It’s become poor form to take the jobs report at face value, and every financial blog out there is doing its best to reveal “the truth” about the misleading numbers. Most recently, the headline is that the 431,000 jobs which were added to non-farm payrolls included 411,000 temporary census-related jobs. Quick arithmetic reveals that this means only 20,000 jobs were actually added.
Should the market care?
The market doesn’t respond to the number of jobs. Instead, it responds to the difference between the reported number of jobs and the market’s expectation. A jobs report of 50,000 could result in either a massive rally or a crash, depending on the expectation. Here, 536,000 jobs were expected to be created, so the market should fall whether it was 431,000 or 20,000 – both miss the mark (albeit by different amounts).
But we all get that the government numbers mess with our employment perception, that’s established. What I really want to know is if the 536,000 jobs that were anticipated already included the 411,000 government jobs? I would expect that they did, because those forecasters aren’t doing their jobs otherwise, in which case it is wrong to compare the survey result to the 20,000 number. Both should include the government boost. However, if forecasters were slacking off – and who knows, they may have been – then it would indeed be correct to point out the emperor’s missing jobs.
And now back to your regularly scheduled market crashes.
This is a huge revision:
The so-called “benchmark revision” that was announced today will not formally be incorporated into the job figures until February, and could be revised. But the figures indicate that last March the government overestimated the total number of jobs by 824,000, or 0.6 percent. Its overestimate of private-sector employment was even greater — 855,000 jobs, or 0.8 percent.
It is most likely that the culprit is the infamous birth/death model that the BLS uses to adjust jobs numbers to account for businesses opening and closing. Unfortunately, the statistical model only works (to the extent that it works at all) in “normal” times; it is counterproductive when dealing with tail events. Sound familiar?
I find this graph very interesting, not just because of any implied political statements, but for how it highlights the absurdity of economic forecasting and the potentially misguided trust we place in such numbers.
The blue lines were circulated by Obama’s economic team when they were pitching the stimulus bill in order to illustrate its beneficial impact on national unemployment. The red line is the realized unemployment rate to date.
There are two ways to read it, depending on your objective:
- Obama’s economic team was overly optimistic, underestimated the severity of the crisis, and the stimulus plan has failed to help as advertised.
- Obama’s economic team was overly optimistic, underestimated the severity of the crisis, but things would have been much worse without the stimulus.
Ultimately, the question is whether the level or the shape of the graph is more important. Personally, I find it surprising that (as with the bank stress tests), a situation which was markedly better than a worst case scenario was used to demonstrate the effects of the stimulus. Nonetheless, the fact that this graph was used for demonstration purposes makes it difficult to fault simply because it was plotted 1% too low.
Perhaps it never should have been circulated in the first place. This raises a very touchy point in forecasting: an expectation is almost never perfectly realized. Unless an audience comprehends that fact, then putting a forecast out there can only lead to critique. In a simple example, if I calculate a distribution of outcomes and know it to be the correct distribution with high certainty, then my forecast will be the mean or expected value. But what are the chances that the mean is actually the realized outcome? To be sure, higher than any other single observation, but relatively small nonetheless. This speaks to the importance of confidence intervals and margins of error; my guess, however, is that the margins of error on this graph (however that might be measured) would have included the “improvement” line, making the difference not statistically significant.
More pointedly, however, the stimulus was supposed to “save or create” 4mm jobs. This means that the area between the two curves equals 4mm, but the implied difference here seems much larger to me.
The White House is praising the stimulus bill, saying that it has created 150,000 jobs in its first 100 days.
Unfortunately, in the last 100 days 9,000,000 people filed jobless claims for the first time. Thus, the stimulus bill has offset less than 2% of the jobs lost, or – to spin it positively – has created almost 2 jobs for every 100 people fired. (Yes, there’s a distinction between layoffs and initial claims. I’m not making it.)
Obama’s advisors expect the bill to ultimately create 3-4mm jobs (i.e. 33-44 day’s worth of losses). Some conservative commentators are using the information that 14% of the stimulus funding has been spent to extrapolate 150k jobs linearly and suggesting only 1mm jobs will be created. I think that approach is a bit ridiculous, but I nonetheless suggest refraining from celebration until there is something worth celebrating.
98% of people who have recently been fired probably agree.
Durable goods +1.9% vs 0.5%! Durables ex-transportation +0.8% vs -0.3%! Initial jobless claims 5k under expectations!
The headlines looked pretty good at 8:30 am, until you took a look at the revisions. Indeed, today’s economic numbers serve as yet more examples of creating the false perception of growth by changing the past.
Durable goods orders were expected to rise 0.5% in April after falling -0.8% in March. The April number blew by the expectation, coming in at +1.9%, but the March reading was revised down to -2.1%. Let’s re-normalize the durable goods index at 100 in February and follow the chain: analysts thought the index dropped to 99.2 in March and expected it to rise to 99.70 in April. Today’s report demonstrates that the actual path was 97.9 in March and 99.76 in April. Thus, the endpoint of the actual path ends up being only 0.06% higher than the expectation, not 1.4% as the headline would suggest.
Durable goods ex-transportation is even more dramatic – the April number was 0.8% vs an expected -0.3%; March was revised from -0.6% to -2.7%. I won’t bore you with the index path, but the result is that the endpoint is 1.33% lower than the expectation, not 1.1% higher!
Initial jobless claims beat the expectation by 5,000… after the previous month was revised to be 5,000 worse.
The 10:00 am headlines weren’t much better. New home sales missed expectations by 8k and the March figure was revised 5k down as well. February sales were revised upward 4k, which only made the March number worse. The MoM % difference statistics that go hand-in-hand with the absolute numbers are poor, as you might expect, especially given the February revision. But the most amazing thing to me is the confidence interval around these statistics: about plus or minus 15% (pdf link)!
It continues to amaze me that the market takes these numbers so seriously, given their inherent statistical variability. Moreover, the tendency of the market to look at headlines and ignore revisions speaks to investors’ myopia, or at least the overwhelming influence of business commentators on TV, who never seem to grasp the revision concept at all.
And I haven’t even addressed mortagage delinquencies, which continue to skyrocket…
A lot of headlines this morning are noting that the jobs report is “in line with expectations”, with a change of 663,000 vs the survey median of 660,000. What gets ignored is the revised number. For 13 straight months, the prior month’s number has been revised lower by a significant amount after the fact – ranging from 50,000 a year ago to more than 100,000 in recent months. The following is a chart of the reported number (white) vs the revised number (red) (click to zoom). The green line represents the survey.
How this trend gets ignored by the market is beyond me. I’m not sure if it’s systemic over-optimism built in to the reported number (read: arbitrary massaging of the data) or what, but it’s a bit ridiculous.
We generally assume that the market “prices expectations”, meaning if data is better than expected the market rallies; if not, if falls. But one thing the market fails to do is price expectations of back-dated revisions. If today’s number were announced as 150,000 more jobs lost than expected, we’d have a full scale rout on our hands. But hold off a couple months and make that announcement as part of an after-the-fact revision and the reaction is notably dampened. Indeed, in February the January number was revised down by around 60,000 and this month it was revised down almost another 100,000.
The sum of the failures to account for revisions like these, in my opinion, is an enormous sword hanging over the market. It may not take much to release it.
