The Identification Problem - The Big Sort: Why the Clustering of Like-Minded America is Tearing Us Apart

Self-sorting complicates the attribution of group outcomes because selection, group influence, and shared circumstances can produce similar observed differences. The identification problem is determining which causal effect the available evidence supports.

When people choose their groups, a difference between groups can reflect who joined, what happened after joining, or causes affecting both. Similar members may also influence one another while sharing an environment. An observed group difference therefore need not identify any one of these mechanisms.

The useful question is a counterfactual: what would comparable people have experienced under a different group exposure? Longitudinal data, natural experiments, and instruments can help only with their own identifying assumptions. Before-and-after observations alone do not remove changing circumstances, and statistical adjustment cannot guarantee that every relevant selection factor was measured.

When to use it

When community-level outcomes need causal analysis that accounts for self-selection; when the Big Sort is confounding policy evaluation; when selection effects and treatment effects need disentangling; when organizational performance attribution needs to account for talent self-selection.

How it can help

Define the counterfactual comparison, examine how people entered groups, and choose a design with explicit assumptions that could distinguish selection from group effects.