I have ran an instrumental variable regression using two instruments on two explanatory variables.
Wanted to check
1. I have performed this correctly
Used command
xtivreg2 CP GVA NED SK4 AF WM FLOW (LNXX LNYY = XXLAG YYLAG), fe
With the aim of instrumenting LNXX by XXLAG and LNYY by YYLAG in a fixed effects model
Results are below and wanted to check
- Due to low Cragg-Donald statistic, does this mean my instruments are weak/invalid?
- Also cannot reject null due to Sargan statistic, does this also imply my instruments are weak/invalid?
FIXED EFFECTS ESTIMATION
------------------------
Number of groups = 33 Obs per group: min = 7
avg = 7.0
max = 7
IV (2SLS) estimation
--------------------
Estimates efficient for homoskedasticity only
Statistics consistent for homoskedasticity only
Number of obs = 231
F( 8, 190) = 14.11
Prob > F = 0.0000
Total (centered) SS = 11.83597205 Centered R2 = 0.6908
Total (uncentered) SS = 11.83597205 Uncentered R2 = 0.6908
Residual SS = 3.65968029 Root MSE = .136
---------------------------------------------------------------------------------
CP | Coefficient Std. err. z P>|z| [95% conf. interval]
----------------+----------------------------------------------------------------
LNXX | -.6982269 .5969966 -1.17 0.242 -1.868319 .4718649
LNYY | .7833251 .4174641 1.88 0.061 -.0348895 1.60154
GVA | -.0000191 6.35e-06 -3.01 0.003 -.0000316 -6.70e-06
NED | .0371381 .026387 1.41 0.159 -.0145795 .0888557
SK4 | -.0165385 .0180644 -0.92 0.360 -.0519442 .0188671
AF | .0126059 .016594 0.76 0.447 -.0199176 .0451295
WM | -.3463884 .6089186 -0.57 0.569 -1.539847 .84707
FLOW | -.0388021 .0861104 -0.45 0.652 -.2075753 .1299712
---------------------------------------------------------------------------------
Underidentification test (Anderson canon. corr. LM statistic): 1.272
Chi-sq(1) P-val = 0.2594
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic): 0.614
Stock-Yogo weak ID test critical values: 10% maximal IV size 7.03
15% maximal IV size 4.58
20% maximal IV size 3.95
25% maximal IV size 3.63
Source: Stock-Yogo (2005). Reproduced by permission.
------------------------------------------------------------------------------
Sargan statistic (overidentification test of all instruments): 0.000
(equation exactly identified)
------------------------------------------------------------------------------
Instrumented: LNXX LNYY
Included instruments: GVA NED SK4 AF WM FLOW
Excluded instruments: XXLAG YYLAG
When I ran first-stage regressions, my lagged instruments returned significant results, but the above results suggest they are invalid?
Thanks
Wanted to check
1. I have performed this correctly
Used command
xtivreg2 CP GVA NED SK4 AF WM FLOW (LNXX LNYY = XXLAG YYLAG), fe
With the aim of instrumenting LNXX by XXLAG and LNYY by YYLAG in a fixed effects model
Results are below and wanted to check
- Due to low Cragg-Donald statistic, does this mean my instruments are weak/invalid?
- Also cannot reject null due to Sargan statistic, does this also imply my instruments are weak/invalid?
FIXED EFFECTS ESTIMATION
------------------------
Number of groups = 33 Obs per group: min = 7
avg = 7.0
max = 7
IV (2SLS) estimation
--------------------
Estimates efficient for homoskedasticity only
Statistics consistent for homoskedasticity only
Number of obs = 231
F( 8, 190) = 14.11
Prob > F = 0.0000
Total (centered) SS = 11.83597205 Centered R2 = 0.6908
Total (uncentered) SS = 11.83597205 Uncentered R2 = 0.6908
Residual SS = 3.65968029 Root MSE = .136
---------------------------------------------------------------------------------
CP | Coefficient Std. err. z P>|z| [95% conf. interval]
----------------+----------------------------------------------------------------
LNXX | -.6982269 .5969966 -1.17 0.242 -1.868319 .4718649
LNYY | .7833251 .4174641 1.88 0.061 -.0348895 1.60154
GVA | -.0000191 6.35e-06 -3.01 0.003 -.0000316 -6.70e-06
NED | .0371381 .026387 1.41 0.159 -.0145795 .0888557
SK4 | -.0165385 .0180644 -0.92 0.360 -.0519442 .0188671
AF | .0126059 .016594 0.76 0.447 -.0199176 .0451295
WM | -.3463884 .6089186 -0.57 0.569 -1.539847 .84707
FLOW | -.0388021 .0861104 -0.45 0.652 -.2075753 .1299712
---------------------------------------------------------------------------------
Underidentification test (Anderson canon. corr. LM statistic): 1.272
Chi-sq(1) P-val = 0.2594
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic): 0.614
Stock-Yogo weak ID test critical values: 10% maximal IV size 7.03
15% maximal IV size 4.58
20% maximal IV size 3.95
25% maximal IV size 3.63
Source: Stock-Yogo (2005). Reproduced by permission.
------------------------------------------------------------------------------
Sargan statistic (overidentification test of all instruments): 0.000
(equation exactly identified)
------------------------------------------------------------------------------
Instrumented: LNXX LNYY
Included instruments: GVA NED SK4 AF WM FLOW
Excluded instruments: XXLAG YYLAG
When I ran first-stage regressions, my lagged instruments returned significant results, but the above results suggest they are invalid?
Thanks
