A few thoughts on Georgia’s apparently conflicting employment data

There has been a lot of commentary lately about the seemingly contradictory employment data in Georgia and in other states, primarily in the South.

Buzz posted most recently about the Wall Street Journal article Rising Jobless Rates Are a Southern Mystery. From the WSJ:

Georgia’s unemployment rate has soared [to 8.1 percent] from a postrecession low of 6.9% in April at a time when the national rate, which was 5.9% in September, has been steady or falling. Tennessee’s jobless rate bottomed out at 6.3% in April but rose back up to 7.4% by August. Alabama, Louisiana and South Carolina have seen similar surges this year.

Before considering the data from Georgia, take note of this additional passage from the WSJ:

Jobless rates at the state and local level are determined by a Labor Department model that blends a monthly survey of households with other data and past trends. Their short-run movements can reflect statistical noise more than any underlying trend, and many economists prefer to look at a larger Labor Department survey that estimates nonfarm payrolls to gauge regional trends.

Much of the commentary in Georgia over the last few weeks seems to have missed the key point that the employment statistics that are endlessly analyzed come from two separate surveys. The survey of nonfarm establishments with payrolls is used to determine the headline jobs numbers. The separate survey of households is used to estimate the unemployment rate, the labor force participation rate, and other information about the labor force.

Typically, we see the two surveys moving in more or less the same direction. As the recession took hold a few years ago, we saw the numbers collapse. And then we saw both of them steadily recovering, if fitfully and sometimes inconsistently across the state. In the last few months, the two surveys have painted conflicting portraits of employment in Georgia — the establishment survey looks strong, while the household survey looks really weak.

I blame the Georgia Department of Labor for some of the confusion among commentators in the press. Routinely, the DOL press releases assert a direct causal connection between the payroll estimates and the unemployment estimates. Commissioner Mark Butler has also on a number of occasions blamed weaker data on seasonal changes, even when the data have already been adjusted for seasonality. I sure wish there were more journalists willing to call shenanigans on such lazy characterizations rather than lazily quoting the press releases each month.

Before I move on, it’s worth remembering that these are both surveys that produce estimates subject to revision and error.

While the payroll survey is considered more accurate, the household survey, according to the U.S. Bureau of Labor Statistics, “includes self-employed workers whose businesses are unincorporated, unpaid family workers, agricultural workers, and private household workers, who are excluded by the establishment survey.”

Also, as we wait for the next data sets, it’s worth keeping in mind that on Oct. 16 we’ll see the Sept. payroll survey estimates for Georgia (the state as a whole and individual metros), plus the number of initial claims for unemployment in Sept., plus the seasonally adjusted unemployment rate for the state. It will be another week before we see the numbers used to compute the statewide unemployment rate and before we see the not-seasonally adjusted estimates from the household survey, including the city/metro/county unemployment rates. Last month, too many commentators started speculating about the change in the unemployment rate before the key data had even been released.

Click here to see all the latest Georgia data for yourself.

Click here to see the establishment survey estimates of nonfarm payroll employment in Georgia in August. The numbers there are not seasonally adjusted, so it’s no surprise that we would see a jump from July (a weak month for employment) to August (a better month). We’ll almost certainly see another increase in the September numbers (an even stronger month for employment).

According to the establishment survey estimates, the number of nonfarm payroll jobs in Georgia was 4.13 million in Aug. 2014, up a strong 2 percent from Aug. 2013. (Private employment was up 2.3 percent over the year, while public employment stagnated.)

But the household survey data for August paint a very different picture. According to those seasonally adjusted estimates, we had 4.38 million employed Georgians in Aug. 2014, a year-over-year increase of about two-tenths of 1 percent. Also, contradicting much of the speculation over the last few weeks, the estimates say that Georgia’s labor force is not growing — the number of Georgians counted as part of the labor force (those employed + those unemployed) was up only 4.3 thousand from the year before.

The seasonally adjusted employment rate in Georgia in Aug. 2013 was 8.2 percent; in Aug. 2014, it was 8.1 percent. Throughout the sluggish recovery, Georgia has pretty much been in or near the worst quintile in the nation in terms of unemployment, but the spike in the rate in recent months has vaulted us to number 1.

So which is it? Are we experiencing a fairly robust jobs recovery across the state? Or has the household survey picked up on some new negative trend?

If I had to guess — and it’s just a guess — I’d say that the household survey might have overstated improvement in Georgia’s unemployment rate over the winter and spring, and it now might be understating the improvement. If anything, I’d expect the payroll numbers to be revised upward and for the rate to fall back closer to 7 percent over the next few months. Just guessing.

I’m also wondering if regional variation might be impacting the estimates. I’ve occasionally written here about the uneven recovery across the state. Even though the state as a whole saw strong job growth of 2 percent over the last 12 months according to the payroll survey, nine Georgia metro areas had growth of less than 1 percent (probably not enough to keep up with population growth) or experienced actual declines in payroll employment.

According to the household survey estimates, Georgia’s statewide not-seasonally adjusted unemployment rate was 8.3 percent in August, but the rate was 10 percent or higher in 48 counties and 9 percent or higher in 31 counties. From the Ga. DOL:

Screen shot 2014-10-12 at 1.14.26 PM

Keep in mind that the map will look a little better in September, which is just a better month for employment than August. Also, it’s worth noting the bizarre color scheme — green is worse than pink on this map.

Look at the huge swath of the southern part of the state where unemployment was over 10 percent in August. Under-sampling and/or over-sampling of rural areas of Georgia conceivably could have contributed to swings in the unemployment rate. Also, we have seen payroll jobs lost on federal installations, but those facilities have also had to cut independent contractors — perhaps some of those sharp changes at the local and regional level have contributed to the statistical noise too.

So what will happen later this week when the September unemployment rate is released?

I’m betting that Georgia’s rate will fall, maybe by a few tenths of a point. I’m also guessing that we’ll still have one of the highest unemployment rates in the nation, as we have for a number of years.


  1. greencracker says:

    I love a good post with some nerdy explanation of stats.

    I look forward to when everybody pays for the news they read so that we won’t have to depend on volunteer journalists to report such stuff.

    After all, one man’s “lazy” journalists is another man’s “you know there are less than 50 reporters statewide covering state government, and most of them don’t even do that beat full time, right?”

    I’m glad we have volunteer journalists, especially those well versed in statistics. I’ll be gladder when people get paid for such work.

  2. objective says:

    this is one of the more wonkiest post i’ve ever read on PP. i like it a lot.
    i think you’re right to say regional variations account for some of the discrepancies.
    overall, it also just paints a mixed, middle-of-the-road picture of employment in GA. there’s a noticeable but not so big increase in firm growth, but also a noticeable amount unsatisfied job seekers. these two numbers can rationally coexist.

Comments are closed.