[Guest]

Hi Nick,

I have attached a sample of my data with this post.

I'd guess that multiple entries per month for some individuals are the key here.

There are a few individuals that ended their previous job for ex. in 2001m6 and started a new job in 2001m6.

* Example generated by -dataex-. To install: ssc install dataex
clear
input long pid byte(jobwave jobseq jobtype) int jobstarty byte jobstartm int jobendy byte(jobendm j501)
     101  1  1 1 1988 10    .  .  1
     101  2  1 1 1988 10    .  .  1
     101  3  1 1 1988 10    .  .  1
     101  5  1 1 1988 10 2001  6  1
     101  5  2 1 2001  6    .  .  1
     101  6  2 1 2001  6 2003  1  1
     101  8  3 2 2004 10    .  .  .
     101  9  3 2 2004 10    .  .  .
     101 10  3 2 2004 10 2006  8  .
     101 10  4 1 2006  8    .  .  6
     101 11  4 1 2006  8    .  .  6
     101 12  4 1 2006  8    .  .  6
     101 13  4 1 2006  8    .  .  6
     101 14  4 1 2006  8    .  .  6
     101 15  4 1 2006  8    .  .  6
     101 15  5 2 2012  2    .  .  .
     102  1  1 1 1989  3    .  .  1
     102  2  1 1 1989  3    .  .  1
     102  3  1 1 1989  3    .  .  1
     102  5  1 1 1989  3    .  .  1
     102  6  1 1 1989  3    .  .  1
     102  7  1 1 1989  3 2004  6  1
     102 10  2 1 2006  2    .  .  6
     102 11  2 1 2006  2    .  .  6
     102 12  2 1 2006  2 2008 12  1
     102 13  3 1 2010  2    .  .  1
     102 14  3 1 2010  2    .  .  1
     102 15  3 1 2010  2    .  .  1
     201  1  1 1 1975  3 1986 10  .
     201  1  2 1 1986 10    .  .  3
     201  2  2 1 1986 10    .  .  3
     201  3  2 1 1986 10    .  .  3
     201  4  2 1 1986 10 2000 10  1
     201  4  3 1 2001  3    .  .  1
     201  6  2 1 1986 10 2002  5  1
     201  6  4 1 2003  6    .  .  1
     201  7  4 1 2003  6    .  .  1
     201  8  4 1 2003  6    .  .  1
     201  9  4 1 2003  6    .  .  1
     201 10  4 1 2003  6    .  .  1
     201 11  4 1 2003  6    .  .  1
     201 12  4 1 2003  6 2008 12  1
     201 13  5 1 2009  7    .  .  1
     201 14  5 1 2009  7    .  .  1
     201 15  5 1 2009  7    .  .  1
     202  5  1 1 2002  3    .  .  1
     202  6  1 1 2002  3    .  .  1
     202  7  1 1 2002  3    .  .  1
     202  8  1 1 2002  3    .  .  1
     202  9  1 1 2002  3    .  .  1
     202 10  1 1 2002  3    .  .  1
     202 11  1 1 2002  3    .  .  1
     202 12  1 1 2002  3    .  .  1
     202 13  1 1 2002  3    .  .  1
     202 14  1 1 2002  3    .  .  1
     202 15  1 1 2002  3    .  .  1
     203  5  1 1 2001  9 2002  5  1
     203  6  2 1 2003  5    .  .  1
     203  7  2 1 2003  5 2003 12  1
     203  7  3 1 2003 12    .  .  1
     203  8  3 1 2003 12    .  .  2
     203  9  3 1 2003 12    .  .  1
     203 10  3 1 2003 12    .  .  1
     203 11  3 1 2003 12    .  .  1
     203 12  3 1 2003 12    .  .  1
     203 13  3 1 2003 12    .  .  1
     203 14  3 1 2003 12    .  .  1
     401  1  1 1 1996  4    .  .  3
     401  2  1 1 1996  4    .  .  3
     401  3  1 1 1996  4    .  .  3
     401  4  1 1 1996  4    .  .  1
     401  5  1 1 1996  4    .  .  1
     401  6  1 1 1996  4    .  .  3
     401  7  1 1 1996  4    .  .  3
     401  8  1 1 1996  4    .  .  3
     401  9  1 1 1996  4    .  .  3
     401 10  1 1 1996  4    .  .  1
     401 11  1 1 1996  4    .  .  3
     401 12  1 1 1996  4    .  .  3
     401 13  1 1 1996  4    .  .  3
     401 14  1 1 1996  4    .  .  3
     401 15  1 1 1996  4    .  .  3
     402  1  1 1 1992  6 1997  6  1
     402  5  2 2 2001 10    .  .  .
     402  6  2 2 2001 10    .  .  .
     402  7  2 2 2001 10    .  .  .
     402  8  2 2 2001 10    .  .  .
     402  9  2 2 2001 10    .  .  .
     402 10  2 2 2001 10    .  .  .
     402 11  2 2 2001 10    .  .  .
     402 12  2 2 2001 10    .  .  .
     402 13  2 2 2001 10    .  .  .
     402 14  2 2 2001 10    .  .  .
     402 15  2 2 2001 10    .  .  .
end

[Nick Cox]

So, good; you've identified the problem. You need to identify a solution; it's about what makes substantive sense for your purpose.

[Guest]

If job end and job start collides with each other for few individuals then I could move the job "end" a month before or job "start" a month after.

Is it possible to recode it like that?

[Nick Cox]

I am not an economist and I don't work for your institution to know what kind of solution makes sense.

But this one doesn't appeal to me as a double outsider. You have arbitrariness because of ties so you propose to break ties arbitrarily?

[Guest]

You're right. That wouldn't solve the problem But I am having trouble understanding how to fix the multiple entries issue so that Stata doesn't produce arbitrary results.

[Guest]

Hi,

I am still having trouble fixing the multiple entries issue. How can I code it so that Stata does not produce the arbitrary results every time? The number of observations that have multiple entries are very few but still it would be better to interpret a stable result.

Thank you.

Best,
Bibek