More Detail on Working the Refs

So there are several comments to my previous post. Ignoring the a good one from Dr. DeLong, several people are taking umbrage at my unsubtle suggestion that the effect on employment being suggested is, to be polite about it, rather creative.

kharris begins, “So let me see if I have this right. If anybody tries to figure out what the impact of snow on economic data might be, they are big fat liars? But those who know that the economy is in bad shape, without reference to actual events, is a stand-up kind of hack?”

Following is an expansion of my comment in that thread, with data:

To the second question, well, I may be a hack, but my stand-up days are in the past. But given the choice between believing that the recovery is in full swing and that long-term unemployment is getting worse and jobs are not and will not be created, well, I’ll take the CBO projection as the baseline:

CBO expects the unemployment rate to average a little over 10 percent for the first half of 2010, and it will probably not dip below 9 percent until 2012.

and note that if we’re calling that a recovery, our definitions have become Very Generous. So bold claims of recovery need to be tempered by the prospect of worse headline unemployment (U-3) for the next five months (including February) and no significant recovery for the eighteen after all.

Sorry I’m not doing handstands that GDP might be slightly positive for a few quarters of sub-replacement level employment increases, but I didn’t cheer the “recovery” of 2002 either, so at least I’m a consistent hack.

To the first: Not at all; trying to figure out the effect is fair game and perfectly reasonable. But the declarations so far are all running in one direction: we believe the economy is better than the data will be, so we need to wait if it looks bad. (See Ms. Caldwell as quoted by CR or Catherine Rampell, for example.) Rampell:

That report will probably be very, very ugly. I have seen some forecasters project job losses as high as 100,000.

The main culprit behind the expected jobs plunge is the blizzard, which closed businesses and kept people from going to work or even seeking work for days and sometimes weeks. These work stoppages probably occurred precisely when the government was collecting data for its February jobs report.

So the current estimates are all that (1) demand was down and (2) employment was down.

And (3) deliveries were down: see the ISM data.

Put it all together, and you can tell a story of heavy snow snarling shipments to and from manufacturers, slowing down production growth.

But at least in this case, we have a clear indicator: the increase in backlogged orders.

Finally, (4)savings.

The reasons for the stall are twofold: For one, rebounding wealth since the recession’s depths has helped provide some support for consumer spending. Secondly, weak income growth has left other consumers with little choice but to spend proportionally more of their incomes, particularly in light of [5] still-tight credit conditions.

So demand, supply, savings, credit, and employment are all down. The first and second are aberrations of snow (and equilibrium), the second and third abide.

Which leaves employment, which is discussed in more detail than most sane people would want below the fold.

Now, it is clear that people who are employed did not work in the week. But they are not likely to have reported themselves as “unemployed” or (except in a very literal sense) “out of work.” True, they did not produce—but what they would have produced was not bought, and hence there is a backlog of orders.

But companies that now have backlogs of orders know that this was because they did not have their current workforce. Accept an order to produce, say, 200 units (which takes a month to produce) and lose five to eight business days and you’ll be 50-80 units behind.

But you’re not going to go out and hire a new person to fill the backlog.

Yes, there was an effect on production and sales. But the idea that 100-200K jobs went unfulfilled solely because of weather conditions that were aberrant primarily in the mid-Continent is either (1) rather optimistic or (2) ignoring that the excess snow effect was mostly in the areas that are least underemployed. (See this nice map from Catherine Rampell)

So in the best case scenario, the recovery was muted because things were not delivered or sold—though money (savings) was (were) spent. And the only reason firms didn’t hire was the snowstorm that closed D.C. and delayed Philadelphia. (Though there was no snow in NYC and, as noted, nothing unusual about the fallings in the Midwest.)

The worst case scenario is that demand wasn’t filled solely because supply wasn’t available because existing workers could not produce. Working on the “nine women pregnant for a month don’t produce a baby and you have a real problem eight months thereafter” rule, employers will (generally correctly) view their February backlog as a result of existing labor not working, not as a need to hire new workers.

If you’re balancing the effects of those two—standard Slutsky analysis, as it were—there is a high likelihood that hiring will be dampened going forward by the snowstorm as firms underestimate actual demand. It is less likely that actual hiring was significantly reduced by it.

But that’s not the way the discussion is going. So a bad (negative) number has excuses, a poor number (positive, but less than replacement rate) has excuses and should be seen as “good,” and a good number (replacement rate or better) will mean “all ahead full.”

So I tried looking at ancillary data. Looking at power usage, for instance, indicates a major decline that would correspond to less activity(Table 1.6.b; Commercial usage YOY down 3.6%; Industrial usage YOY down 5.6% with declines in all areas; total usage down 4.3% YOY [Table 1.1])—but that’s only through November.

Maybe the past three months have been part of a miraculous recovery. But it’s not in employment, its not in the available energy usage data, and it doesn’t follow from the ISM data, which indicates slow growth at best.

Those who want to claim the economy is recovered have been, as noted, “working the refs.” So a bad number (by Rampell’s apparent reasoning) will kill health care reform, but not mean that we need a second stimulus—even though the states are hemorrhaging money and, soon, jobs. (Teachers, police and fire–you know, all the nonessential personnel.)

It’s a heads-we-win-tails-we-win-more situation being set up.

If we pretend that all of the argument are true: that the snowstorm was a once-in-a-lifetime event and that it really did produce a major skew though, we might want to look at what happened the last time a “once-in-a-lifetime event” occurred near the end of a recession.

The vertical lines are at September and December of 2001. For a week in September, everyone—and this time I mean everyone, not just the bottom third of the Bos-Wash corridor—stopped shopping for a week. As predicted above, the employment effects abided for at least the next few months. (Recall, after all, that that recession officially ended in November.)

Given the choice between (1) assuming that there will be a one-off decline in employment due to the snow and that everything will return to recovery next month or (2) that there will be a lingering, negative employment effect from the snowstorm and attendant business slowdowns, there appears to be only one way to bet, given the data and the history.

Yet the calls right now—absent evidence—are going the other way.

If we’re working from anecdotal evidence, then certainly there is a recovery. It’s the extant data that doesn’t support any recovery that is not also described as “jobless and uncertain.” That may change on Friday. But it’s not the way to bet, no matter how much the refs are worked.