Understanding Society User Support

Fergus,

You will need a longitudinal weight for this as for some TSMs you will have only 1 wave observation and they wouldn't be in your model.
Also, please carefully consider how you define your population. The definition should be substantive, such that your population should exist outside of UKHLS data, and therefore be not dependent on participation. So, you can't represent people for whom we observed unemployment spells - there are no such people as a subgroup in a population outside UKHLS data. But you could represent people who were unemployed and found a job at the time of wave 3, for example. Or people who were unemployed for at least 3 years and found a job (pooled over years). This definition would define which waves you use and will help you with the choice of the weight.

Hope this helps,
Olena