Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Blinder-Oaxaca decomposition with estimated endowments

    Dear Stata users,

    The Blinder-Oaxaca decomposition typically uses the observed endowments (i.e. group-means on covariates) and estimated coefficients (i.e. from some kind of regression model). In this approach, endowments can be manually set to certain values. For instance, we could set the mean value on X to a for group 1, and to b for group 2:

    Code:
    oaxaca y x, by(group) x1(x a) x2(x b)
    One consequence is that standard errors become very small. This is usually fine, e.g. if you want to correct for group differences due to design.
    However, I want to estimate the endowments in a regression model, to control for spurious group differences in endowments. Simply plugging in the mean estimated value leaves out the uncertainty of this estimate, and thus results in standard errors that are way too small.

    Is there any way of accounting for uncertainty in estimated endowments? I can think of bootstrapping the entire procedure, but any alternative suggestions are highly appreciated since my dataset is very large.
Working...
X