Test | Explanation | Results | Comment |
---|---|---|---|
Durbin-Wu-Hausman test for choice of model | H0: Random effects is preferred | Chi-sq (15) = 14.70 p-value = 0.4733 | Did not reject null hypothesis hence random effects is preferred. |
Endogeneity test: Wu-Hausnam F test Durbin-Wu-Hausman chi sq-test | H0: variable is exogenous. Rejection of H0 means that the variable is endogenous and instruments are needed | 12.58 F (1,2283) | Insurance is endogenous. IV estimates appropriate. |
p-value = 0.00040 | |||
12.61 Chi-sq (1) | |||
p-value 0.00038 | |||
Kleiberg-Paap test for under-identification (Anderson canon. correlation LM statistic) | H0: model is under-identified. Rejection of H0 implies that the model is well identified. | LM statistic 225.565 | Instruments well identified. |
Chi-sq (4) p-value = 0.000 | |||
Weak instrument test (Cragg-Donald Wald F statistic) | H0: weakly identified. Rejection of H0 & t statistic >10 (=rule of the thumb) implies that the model is not weak. | F statistic 61.976 5 % maximal IV relative bias = 16.85 | Instruments not weak. |
Sargan-Hansen J statistic | Test if instruments are uncorrelated with error term. Rejection of H0: instruments are invalid. No problem | Statistic = 2.548 | Null hypothesis is not rejected implying that the instruments are valid |
Chi-sq = (3) | |||
P-value = 0.4667 |