0
$\begingroup$

I am working with a panel data set in which I can observe not only if you were unemployed in the past year but also how many times you were unemployed and the duration of each unemployment spell. I would like the run a regression to identify how variables (such as job-search methods used each unemployment spell) affect the length of the unemployment spell.

The difficulties are threefold. First, each individual can report having multiple spells within a year, and the number of spells within a year differ person to person. Second, it is likely that there is some sort of heterogeneity that makes it impossible to treat each unemployment spell as a single observation and use individual fixed effects to denote the same individual. Finally, in the panel data, not every person is interviewed every year.

Basically, I was wondering if there is a standard method for dealing with this kind of multiple spell regression. I have done some searching, but all the papers I have found are fairly old, so I do not know if they are still relevant. Any guidance or advice would be greatly appreciated!

$\endgroup$
0
$\begingroup$

I have an idea, but I haven't tried it on your data: Consider

  • employment as an "event",
  • length of unemployment as a "time to event",
  • persons who are not interviewed in a year as
    • "censored" if you never see them again, or
    • "missing" if they come back.

This may allow you to use Survival Analysis with Recurrent Events. The epidemiologists have methods for dealing with "events" that happen to people multiple times, and they are well used to missing and / or censored observations. I think that their literature may be useful to you. I think something similar to this study may help you deal with the missingness in the data (as long as you don't have reason to suspect the unobserved years for a person are deterministic).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.