The ongoing survey program by the US Census Bureau, the American Community Survey (ACS), provides a great opportunity to examine detailed and current trends in the USA. The American Community Survey has been said to contain “warm” (more current), and “fuzzy” (less precise) data, compared to the April 1st of each decade’s Decennial Census. (The Cold, Hard Facts!)
You won’t see mention of “standard errors” and “margin of error” in the Decennial Census data. It’s a 100 percent “sample” (that is, a “census”) without standard errors. (There are other errors associated with Decennial Census data, but that’s another story.)
I didn’t invent the metaphor of the ACS as a “warm and fuzzy” database. I think I first came across this in an article:
- Heather MacDonald. “The American Community Survey: Warmer (More Current), but Fuzzier (Less Precise) than the Decennial Census” In Journal of the American Planning Association, Vol. 72, No. 4, Autumn 2006, pp. 491-503.
Unfortunately articles in scholarly journals tend not to be “free”, so I can’t provide a link to a free PDF version of this article. It might be available at a local University library.
I do like enjoy the concept of the American Community Survey as “warm and fuzzy.” The data is “warm” (sometimes even “hot” news!) since the most recent data (as of today) is from 2015. The most recent decennial census (April 1, 2010) seems stale in comparison!
But “fuzzy” is a little harder to appreciate. The US Census Bureau takes extreme care in presenting the standard sampling error associated with the American Community Survey. Most of the estimates reported in the ACS have a “MOE” (Margin of Error, 90 percent confidence level) provided alongside the estimated value. These MOEs are critical in determining whether or not Area “A” is better/worse/higher/lower than Area “B”, or if Area “A” has significantly changed between year “20xx” and “2015”. Well, they could be critical, but sometimes it might get out of hand — “eyes glaze over” with too much talk of tests of statistical significance.
I’ll try to be careful, and just reach into the statistical bag-o-tricks when trying to make a point about comparing data: year-to-year changes, or differences between areas.