Follow-up question to an answer you provided

Hi Joe,

I saw your response for this question: Can you use PERWT in combination with MULYEAR to obtain weighted data for a single year from a multiyear file?

Could you please explain the rationale for multiplying PERWT by 3? Is it also, then, save to assume that I would have to multiply PERWT by 5 if I were to use a 5-year dataset?



The multi-year files are actually just concatenations of the single year files, containing exactly the same cases and mostly the same information. Every single person in the 2008 1-year ACS sample is also in the 2010 3-year file, the 5-year 2011 sample, the 5-year 2012 sample, etc. When these files are combined the Census Bureau adjusts each person and household weight appropriately. These new weights do include sophisticated adjustments for specific characteristics, which you can read about further in the ACS multi-year Accuracy Statements, but because these methods are similar to the 1-year methods, the weights end up being roughly equal to 3 of 5 times the 1-year weights.

Again, I would emphasize that the best way to get single-year estimates is to use the 1-year files, applying the weights that were created to represent that specific year.

I hope this helps.