Table 2.

Continuity-based RD analysis: effect of redlining on grandchildren’s average household wealth using covariate-adjusted local polynomial regression

N = 173	RD estimator ($)	MSE-optimal bandwidth [meters]	Robust inference			CER-optimal bandwidth [meters]	Nl	Nr
			95% CIrbc	SE	p-value
Outcome: Wealth (mean)
Model 1a	− 96,104	828.83	[− 243,710, – 23,539]	56,167	0.017**	685.95	19	90
Model 2a	− 105,199	807.05	[− 285,228, – 3432]	71,888	0.045**	667.93	19	86
Model 3a	− 95,124	984.23	[− 260,168, – 23,875]	60,280	0.018**	814.56	22	94
Model 4a	− 35,419	399.49	[− 37,423, – 7615]	7604	0.003***	331.79	7	65

Discrete analysis using cluster standard errors; bandwidth is the distance (in meters) from yellow/red border to grandparent’s census block; adjusted models: 2a — age, year; 3a — age, year, gender; 4a — age, year, gender, race; all models use a first-degree polynomial; Nl (left) and Nr (right) indicate the effective number of observations within the MSE bandwidth used for estimation; wealth is rounded to the nearest whole dollar.

^*p < 0.10;

^**p < 0.05;

^***p < 0.01