State of the State Poll (Georgia)

So Georgia College has released its "State of the State" poll. I got there by this story via Peach Pundit. It's a nice summary. You can see the actual poll here.

Before we get to the findings -- which you can read on the Pundit site -- let's look at the methodology. This is a survey of 500 Georgia adults between February 9-16 using landline and cell phones. The data were weighted to reflect the accepted population parameters. In other words, it looks like a good piece of work.

Here's the deal. The poll breaks down some questions by race and ethnicity (white, black, Hispanic). But it reports a 4.4 percent margin of error.

No. No. No.

That N=500 is far too small. Figure you have, at most, a hundred or so blacks (and fewer Hispanics). The margin of error for 100 respondents is about 10 percent, rendering analysis kinda uselessness. This makes that "most trusted level of government" chart that breaks it down this way kinda hard to interpret. In most cases, it's a statistical tie. The same caution goes for examining differences between self-described liberals and conservatives.

Okay, but how about the results not broken down that way? The poll notes it's pretty much a tie. Peach Pundit, however, says:

Most respondents said they trusted the federal, county and city governments more than the state government.

But look hard at the numbers. Asked the most trusted level of government, all respondents, 25.6 percent identified the federal government while 19.9 percent said state government. Add 4.4 to that 19.9, subtract 4.4 from that 25.6, and, yup, they overlap. That means the results are within the margin of error. Easily. And the other levels are even closer. You have to be careful not to over-interpret results.
 
I'd love to break this survey down more, because there's some interesting stuff, but gotta go do a hospital thing. Sigh.