ADP Jobs Number Increasing Divergence From Reality

It’s always like that.

Just as the thing you lost is found in the last place you look so economic statistics start to go a bit funny just when you want to use them. We even have Goodhart’s Law about this – when a measure becomes a target it ceases to be a good measure. Sure, not entirely the same but containing the same exasperation with the universe as the first two.

So it is with certain of our estimations of what unemployment is going to be. We have the weekly initial claims numbers, allied with that we have an estimation of the unemployment rate. Then we’ve the monthly jobs numbers which we take to be the accurate ones. The monthly ones are preceded by the ADP numbers which are projections from their payroll customers over the rest of the working economy. All of our different numbers use different sources of information, calculate things a little differently and if we’re honest about it they’re also counting different things. Definitions really do matter.

In normal times we often get pretty much the same answer though – the calculation methods and definitions of each have been tweaked over the years so that they do broadly agree. The time when we really want to use the advance versions of these numbers is like right now, when matters are going all over the place. Unfortunately this also seems to be the time when the numbers themselves are doing just that. It’s difficult, for example, to see the ADP numbers as being in concordance with the monthly official jobs numbers, or even the weekly unemployment claims ones. If that’s so then they’re much less of a guide than they normally are.

ADP jobs numbers

We have those ADP numbers for this month:

Private sector employment increased by 428,000 jobs from July to August

You what? Seriously? But, but…

Quite, but, but. We’ve had weekly claims numbers showing greater decreases in unemployment than that, like here and here. That second one had a reduction in continuing claims of 600,000 for just the one week. OK, sure, that’s for week ending 1 August. And yes, agreed, not claiming unemployment insurance is not the same as getting a job. And yet 600k in a week and 428k in a month?

No, there’s something wrong here. No, this doesn’t mean that someone is doing something wrong. It means there’s something in the system that makes the numbers not comparable. They’re not, for some reason, counting the same thing any more.

This isn’t new

As varied reports are telling us the ADP and the official numbers have diverged recently:

Most recently, the first print on July private payroll gains came in at 167,000 according to ADP, versus the more than 1.4 million private payroll gains the Labor Department reported thereafter.

That’s you know, a pretty big difference. mAnd it does mean that we shouldn’t be taking the ADP number as a good guide to what Labor is going to tell us in a day or two.

So, what’s the difference?

Assuming that we’re going to use these sorts of statistics as a guide then we need to work out what that difference is. Then we can apportion the right amount of our belief to each of the numbers. Again, let me note that I don’t think that anyone is doing anything wrong. It’s that both numbers are proxies for what we don’t in fact know and never will, the exact number of jobs out there. Proxies can wander around in how accurate they are.

This is also not the definitive proof of anything but my opinion is that the difference is here:

One difference is that the ADP NER includes any active employee within the company, whereas the BLS tracks employees only if they are paid during the applicable month.

ADP is counting people on the books. BLS (for the Labor Department) is counting those who got paid. So, the ADP numbers increase by the number of new employees that are taken onto those books. The BLS numbers increase by the number who got a paycheck in the period and didn’t in the one before.

So, what’s happening out there in the labour market at present? There’s that pandemic unemployment insurance program. Which may or may not be captured in both sets of numbers. There will be people laid off and truly laid off. there will also be people who a company says, well, there’s no work this month, sorry, but as soon as there is back you come. So, they don’t wipe them from the systems but there are no checks being run for them.

How much of each is going on, how official is all of it? Who knows?

But the different things being counted would cover several of those situations. Thus we can look on the ADP and Labor numbers as representing rather different things.

The ADP numbers are people newly entered into a payroll system. These are new jobs being created that is. Proper new jobs, not just temporary – even if unofficially – laid off people not expunged from payroll systems. And that ADP number for that is actually a pretty good number.

The Labor numbers are, to some extent at least, both those and also, on top, those being taken back into jobs that they were only out of because of lockdown. Where they didn’t get wiped out of the computer system.

So, ADP is talking to us about jobs churn and new jobs, Labor is that plus people coming back from furlough. Or sorta, because there won’t be sharp dividing lines between each of these classes.

Please note I do not say that this absolutely is what is going on. Rather, that’s the best explanation I can see for what is happening here, given the difference in numbers, they must be counting different things.

The use of the numbers

If the above is indeed true, as I think it is, then the best number for us to use to measure the economy as a whole is the Labor one. That’s what’s showing the total effect of people getting back to work. Sure, we’d like it to be more than a million a month doing so but even then if the rate stays at that level then we’ll be done with the Covid-caused unemployment in 6 to 9 months.

My view

As ever I am thinking about a swift and V shaped recovery. And looking for evidence that tells me it’s not going to happen. The ADP numbers, if they were the full re-employment rate, would be the beginnings at least of that proof. But, as above, I don’t think they’re representative of what’s really happening out there. Again, not because of any mistake but just because of the details of what is actually being counted.

The investor view

Assume that the Labor numbers are correct and the markets are about correctly priced. For they are priced for a reasonably swift and V shaped recovery. If the ADP numbers are the true employment situation, including those returning after lockdown to their previous jobs, then recovery is going to be a long slow grind and the stock indices are over-valued.

At which point each investor has to make up their own mind about which set of numbers to believe and then adjust investment strategy to suit. Good luck and you know from the above which way I lean.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.