The December unemployment rate was released this morning.  The measured rate dropped to 8.5%.  As always, a good part of this drop was caused by the ongoing decline of the number of workers in the labor force.  The labor force participation rate dropped sharply to 64.1%.  This is the lowest participation rate since 1983′s 64.0%.  Anyone who argues that the drop in the unemployment rate signals an improving economy should be forced to recycle their Ph.D. in economics into aluminum beer cans.  The graph below tells the gruesome story.

U.S. Labor Force Participation Rate

Unemployment Rates Revised for 2011

In other news, the Bureau of Labor Statistics issued revisions to the unemployment rate for the twelve months in 2011.  The table below is from the BLS employment report.

As I wrote last month, the November unemployment rate seemed too low.  Frankly, however, I’d chalk that up to a lucky guess rather than any particular skill on my part.

As my fellow economist Dean Baker (co-director, Center for Economic Policy Research) noted in his Twitter feed (@DeanBaker13), some 42,000 of the new jobs were courier jobs, presumably seasonal hiring by FedEx, UPS, and other companies whose business increases sharply during December.  However, the unemployment rate and most other statistics are seasonally adjusted.  Let’s see if we can sort this out.  (Even the BLS commented on this in their report, stating “Employment in transportation and warehousing rose sharply in December (+50,000). Almost all of the gain occurred in the couriers and messengers industry (+42,000); seasonal hiring was particularly strong in December.”  The actual change from November to December in courier employment was an increase of 85,900 workers.  The 42,200 figure cited by Baker uses the seasonally adjusted (SA) numbers.  The 85,900 figure uses data that is not seasonally adjusted (NSA).  Allow me a small digression.

Seasonal Adjustment

Most monthly and quarterly data is seasonally adjusted.  The idea is to adjust for regular cyclical changes that occur every year.  Two big causes of seasonality are holidays (Christmas, Kwanzaa, Hanukkah) and, well, the season.  Ice cream sales rise in the summer.  More products are shipped in November and December.  Seasonal adjustment is designed to remove the regular changes that occur every year.  These days seasonal adjustment is tricky because of the rapid changes in technology, demographics, and the economy itself.  So the real question is how the difference between the seasonally adjusted and not seasonally adjusted figures for December, 2011 compare with historical data.  I’ll add a technical note at the end of this post to explain how to find the exact data.

There’s another issue that came up looking at the data.  The difference between the NSA and SA figures for December rose sharply beginning in 2006.  The average difference between 2001 and 2005 was 15,620.  From 2006 to 2011 the average was 40,200.  Thus the NSA – SA figure more than doubled on average starting in 2006.

Statistical Results

Before I go any further, if you want the data, click here to download an Excel 2003 workbook.  The last tab includes data sources and an explanation of how to extract this data from the bls.gov website.

What we really want to determine is whether the 60,900 difference between NSA and SA in December, 2011 is statistically different from the average.  To do this we use a t-test.  Calculating the standard deviation, then calculating (December – average)/standard deviation gives us a t-statistic.  (Yes, I calculated the sample standard deviation.)  Using data from 2006 – 2011 the t-statistic is 1.68.  Not statistically significant at the five percent level.  According to Stat Trek online the actual significance level is about 7.74%.  Nice grey area result.  (Naturally, using the entire sample period 2001 – 2011, the t-statistic improves to 2.06.  Until someone can explain what happened in 2006, I’m sticking to 1.68.)  For the masochists who like to see the numbers, here they are:

Another way of looking at this issue is to compare job gains from November to December with job losses from December to January.  We’ll have to wait another month to get the full comparison for this season, but here’s the seasonally adjusted data from 2006 – 2010:

For the last two years job losses in January have been greater than job gains in December.  And remember, these are seasonally adjusted data.  Unless the labor force continues to shrink at an alarming rate, the unemployment rate for January should tick up to about 8.7%.  Remember, you read it here first.

One more item of note.  The NSA – SA difference in December, 2010 was 50,100.  In the four years before that, the difference was in the neighborhood of 35,000.  Something is going on here.  If I had to guess, I’d speculate that it’s the ongoing steady increase in online shopping.  But wait — these are courier services, not UPS and FedEx.  I look forward to comments from those smarter (and/or more imaginative) than me to clarify this mystery.

Note on Retrieving BLS Data

Start at http://www.bls.gov/data.  Select the link to Employment data, then select the multi-screen data search column in the “Employment, Hours, and Earnings – National” row.  On the next screen, check both the seasonally adjusted and not seasonally adjusted boxes, then click next form.  Scroll down the “Supersector” list box until you see sector 43: Transportation and Warehousing.  Click that, then next form.  In the Datatype list box, click All Employees, Thousands, then click next form.  Scroll down the Industry list box until you find 434920000 Couriers and messengers.  Click that, then next form.  The list box should have two series in it: CES4349200001 and CEU4349200001.  CES is the seasonally adjusted data and CEU is not seasonally adjusted.  Click retrieve data, then download the two Excel files that are generated.

There’s probably an easier way to get this data, but I’ve spent enough time trying to figure out how the BLS hides data.