I pooled IPUMS-CPS data from 1996 to 2012 to obtain a large enough sample for my research paper because I am interested in small immigrant groups. I am doing a person-level analysis, and to avoid double counting I am using the March supplement of 1996, 1998, 2000, 2002, 2004, 2006, 2008, 2010 and 2012.
However, I am a bit confused about how to properly set up the weights for my analysis. I want to produce means/proportions and a regression analysis.
I read that replicate weights were added beginning in 2005, and the code to use the weights is the following:
Whether or not to use replicate weights is a choice best made by the researcher. The IPUMS-CPS FAQ page on Replicate Weights (which you may have already seen) lists some of the benefits and disadvantages. If you do decide to use replicate weights, the provided equation represents the STATA equivalent of the SAS procedure described by the Census Bureau (in their document “Estimating ASEC Variances with Replicate Weights” available on the IPUMS-CPS FAQ page) for working with Replicate Weights. Also, since you are you using years prior to 2005, you would have to separate those years into a different pooled group because they do not have replicate weights.
If you choose not to use Replicate Weights, the standard weight to use with supplement data is WTSUPP (the description also mentions other weights that you should consider, depending on the variables you are interested in). The description also contains a brief explanation of how the weights were calculated. All of these weights are representative of the year in which they were drawn, so combining 9 years will give you a population roughly nine times the yearly average. I hope this helps.