Saturday, 2 February 2008

NFP Revisions of Revisions of Revisions

Whilst the market obsessed over Microsoft's grossly overvalued bid for Yahoo, the most anticipated economic number of the month came out and it wasn't pretty. From

U.S. Economy: Payrolls Fall for First Time Since 2003
The U.S. unexpectedly lost jobs for the first time in more than four years, increasing the odds the economy will fall into a recession and making it likely the Federal Reserve will cut interest rates another half point next month.

Payrolls fell by 17,000 in January after an 82,000 gain in December that was larger than initially reported, the Labor Department said today in Washington. None of the 80 economists surveyed by Bloomberg News predicted a decline.

You've got to love that last line, not one in 80 economists forecast a decline. However that won't stop the media from asking them next month what they think the number will be when they could just as easily throw darts at a piece of paper and achieve the same level of accuracy.

The drop in payrolls in January was the first since August 2003. The median forecast was for a payrolls gain of 70,000, compared with an initially reported increase of 18,000 in December. Forecasts ranged from gains of 5,000 to 160,000.

Revisions for November and December brought total job gains for the two months to 142,000, versus a previously reported 133,000.

As noted previously on many occasions, more attention should be paid to the revisions. December's number was revised upward from the weak 18k intially to a more decent 82k.

We also need to pay attention to the revisions of the revisions. In November NFP growth was intially reported at 94k, was subsequently revised upward to 115k in December and now has been revised sharply downward to 60k. Now if you are feeling all revised out take a look at the annual revision, which makes revisions of the revisions of the revisions.

With today's report, the Labor Department revised the payroll numbers after reviewing more complete tax data not available earlier from state unemployment insurance programs and making adjustments to its estimates of seasonal hiring patterns.

The revision subtracted 376,000 jobs from the previous estimate for the year ended December 2007, bringing total job growth for the period to 1.137 million.

Goods-producing industries were revised down by 233,000.
Construction payrolls were revised down by 63,000.
Factory payrolls were revised down by 176,000.
Services producing industries were revised down by 125,000.

And of course the NFP report would not be complete without an eyeballing of the controversial Birth/Death adjustment. The B/D adjustment is especially important this month as it also undergoes an annual revision.

In one fell swoop 378k jobs have been wiped away. That's almost 30% of the jobs estimated by the B/D model in 2007. As mentioned many times before the B/D adjustment is not a good predictor of job growth at inflection points in the economy and it is likely that job growth for 2007 has still been overstated.

So with all these revisions it's best to take a look at the 3 month moving average of job growth as represented in the chart at the beginning of this post. That average is now running at a very lacklustre 42k per month. That average was 169k just 1 year ago. Clearly the labor market has slowed, but is it recessionary?

There were a parade of economists on CNBC today arguing that whilst job growth has slowed it is not yet recssionary. What you see in a recession, they argue is several months of outright declines in employment. Whilst that is partly true, it is of little practical value.

As the chart at the beginning of the post shows, every single recession in the last 40 years began whilst the 3 month moving average of job growth was still positive. Consecutive months of sharp declines in payroll numbers do not come until well after a recession has gotten underway and often toward the end of a recession.

The same worthless observation is made about the unemployment rate. Many an economist argued that a 4.9% unemployment rate is not recesionary. However, as the graph below illustrates the unemployment rate does not peak until after the official recession has ended.

As an example, the last US recession offically troughed in October 2001 with the unemployment rate at 5.3%. However the unemployment rate kept climbing for 20 months until June 2003 when it peaked at 6.3%.

So basically these economists will only admit payroll numbers and the unemployment rate are looking recessionary when the US economy is well into a recession. That's not particularly useful to anyone, just as if I tell you that you're about to be hit by a car after the car runs you down.