Hansen J Test in ivregeh command

[LEE CHAN]

Hi everyone,
I performed the IV-2SLS Lewbel model by Ivreg2h, and as I understand, when the Kleibergen-Paap rk Wald F statistic values are all greater than the Stock-Yogo critical values, it means that the internal instrument variable is valid. But in the Hansen J test, the p-value=0.000. What does this value mean, that the instruments are not valid? and is my model then reliable?

IV (2SLS) estimation
--------------------

Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity

                                                      Number of obs =     6702
                                                      F(  9,  9375) =   121.33
                                                      Prob > F      =   0.0000
Total (centered) SS     =  839.4454981                Centered R2   =   0.1985
Total (uncentered) SS   =          932                Uncentered R2 =   0.2781
Residual SS             =  672.7901178                Root MSE      =    .2677

------------------------------------------------------------------------------
             |               Robust
  indep_var | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
   age_meanhh  |  -.0003957    .0002584    -1.53   0.125    -.0009022     .0001108
   city_reside |   .4876124    .0212279    22.97   0.000     .4460087     .5292161
   edu_share   |  -.0684519    .0114893    -5.96   0.000    -.0919708    -.0449330
   ...........|
         _cons |   .3498312    .0296652    11.79   0.000     .2916881     .4079743
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic):            612.842
                                          Chi-sq(8) P-val =    0.0000
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic):            5421.317
                (Kleibergen-Paap rk Wald F statistic):        4178.901
Stock-Yogo weak ID test critical values:  5% maximal IV relative bias    20.25
                                         10% maximal IV relative bias    11.39
                                         20% maximal IV relative bias     6.69
                                         30% maximal IV relative bias     4.99
                                         10% maximal IV size             33.84
                                         15% maximal IV size             18.54
                                         20% maximal IV size             13.24
                                         25% maximal IV size             10.50
Source: Stock-Yogo (2005).  Reproduced by permission.
NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
------------------------------------------------------------------------------
Hansen J statistic (overidentification test of all instruments):       24.612
                                          Chi-sq(7) P-val =    0.0000
------------------------------------------------------------------------------

[George Ford]

Strong instruments but not good ones (at least one instrument is likely correlated with unobserved factors affecting the outcome).

[LEE CHAN]

George Ford , many thanks Prof. But should I use this model to present the results, will it be reliable? Or is there any way to fix the p-value in J test?

[George Ford]

Not reliable. Need to work on your instruments. Try a transformation, interactions, other exclusions, etc...