Rosenberg Deconstructs The Unemployment Number

From David Rosenberg of Gluskin Sheff


There were so many cross-currents in the January employment report that it is next to impossible to make book on the data. Bullish or bearish, there was something for everyone. All I can say is that the data were not as weak as the disappointing headline would suggest, but there was nothing here to suggest that the U.S. labour market is progressing at anything remotely close to resembling a normal post-recession recovery, even when benchmarked against the past two jobless recoveries in the early 1990s and again coming out of the tech wreckage a decade ago. Fed Chairman Bernanke highlighted the lack of impetus in the jobs market as a chief source of concern and there was nothing here that will help alleviate that, sad to say.

The headline nonfarm payroll report came in light at +36k, well below consensus views of 146k and whispered numbers ahead of the report that were bordering +180k. Not only that, but adjusting for our estimate of what the Bureau of Labor Statistics (BLS) birth-death model artificially added, the headline would have been -52k! (Of course, because this was such a give-and-take report, the bulls would retort that there was also +75k of upward revisions to the back data).

The Household survey also showed a more modest 117k jobs being created last month versus +297k in December (all of the gains and then some were in full-time positions — for a change!). So at face value, the labour market hardly strengthened in January despite all the anecdotal evidence to the contrary (though the level of jobless claims is still at levels that would characterize fairly tepid job market conditions).

There is no doubt that the weather exerted a major impact — wouldn’t the consensus have realized that ahead of time? It’s not new news that January was a terrible weather month and that the data would be impacted. Then again, yesterday all we heard on CNBC was how the chain store sales data were unaffected by the inclement weather, but somehow the labour market was! Go figure. Maybe instead of looking for work, people were choosing to stay warm in the malls and spend their extended unemployment insurance cheques and newly received payroll tax deductions. What an economy!

Back to the weather. First sector to gauge here would be construction and it posted a 32k decline. Then again, this area of the economy is still stuck in the mud and not all of this was due to the snow, ice, and sleet. Transportation was off 38k and this too is likely reflected by the bad weather.

In total, the BLS reported that 900k people did not make it to work at some point during the month due to the weather. This is well above the norm for January (it is typically 450k) although many of these folks would still have technically been counted on the payroll even if they missed a few days of work. It does pay to note that in the past, when the number of people not at work due to weather was this large in January, we almost always saw a February bounce back — the only time this did not happen was in the early-1980s recession:

  • January 1996: over 1.8 million people not at work due to weather and NFP dropped 19k that month (but in February there was a 434k bounce back in NFP)
  • January 1982: nearly 1.1 million not at work due to weather; January NFP was -327k (February was -6k)
  • January 1979: close to 800k not at work due to weather; January NFP was 137k (February was 243k)
  • January 1978: over 800k not at work due to weather; January NFP 187k (February was 353k)
  • January 1977: 1.1 million not at work due to weather; January NFP was 244k (February was 295k)

On average, February saw a rebound in the headline number of 260k and on a median basis, 295k. So the real picture will be painted when the February data are released — the onus will be on a hefty rebound. Absent that, it will be safe to say that what we have on our hands is still a very lacklustre labour market backdrop.

The drop in the workweek to 34.2 hours from 34.3 hours was also clearly weather related and should rebound when the February data roll out. Once again, the decline here was largely concentrated in the construction sector, which obviously would be hardest hit by the snow storms — the workweek shrank here by 2.6% which last happened in January 2007 (rest assured declines like this tend to occur this time of year).

The bulls will point to the one sector that was relatively unaffected by the weather, which was manufacturing. The conditions in this sector were quite buoyant, confirming what we saw in other indicators through the month. Factory payrolls were up 49k, the third increase in a row and even skeptics like us would have to sit up and take notice that this was the best tally since August 1998. Overtime hours on manufacturing also edged up.

The manufacturing diffusion index soared from 61.7 in December to 69.1 (it was 45.1 in October!), the highest level since the tech boom was in full gear back in January 1998. Too bad this sector today only makes up about an 11% share of the economy because it is one of the few that is really humming right now. Based on our back-of-the-envelope calculations, the diffusion index in the rest of the private sector would have declined around 5% last month.

There are still pockets of weakness across many sectors. While an M&A boom may be underway and banks are now willing to reopen the lending spigots, the financial sector shed 10k jobs in January. Government employment was down 14k, largely in the state and local segment, and this can be expected to persist and cast a cloud on the jobs landscape for some time to come. And if temp agency employment is indeed a forward-looking labour demand index, it may be worth noting that this sector let go a net 11k employees in January, the first decline since July 2010.

As we mentioned, the Household survey showed a modestly positive result though smaller than what we saw in December. But there were all sorts of bizarre gyrations in this part of the employment report, which contains all the crucial labour force ratios.

First, the labour force contracted a massive 504k and the participation rate continues to slide fully 19 months into the economic expansion, which is just an off-the-charts development. It shrunk from 64.7% in September to 64.5% in November to 64.3% in December and to 64.2% in January. The participation rate has not been this low since March 1984. So in what way can we really characterize the labour market as being anything remotely close to normal?

While the level of unemployment did take a big dive in January — down 622k — these folks seemed to have left the labour force altogether rather than maintain a job search. Not only that, but the working-age population sank 185k in January. Surely the level of the working-age population and the level of the labour force would not be affected by bad weather in January.

In any event, absent the slide in the labour force last month, the unemployment rate would have gone to 9.3% as opposed to the huge decline to 9.0% from 9.6% that was actually posted. Call it an aberration or certainly something that will need validation in coming months.

Yes, the wage number was pretty solid at +0.4% MoM in January but a good part of that likely reflects the shift in the mix of employment towards higher-paying manufacturing jobs. In other words, it may be good for income — though weekly earnings lagged behind due to the weather-induced decline in the workweek — but it is not inflationary. Also, there may have been some impact from the payroll tax cut in the wage data as well. We did some digging and found that a lot of the tax preparation software companies were mentioning this was a development that would impact wage growth for a little while. Either way, this is not something for the bond market to get ‘fussed’ about and a nice buying opportunity seems to be opening up right now — shades of March-April 2010.

In the final analysis, if you were to ask us to pick just one item out of today’s jobs report that accurately depicted the current state of the labour market, given all of the inconsistent messages, it would be the Household employment data-point on a payroll- and population- adjusted basis. This aggregate showed a decent but hardly mind-blowing +153k print in January — far from the 498k surge in December but right in line with the average of the past 12 months. For the growth bulls, this is confirmation of a continued labour market recovery; for the bears, it shows that there is still nary a sign of acceleration in employment growth at a time when the economy has likely seen its best growth days behind it with the inventory cycle running its course and government austerity soon to replace the past three years of unrelenting stimulus.