For many years I have written a regular monthly preview of the Employment Situation Report. I have done extensive research on all of the methods and even visited the stat guys at the BLS to discuss their approach.
My preview gives appropriate respect to the BLS, but also to the leading alternative methods. Two months ago I explained this with a bean-counting contest. The winner was NOT the contestant who was closest to the correct answer. Instead, the winner had to predict the guess of a fellow contestant.
This is what we do every month. We want to know the truth about the economy. Instead of recognizing that there are several good estimates, everyone tries to guess what the BLS will report.
Please read the September preview both for a laugh and a deeper look.
We rely too much on the monthly employment report. It is a natural mistake. We all want to know whether the economy is improving and, if so, by how much. Employment is the key metric since it is fundamental for consumption, corporate profits, tax revenues, deficit reduction, and financial markets. Whenever there is an important question, we all seize on any available information. While we might know the limitations of the data, any concern is briefly acknowledged -- if at all -- and then swiftly put aside.
We would like to know the net addition of jobs in the month of November.
To provide an estimate of monthly job changes the BLS has a complex methodology that includes the following steps:
- An initial report of a survey of establishments. Even if the survey sample was perfect (and we all know that it is not) and the response rate was 100% (which it is not) the sampling error alone for a 90% confidence interval is +/- 100K jobs.
- The report is revised to reflect additional responses over the next two months. This is especially important this month since the official survey response date has been moved forward by a week. This is a routine adjustment for the Thanksgiving holiday, but it increases the potential for error and later revision.
- There is an adjustment to account for job creation -- much maligned and misunderstood by nearly everyone.
- The final data are benchmarked against the state employment data every year. This usually shows that the overall process was very good, but it led to major downward adjustments at the time of the recession. More recently, the BLS estimates have been too low, as revealed in the most recent report. For the year ending in March, 2012, the BLS estimate was off by about 30K jobs per month overall, and 35k jobs per month on private employment.
The BLS report is really an initial estimate, not the ultimate answer. The BLS is actually like one of the contestants, with the full report coming later. The market uses this estimate as "official" and declares winners and losers on that basis. No one pays any attention to the final data, which we do not see for eight months or so.
- ADP has actual, real-time data from firms that use their services. The firms are not completely representative of the entire universe, but it is a different and interesting source. ADP reports gains of 118K private jobs on a seasonally adjusted basis. In general, the ADP results correlate well with the final data from the BLS, but not always the initial estimate. In recent months ADP is using an improved methodology with a stronger sample. The objective is to improve the correlation with the final print of the employment data.
- TrimTabs looks at income tax withholding data. The idea is that this is the best current method for determining real job growth. I regularly feature the TrimTabs forecast, but at the time of writing this post they have not yet issued a release. Their site is here, and perhaps those commenting can help if there is an update.
- Economic correlations. Most Wall Street economists use a method that employs data from various inputs, sometimes including ADP (which I think is cheating -- you should make an independent estimate).
- Jeff Method. I use the four-week moving average of initial claims, the ISM manufacturing index, and the University of Michigan sentiment index. I do this to embrace both job creation (running at over 2.3 million jobs per month) and job destruction (running at about 2.1 million jobs per month). In mid-2011 the sentiment index started reflecting gas prices and the debt ceiling debate rather than broader concerns. When you know there is a problem with an input variable, you need to review the model. For the moment, the Jeff model is on the sidelines. The recent uptick in consumer confidence-- despite gas prices, the fiscal cliff, and Europe-- is encouraging for jobs. It remains difficult to account for the effect of headlines about Europe and the fiscal cliff. The model inputs have improved a lot. The layoffs catch the headlines, because these are big visible chunks of jobs. I do not think we have a good grasp on job creation. The BLS tries hard, but their approach lags on this front.
- Street estimates are generally similar to my method, but few reveal much about the specific approach. These estimates usually adjust for the ADP report, but there was little reaction this week.
- Briefing.com cites the consensus estimate as 90K, while their own forecast is for 80K. Their private jobs forecast is about 10K higher, since the loss of public jobs is well known.
- Gallup sees unemployment as increasing dramatically to 7.9% on a seasonally adjusted basis. This is interesting since they have a different survey from the government, a relatively new approach to seasonal adjustment, and an extremely bearish and political approach in past commentaries. Gallup's methods deserve respect, so I am watching closely. Here is their chart:
Despite this result, Gallup sees net job creation -- yet another illustration of why employment is so difficult to understand.
Failures of Understanding
There is a list of repeated monthly mistakes by the assembled jobs punditry:
- Focus on net job creation. This is the most important. The big story is the teeming stew of job gains and losses. It is never mentioned on employment Friday. The U.S. economy creates over 7 million jobs every quarter.
- Failure to recognize sampling error. The payroll number has a confidence interval of +/- 105K jobs. The household survey is +/- 450K jobs. We take small deviations from expectations too seriously -- far too seriously.
- False emphasis on "the internals." Pundits pontificate on various sub-categories of the report, assuming laser-like accuracy. In fact, the sampling error (not to mention revisions and non-sampling error) in these categories is huge.
- Negative spin on the BLS methods. There is a routine monthly question about how many payroll jobs were added by the BLS birth/death adjustment. This is a propaganda war that seems to have ended years ago with a huge bearish spin. For anyone who really wants to know, the BLS methods have been under-estimating new job creation. This was demonstrated in the latest benchmark revisions, which added more jobs.
It would be a refreshing change if your top news sources featured any of these ideas, but don't hold your breath!
And most importantly, it would be helpful if anyone would realize that the BLS is just one estimate among others -- and perhaps not the best. The bean counter example illustrates this.
This week's employment situation report involves major risk factors:
- There will be effects from Sandy, but these will be difficult to quantify. The shift in the reporting date will magnify the distortion.
- Post-election skepticism will dominate the analysis. Any weakness will be seized upon as evidence that the last two reports were artificially inflated.
- Those claiming that a new recession has begun will pounce on any weakness.
- Traders no longer see the Fed as a viable source of support. We have a new era of good news is good, and bad news is bad.
I expect some temporary weakness from business anticipation of the fiscal cliff and the effects from Sandy. There is a real danger that the +/- 100K confidence interval could result in a very bad number -- possibly even a negative print.
Despite the various adjustment factors cited here, there would be a huge reaction to a bad number. When the report comes out, no one will take note of the wide confidence interval, or the fact that the BLS methods have understated job growth in the last year.
As usual, even if the report is surprisingly strong we can expect a spin-induced opportunity to buy.
For traders, it is wise to be cautious and short-leaning.
For investors, there may be a chance to buy at irrational prices.
Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.