bootstrapping manually

[Guest]

Dear Stata experts,

I decided to run IV regression by hand. In the first stage, I regress my endogenous variable y on the instrument z and dependent variables a,b,c

In the second stage, I want to run the regression on my dependent variable and predicted endogenous variables, variables a,b,c, and lags of a,b,c.

I think it isn't like in econometrics books, but in line with the economic theory, it makes sense.

I realized that standard errors obtained in the second stage might not be correct. As a result, I cannot interpret p-values. I read something about bootstrapping that I can do it so that I obtain the right standard errors. Can someone tell me how to do it in Stata?

I would also be grateful for some scientific articles about this issue

[Rich Goldstein]

try

h bootstrap
h bs

[Carlo Lazzaro]

Guest:
see the wonderful 2nd edition of Microeconometrics using Stata (see Stata bookstore website for more details).

[Guest]

Thank you so much Carlo for this good recommendation. It looks like it is quite a complicated task and my time capacity doesn't allow me to do it. Does bootstrapping change p-values substantially? I decided not to do it in my study. Could somebody recommend me an article stating that p-values generally don't differ after bootstrapping? I hope such an article exists

[Carlo Lazzaro]

Guest:
you may want to take a look at:

. use https://www.stata-press.com/data/r17/hsng
(1980 Census housing data)

. ivregress 2sls rent pcturban (hsngval = faminc i.region), vce(bootstrap, nodots seed(123) rep(200) saving(stuff))

Instrumental variables 2SLS regression            Number of obs   =         50
                                                  Wald chi2(2)    =      33.60
                                                  Prob > chi2     =     0.0000
                                                  R-squared       =     0.5989
                                                  Root MSE        =     22.166

------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
        rent | coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
     hsngval |   .0022398   .0006422     3.49   0.000     .0009811    .0034986
    pcturban |    .081516   .4220788     0.19   0.847    -.7457433    .9087753
       _cons |   120.7065   18.01891     6.70   0.000     85.39011    156.0229
------------------------------------------------------------------------------
Instrumented: hsngval
 Instruments: pcturban faminc 2.region 3.region 4.region

. bstat using stuff

Bootstrap results                                          Number of obs =  50
                                                           Replications  = 200

      Command: ivregress 2sls rent pcturban (hsngval = faminc i.region)

------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
             | coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
     hsngval |   .0022398   .0006422     3.49   0.000     .0009811    .0034986
    pcturban |    .081516   .4220788     0.19   0.847    -.7457433    .9087753
       _cons |   120.7065   18.01891     6.70   0.000     85.39011    156.0229
------------------------------------------------------------------------------

.