What weights should I use to calculate state*year statistics combining info from the March & Oct supplements?

I am interested in looking at the share of young adults who have health insurance and who are enrolled in school, so I would like to combine variables from the March supplement & Oct Education supplement. I am only interested in statistics at the state*year level, so I do not need to match individuals over time. Is there a way to weight data from combined these two supplements to yield representative state*year estimates? Thanks!

Within each March and October sample, individuals are weighted to be nationally representative (using WTSUPP). Combining a single March and a single October sample and summarizing the data by year would result in a weighted population roughly twice the size of the actual U.S. population.

For your particular interests, looking at a State*Year analysis, I would recommend creating two datasets. One dataset with data from March Samples only summarized to the State*Year level using WTSUPP. Then create a separate dataset with data from October samples only summarized to the State*Year level using WTSUPP. This should result in two datasets, each with a single record foreach State Year combination (i.e. total number of Young Adults with Health Insurance in New York 1999 in one dataset, and total number of Young Adults Enrolled in School in New York 1999 in the second dataset). You could then merge these two datasets on State and Year to get State*Year summary data.

I hope this helps.

1 Like