Endogeneity & the IV Idea

Ability bias and measurement error push OLS off the truth; an instrument's group means recover the return to schooling.

Day 61: Endogeneity & the Instrumental-Variables Idea

Ability confounding (ability → schooling)1.5Instrument strength (years added by proximity ÷ 2)0.8Measurement error in schooling (sd)0Sample size800

OLS

0.099

IV = RF ÷ FS

0.073

0.114 ÷ 1.55

truth

0.080

OLS slopeIV slope through group meansnear a college (Z = 1)far (Z = 0)

Simulated workers: ability raises both schooling and log wages (true return 0.08). Living near a college (the instrument) raises schooling but has no other route to wages. OLS (orange) absorbs ability bias and measurement-error attenuation; IV (green) uses only the schooling differences created by the instrument: the slope through the two group means (large dots).

← Single-World Intervention Graphs The Wald & Grouping Estimator →