That Can’t Be Right: Jobs Report Short Of Expectations

The monthly JOLTS (labor turnover) survey isn’t due until next week, but we got a big jolt anyway on Friday from the December jobs report. There’s an extensive write-up in the Economic Beat below, but the gist of it is that December probably was on the low side and November on the high side, due to fluctuations in weather and weekly claims data.

That said, even a 100% upward revision would leave the 74,000 result well short of expectations, and so doubts were raised in the thinking of the stock market. Despite many attempts at a jobs report rally that is virtually legally mandated, the market finished mixed, with the Dow off a bit and the S&P getting just rescued (as usual) by the Friday last-minute tape job. Ironically enough, I had written two days earlier that a good-sized shortfall could set off a QE-hope rally – so long as it wasn’t shocking. 74K was shocking.

But not quite shocking enough, ironically. Stocks would have had their usual rally on a big print, a modest print, a consensus print, a slightly short print – but also on an out-of-the-blue actual decline, the kind of thing that gets the market to put on a reverse-riptide defiance rally, convinced that the Fed will have no choice but to rescue them. Of course that’s crazy, but it’s what the market does.

How the Fed might rescue them is something that drifted into the discussion Friday, and probably weighed on the market’s attempts to go anywhere. If the Fed is really having doubts about the efficacy of further QE, as the latest meeting minutes suggested, then what else is left in the Fed arsenal? More guidance? It’s a question I’ve raised in the past – the Fed’s aggressive pursuit of an extended zero-interest rate policy, combined with its enormous balance sheet expansion, has left it with little other ammunition should things turn sour. For the moment, the thinking is that the drop to 6.7% on the unemployment rate, however artificial, and the Fed’s own momentum means no change in the planned reduction rate in QE. But the Fed has often done the unexpected.

The stock market is still down on the month, with earnings season coming up and a late Fed meeting at the end. Either could still rescue January, which is important for its symbolic value. Alcoa (AA) started things off with another disappointment this week, but things won’t really get started until the back half of the week ahead. The estimate game that guarantees a majority of companies “beat” on earnings results means that prices will typically rally as the season gathers steam. Don’t count out the month yet, though a tepid retail sales report next week could cause more uncertainty.

The S&P isn’t down by much anyway, it just feels that way seasonally adjusted. It’s supposed to be up at this point, but failed the “First Five Days” test, an indicator that notes the fact that when the S&P is up through the first five days of the year, that year has been positive 88% of the time. If it’s down, a negative year has followed 50% of the time.

Seasonal adjustment factors are good, though often as much black art as science, but I sometimes wonder if people in the industry are even aware of them anymore. When I hear economists saying things on the radio like, “the important thing is that we still added more jobs in the month,” it feels a bit surreal.

The numbers that we see in the media or hear on the radio aren’t estimates of actual job tallies – they are normalized monthly estimates of the underlying annual trend. We don’t ever actually add jobs overall in December – there is some seasonal retail hiring in the beginning, but apart from that it’s a layoff month, with nearly all of the seasonal help out the door by the 25th of the month.

The real question is figuring out how much the annual shrinkage compares with other Decembers and what the special factors might be, if any. Better to say, “the annualized jobs growth rate was still positive,” rather than conjuring up some image of more people working every month. If one automaker keeps on an extra shift instead of doing a usual annual layoff, it may be accurate to say the underlying manufacturing jobs growth rate has improved by 3,000, but it’s not accurate to boast that “manufacturers added 3,000 more jobs last month,” as journalists are wont to do.

So the usual seasonal stock market addition of a percent or two to prices from earnings season could reposition it favorably again in time for the FOMC statement on the 29th. If it doesn’t, that’s a flashing yellow light so far as sentiment is concerned. It’ll probably all come down to the Fed statement on the 29th again – but I wonder how much longer the central bank’s juju can work.

The Economic Beat

I was watching CNBC Friday morning when the jobs report number was released, and the surprise figure of 74,000 was instantly dismissed by the panel, most emphatically by Moodys.com economist Mark Zandi. After the ADP reported 238,000 December jobs on Wednesday, the consensus had moved up to 200,000, with a much higher whisper number of as much as 300,000. Most of the Bloomberg and CNBC folk were calling for about 250,000. So who was at fault, the Bureau of Labor Statistics (BLS), or ADP, Mark Zandi and the media panels?

They may all be wrong, but in different ways and not for the reasons that appear evident. To begin with, ADP doesn’t even try to produce a proprietary calculation intended to compete with BLS. The goal of the ADP release is to anticipate the BLS call – they want it to be as close as possible.

CNBC economics reporter Steve Liesman made a remark in the immediate wake of the release that might be said to symbolize what often happens with celebrity punditry – one is so busy talking to media and clients that the nuts-and-bolts analysis is handed off to others, and summaries and headline numbers are used in their place. Fortunately I do not have this distraction.

Liesman said that there was nothing in recent data – the PMI surveys and weekly claims, to name two – that would support a weak report, but that isn’t quite so. Recent volatility in the seasonally adjusted claims figures had led many to tune out the reports to some degree, but something in the raw data did get my attention, namely the-year-on-year change in total claims. My table, which doesn’t adjust for holidays and calendar slippage, showed that the average year-over-year decline for monthly four-week reporting periods went a bit crazy in November and December. The average four-week decline for the year was about 120,000, but in November the decline soared to 361,000 (four weeks ending 11/30/2013, vs. 12/1/2012), which is very good, and then nearly vanished in December, to only 32,000 for the lowest four-week decline since – rather suspiciously – that same week of December 1st, 2012.

The volatility is more likely due to changes in weather and the Thanksgiving holiday period, which are not the same every year, than to some episode of “Hiring Managers Gone Wild.” Data for the rest of the year were reasonably dispersed about the mean. So the precipitous drop in holiday-related November claims could have led to an outsized November jobs estimate – and in fact, the original November tally of 201,000 was revised up to 241,000, the largest gain of the year (bear in mind that the cut-off date for the BLS initial monthly estimate is around the second Friday of the month, so the late-in-the-month Thanksgiving holiday can disproportionately swing both numbers).

It may well be the case then we ought to do as Liesman quickly suggested – average out the two numbers. Supposing that December gets revised up to say, 125,000 – admittedly hefty, but possible in light of the claims data and still less than the household estimate of 143,000 – that would take the two-month average up to about 180K, which may be where it should be.

Mr. Zandi would have none of that, insisting that the number was just plain way off and that this time the economy has truly turned. He has been emphatic on that point of late, so he has some reputational skin in the game, and went so far as to say that GDP will be up 4% in the second half, not just the third quarter (Q3). Seemingly a bit daunted by this, Liesman also talked about GDP running up to 4%, and of such things are media perceptions built.

I’m skeptical of the claim. When the first (hefty) revision to Q3 GDP came out, raising it to 3.6% annualized, the reactions of analysts were restrained by the fact that it was due to a big inventory build, a category where the saying “what goes up, must come down” carries much weight. Many revised downward their fourth quarter estimates for that reason.

The second revision carried the estimate up to 4.1%, based on increased consumption and some of the softer, new category expenditures, like intellectual property, or research and development, as well an increase in fixed non-residential structures. That seemed to launch a new wave of enthusiasm that forgot about the inventory build.

But the wholesale sales and inventory data released Friday morning belies that thinking. Start with the fact that wholesale sales fell by 0.64% in the third quarter (unadjusted). The average for the third-quarter going back to 1992 is for a gain of 0.63% (2012 was also a loss, at the time blamed on the budget-default drama). The twelve-month growth rate in sales fell to 2.6% in May 2013, ramped back up to 3.9% in September, and has since fallen back to 3.5%, despite what should be soft comparisons with the 2012 fourth-quarter slowdown. The inference is that the inventory build is just that, another episode of restocking rather than a new trend. The inventory-to-sales ratio is a bit lower than it was this time a year ago.

It’s also possible that the surge in non-residential fixed investment was Fed-related, being a surge to get projects underway while interest rates were still low. Other data in the jobs report that shouldn’t have been influenced by the claims data were also weaker than expected: the average workweek fell back to 34.4 hours (it was 34.5 a year ago), and the aggregate weekly payrolls index declined on the month as well. Both average hourly and weekly earnings declined. The weather could have dinged both numbers and probably did, but it isn’t turn-the-corner stuff, either.

Temporary help increased by 40K (seasonally adjusted), which could be a positive. Transportation and warehousing fell, which isn’t good, but the ISM services survey had the sector reporting positive employment (unlike health care and social assistance, which fell in both reports for the first time in a decade). I don’t know which one is correct, but I do know that if company A lays off 100 employees, and companies B and C each hire ten of them, the net result is a loss of 80 jobs in the jobs report, and a “rapidly expanding” result in the ISM’s diffusion-style result (two of three respondents report expansion). Such is the limitations of the survey, which also reported an 8-8 score for industries reporting expansion versus contraction. That last result was a warning, not a validation that the economy has finally turned.

Mr. Zandi also remarked more than once that the impending annual benchmark revision due next month from the BLS was going to add a slew of jobs and correct the obvious December mistake. I wonder. Goldman Sachs economist Jan Matzius called it simply a “bad report,” but added that the momentum of the economy had changed as well, also talking up 4% GDP in the second half. I have to say that the labor force participation rate falling to a new 36-year low, along with the declines in hours worked and hourly earnings, don’t tell a tale of escape velocity. Neither does the 8-8 score in the ISM survey.

We have had these GDP pops before in the post-crash past. They too were accompanied by great fanfare about “escape velocity,” yet came to nothing in the end, being only restocking episodes with some data lumpiness. In 2009, the fourth quarter grew at 3.9%, invoking the first chorus of escape velocity, yet the next quarter came in at 1.6%. In the second quarter of 2010, it again grew by 3.9%, then 2.8% in the following quarter, the best six months since 2006. That was followed by a decline of 1.3% (fortunately for stocks, not originally reported that way).

In the fourth quarter of 2011, GDP was reported at 4.1%, old style, now 4.9%, followed by a 3.7% first quarter of 2012. The rest of the year stunk. So even if GDP does manage to print above 3% this quarter, it could well be what it’s been in the past – deceleration back to the normal trend of 2% (real). We don’t need more “confidence” to turn the corner, we need more demand.

As for the impending employment benchmark revision, it’s certainly true that previous revisions have often added lots of jobs, so there’s good reason on that basis to expect the same. Yet after the last benchmark revision, the monthly job numbers (unadjusted) started coming in with year-on-year growth that was nearly identical to 2012. The dispersion was so tight, in fact, that through the first ten months of 2014, total percentage job growth of 1.46% was identical to the first ten months of 2012. That’s quite unusual, so much so that I have written several times since that it would appear that the BLS is trying hard to avoid those big revisions by matching up incoming samples to the revised 2012 data and retrofitting them.

Another reason to suspect that the benchmark revision won’t be as big comes from the household survey. The establishment survey comes from a sampling of actual payroll data, while the household survey (used to calculate the unemployment rate) is from people filling out surveys. It’s got a lot more volatility because of that, but over time the two do converge.

They did not in 2013, not yet anyway. In fact, the divergence is still startlingly large. The provisional increase in jobs for 2013 from the establishment survey is 2.2 million, but from the household survey, only about 1.5 million (1.3mm using December as the comparison point). If the benchmark revision does indeed add “a whole bunch of jobs” as Zandi claims it will, then the household survey data is going to have be revised upwards of 70%. Anything is possible, but I confess that I don’t see that as being especially likely, given the past.

The ISM services survey, besides showing that expanding and contracting industries were evenly matched, reported a result of 53.0, still decent but a bit less than expected. It wasn’t as good as December 2010 (57.1) or 2012 (55.7), but ever so slightly better than 2011 (52.7). The cold weather probably played a role, as it seemed to have with the jobs report, which reported an apparent weather casualty of a loss of 16,000 construction jobs (compared to +19,000 in ADP).

The factory orders report was good, with new orders up 1.8% (seasonally adjusted) and +0.6% excluding transportation. The business capital goods segment was up 2.7%, seasonally adjusted, after two months of decline. Shipments rose after two months of nearly flat data. However, trade data continued to show weakness, with the flow of imported goods continuing to ebb.

Who knows what data incoming Fed chair Janet Yellen will think is most important – the unemployment rate falling to 6.7%, the participation rate falling to a 36-year low, the weak December jobs number, the 4% third-quarter GDP print or the $4 trillion Fed balance sheet? The FOMC minutes suggest waning faith in the efficacy of more QE.

Over in Europe, the data was mixed, with strength in France and weakness in Italy. One item that did catch my eye was that eurozone GDP was (-0.4%) for the year.

Next week brings the next leg of big data, with retail sales data for December coming on Tuesday. There is a full slate of inflation data, with import-export prices Tuesday, producer prices on Wednesday and consumer prices Thursday. There’s also a full slate of manufacturing numbers, including the New York Fed survey Wednesday, Philadelphia Fed on Thursday, and industrial production on Friday (look for a pickup from utility production). The homebuilder index is on Thursday followed by housing starts on Friday, and the JOLTS labor survey wraps it all up on Friday.