Why money helps strangers co-operate at scale

Co-operation in a world of strangers

Globalisation allows people to trade, collaborate and create value with vast numbers of individuals they do not know. Yet the same scale that creates these opportunities also makes co-operation harder. In small communities, misconduct can be disciplined through reputation and social sanctions. In large, anonymous networks, keeping track of who behaved well, deciding whom to punish, and coordinating on a common response becomes increasingly demanding.

Our paper asks a simple question: which institutions best support co-operation as a society grows? We compare 2 forms of social exclusion – permanent ostracism and temporary exclusion – with monetary exchange. The central finding is that forgiveness can work well in small groups, but money is uniquely effective at sustaining co-operation at scale.

Three institutions for dealing with defection

We study this question in an indefinitely repeated helping game. Participants are placed in anonymous networks and alternate between playing producers and consumers. Producers decide whether to help all their connected consumers. Helping creates social surplus, while refusing to help benefits the producer at the consumers’ expense. Consumers observe how their producer neighbours behaved and can respond.

The experiment varies both the size of the network and the institution available. In the Ostracism treatment, consumers can permanently cut a link with a producer. In the Time-out treatment, they can also suspend a link for one round, allowing punishment followed by reconciliation. In the Tokens treatment, participants receive intrinsically worthless digital tokens. Producers can choose to trade: consumers with a token receive help and transfer one token to the producer, while consumers without tokens do not receive help.

The study involved 15 laboratory sessions with 480 participants. The design allows us to observe not only whether participants co-operate, but also whether punishment fragments the network and destroys future opportunities for mutually beneficial interaction.

Why theory does not settle the question

Standard theory alone does not give a clear ranking. Full co-operation can be sustained in all 3 institutional settings. The important difference is how difficult the strategies are to use.

Social exclusion is history-dependent. Participants must monitor what their neighbours have done and decide which links to cut or suspend. That burden rises with the number of connections. Monetary trade is simpler: the same rule – trade help for a token – can be followed regardless of a neighbour’s personal history. A token also targets punishment automatically. Someone who previously refused to help does not earn a token and therefore cannot buy help when their role changes.

In small networks, forgiveness is enough

In small societies, both temporary exclusion and tokens substantially outperform permanent ostracism over time. Realised efficiency, as a proportion of the total potential gains from co-operation, reaches 78.9% with Time-out and 75.5% with Tokens, compared with 62.8% under Ostracism.

The key is network preservation. Permanent exclusion may punish a defector, but it also removes the possibility of future co-operation between the 2 people. As cuts accumulate, the network becomes sparse and the society loses productive opportunities. A one-round suspension is more forgiving: it imposes punishment without permanently destroying the relationship. In a small group, where monitoring is still manageable, this is enough to perform about as well as a monetary system.

In large networks, money is the best institution

The picture changes when the network doubles in size. Initially, temporary exclusion performs best: in the first repeated game, efficiency is 45.9% with Time-out, compared with 41.1% with Tokens and 38.2% with Ostracism. But as participants gain experience, the monetary system pulls decisively ahead. By the fourth repeated game, efficiency reaches 85.1% with Tokens, versus 63.4% with Time-out and 59.2% with Ostracism.

Tokens improve both sides of the problem. They keep the network connected and make defection rare. In the long run, efficiency losses caused by missing links fall to 8.9% with Tokens, compared with 19.3% under Time-out and 25.8% under Ostracism. Losses caused by defection fall to just 3.3%, compared with 15.0% under Ostracism and 17.2% under Time-out.

How tokens change punishment

The experiment also reveals how money changes punishment. When tokens are available, participants overwhelmingly use trade rather than cutting links or retaliating by refusing to help everyone. The monetary strategy accounts for more than 80% of observed punishment responses.

This matters because trade is temporary, targeted and predictable. A defector is denied help when they lack a token, but co-operative neighbours are not harmed and the network remains intact. The rule does not require participants to remember a long history of individual behaviour. It therefore becomes relatively more attractive as the network grows.

Money also creates a positive network externality. The more people who accept and use tokens, the more useful each token becomes. Ostracism has the opposite tendency: every additional cut reduces the set of possible future interactions.

What the findings mean

Our results do not imply that social sanctions are unimportant. Temporary exclusion performs very well in small networks, and real institutions often combine market exchange with reputation, law and social norms. The experiment instead identifies a limit to punishment systems based on monitoring and exclusion: even when they can target individuals, they may become difficult to coordinate and costly to use in large societies.

This is especially relevant in an era of digital platforms and public ledgers, which make individual behaviour easier to record and may appear to make highly personalised sanctions feasible at low cost. Our findings suggest that more information does not automatically make exclusion the best institution. The old institution of money can still outperform sophisticated social punishment by aligning incentives without breaking valuable connections.

Money, in this sense, is not merely a record of past behaviour or a convenient object for exchange. It is an infrastructure that allows strangers to co-operate under a simple, shared rule. That role becomes more – not less – important as economic and social networks expand.