Survey weights and panel data

[Giulia Ferrero]

Dear all,

I am currently working on a project using a panel dataset where the panel variable is woman and data are at yearly level. I built the panel dataset starting from DHS surveys and reconstructing birth histories for each woman.
Since we are using panel data and the main outcome is a binary variable that takes value 1 if the woman gives birth to a child in year t, I would like to use xtlogit. However this last command doesn't support the inclusion of survey/probability weights and this would bias my standard errors.
The only xt-command that seems to support pweights is xtmlogit but it's used with categorical variables as outcomes.

What would you suggest? should I use xtmlogit with a binary outcome? should I ignore the issue with survey weights?

Thanks!

Giulia

[Andrew Musau]

xtmlogit with a binary outcome is equivalent to xtlogit. Another option is to use clogit which is xtlogit, fe. Here, the weights must be constant within groups.

Use of weights
With clogit, weights apply to groups as a whole, not to individual observations. For example,
if there is a group in your dataset with a frequency weight of 3, there are a total of three groups
in your sample with the same values of the dependent and independent variables as this one group.
Weights must have the same value for all observations belonging to the same group; otherwise, an
error message will be displayed.

See https://www.stata.com/manuals/rclogit.pdf. You could also use the svy prefix with either clogit or xtmlogit. That is for the practical part. There may be a reason why xtlogit with option -fe- or -re- does not accept probability weights. I know that there is some literature which states that conditional logit is biased and inconsistent in the presence of serial correlation, which may explain why the estimator does not allow clustering of the standard errors with option -fe- or -re- (see https://www.degruyter.com/document/d...2021-0005/html), but I have nothing on probability weights.