Working Paper: CEPR ID: DP15210
Authors: Massimiliano Marcellino; George Kapetanios; Liudas Giraitis
Abstract: We develop non-parametric instrumental variable estimation and inferential theoryfor econometric models with possibly endogenous regressors whose coefficients can varyover time either deterministically or stochastically, and the time-varying and uniformversions of the standard Hausman exogeneity test. After deriving the asymptotic propertiesof the proposed procedures, we assess their finite sample performance by meansof a set of Monte Carlo experiments, and illustrate their application by means of anempirical example on the Phillips curve.
Keywords: Instrumental Variables; Time-Varying Parameters; Endogeneity; Hausman Test; Nonparametric Methods; Phillips Curve
JEL Codes: C14; C26; C51
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Unemployment (J64) | Inflation (E31) |
| Time-varying IV methods (C36) | Time-varying coefficient of unemployment in Phillips curve (J64) |
| Unemployment coefficient was statistically significant from early 1990s onward (J64) | Causal relationship between unemployment and inflation (E31) |
| Unemployment was endogenous until end of 1970s (J64) | Unemployment coefficient in Phillips curve (J64) |