Social Learning with Endogenous Information and the Countervailing Effects of Homophily

People learn about opportunities and actions by observing the experiences of their friends. We model how homophily – the tendency to associate with similar others – affects both the endogenous quality and diversity of the information accessible to decision makers. Homophily provides higher-quality information, since observing the payoffs of another person is more informative the more similar that person is to the decision maker. However, homophily can lead people to take actions that generate less information. We show how network connectivity influences the tradeoff between the endogenous quantity and quality of information. Although homophily hampers learning in sparse networks, it enhances learning in sufficiently dense networks.

💡 Research Summary

This paper presents a theoretical model examining the dual role of homophily—the tendency to associate with similar others—in social learning processes where information is endogenously generated by agents’ actions. The model features overlapping generations of agents from two groups (“blue” and “green”). In each period, agents choose between a “safe” action with a known payoff of zero and a “risky” action with an unknown payoff v (which can be 1 or 0). Agents know their personal cost of taking the risky action but must learn the true value of v by observing the actions and outcomes of their friends from the previous generation. The network is characterized by the number of friends (d) and the homophily level (h), the probability a friend is from one’s own group.

Homophily creates a fundamental trade-off. On the positive side, it improves the quality of information: observing the success or failure of a friend with similar costs provides a more relevant and precise signal about the value of v for the observer. On the negative side, it can reduce the quantity and diversity of information: if homophily leads a group to herd on the safe action, future members of that group may never observe the risky action being taken, even if it is beneficial. This results in “sample herding,” where a group is stuck with an inefficient action due purely to a lack of observational data, distinct from traditional “inference herding.”

The paper’s key finding is that network connectivity critically shapes this trade-off. In sparse networks (low d), the negative effect dominates. With few observation samples, high homophily makes it likely that an entire group fails to see any examples of the risky action, leading to permanent learning failures and inefficient herding. Conversely, in dense networks (high d), the positive effect dominates. The abundance of observations mitigates the sample bottleneck; agents are likely to see the risky action regardless of homophily. Therefore, the ability to glean higher-quality signals from similar friends becomes the primary advantage, and homophily enhances the speed and accuracy of social learning.

The analysis yields nuanced insights. For instance, a higher correlation of costs across groups can unexpectedly increase the benefits of homophily. While this correlation itself facilitates cross-group learning, it also makes group actions more similar in equilibrium, which reduces the negative sample-bottleneck effect of homophily and lets its quality-enhancing effect shine. The model also offers policy implications, suggesting that promoting cross-group ties may be most valuable in early learning stages to kickstart information generation for disadvantaged groups, while within-group ties (homophily) become more beneficial later for refining beliefs.

By endogenizing both the generation of information and the network’s role in filtering it, this paper provides a more complete framework for understanding how social structure impacts collective learning, economic mobility, and the persistence of inequality.