Immigration, integration and cooperation: experimental evidence from a public goods game in Italy

ABSTRACT

Immigration has rapidly changed the demographic profile of most Western European societies, increasing their ethnic diversity. Some cross-disciplinary literature comparing homogeneous and diverse communities provides observational evidence of a link between high levels of ethnoracial diversity and lower levels of public goods provision. While these results are often interpreted as driven by context/interaction effects, whereby individuals lower their cooperativeness in response to the presence of non-coethnics, they could also be explained by composition effects – immigrants having different baseline levels of cooperativeness, and thereby lowering average cooperation rates. To disentangle these effects, we conducted a lab-in-the-field experiment with a sample of Italians and immigrants from Morocco and the Philippines residing in Milan. In our public goods experiment, participants were randomly assigned to either homogeneous or ethnically mixed groups. We find that Italians behave similarly in both homogeneous and heterogeneous groups, thus contradicting arguments about the negative effects of diversity on the native population, while there are both compositional and interactional effects when considering the behaviour of Moroccan and Filipino immigrants, respectively. Moreover, differences largely disappear when we consider only the behaviour of more socio-economically integrated immigrants, highlighting the need for a more processual understanding of cooperation in multiethnic communities.

KEYWORDS:

• Ethnic diversity
• immigration
• cooperation
• public goods game
• experiment

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 For a recent meta-analysis on the link between ethnic diversity and social trust, see Dinesen, Schaeffer, and Sønderskov (2020).

2 Prior to data collection, the study received ethical approval from the ethics committee of the grant hosting institution (Bocconi University).

3 More information on our sampling strategy and recruitment can be found in Section 1 in the Appendix.

4 At the beginning of each session, all participants provided written informed consent.

5 Participants final cash payment was determined by one randomly selected individual task and one randomly selected round of the PGG. The average hourly wage in Italy was 21 euros in 2017. While there is no official minimum wage in Italy, most of the minimum wages determined by unions are 7 euros per hour. We therefore provided fair compensation for participants’ time and effort.

6 Non-excludable means that everyone benefits from the public good regardless of their own contribution. Non-rivalrous implies that the usage by one individual does not reduce its availability to others in the group.

7 We also tried to ensure that out of the six participants, there was a relatively balanced gender ratio of 3–3 or 2–4.

8 The majority of our participants, 72% successfully answered most control questions correctly. The control questions were: the maximum amount a player could contribute to the PGG, the minimum amount a player could contribute, and the total number of players in the game. However, there were some differences in the understanding of the game across ethnic groups as shown in detail in the Summary Statistics Table A1 in the Appendix.

9 In total, the PGG was played for 8 rounds. Participants only knew that they will be interacting with the same group for several rounds, not the exact number of rounds. All participants played two initial rounds of the standard public good game. Afterwards some sessions were introduced to different punishment treatments. The PGG was the first group activity in each session, which always started with a short survey on socio-demographic characteristics and several individual decision tasks.

10 Note that participants had engaged in no prior communication or interactions before making their decisions in the PGG. We therefore use heteroskedasticity-robust standard errors rather than standard errors clustered at the session level. That said, clustering the standard errors does not appreciably change our results (not shown).

11 These contained on average four Italians and two immigrants.

12 In the Appendix, Table A7, we examine whether Italians behave differently when matched with Filipinos or Moroccans. We find that Italians contribute less to the PGG when mixed with Moroccans than Filipinos. This difference is marginally statistically significant (10% level). Lower cooperation rates seem to be driven by Italians in mixed groups matched with two or more Moroccan men. This is consistent with other research on the effect of the size of the outgroup (Adida, Laitin, and Valfort 2016b) and their gender (Gereke, Schaub, and Baldassarri 2020).

13 This is true even after controlling for whether individuals knew anyone else who had participated in the study before. See Model 6 in Table A6 in the Appendix.

14 See Table A10 in the Appendix for a breakdown of contributions by size of Moroccan participant share in mixed Moroccan-Italian groups.

Additional information

Funding

This work was supported by European Research Council [grant number 639284].