Some Deep Wonkery on Moving Seasonality

January 7th, 2012 at 6:02 pm

The BLS noted something a little unusual in the jobs report data from yesterday, and it got me thinking about how seasonal employment patterns are evolving over time.

There was a 42,000 spike in the number of couriers/messengers last month, and the Bureau pointed out “that seasonal hiring was particularly strong in December. This may reflect increased online purchasing during the holiday season.”

Now, the data come adjusted for seasonal effects, meaning that since we expect more delivery people to be hired in the December, the Bureau subtracts what we’d expect in terms of seasonal hires, leaving any gains (or losses) above (or below) that to counted as net new jobs.

But what if a lot more people start shopping online instead of in stores?  In that case, you’d want to crank up the seasonal factor for messengers and probably take down the one for retailers (i.e., since retailers are making fewer seasonal hires under this scenario, you’d want to be careful not to over-adjust by subtracting too many December jobs from the brick-and-mortar establishments).

This is called “moving seasonality”, reflecting the reality that seasonal patterns that affect different industries can change significantly over time.  Now, the BLS seasonal adjuster does allow for moving seasonality, but it hasn’t picked it up yet for the delivery persons’ sector.

A number of time-series analysts like to use STAMP (structural time-series modeling…look it up if you’re interested) to capture these types of movements in the seasonals, and in fact, that method applied to the messengers series works well.

The figure below shows a) the series as published by the BLS yesterday and b) adjusted for the moving seasonality—i.e., the increased December hiring by all those gift-bearing delivery folks that we’re all so happy to see!

As you can see, the BLS seasonally adjusted series has developed a bad case of the spikes over the past few Decembers.  The alternative adjustment which allows for faster moving seasonality takes away most of those spikes.

Bottom line #1: a more accurate read on yesterday’s jobs report would subtract about 30K from the 200K topline number.

Bottom line #2: I wonder what’s going on in other sectors?  Are they now over-adjusting in retail?

Bottom line #3: If you’ve read this far, I salute you, fellow nerd!


Print Friendly, PDF & Email

6 comments in reply to "Some Deep Wonkery on Moving Seasonality"

  1. Nhon Tran says:

    Thank you. I have read. You have explained well.
    Seasonal adjustment is always fraught with difficulties.
    I wonder if the trend numbers should also be analysed.

  2. JGBellHimself says:

    JB, there was a different shift this year that we suggest was also not picked up by the BSLabor (don’t ask) numbers – the well known fact that Halloween started before Labor Day and Christmas started immediately after the Halloween party ended.

    PPss. apply your statistics to my point above and you will “interpret” more correctly what we mean.

    While Black Friday and Cyber Monday did set records, the brick N mortars tried to get the shopping season off earlier than normal. Their goal might have also been to smooth out the holiday hiring. Either they hired more people earlier to stock and sell, OR they hired less later to re-stock and sell. Might have done both, but with Amazon flooding the streets, we think not.

    Either way, the normal shopping season was not as normal this year. Sorta like being more like “politicians”. Although one might argue that those people are never, ever, real-ly normal.

  3. wkj says:

    A year or two ago, there was extensive discussion of the “birth-death” model of new business hiring & the problems/distortions that it created in BLS estimates, as evidenced by subsequent adjustments of employment figures. Has that problem gone away or just lost popularity?

  4. rjs says:

    although retail hiring this Dec was about the same as 2006 & 2007, this month’s seasonal adjustment for retail employment is skewered by lousy the decembers in 2008 & 2009…so if anything, 2011 SA overreports…

  5. readerOfTeaLeaves says:

    Wonkery, but stepping back from this post to a larger context for wonkery — to ask about the paradigms we use to think about the economy: is it a rigid, non-adaptive machine (i.e., a closed system)? Or is it a series of adaptive gardens connected by networks (i.e., an open system)? And what does that mean for collecting and disseminating employment statistics?

    Last Friday, Dr JB was on Ratigan’s program and at one point the host mentioned the economy being ‘like an ecosystem’. That struck me as a harbinger.

    At a later point in that interview, Dr JB mentioned that the GOP conversations so far this year are pretty much a continuation of the same old ‘trickle down’ economics we’ve heard for a generation.

    Unfortunately, the ‘trickle down’ framework does not explain how (nor why) formerly retail jobs are now morphing into network-based, logistics-enabled delivery jobs.

    Last week, Ratigan also hosted entrepreneur Nick Hanauer on his program, talking about how our very concepts of what an economy ‘is’ needs to adapt to new scientific insights about about individual and social behavior.

    Having just finished Hanauer’s “The Gardens of Democracy”, what strikes me about this post is how Dr JB’s “… thinking about how seasonal employment patterns are evolving over time…” is exactly the kind of adaptive, real-time thinking we need. What underlies the shift in seasonal employment patterns? Probably the emergence of digital technologies, eCommerce, logistics (i.e., UPS, FedEx, DHL), vast server farms, and networked communications. Top-down, trickle-down simply does not map to this kind of economic activity.

    Apart from the specifics of employment patterns, lurking under the surface of this post is the bigger topic of economic paradigms, and how ‘trickle-down’ is not a realistic roadmap given the deeper economic shifts we are seeing. The employment stats seem like more evidence of the increasing disconnect between outdated economic paradigms (i.e. ‘trickle down’), the digitizing economy, and the sclerotic effects on economic behavior when policies are in thrall to outdated thinking.

  6. Fred Donaldson says:

    What is more concerning than BLS seasonality is the media ignorance of two factors:

    1. People “leaving” the workforce (as though leaving means they decided to just live on their trust fund, instead of hitting the office twice a week).

    2. Population growth, which has to add 110,000 new workers to the workforce – a workforce that is far smaller today (4 million) than five years ago.

    Instead of economic analysius, we are faced with media blather about whether each number is a winner or loser. Trends are ignored and numbers are treated like scores in the football playoffs.

    It has all digressed to: is my team (Dem or Rep) winning, not whether or not our incomes and futures will see improvement.

    Such plays well to us accepting the financial takeover artists (Bain?), who control an increasing number of American companies, and whose management style is just winning the latest quarterly goal, and the Hell with the future.