**Q: **

I am using the new 5 year ACS data to look at the number of vacant homes in Fruitport Charter Township. Table B25004 shows 111 housing units classified as “other vacant” which I have been told includes boarded up, foreclosed homes. The margin of error for this category is +/- 97 housing units. What should I do?

**A:**

The ACS is a sample so you should start out by looking at how many households or people are actually included in your sample. You only have 458 in your sample (B00002) and you are looking at the tail of the distribution as most of the housing units are occupied. There are appendices at the back of the Compass Series that illustrate how you can combine cells and re-calculate standard errors. So, in your case there might be several townships within Muskegon County that make sense to combine. Here’s a link to the appendix snagged from one of the Compass handbooks:

http://www.psc.isr.umich.edu/dis/acs/handouts/Compass_Appendix.pdf

The formula for this is on page 14 of the appendix. It is something you can do in Excel pretty easily.

The Census Bureau also recommends that you keep your coefficient of variation below 10%. The coefficient of variation is:

Standard Error/Mean [page 13 of the appendix shows how to calculate your SE from MOE]

If you are using a proportion, as it gets closer to 0 the more unstable your coefficient of variation will be. You’ll be safer with proportions that are in the neighborhood of .35 rather than .09.

I’ll check on the definition of “other vacant.” It could well be where foreclosed homes fall. I’m not real familiar with that item. Looking at a place like Las Vegas with high foreclosure now vs 2000 might help you clarify that this code is indeed measuring foreclosure.

Something else you might find useful is an annotated link to the ACS website.

http://www.psc.isr.umich.edu/dis/census/webinars/web1/annotated.html/