Monday, May 12, 2008

Death, Taxes, and A Ridiculous Employment Report

I’m finally getting around to reviewing the latest monthly employment report. I would have tackled it sooner, but I don’t type well when I’m rolling on the floor laughing and coffee is coming out of my nose. Now that my sides have stopped hurting and my nose isn’t burning, let’s take a look at the report.

To recap, the market was looking for a loss of 80,000 jobs in April, and the “official” number came in much better at a loss of only 20,000. The media pretty much universally applauded the report with headlines such as:

“US Stocks Rally on Jobs Report” – International Herald Tribune

“Jobs: Glimmer of Good News in April” – BusinessWeek

“Fur Seal Caught Trying to Have Sex With Penguin” – FoxNews

“Stocks Seeing Strong Gains On Employment Report” – RTT News

“Job Loss Far Below Expectations” –

This employment report reminded me of the recent Google and Intel quarterly earnings reports. If you keep bringing down the estimates prior to the report to a level that’s easy to beat, you look like a hero when your report is just bad instead of apocalyptic. The purported loss of 20,000 jobs should look very suspect to anyone willing to pop an extra Ritalin, take a few minutes, and look beyond the headlines (this rules out most portfolio managers, government employees, and the breathing).

To really understand the employment numbers we have to look at what the Bureau of Labor Statistics (BLS) calls its Birth/Death model. According to the BLS,

There is an unavoidable lag between an establishment opening for business and its appearing on the sample frame and being available for sampling. Because new firm births generate a portion of employment growth each month, non-sampling methods must be used to estimate this growth.

What’s going on here? Basically, the BLS samples businesses and government agencies for its employment report. These sampled entities account for about 1/3 of all nonfarm payroll jobs. The BLS should be able to accurately measure the change in employment for those firms that remain in the sample from one month to the next. After coming in late for work on one of his 27 annual non-holiday work days, catching up on the latest Hollywood gossip online, and taking a lunch break that would make even the Italians blush, your average government employee can still probably muster up the simple arithmetic needed for this calculation. The picture is a little fuzzier for those firms that die (go out of business) since some of them may be too busy hocking their nail guns and backhoes at the local pawn shop to respond to the survey, and others will have already fired everyone, including whoever was responsible for reporting their data to the BLS. The final factor is the birth of new firms. This one isn’t directly measurable in any given month since it takes about 7 months before a newly spawned firm is available for sampling.

So, the BLS created a statistical model to estimate the net employment effect of the recently deceased and the newly hatched. Unfortunately, their model doesn’t appear to be very effective at times when a significant change in trend is occurring. This is hardly surprising. They’re no better than any other government agency, circus monkey, Federal Reserve chairman, or economist at knowing when a recession is underway (no offense to the monkey), so we shouldn’t expect their model to accurately account for this. So there’s a lag between what the birth/death model calculates and what’s happening in the real world at significant turning points in economic activity. When we’re headed into a recession, the model is spitting out fairy tale numbers based on much more robust historical comparisons while real jobs are disappearing. When the economy is turning up and creating more jobs, the model is finally reflecting the preceding deterioration and predicting fewer birthed businesses. Interestingly, they used to admit to this shortcoming on their website but no longer do.

With that background let’s turn to the most recent employment data.
According to the bullish BLS report, the economy only lost 20,000 jobs last month. Virtually everyone in the media focused on this figure (including the experts dropped off at CNBC daycare). What they should have been discussing (had they been aware of it) is that this figure includes 267,000 jobs assumed to have been created by the birth/death model! Unless Obama infiltrated the BLS with his cronies who secretly adjusted the numbers for his 7 new states, this number is ridiculous. Ignoring those assumed jobs, the economy would have registered a loss of 287,000 jobs. Can you imagine how the market would have reacted to that headline?!

So, GDP is flat-lining, the financial and retail industries are suffering, manufacturing remains a basket case, and housing is in freefall, but somehow this model assumes that new businesses created 267,000 jobs last month. It stretches the limits of one’s imagination. Where exactly is this job growth? Not everyone can be a repo man, auctioneer, or Wal-Mart greeter. The model tells us that 72,000 jobs were added in the “Professional & Business Services” category. Maybe all of those ex-Realtors and mortgage brokers have reinvented themselves as credit repair specialists. 83,000 jobs were supposedly created in “Leisure and Hospitality”. I guess it’s possible if they’re counting all the new stay-at-home Dads. 8,000 new “Financial” jobs were supposedly created despite the carnage in banking. Maybe they recategorized all of the lawyers now lining up to sue the banks?

I saved the best category for last. The model assumes that 45,000 jobs were created in April in the construction industry. The construction industry! Seriously, why are we paying these folks? Why do we have or need a BLS if this is the quality of their work? How do they explain this? We all know that the construction firms are going out of business and laying people off. Residential building is screeching to a halt and commercial is now being impacted as well. I suppose all of those people who were fired from their construction jobs turned right around and started their own construction businesses to take advantage of all the business the firm that fired them didn’t have. Before hocking his nail gun I think our pawn shop friend used it to put the last nail in the coffin of common sense.

Here’s another point I love. The actual full release of the employment report runs to 28 pages. Guess how many of those pages make mention of the 267,000 birth boost? Zero. There is a reference to the existence of the birth/death model in the “Frequently Asked Questions” and the “Reliability of the Estimates” sections, but if you want to actually find the number you have to go to a separate web page (

Furthermore, we all know that the employment figures are often significantly revised. The birth/death figures are no different. So, why does anyone pay attention to these releases, and why does the market move on this “news”? First of all, most participants don’t take the time to delve into the details of these reports. As I mentioned, the BLS doesn’t even provide the birth/death figures in its 28-page release. Second, even if you know these numbers are inaccurate at best, if you think that the rest of the market is going to pay attention and trade off of them then you need to be interested in them, too. In this case, if Trader Timmy is going to jump off a bridge, you’d better at least go and dangle your toes off the edge.

As for the BLS statistical methodology, I’m not one to rush to judgment. In the interest of fairness, I decided to use their random sampling methodology on the BLS itself. The results were as expected. I just came up with the B and the S.