xtabond2 with missing Hansen test

[Mina Wu]

Dear all,

I have applied xtabond2 system gmm command to my dataset, which consists of 45000 observations- that is 6800 individuals tracked over 9 years. The panel dataset is unbalanced, and some individuals count less than 3 year observations.

When I apply the command (28 instruments in total), trying to control for the number of instrument using gmm(lag() collapse) option, i get the following diagnostics.

 
 ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z =  -4.71  Pr > z =  0.000 Arellano-Bond test for AR(2) in first differences: z =  -0.71  Pr > z =  0.480 Arellano-Bond test for AR(3) in first differences: z =  -0.30  Pr > z =  0.764 Arellano-Bond test for AR(4) in first differences: z =   0.17  Pr > z =  0.863 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(2)    = 164.84  Prob > chi2 =  0.000   (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(2)    =  10.51  Prob > chi2 =  0.005   (Robust, but weakened by many instruments.)  Difference-in-Hansen tests of exogeneity of instrument subsets:   gmm(L.TFP1, collapse lag(1 1))     Hansen test excluding group:     chi2(0)    =   8.63  Prob > chi2 =      .     Difference (null H = exogenous): chi2(2)    =   1.87  Prob > chi2 =  0.392

the command is restricted to seven years span, instead of available nine years. Do you know why the Hansen test excluding group ismissing? What could I do to "fix" this?

Thank you.

[Mina Wu]

------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z =  -4.71  Pr > z =  0.000
Arellano-Bond test for AR(2) in first differences: z =  -0.71  Pr > z =  0.480
Arellano-Bond test for AR(3) in first differences: z =  -0.30  Pr > z =  0.764
Arellano-Bond test for AR(4) in first differences: z =   0.17  Pr > z =  0.863
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(2)    = 164.84  Prob > chi2 =  0.000
  (Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(2)    =  10.51  Prob > chi2 =  0.005
  (Robust, but weakened by many instruments.)

Difference-in-Hansen tests of exogeneity of instrument subsets:
  gmm(L.TFP1, collapse lag(1 1))
    Hansen test excluding group:     chi2(0)    =   8.63  Prob > chi2 =      .
    Difference (null H = exogenous): chi2(2)    =   1.87  Prob > chi2 =  0.392

[Sebastian Kripfganz]

You have 2 overidentifying restrictions (the degrees of freedom for the Hansen test). The gmm() option creates one instrument for the first-differenced and one for the level equation, respectively. Excluding these 2 instruments leaves you with an exactly identified model for which you cannot compute a test of the overidentifying restrictions (2-2 = 0 degrees of freedom for the Hansen text excluding these gmm() instruments).

You might want to consider specifying the instruments separately for the first-differenced and the level equation with the options gmm(L.TFP1, collapse lag(1 1) equation(diff)) and gmm(L.TFP1, collapse lag(0 0) equation(level)), respectively.

[Mina Wu]

That is a good advice. Thank you!