« The February Employment Report | Main | "Rewriting the Macroeconomists’ Playbook in the Wake of the Crisis" »

Friday, March 04, 2011

Spiky Population Changes?

I need your help. I was going to post the change in population per month against the change in employment to show how far job creation has fallen short of changes in population during the recession.

So I grabbed population data and employment data from the household survey in FRED, and plotted one against the other. Here's the result (these are differences, the data are not logged):


Here is a smoothed version highlighting the job creation shortfall (that hides the spike problem):


But back to the first graph, here is the population series by itself (the same pattern is evident in the full series, i.e. the series since 1948):


I feel like I should know the answer to this, but don't. So my question is, what is causing the spikes? The population series is not seasonally adjusted, but the irregularity of the spikes means that seasonal adjustment wouldn't take care of it anyway (plus, the spikes are both positive and negative). They are some type of adjustment since it always occurs in January (and particularly big spikes seem to occur at decade changes where there is a census), but what?

The reason I'm asking is that echoes of these spikes show up in the seasonally adjusted employment series (see the first graph, for example, around January of 2003 or January of 2004). More importantly, there is a negative spike at the trough of the current recession -- did that affect the employment numbers and make things look worse than they actually were?

Any help would be appreciated.

Update: Via email, and also in comments, Ben Herzon of Macroeconomic Advisers, LLC provides an answer:

The spikes reflect incorporation of new population controls (every January).  They effect the levels of all the major series from the household survey (population, labor force, employment, unemployment), but not so much the ratios (employment ratio, unemployment rate, labor-force participation rate).  When BLS gets new estimates of the population, they scale up or down their aggregates to reflect them, but they do not revise the data back.  This leads to the discontinuities observed.

The first thing I checked after noticing this was the employment-population ratio. As noted, these spikes are not evident in this or other ratios, but they do appear in the levels.

    Posted by on Friday, March 4, 2011 at 12:24 PM in Economics, Unemployment | Permalink  Comments (13)


    Feed You can follow this conversation by subscribing to the comment feed for this post.