Missing values for some years - dummy adjustments?

[Flor Carosa]

Hi, everyone,
I'm conducting a multinomial logit analysis on occupational distribution, and I would like to include work status one year before the survey as a control variable. However, I've encountered a challenge: Eurostat stopped collecting this particular variable from 2021 onwards. Consequently, I have data available for this variable up to 2020, but it's missing for 2021 and 2022.

Given this situation, I guess there should be two options:

• Exclude the variable from the analysis entirely.
• Account for its unavailability in some years (kind of 'ignore it' for the years in which it is not available) and use it for the years where data is available, all without dropping observations.

Is the latter approach even possible? If it is, what would be the best method to handle it? I've read about the dummy adjustment approach, but I've also seen some of its critiques, even in this forum - would it make sense in this case?

I would greatly appreciate any suggestions and/or literature recommendations on how to appropriately address this missing data challenge.

Thank you!

[Carlo Lazzaro]

Fior:
leaving things as they are, Stata will omit all the observations with at least one missing value in any variable (listwise deletion). Therefore, no action is needed from your side.
If you want to impute the variable for the years it is missing, you may want to consider -mi- but get rid of the so called dummy adjustement.
Eventually, you can exclude the variable from the analysis entirely justifying your choice in the research report/working paper/article.

[George Ford]

Run the model through 2020 with and without the variables that are missing observations to see how much it affects the stats.