command for Wooldridge Test for Strict Exogeneity

[Aleem Akhtar]

Hello..i am stata 14 user. I have to run "The Wooldridge Test for Strict Exogeneity", kindly share me the command for it and guide me on how to perform this test.Thank you.

[Jeff Wooldridge]

In my view, this only makes sense when using fixed effects or first differencing to remove the main source of endogeneity: the unobserved heterogeneity. You can use the same test in RE, but first you'd have to pass the Hausman test comparing FE and RE.

If you are using FE and have more than three time periods, this will do it. I'll assume w1, ..., wJ are the suspect variable, x1 ... xK are thought to be strictly exogenous. d2, ..., dT are the time dummies. For the test, the last period is dropped because of the lead (forward) values. With a sufficiently small p-value you reject the null that the wj are strictly exogenous. Then you probably look for an IV solution, such as Arellano and Bond.

xtset id year
xtreg y d2 ... d(T-1) dT x1 ... xK w1 ... wJ, fe vce(cluster)
xtreg y d2 ... d(T-1) x1 ... xK w1 ... wJ F.w1 ... F.wJ, fe vce(cluster)
test F.w1 F.w2 ... F.wJ

[Aleem Akhtar]

Thank you so much @jeff wooldridge for your feedback.

[Aleem Akhtar]

Jeff Wooldridge sorry for this naive question.. what is the function of F in the 2nd equation?

[Eric de Souza]

If you have a variable, x, which is a time series, then L.x is the backward operator;: L.x = x_t-1, and F is the forward operator: F.x = x_t+1

[Sinem Ates]

Dear Jeff Wooldridge,

I have just tried to apply the steps you explained in the post #2 as below. I want to determine if my regressors (ESGSCORE SIZE_w LEV_w ROA_w) are exogeneous or not. Could you please control whether my syntax is correct? Also, I am wondering if there is a test to classify variables as predetermined and endogeneous?

xtreg TOBINSQ_w YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10 ESGSCORE SIZE_w LEV_w ROA_w , fe vce(r)

xtreg TOBINSQ_w YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 ESGSCORE SIZE_w LEV_w ROA_w F.ESGSCORE F.SIZE_w F.LEV_w F.ROA_w , fe vce(r)

 test F.ESGSCORE F.SIZE_w F.LEV_w F.ROA_w

 ( 1)  F.ESGSCORE = 0
 ( 2)  F.SIZE_w = 0
 ( 3)  F.LEV_w = 0
 ( 4)  F.ROA_w = 0

       F(  4,   624) =   26.71
            Prob > F =    0.0000

[Sinem Ates]

Dear Jeff Wooldridge,

I have just tried to apply the steps you explained in the post #2 as below. I want to determine if my regressors (ESGSCORE SIZE_w LEV_w ROA_w) are exogeneous or not. Could you please control whether my syntax is correct? Also, I am wondering if there is a test to classify variables as predetermined and endogeneous?

xtreg TOBINSQ_w YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10 ESGSCORE SIZE_w LEV_w ROA_w , fe vce(r)

xtreg TOBINSQ_w YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 ESGSCORE SIZE_w LEV_w ROA_w F.ESGSCORE F.SIZE_w F.LEV_w F.ROA_w , fe vce(r)

 test F.ESGSCORE F.SIZE_w F.LEV_w F.ROA_w

( 1) F.ESGSCORE = 0
( 2) F.SIZE_w = 0
( 3) F.LEV_w = 0
( 4) F.ROA_w = 0

F( 4, 624) = 26.71
Prob > F = 0.0000

Dear Professor Wooldridge,

Should I test the endogeneity of the explanatory variables one by one or based on the joint significance test of the lead variables above can I conclude that explanatory variables are not strictly exogeneous?

Thank you in advance