What happened to the jobs? Did the economy really crash down on job creation in March? If so, what’s to blame – higher gasoline prices? Is seasonal adjustment playing a role? These are the questions that the mainstream press is asking after a weak addition of 120k jobs were added in the March report from the Bureau of Labor Statistics (BLS).
For all the hand wringing, one possibility never comes up – that the data are simply whacked. That may seem unlikely, but given that a different labor report comes out with a potentially better method of collecting the data you never know. Let’s be fair and call it a bit of a mystery for now. Here’s why:
First, there is the BLS report and its methods. This is created by a survey of 140k businesses, which is to say that an army of BLS employees does nothing but call up employers and ask, “How ya doin’?” In March they averaged less than one job gained per phone call. More interestingly, they say a decline in retail employment of –34k that was offset by a very big gain in manufacturing of +37k. According to their figures, one third of the net gain in jobs came in manufacturing, a big change.
Two days before this report came out we had the monthly report from ADP, a payroll processing company. Their methodology is complex, but the short version is that it is based on actual payrolls that they are handling, combined with a few corrections. Their report showed a net gain in March of +210k, about the same as February. They do not report out retail separately, but they showed a net gain of only +23k in manufacturing, its allotted 10% of the economy. What they do separate out is the net gains by company size, with large firms (more than 500 employees) adding 22k, medium (50-499) 87k and small (>50) +100k.
Why the discrepancies between the two? First of all, this divergence between the two reports happened last year at the same time. A strong Winter quarter led to a weak Spring, according to the BLS, but a continuation through in the ADP report. The chart below from the Federal Reserve of St Louis shows the monthly net change in both reports since the bottom in early 2010, BLS in red and ADP in blue:
You can see right away that both of them are a bit jumpy, but the BLS report (red) is by far the noisiest of the two. Where the ADP report generally shows a net gain of 200k over the last 18 months, excepting last summer, the BLS figures are all over the place.
Why might this happen? I think the answer is in the ADP data itself and the BLS methodology. ADP shows that half of the job growth came at small businesses, exactly what you would expect during an economic restructuring at the end of a Depression. Are these the firms that the BLS calls every month to find out how things are going? It’s very likely that they miss a few of them, since some are probably very new. And that’s the likely source of the noise in their data – a survey, even a very large one, is not catching as much as a big payroll run can.
This means that the net gain of a measly +120k that sent the stock market swooning and sent reporters scrambling to explain what’s up is probably just a statistical blip. While it is possible that ADP is over-stating job growth a bit and may have its own correction coming, job growth has likely not slowed as much as everyone fears. It’s just hard to find where the new jobs are coming from because, as you might expect, it’s not from the same places that the BLS is used to looking.
No matter what, however, we can say that job growth – even at +120k – is leading economic growth dramatically. By any measure we should be lucky to hold even right now with a projected gain in our Gross Domestic Product of 2.0% this last quarter. Job growth almost never leads economic growth, so it goes without saying that something unusual is happening. That fits with the observation that growth is happening in the smallest companies and away from the BLS survey’s usual cast of characters.
So what does the jobs report mean? There is continued weakness among established employers and there could be some trouble in retail especially. But overall we are probably still gaining through economic restructuring and growth among new companies. It’s a good thing.
So basically you think it’s a statistical blip?
Yes, that’s the short version. But would you believe me if I just said that? 🙂
I read so many different reasons why there was a downturn. Every article had a few more. Gas prices was probably the most often blamed but seasonal adjustment came up a lot too. No one ever explained anything. The idea that it might be just a fluke is really funny when I think about it because it would make as much sense as anything.
All that speculation is pure crap. Reporters never credit it as speculation, which is all it is, when the market falls or some piece of data comes it differently than expected. It’s just nuts. Here’s the inner workings of the two, you decide for yourself, I say. The most logical conclusion is that we’re gaining jobs but it may be softening up some so watch out.
The blue line does look like a trend upward with some noise where the red line looks like a lot of noise and maybe a small up trend. I see the frustration in having both of these.
You picked a point where they both start at zero which is much better than most of these Federal Reserve graphs. We have seen a statistically significant improvement in job growth it seems by either measure. Can you run a regression?
That’s my call as well on the two. I didn’t run a regression because there is a very real up and down to job growth with seasons and what-not, so I don’t know that the variation would mean anything at all. It makes sense to me to start at the low point, Jan 2010, but that’s not to say there haven’t been several inflections since then. I don’t think a regression would tell us an awful lot – especially since, as you can see, the two charts would almost certainly give us different trends.
I’m thinking about it. Just wasn’t an argument I wanted to lead with or spend a lot of time on.
So you say there is job growth and the mainstream media is just wrong?
Well, yes. They are wrong to the extent that they say there is some kind of trend from this one month of datum that might not even be a good number in the first place. It’s a lot like reporting on public survey polls – if the margin or error is 4% and there is movement of 2% from one poll to the next, there is not actually any movement at all as far as you know. This just seems to be well within the noise, given the methods and the history, and the other method gave a different result.
So it’s “wrong” to say that job growth is slowing, yes.
If the ADP report is so much better why don’t we use that one? Why are there two jobs reports anyway?
The ADP report is less than 6 years old and not “official”
Why Wall Street doesn’t focus on it more is a bit of a mystery, but the “official” number is probably always going to trump anything. That “Establishment Survey” has been going on since 1939 and is traditional at this point. We have a lot of data through many recessions and it’s kind of handy. But … the modern ADP number is clearly a better method and the data seems to be much more consistent from month to month, as you would expect.
I hope that puts this to rest. We’ll see, tho. 🙂
OK, I ran a regression, fitting a straight line to each. The BLS has an R squared of 0.07, which is to say it really is just about all noise – job creation in thousands is y = 4.9x + 66 (where x is the months after Jan 2010). That it doesn’t intercept at zero is a problem, since that month was indeed the inflection point in job creation / loss. But we’re accelerating 4.9k jobs per month.
Doing the same for ADP we have an R Squared of 0.51, which is still lousy but much better. y = 7.0x + 33 in the same way, closer to a zero intercept, and we’re accelerating by 7k jobs per month.
By the ADP report, we’ll be creating about 270k jobs per month by the election if this trend continues (which, given the quality of the fit, seems unreasonable as all Hell).
do you ever get the feeling that no one knows what the hell they are talking about?
Pingback: Five Years On | Barataria – The work of Erik Hare
Pingback: Job Dynamics | Barataria – The work of Erik Hare
Pingback: GDP: The Bet | Barataria – The work of Erik Hare
Pingback: Tale of Two Reports | Barataria – The work of Erik Hare