When the official data isn’t good data

As I was retiring, the American Community Survey(ACS) was replacing the long-form Census questionnaire.  There is merit to the argument that a survey can provide data that is as good as a form that one out of six people fill out – both are, after all, actually surveys.  Still, as a rural sociologist whose primary duties were rural demography, I wasn’t comfortable with the American Community Survey results – the sampling size was too small.

Now, I can access data that compiles five years worth of estimates – so here is some data on Rexford, Eureka, Fortine and Trego, by zip code, in two separate five-year conglomerates:

2012-2016	Rexford 59930	Eureka 59917	Fortine 59918	Trego 59934
Total Population	566	4,425	584	562
5 to 9	21	282	34	52
10 to 14	35	305	22	49
Median (Average) Age	58.0	43.6	50.7	49.2
Per capita income	$22,377	$18,799	$21,203	$25,999

2015-2019	Rexford 59930	Eureka 59917	Fortine 59918	Trego 59934
Total Population	658	4,769	747	476
5 to 9	–	532	12	–
10 to 14	50	216	35	20
Median (Average) Age	56.0	46.8	47.2	60.4
Per capita income	$13,438	$22,867	$28,753	$26,671

The American Community Survey is a well-conducted survey.  The data is correct, in both cases, within the limits of the survey.  The small samples, however, can create some large swings and make the data less useful.  I have been looking forward to reviewing the Trego data since I was selected to return the ACS survey.  Trego’s median age went up 11 years.  The population dropped by 15%.  The youth population plunged.  Meanwhile Fortine incomes increased by 36%, as Rexford plunged into the depths of poverty.  All the survey data is correct – but sampling bias, due to the small number of participants, has given us data we can’t use.

I still prefer the old, time-consuming long form results over the ACS.