I am attempting to study intermetro migration rates longitudinally, from 2005 till the most recently available data. Of course, in 2012 they discontinued releasing the MIGMET variable, but I was able to generate one using the crosswalk IPUMS provided.
Looking at my migration variable though across time, there seems to be a shift in the values in 2012 from 2011, which is likely either generated by something in the data or something I’ve done wrong. I’m mostly asking whether intermetro migration should be comparable before and after 2012, or whether something changed in the new definition of PUMAs that would make data from those two periods not directly comparable.
I define intermetro migration by assessing whether the individual 1. Moved, 2. Weren’t in an unidentified area in either year and 3. Whether their current metro does not match their past metro. Using that definition for all years, the rate of intermetro migration is consistently around 1 percent lower after 2012, making me worry there’s something different in the denominator I’m not accounting for.
Saying the two periods aren’t comparable would actually be a welcome answer, because I’m pulling my hair out trying to figure out what I could have done wrong.
In short, the two periods aren’t comparable (so you can stop pulling your hair out!)
The 2012 ACS PUMS was the first to use 2010 Migration PUMA definitions, which include many significant differences from the previously used 2000 Migration PUMA definitions. In general, where the two definitions differ, the 2010 Migration PUMAs are larger, with some extreme changes in New England definitions and in a few other areas. See the IPUMS USA 2010 Migration PUMA Definitions page for more discussion and a map comparing 2010 and 2000 definitions.
In many cases, it’s still possible to identify metro areas with reasonably consistent boundaries from both definitions of Migration PUMAs, but in many cases, it’s not, so I think it’s likely the 1-percent difference in migration rates you identified is due mainly to the changes in PUMA definitions.
By the way, we’re currently working on a new update to extend MIGPUMA1 through 2017, and to add a new variable that will identify relatively consistent MIGMET definitions across all ACS years. I currently expect we’ll release it in the next month or two. This still won’t completely solve your problem–there’s really no way to match metro areas exactly across the 2011-2012 change–but it may be adequately reliable.