I am using the online SDA frequency tool trying to recreate a chart I saw in a Pew Research study from 2015 (U.S. Labor Force by Generation, 1995-2015). I set my variables as follows: row=Year, column=age, selection filter=labforce(2), weight=wtsupp. The table that results looks in line with what I would expect, except the year 2014. The year 2014 looks like it is about double the average of the years before (2013) and after (2015), for every age break and the total. The unweighted data does not appear to mimic this outlying result for year 2014. Any suggestions?
Yeah, this is a funny quirk of the CPS ASEC sample for 2014. In this year, the Census Bureau fielded an experimental redesign of the income questions in the ASEC. To preserve comparability, they split the 2014 ASEC sample into two groups: 3/8ths received the redesigned income questions and 5/8ths received the old income questions. Importantly, both groups were given WTSUPP values so that each independently could calculate representative statistics. In SDA you are pooling these two groups together, so that is why you are calculating statistics that are roughly double what you are expecting. Unfortunately, there is currently no way around this in the online analysis tool. The HFLAG variable exists to identify which group households in the 2014 ASEC sample belong, but it is currently not available in SDA. The IPUMS CPS team hopes to add this variable to SDA soon. More information about the 2014 3/8ths file is available here.