st: Strange merge behavior

[Danielle H Ferry <dferry@nber.org>]

Dear Statalisters,

I wonder if anyone can explain to me the bizarre behavior I am  
getting on a match merge. The master dataset is a panel (12 years),  
and the data is not unique on the match variable (placefip). The  
using dataset contains just 2 variables: the match variable  
(placefip) and another (macci), which is what I am after here. The  
using dataset is unique on the match variable (placefip). There is a  
1-1 correspondence btween placefip and macci in the using dataset.  
When I merge on placefip, however, I get the result that macci  
differs by year within placefip. For example:

. ta year macci if placefip==5553000, mi

           |      Center city
      year |         1          . |     Total
-----------+----------------------+----------
      1990 |    12,818          0 |    12,818
      1991 |    12,607          0 |    12,607
      1992 |    12,437          0 |    12,437
      1993 |    12,187          0 |    12,187
      1994 |    11,749          0 |    11,749
      1995 |    11,183          0 |    11,183
      1996 |    11,237          0 |    11,237
      1997 |         0     10,741 |    10,741
      1998 |         0     11,002 |    11,002
      1999 |         0     11,254 |    11,254
      2000 |         0     11,150 |    11,150
      2001 |         0     11,196 |    11,196
-----------+----------------------+----------
     Total |    84,218     55,343 |   139,561

I don't see how this can be! (Please note that for placefip==5553000,  
_merge==3 in all cases.) After the merge, macci should be constant  
within placefip (and _merge==3), right? Perhaps I have been working  
at this too long, and the answer is obvious. If so, I apologize. Can  
anyone please help?

Thanks,
Danielle H Ferry


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/