Each month we ask the question, "What change in payroll employment would be consistent with other economic data from the same time period (the middle of the prior month)?
This is not a forecast, per se, since we do not posit any causal relationship among these variables. They are all concomitant indicators of economic activity. We use the four-week moving average of initial unemployment claims, the University of Michigan sentiment survey, and the ISM manufacturing report. We carefully choose data from the correct time period. Even though the ISM report was released today, the survey is obviously from earlier in the month.
None of these indicators have improved very much, so we continue our negative outlook on employment. We were surprised last month when the job losses were less than we (and nearly everyone else) expected. We are still looking for losses over 550K, much worse than the consensus loss of about 400 K.
Since our analysis is based upon the final data, after all revisions, the ultimate accuracy may not be known until next year! That is when final benchmark revisions are done. Also, the sampling error (90% confidence interval) alone on the payroll survey is more than +/-100K jobs.
In addition to the consensus forecasts, there are various predictions using proprietary data. These are all interesting.
TrimTabs uses data from income tax deposits of salaried employees. They expect job losses of 472,000.
ADP uses data from their payroll administration business, information that no one else has. They have attempted to gear their results to the "official" government report. They forecast a loss of 473,000 jobs, amazingly close to TrimTabs.
New entrant Wanted Technologies uses an algorithm reflecting online job ads. They have a startling forecast: a loss of "only" 260,000 jobs. Furthermore, they made their call on June 19th. And why not? That was the right time frame to match the payroll survey, and their online job data is more readily available in real time.
Our own prediction of the jobs report has no special inputs -- just the analysis of concurrent economic data. We are surprised to be the most bearish of the group. As noted, the error band is wide. The market will react wildly without regard to the sampling error or other issues.
We do have a few predictions that we can make with more confidence:
- Whatever the job loss, the unemployment rate will move higher. The demographic factors at work require job gains of at least 150,000 (and probably more) just to maintain current unemployment levels. The unemployment rate is an important social and political indicator, but it will lag in reflecting an economic change.
- The assembled punditry will state, whatever the number, that it should have been worse because the government is incorrectly projecting job creation.
- If the result is really good, the rumor mill will start, as it did last month. When the market spiked on a better-than-expected report for May, the rumors quickly circulated that it was an error -- a government worker had a "fat finger." Those circulating this rumor (and those believing it) have absolutely no concept about how government reports are assembled, how many people are involved, and how many check points there are.
It just shows that if you want to be short going into this report, you can have confidence that the Bearish Blogging Network (TM OldProf) will have your back. They will take advantage of the blogosphere to spin at high speed. The official sources have to wait for a news conference or an interview to reply. This is plenty of time to cover your shorts.
It is an attractive trade for any hedge fund manager. Take last month as an example. You could come in short and be an instant winner on a bad number. If the report was positive, you sell more on the spike (averaging up in price). You then cash in on the silly "fat finger" rumor and the expected monthly spin on the birth/death adjustment.
How Can this Work?
It is amazing. Take a roomful of traders. Ask them whether government or a trading desk is more efficient. We know what they would say. Trading desks can execute baskets with a keystroke. There are "fat finger" examples and also stories about interns sitting on keyboards.
Does anyone really think that a very complicated government report is generated in the same way? Well the silly story was good enough to move the market last month.