October's Job's Report Is Bogus

When October's jobs report came out on Friday November 2nd, there were some questions as to the validity of the numbers. Maria Bartiromo asked CNBC senior economics reporter Steve Liesman to comment on the numbers. His response was something on the lines of the following, "I take the numbers at face value." I sure as heck don't!

There are two main points that led me to question the validity of the numbers. The first is the discrepancy between the establishment survey (which is the most widely followed and quoted employment survey) and the household survey. The second point is where the strength in the establishment survey is coming from.

Establishment Survey vs. the Household Survey

Before commenting on the different employment numbers from these surveys, a detailed description as to their construction is in order. This breakdown of the construction for the two surveys was done recently by John Mauldin in his weekly commentary on the day the jobs report came out and is presented below:

Thoughts from the Frontline: (11/02/07): Payroll Survey Sausage
The Bureau of Labor Statistics actually does two different surveys. One is called the payroll or establishment survey, which is comprised of calling approximately 160,000 businesses (out of 9,000,000) and seeing how many workers they have that month. They survey enough businesses to cover about 1/3 of non-farm employees. And that should be enough to get a good idea of where things are going, right?

Close, but not exactly. They do not contact very many small businesses, and of course cannot call new businesses. And since small and new businesses are the engine of job growth in the US, it is important to include an estimate for them. And they do this by estimating the number of new jobs in various categories that are created or lost in something called the birth-death (BD) ratio.

The BD ratio estimate is based upon past history. While estimating the most recent month's employment picture is quite difficult, you can do a fairly accurate job when you go back a few years, using other government data, tax information, etc. And so you can create a trend for how many jobs you miss due to the birth and death of jobs in the small business area. Now, remember, that number is an average of many years of history.

But there is one flaw in this methodology: it will tend to underestimate new jobs when the economy is recovering from recession and overestimate them when the economy is slowing down (emphasis added). Thus, in 2003-4, the Democrats were beating up Bush about the jobless recovery. As it turns out, those employment numbers were massively revised upward a few years later. There was in fact a powerful recovery going on, just not in the statistics. However, nobody but a few economic geeks paid attention, as it was last year's news.

Now, I mentioned that the Bureau of Labor Statistics does two surveys. The other one is the household survey, where they simply call 60,000 homes (at random) and ask how many people are in the home and who has jobs (part-time or full-time), does anyone want a job who doesn't have one, and so on. This survey covers people who are employed both by large and small employers, illegal immigrants, etc.

These surveys tend to parallel each other, except at turning points in the economy. Then there can be some large discrepancies (emphasis added).

Thoughts from the Frontline
John Mauldin, John[at]FrontlineThoughts[dot]com

As John Mauldin pointed out above, the widely reported establishment survey lags the household survey at major turning points in the economy, either when a recovery or a contraction in employment is underway. This can be illustrated in the figure below that highlights the lagging behavior of the establishment survey to the household survey coming out of the 2001 recession. Also shown below is the establishment survey currently lagging again, but this time to the downside.

Figure 1

Source: Moody's Economy.com

The above figure graphically shows the divergences between the two surveys at turning points in the direction of employment growth and Brian Pretti does an excellent job on commenting on this divergence with numerical values. His commentary is provided below:

You might also remember that a number of months back we really delved into the history of these two payroll employment series in a major part of a discussion and offered the comment that at major turning points in cyclical payroll employment trends, there is very wide divergence between the headline numbers (that are heavily dependent on estimates) and the household series. Well guess what? When we look at the numbers for these two on a YTD basis, it really looks like we are seeing very serious divergence. Divergence of a magnitude suggesting that we are at a serious inflection point for the US labor market. Have a look at the following table and we'll have a few comments.

Payroll Employment: YTD Headline and Household Survey History (000)

Source: ContraryIvestor.com, Market Observation (11/06/07)

Although there can certainly be differentials between these two each year, it's the major divergences that deserve attention. You'll remember that the current payroll employment growth cycle started in late 2003. As you can see above, for that total year, a little over 100,000 jobs were created as per the headline numbers. But the household survey was wildly stronger than the headline numbers, indicating that perhaps we were at an important upward inflection point for payroll growth. As it turned out, that was to be the exact case, only clear in hindsight. The household survey was pointing the way. But now the shoe is on the other foot. Although the headline payroll numbers continue to show us reasonable growth year to date, the household survey over the past ten months is giving us a completely different story. It is warning us that a meaningful payroll employment slowdown is upon us, if indeed historical precedent in the lead and lag nature of these two data points is to remain intact. We suggest heeding this warning. Remember, only in hindsight as these headline numbers are revised some years ahead will the actual truth be revealed. Is the 2003 and 2007 experience in terms of this divergence bookends for the current payroll cycle? For now, we personally trust the message of the household survey and the year over year contraction in growth rate it is showing us. When the household survey bottoms on a rate of change basis, we'll be more than willing to change our minds.

So there you have it. Both John Mauldin and Brian Pretti have painted a clear picture as to the differences in construction and the different messages both surveys are saying regarding the employment picture. Brian Pretti said above that he personally trusts the message of the household survey. Below will be shown another reason why to put more faith in the household survey versus the establishment survey.

The Notorious Birth Death Model (BDM)

The biggest point to be made regarding the BDM to the establishment's employment survey is its relative role in overall job creation accounting. This is shown below in two different graphical presentations. The first figure shows the headline number reported from the establishment survey and what the employment number would be if the BDM adjustment was subtracted out. Figure 3 illustrates the weight that the BDM adjustment accounted for total payroll growth over a twelve month period. The figure shows this along with the National Federation of Independent Business (NFIB) survey of small business reporting now is a good time to expand.

Figure 2

Source: Moody's Economy.com

Figure 3

Source: Moody's Economy.com

As can be seen from the figures above, the role of the BDM to the total employment figure reported has been substantial and now accounts for nearly 70% of job creation over the last twelve months, and at a time small businesses are turning sour on thoughts of expanding their businesses. Any accounting number that has 70% of the total number coming from a model is highly suspect.

Case in point, just look at the market turmoil with what we have seen with the accounting of asset-backed securities (ABS) and collateralized debt obligations (CDOs) on the balance sheets for major banks and brokerage houses who reported the value of these instruments based on their models, not market pricing. The same can be said of the two employment surveys where the establishment survey is more of a 'mark-to-model'report of employment while the household survey is a reality based number that is more akin to 'mark-to-market' as it is based on real data.

Not only should the establishment survey be questioned due to the fact that the BDM accounts for nearly 70% of job creation, but the validity of the BDM itself should be questioned as areas of strength within the model do not stack up to present day economic reality. For example, the October BDM added 103,000 jobs for the month and the real kicker comes from the breakdown of this total. Of that total, financial activities employment added 25,000 jobs (24.3% of total BDM) and construction added 14,000 jobs (13.6% of the total BDM). Does this sound plausible? Would you trust the BDM adjustment when the areas of strength are from sectors that are in a deep recessionary territory or are undergoing massive layoffs?

We all know that housing is in the midst of one of its worst recessions ever with residential construction dropping sharply over the past year. If this is the case, then how is it that the twelve month cumulative BDM adjustment for construction employment has added nearly 140,000 jobs? Removing the BDM adjustment component from total construction employment paints a more accurate representation of construction employment relative to residential construction spending.

Figure 4

Source: Moody's Economy.com

Figure 5

Source: Moody's Economy.com

What is also puzzling in the BDM adjustment is the share that the financial activities sector has played over the last several months in the midst of a credit crunch in the markets and widespread layoffs of mortgage brokers. The number of mass layoff events for the credit intermediation employment sector has surpassed that of the telecommunications sector during the dot.com bust and 2001 recession as a result of a freezing up of available credit in the mortgage markets.

Figure 6

Source: Moody's Economy.com

Despite the surge in mass layoff events in the financial sector of the economy recently, the financial activities employment BDM adjustment's share of the total BDM has jumped considerably over the last few months. This at a time of a seizure in the credit markets and rising mass layoff events for the industry! Does this make any sense?

Figure 7

Source: Moody's Economy.com

So, at a time of credit market instability and a deepening housing recession, both financial activities and construction's share of the total BDM adjustment have risen, with construction accounting for nearly 71% of the BDM adjustment for September. These results should have everyone questioning the validity of not only the BDM adjustment, but also the establishment employment survey that derives its results from the model's adjustment.

Figure 8

Source: Moody's Economy.com

Today's Market

The markets continued their downward movement after yesterday's respite on news of more credit write-downs. The Dow Jones Industrial Average fell 83.16 points to close at 13223.93 (-0.62%), the S&P 500 lost 10.53 points to close at 1470.58 (-0.71%), and the NASDAQ posted the largest decline, falling 29.33 points to close at 2644.32 (-1.10%).

Treasuries fell with the yield on the 10-year note rising 1.3 basis points to close at 4.269%. The dollar index was down on the day, falling 0.02 points to close at 75.83. Advancing issues represented 39% for both the NYSE and NASDAQ.

The best performing sectors were the reflation sensitive materials sector (+0.47%) and the defensive consumer staples sector (+0.04%), while consumer discretionary (-2.02%) and technology sector (-1.50%) were the sectors putting in the weakest performance on the day.