ACS Data Users Group

Wondering if it is statistically valid to average data from two or more 1-year supplementals. If so, can I also average the MOE or would I use another approach to find the combined MOE?

Purpose: I'm doing data visualizations of socioeconomic changes over time in a 5-county area, and each of the counties are 20-30K population.

I'd like to use data from five periods:

2000 Census

2005-2007 3YR

2008-2010 3YR

2011-2013 3YR

And ... since the 3YR was discontinued... in October when the 2016 supplemental is released, I'd like to combine the 1 YR supplementals for 2014-2016 to make a 3YR for that period.

Thanks for any feedback on this.

What you are proposing is a viable solution, that is combining three one year ACS values into a single value for the period. I believe the Census Bureau has explored this approach and found that, while not ideal solution, doing so using the proper techniques will produce at least moderately reliable MOEs. I don’t have the reference handy, but I think that guidance about how best to go about this exists somewhere on the web.

Cliff Cook

Thank you, Cliff! Much appreciated. So far I haven't found guidance online, but I asked the Census PIO for help too.

I think David Hubble gave a presentation on this at the recent ACS Data Users conference. A PDF and video can be found at https://acsdatacommunity.prb.org/p/2017_acs_conference

Thank you, David! I just went through Hubble's slides and it focuses on the problems with aggregating supplementals. Although he made suggestions for solutions, it appears that, for now, what I want to do has too many problems to be viable.

Hi Lisa -

There is also a more focused webinar by Alicia VanOrman of PRB, Nate Ramsey of US Dept of Education, and Dave Hubble... That webinar left me feeling more confident in options for aggregating multiple years of data:
acsdatacommunity.prb.org/.../custom-multiyear-estimates-webinar

Also, for combining the margins of error across three years of estimates, I'd follow the instructions found on page 2 here:
www2.census.gov/.../2015StatisticalTesting1year.pdf

Hope this helps!

Beth, this is great - just what I needed! Thank you so very much for the links.