Why the April Jobs Report Could Be a Disaster
In a speech by Fed Chief Ben Bernanke the other day - the one that turned any lingering “risk off” thoughts in financial markets decidedly back to “risk on” (at least for the day) as another round of central bank money printing suddenly seemed possible again - the subject of a surprisingly good labor market being out-of-step with other economic indicators was raised when Bernanke commented:
… we cannot yet be sure that the recent pace of improvement in the labor market will be sustained. Notably, an examination of recent deviations from Okun’s law suggests that the recent decline in the unemployment rate may reflect a reversal of the unusually large layoffs that occurred during late 2008 and over 2009.
Though there were no explicit references to seasonal adjustments in the above, a number of analysts have noted that odd seasonal adjustments stemming from the 2008-2009 financial crisis could be one of the reasons why some economic data reported in recent months has been coming in much better than expected.
The thinking goes that, since the late-2008/early-2009 period was so horrific, it has resulted in subsequent late-year and early-year data being adjusted upward as if the economic storm that occurred four years ago was somehow weather related.
Of course, the just concluded unusually warm and dry winter is another reason why recent data has been surprisingly good as people were doing things in January and February that they would normally due in March and April, pulling demand (and job creation) forward. Some analysts say that upwards of 100,000 jobs a month in the recent job growth spurt - nearly half - could be “weather related”.
But, the seasonal adjustment question seems to have garnered little attention and, after looking into this a bit, it’s clear why. It’s a difficult subject to analyze and then have much confidence in drawing conclusions about - not just because of the high seasonality in the labor market in general but due to other factors such as the impact of the “birth-death model” and how dramatically jobs data is revised over time.
For all we know at this point, the lower jobless rate and impressive job growth in recent months could all be revised away someday.
In any event, it seemed worth looking into what, if any, impact the devastating economic data of the late-2008/early-2009 period might have had on the recent economic data and the Labor Department’s NFP (nonfarm payrolls) seemed like a good place to start since they provide both the raw data and the seasonally adjusted data.
Subtracting the former from the latter going back eleven years results in the following:
Note that January payrolls require a huge upward seasonal adjustment due to post-holiday layoffs and the annual auto plant retooling shutdown in July is the other period that economists have seen fit to give the raw job creation totals a big boost. Otherwise, you’d get these huge downward spikes every January and July that would make it more difficult to make sense of the data on a month to month basis.
As expected, if you add up all the bars in the chart above, you get zero (actually, it’s seven, not zero, but that’s likely due to rounding) and this makes good sense because, over the course of the year, what you add in January must be taken away over the next 11 months, otherwise, you’re adding to the total, not just smoothing things out to make better the data more meaningful over the short-term.
So, getting back to the inflammatory title of this post, the question is: How have these seasonal adjustment factors changed from before the financial crisis to after the financial crisis and how might that be affecting the labor market data?
In order to answer that question, the chart below was prepared showing the difference in the magnitude of the seasonal adjustments from the November to April period from before and after the financial crisis. For example, the average 2000-2007 adjustment to the raw NFP in December was +283,000, whereas, the average 2009-2011 adjustment was +406,000, hence the resulting blue bar for December extending by 123,000 above the x-axis below.
Taken collectively, the November to March seasonal adjustments that are in excess of what they were before the financial crisis totals some 201,000, or, about 40,000 per month over that five month period.
So, an alternative title to this post could have been “Odd Financial Crisis Related Seasonal Adjustments Have Accounted for 40,000 Jobs Per Month in Reported Payroll Gains”, however, that’s a rather unwieldy handful of words and not nearly as interesting as what that blue bar for April extending below the x-axis portends.
Now, before getting too worked up about what might happen in April - when job creation that would normally have occurred during that month doesn’t occur because those jobs were created back in February due to the nice weather and seasonal adjustments begin to work against the job creation totals instead of for them - it should be noted that this is not an exact science.
I don’t know what goes into the Labor Department seasonal adjustments and, importantly, the blue bars for the other six months of the year for the image above were all over the place and, in some cases, even bigger.
But, the thing that convinces me that this just might have some important meaning is that that the other data appeared to be random.
In the chart above, there is a clear pattern - five straight months of better seasonal adjustments in recent years than prior to the financial crisis and then comes the month of April.
In short, the one-two punch of a warmer winter and unusual seasonal adjustment factors stemming from the financial crisis could combine to create something of a disaster for those writing the labor market headlines in early-May when the April jobs data is reported.