Estimates and MOE


I am new to IPUMS and was wondering how to use the estimates and the margins of error in ACS 5 yr data 2013-2017 to generate block group level information. Do I need to weight and cluster information?I need to know the number males below 65 years of age in each block grp.

It sounds like you are interested in the population counts of males under the age of 65 at the block group level for the US. You should be able to get this information from IPUMS NHGIS without further processing using weights, etc. if you filter on the following characteristics in the Data Finder:

Geographic Levels: Block Group
Topics: Age and Sex

You can further restrict the list of tables that the data finder displays by entering a year in the selection filters (I will note that in my quick search I only saw relevant tables using the decennial years or 5-year ACS files).

Thanks, Kari,

That is right. I wanted to know the percent of males below 65 in each block group in MA and will be using 5 yr data based. Our study period is from 2008 to 2018. Does that mean I should use 2008-2012,2013-2017,2014-2018 (5 yr datasets). How should one manage the overlap in years?

I also wanted to know the role of the margin of error when we sum across variables.I came across some formulas to be used to create the margin of error for the derived variables, How necessary would this be?


For a good recent discussion on considerations in using overlapping ACS periods, see this post in the ACS Data Users Group discussion forum. In short, overlapping periods can be compared, but they’re harder to interpret correctly, and for small counts, like those in block groups, there’s typically not a strong enough sample for differences between overlapping periods to be meaningful.

Also, for the year 2008 (the first in your study period), the best corresponding 5-year period is 2006-2010, which is centered on 2008.

If you want to compute margins of error (MOEs) for any derived statistics (sums, proportions, etc.) it is very important to use the correct formulas. See Appendix 3 in this Census Bureau handbook for a good explanation with the required formulas for the most common cases.