oaxaca_rif with bootstrap standard errors and probability sample wiehgts

[Aleksandra Anic]

Dear all,

I implement user written command oaxaca_rif, which preforms Oaxaca Blinder decomposition with recentered influence function.

When using bootstrap option weights are not allowed. Is there a way to use survey weights with bootstrap standard errors in this command?

I can eigher run one of the following models:

bootstrap, reps(100): oaxaca_rif y x1 x2, by(female) wgt(0) rif(mean)
oaxaca_rif y x1 x2 [w=weight], by(female) wgt(0) rif(mean)

Firpo et al. 2018 paper uses bootstrap standard errors when presenting emipirical example, some other authors also used bootstrap s.e, but with no information about sample weights although sample microdata where used.

Here are main references:

Firpo, S. P., N. M. Fortin, and T. Lemieux. 2018. Decomposing wage distributions using recentered influence function regressions. Econometrics 6: 28. https://doi.org/10.3390/econometrics6020028.

Rios-Avila, F. 2020. Recentered influence functions (RIFs) in Stata: RIF regression and RIF decomposition. Stata Journal, 20(1), 51-94. https://doi.org/10.1177/1536867X20909690.

I hope somebody can help me.

Best wishes,
Aleksandra

[FernandoRios]

Hi Alexandra.
Check page 64 of my paper, Paragraph 5.
and the last example of oaxaca_rif.
The Key is to create bootstrap weights, which can be done using the command "bs4rw".
F

[Aleksandra Anic]

Hi Alexandra.
Check page 64 of my paper, Paragraph 5.
and the last example of oaxaca_rif.
The Key is to create bootstrap weights, which can be done using the command "bs4rw".
F

Dear Fernando,

Thank you very much for your answer. I am not sure how to properly create bootstrap weights, both commands bsweights and bs4rw are confusing. I use Survey of Income and Living Conditions database, I have variable Primary sampling units, Secondary sampling units, Primary strata, Personal cross sectional weight. So far I only used personal cross sectional weight.

Also, as it is presented in examples in your paper, there is no difference between robust st. errors and bootstrap st. errors, therefore, I question if I could use [pw=weihgt] option without bootstrap st. errors.

If anyone could help me with writing example with bs4rw and bsweight.

Best,
Aleksandra