A Little Blogometrics on the Temp Harbinger Question

December 5th, 2011 at 9:05 am

Temporary work has generally been ticking up this year—up 126K so far—and some accounts of last Friday’s jobs numbers argued that this is a harbinger of more job growth to come.

I’ve always wondered about that.  The logic is that employers sense that demand is picking up—and never forget, OTE’ers, the first principle of labor economics: demand for labor is derived from demand for the goods and services that labor produces. To understand this fundamental principle is to understand why all that austerity stuff is so damn wrong right now.

Anyway, employers sense improved demand, but aren’t sure, so they hire temp workers, who are cheaper and easier to lay off, instead of permanent commitments (wow…I’m flashing back to this girl in high school—I’m pretty sure that was her dating strategy with me…sorry…we’re back live at the analysis).

But that logic doesn’t mean temp work will lead to faster job creation.  Employers are testing the waters and could just as easily find them too cold than just right for permanent hires.  The temp harbinger also works in reverse: prior to downturns, we often see employers switching to temps as their expectations of future demand begin to reflect the worsening economy.

So, I ran a few simple and admittedly inadequate statistical tests (blogometrics) and this is what I found:

–using monthly changes, there’s really not much there.  A “Granger causality test” finds that growth in temp work doesn’t predict growth in private sector jobs.  It’s the other way around: growth in the private sector jobs predicts growth in temp work.

–the figure below shows a result from a beast called a vector autoregression (VAR) which just asks how the dynamics of the two variables affect each other.  This figure shows that if you positively tweak temp work, you get a little boost in private sector (net of temp, of course).  But it’s small (0.03% at its peak, about 30K jobs right now) and barely statistically significant (the numbers on the X-axis are number of months since the tweak).

–using longer changes, say over six months, you do find causality going both ways, but the temp effect is quantitatively small, meaning that increases in temp work over the longer term don’t correlate with anything sizable in terms of overall private sector growth.  And over the longer term, in such a “reduced form” model, I’m probably just picking up the fact that both series usually grow.

Anyway, if someone out there wants to spend a bit of real time investigating this question of how real is the temp harbinger, let me know what you find.  (I think the Goldman Sachs researchers did a nice job on this once, but couldn’t find it.)

Source: BLS, see text.

Print Friendly, PDF & Email