then started. Each step of the game was guided by the experimenter. “Now my colleague will pass among you with this box. Please insert in its opening the two cards you decide to give.”

  “Now I will register your choices on the computer.”

  (Notice that the spreadsheet on which this was done was not shown on the projector screen.)

  “Now my colleague will return you your two cards: you can verify they are indeed the two you gave. We can then proceed to the next round.”

• After the first two rounds, absolute silence was requested:

  “From this moment, and until further notice, please abstain from any talking. If you have any question, please raise your hand an I will come and answer it.”

  Immediately after, the composition of the groups was revealed, one at a time:

  “Please Eeeeee Ffffff, Gggggg Hhhhhh, Iiiiii Jjjjjj and Kkkkkk Llllll now raise your hand. You are members of group X. Please make sure you all identify each other, while remaining silent. Please now lower your hand.”

  (notice participants were called by name and surname, while X, the number of the group, was one of A, B, C, or D).

• After each round starting from the third, information about past contributions was shown on the projector screen (see Fig. 3 for an example referring to round 5).

  “These are the average contributions so far: each line refers to a group”

• After round four, members of each group were invited to discuss among them:

  “From now on, you can talk again. Please, members of group A move to this corner of the room, members of group B to this other corner, members of group D to this other corner, and members of group D to this last corner. I will give you exactly two minutes, starting from now, to freely discuss with your group members, then we will proceed with the next round of the game. Remember that it is still forbidden to show your cards to any other participant.”

• Finally, before round six, subjects were informed that it would be the last round.

  “We are now going to play round six, which will be the last round of this game.”

• A debriefing would follow, in which participants were invited to comment the experiment and the effects of differences across rounds. The experimenter would mention some real world examples of public goods, and comment on the fact that even non-binding mechanisms such as identification and communication can have important consequences on economic decisions made by individuals within societies.

Appendix C: Additional material

Figure 3 features an example of how information about past group contributions was shown to participants (from round 3 onwards).

Table 3 provides some descriptive statistics for participants: for each session, we show the distribution of individual characteristics (geographic origin/gender) based on the assignment of individuals to the treatment. T-tests ran on the each session fail to reject the null of identical distribution between the two categories, with the exception of Session 2 ($p=0.001$), in which homogeneous groups were composed only of female participants (we take this into account in Sect. 5.4).

Table 4 provides information about the 12 prize winners (three for each session). For comparison, the minimum possible earning was 2 and the maximum 32 [see Eq. (2)]. The signs of deviations between the shares of winners and the shares of sample presenting each feature are in line with results presented in the main text (females contribute more, although not significantly, “North-only” groups perform better, although not significantly, homogeneous groups perform better).

Fig. 3

Example of past information, as shown to participants. Information shown to participants of session 1 before the last round (labels translated from Italian)

Table 3 Descriptive characteristics

Table 4 Descriptive characteristics of winners

Appendix D: Supplementary results

1.1 D.1 Analysis of group averages

Table 5 replicates the analysis of phase- and treatment-effects of Table 1 (columns (1) and (2), respectively) at the group, rather than individual, level (analysis of gender and geographic origin is omitted because these characteristics vary within groups).

Table 5 Main results, group averages

As a robustness test for results 1 and 2, we also replicate tests presented in Sect. 5, confirming all findings (with the only—unrelated—exception of (HmIII), for which we now obtain $p=0.137$).

Figure 4 is the equivalent of Fig. 2, disaggregated at the group level. While variability increases significantly (each point only represents the mean of four to six individual contributions), it allows to observe that all groups are generally increasing , and that no obvious outliers emerge—with the exception of the last round (discussed at the end of Sect. 5.2), in which a few groups drop their contribution levels significantly, as compared to the other groups.

Fig. 4

Average contributions per round, disaggregated by group

1.2 D.2 Phase-treatment interaction

Table 6 Additional estimation results

In what follows, we combine Eqs. 2 and 3, interacting phase and treatment dummies with the geographic origin of participants.

$$\begin{aligned} x_{i,t} = \zeta ^f F_i + \sum _{P = I}^{IV} T_{t,P} (\zeta _{P} + \zeta ^h_P HOM_i + N_{i,P} (\zeta ^n_P + \zeta ^{hn}_P HOM_i) ) + \epsilon _{i,t}. \end{aligned}$$

(5)

Hypotheses (HdII), (HdIII) and (HdIV) allowed us to investigate whether being in a homogeneous group (instead of a heterogeneous one) has an effect on contributions. The estimation of Eq. (5) can help us verify if there is a treatment effect conditional on the geographic origin of individuals. Namely, we can answer such question by running the following joint tests on coefficients presented in Table 6.

• ${\mathcal {H}}_0: \zeta ^h_P + \zeta ^{hn}_P > 0$ for individuals from the North,

• ${\mathcal {H}}_0: \zeta ^h_P > 0$ for individuals from the South,

for each phase $P=II,III,IV$. From such tests, no significant differences emerge ($p=0.183, 0.239, 0.259$ for the North, 0.149, 0.517, 0.204 for the South, respectively).

By exploiting the disaggregation along the dimension of geography, we can also compare North-only and South-only groups between them. This is done by testing ${\mathcal {H}}_0: \zeta _P^{n} + \zeta _P^{hn}> 0$ for each phase $P=II,III,IV$.²⁶ Results do not suggest that people from the North act differently from people from the South in homogeneous groups ($p=0.703, 0.191, 0.306$, respectively).

By running the same analysis for mixed groups, we can instead compare the behaviour of Southerners and Northerners subject to the same treatment (i.e. being in a mixed group) in each phase.²⁷ Namely, we test ${\mathcal {H}}_0: \zeta ^{n}_P > 0$ for each phase $P=II,III,IV$: in line with Result 3, in mixed groups we find a higher level of contributions on behalf of Northeners compared to Southeners, for two phases out of three ($p=0.669, 0.055, 0.087$, respectively).

In conclusion, while we confirm the higher level of contributions of Northeners (Result 3) in mixed groups, we find no evidence that the treatment effect is related to the geographic origin of subjects. That is, we cannot explain Results 1 and 2 as the consequence of an interaction between the treatment and the geographic origin. However, we cannot exclude the possibility that such “non-result” is due to the low numerosity of observations in each of the subsamples considered.