Correcting sample selection biases with IPW

[William Smith]

I am examining black-white earnings gaps. I wish to correct the sample selection bias (i.e., being employed) with IPW (inverse probability weight). As the first step, I estimated a logistic regression model (DV: being employed) and generated a variable of IPW. For the second step, I need to estimate the earnings disadvantage of being a black worker. I wonder if I can combine IPW and survey weight. This is how I did:

****Codes
.svyset psu [pw=surveyweight*myIPW], strata (strat_id)
.svy: reg logearnings black age agesq i.edu if employed == 1

Is this correct? Also, I want to impute my data to replace missing values before the first step (i.e., logit). Is there a proper way to apply IPW with imputed data? For example, can I generate IPW for each imputed dataset and use the average IPW for the second step regression?

Thanks for your consideration in advance.