In pooling observations across multiple comparable surveys (e.g. CPS Basic Monthly Jan 2020 and CPs Basic Monthly Dec 2019), is there any need to adjust the sampling weights? I am trying to draw inferences about the population during the period of Dec 2019 to Jan 2020 and was concerned that perhaps something like the size of the overall population changing across both months would prove a problem. Note that I understand that if I wanted the months to be counted equally (i.e. an average of Jan and Feb rather than the pooled analysis), I would THEN need to reweight to ensure the total weight for JAN observations and DEC observations are equal.
The mean of the variable WTFINL varies across the CPS Basic Monthly Surveys. For example, the average for the January 2020 Basic Monthly Survey is 2764 while the average is 2756 for December 2019. Shouldn’t it be constant across months? Maybe it is the share of people in oversamples that is driving this?
Given WTFINL is supposed to be a sampling weight (with some adjustments) shouldn’t sum(1/WTFINL) = 1? I get much larger numbers.
In order to answer your questions, I will provide you some general information about WTFINL and weighting the CPS. Mean WTFINL is different for each sample due to the fact that sample size and general population size changes each month. This means that, in order to pool samples, all you need to do is divide the weights by the number of samples since the population changes have already been taken into account during calculation of the sample weight. For more information about how to use sample weights with IPUMS data, see this blog post and for more information on CPS weights, refer to Chapter 10 of the Technical Paper 66.