Hi: I have been trying to get the full series for the state to state migration flows from the ACS survey from 2000-2013. In order to do that, I looked at the census data website where they report the 2005-2013 data. And then for the remaining years 2000-2004, I used the interface at IPUMS, using the variable migplac1, and then obtaining the state-to-state migration flows by crosstabulating it using the ‘statefip’ variable. I am however very surprised to find out the two datasets do not coincide for the overlaping years of 2005-2013. I do not understadn why that is and which would be then the more reliable source. Maybe I am missing something. Some help would be greatly appreciated.
I have generated a crosstab similar to the one you have described, with current state of residence on the row and state of residence the previous year on the column. This replicates the basic format of the tables the Census Bureau producedusing the ACS. When I look specifically at the 2012 ACS data I am able to generate estimates close to the Census Bureau tables, though I do not match them exactly. However, this is to be expected. The Public Use files produced by the Census Bureau (and used by IPUMS-USA) are a subset of the full samples and are subject to additional sampling error and further processing before being released so it is not likely that you will get the exact same figures that are published in summary tables (such as those on American Fact Finder and on the linked Census page) using the Public Use files. This topic is addressed to a greater extent in the PUMS accuracy statement. Furthermore, summary tables are often updated multiple times as more demographic data is gathered and refined, whereas the PUMS files are static.
So, it is not likely that the tables you produce will match perfectly with the published tables, but they should be close. Looking at the table I produced, all of the values are within the Margin of Error for the estimates from the Census Bureau table.
I hope this helps.