 # Combining Census samples for a given year

Is there a consensus on whether it is appropriate to combine Census samples for the sample year? Given that the jurisdictions and variables of interest are the same, can I combine, say a 1% sample and a 5% sample to yield a six percent sample? Someone asked a similar question here: Questions about 1970 Census Data

I figure since the samples are constructed independently, using both of them without replacement (i.e. including people/households that happen to be in both samples) would pose no problem for constructing accurate population statistics.

If this is appropriate, how should I adjust the weights? I’m thinking of performing a weighted collapse by sample/year and then just performing a simple average of all the samples in a year.

I would not assume that this is going to be the case for all samples. Each census year, the public use microdata samples were selected using different methodologies. The details can be found at the links here.

In cases where two samples were drawn independently, as in the example you gave, you’d generally want to give approximately equal weight to the households in each sample. For the 5% sample, you’d multiply the weights by 5/6 and for the 1% sample you’d multiply the weights by 1/6. 1 Like

This is incredibly helpful! Thank you so much :