Problems with Spell-Data

[Christine Smith]

Good afternoon

Since couple of hours I am struggling with calculating the share of individuals who exit from unemployment and enter employment ("exit ratio").
I have a spell-data (unbalanced panel) which has following structure:

* Example generated by -dataex-. To install: ssc install dataex
clear
input double ID float(Beginn End Interview Spelltype Unempl_dur)
1 555 614 614 1 59
1 555 614 590 1 59
1 555 614 568 1 59
2 590 595 590 1 19
2 609 614 614 0 19
3 540 590 580 1 50
3 540 590 590 1 50
3 540 590 568 1 50
4 540 639 590 1 99
4 540 639 602 1 99
end
format %tm Beginn
format %tm End
format %tm Interview

"ID" describes the ID of individual; "Beginn" and "End" describe the beginn and end of spell; "Spelltype" denotes the type of spell (1 if unemployed, 0 if employed); "Unempl_dur" measures the duration of unemployment.

I am interested in plotting "exit ratio" with respect to unemployment duration. (It should look like positive skew, i.e. the mass of the distribution is concentrated on the left of the figure.)

I would appreciate any comments or suggestions.

Many thanks,
Christine

[Christine Smith]

Does anyone know some possible solutions to solve this problem?

[Clyde Schechter]

I can only speak for myself, but I passed over commenting on this because I have no idea what you mean by exit ratio. When posts don't draw any response it is often because they are unclear. See the FAQ for hints about writing clearer posts.

[Christine Smith]

"exit ratio" is a share of individuals who are leaving unemployment for employment (number of successful jobseekers divided by number of total jobseekers).

[Clyde Schechter]

OK, but I still don't get how you would calculate that from your data, mathematically. How do you know who is a job seeker? And maybe that changes over time? I suppose a jobseeker would be "successful" if he/she transitions from unemployment to employment, but given that there are sometimes gaps between End in one observation period and Beginn in the next, we don't even really know quite when that happens.

Maybe you could do the calculation by hand for the sample of data you provided, explaining the steps involved, and then someone can turn that algorithm into Stata code.