Aug 03, 2012 at 12:56 pm
Any jobs report that comes out a few months before a big election is going to be fodder for politics. And this month’s report is particularly vulnerable because it contained numbers that each party could use to their advantage.
So let’s dive a bit into the weeds and see if there are any good reasons for taking some claims more seriously than others. To telegraph the punch line: the payroll change is statistically significant; the unemployment rate change is not.
Ds and the President are likely to stress the 163K pop on payrolls, while the Rs and Gov Romney will stress the uptick in unemployment from 8.2% to 8.3%.
First, how is it that you can get the two numbers going in different directions (OK, they’re going in the same direction — both up — but you know what I mean)? For that, read Wonkblog’s Suzy Khimm who discusses the two different surveys the BLS uses for the monthly report, one of businesses for the payroll numbers, and the other of households for the unemployment rate.
As I stressed earlier, the payroll survey is more reliable on a monthly basis than the household survey because its sample size is much larger. The household survey goes to 60,000 households; the establishment survey to about 140,000 businesses and government agencies representing about 490,000 establishments.
This means that the statistical reliability of the payroll survey is greater than that of the HH survey. In real numbers, this means, for example, that the change in payroll jobs has to at least 100K for it to be significant. Thus, July’s 163K crosses that bar. (For an employment change from the HH survey to be significant, it has be at least 400K.)
The jobless rate typically has go up by at least 0.2% for it to reach significance. As Khimm points out, if you go out another decimal place, you get the following:
That’s an increase in unemployment of about 3 basis points, or 3 one-hundredths of a percentage point, well short of the 20 basis points (two-tenths of a percentage point) you need for statistical significance. In other words, the change in the unemployment rate from June to July was statistically indistinguishable from no change at all.
Let me be clear and fair. The fact that the 163K on payrolls is statistically significant doesn’t mean we’re into a new, more favorable trend in job growth relative to the last few months. One month does not a new trend make, and these numbers get revised…numerous times!
But I’m sure that campaigns want to be careful in considering statistical confidence intervals.
That’s a joke, son.
Thank you for joining the conversation. Comments are limited to 1,500 characters and are subject to approval and moderation. We reserve the right to remove comments that: