Replicating health insurance estimates from Health Insurance Coverage 2015


#1

Trying to understand why I am not replicating some results in Table A-1, found here:

https://www.census.gov/content/dam/Census/library/publications/2016/demo/p60-257.pdf

I am using the 2001-2015 data set; running 2008-2015 single year IPUMS ACS data.

I have the following specifications:

Row: year

Column: hcovany

Selection filter: empstatd(0-12, 20-34) gqtype(0)

Weight: Person Weight

Row Percentages

Weighted

While the national percentages of uninsured for 2103, 2014, and 2015 match table A-1 in the Census Health Insurance Coverage report, the number of uninsured does not.

Can you explain what I might be doing wrong?I would like to replicate this table so that I know that I am using the data set correctly and can then move on to other restrictions/filters.


#2

It seems there is a simple tweak that can be made to your filters which improves the comparability of the online analysis with the table. In table A-1 the entire civilian non-institutionalized population is included. In the ACS data this subset of the population is noted, not only by GQ==0, but also ==5 and 9. Once these additional codes are included you should generate a table with estimates that are within the margin of error of the official published table.


#3

Thanks very much for the quick reply. I am still not quite there. Do you have any other thoughts?

Row: year

Column: hcovany

Selection filter: empstatd(0-12, 20-34) gqtype(0,5,9)

Weight: Person Weight

Row Percentages

Weighted

Percents still match and I am closer on the counts. Here is what I get for national counts:

2013: 45,102,311 A-1 table 45,181 thousand

2014: 36,574,030 A-1 table 36,670 thousand

2015: 29,705,995 A-1 table 29,758 thousand


#4

These are the numbers I calculate as well. It should be noted, that we generally don’t expect to exactly match official statistics with public use data. The numbers calculated, however, do fall within the margin of error of the official estimates.