Thursday, August 22, 2013

Gallup Unemployment Rate Spikes from 7.9% on Aug 1 to 8.9% on August 20! Sampling Error or Seasonal Effect?

Inquiring minds note the steady rise all month in Gallup's Daily Employment Survey.

  • 7.9% August 1-2
  • 8.0% August 3-6
  • 8.1% August 8-8
  • 8.2% August 9-11
  • 8.3% August 12
  • 8.4% August 13-16
  • 8.6% August 17-18
  • 8.8% August 19
  • 8.9% August 20

Is This a Seasonal Effect?

That's quite a trend. The unemployment rate steadily rose a full percentage point in a mere 20 days.

Gallup notes "Because results are not seasonally adjusted, they are not directly comparable to numbers reported by the U.S. Bureau of Labor Statistics, which are based on workers 16 and older. Margin of error is ±1 percentage point."

My first inclination was the rise was part of a seasonal trend: "Each result is based on a 30-day rolling average; not seasonally adjusted".

To test seasonal bias, I downloaded the data which dates back to 2010. Let's take a look at an apples-to-apples comparison of August 2013 vs. August in prior years to eliminate any seasonal bias.

August 2010

Date% Underemployed% Unemployed
8/1/201018.48.9
8/2/201018.48.9
8/3/201018.68.9
8/4/201018.79
8/5/201018.58.8
8/6/201018.68.9
8/7/201018.58.9
8/8/201018.58.9
8/9/201018.49
8/10/201018.49.1
8/11/201018.59.2
8/12/201018.69.2
8/13/201018.69.2
8/14/201018.59.1
8/15/201018.39.1
8/16/201018.29
8/17/201018.19.1
8/18/201018.19.1
8/19/201018.39.3
8/20/201018.49.4
8/21/201018.49.4
8/22/201018.49.4
8/23/201018.49.4
8/24/201018.69.6
8/25/201018.69.6
8/26/201018.69.6
8/27/201018.79.5
8/28/201018.79.5
8/29/201018.59.4
8/30/201018.59.3
8/31/201018.69.3


August 2011

Date% Underemployed% Unemployed
8/1/2011188.8
8/2/201117.98.8
8/3/201117.98.9
8/4/201117.98.9
8/5/201118.19
8/6/2011188.9
8/7/201117.98.8
8/8/2011188.8
8/9/2011188.8
8/10/2011188.8
8/11/201118.18.8
8/12/201118.18.8
8/13/201118.29
8/14/201118.18.9
8/15/201118.29
8/16/201118.19
8/17/201118.29
8/18/2011188.9
8/19/201118.29
8/20/201118.29.1
8/21/201118.19
8/22/201118.19.1
8/23/201118.19.1
8/24/201118.19.1
8/25/201118.29.1
8/26/201118.29.1
8/27/201118.29.1
8/29/201118.49.1
8/30/201118.59.1
8/31/201118.59.2


August 2012

Date% Underemployed% Unemployed
8/1/201217.18.2
8/2/2012178.2
8/3/201216.98.2
8/4/201216.88.2
8/5/201216.68
8/6/201216.68.1
8/7/201216.68.2
8/8/201216.68.1
8/9/201216.88.2
8/10/201216.88.2
8/11/201216.98.2
8/12/2012178.3
8/13/201216.98.3
8/14/2012178.4
8/15/2012178.3
8/16/201217.18.4
8/17/2012178.3
8/18/201217.18.3
8/19/201217.18.4
8/20/2012178.3
8/21/201217.28.3
8/22/201217.38.3
8/23/201217.28.2
8/24/201217.28.3
8/25/201217.28.2
8/26/201217.28.2
8/27/2012178
8/28/2012178
8/29/201216.97.9
8/30/201217.18
8/31/201217.28.1


August 2013

Date% Underemployed% Unemployed
8/1/201317.37.9
8/2/201317.37.9
8/3/201317.48
8/4/201317.58
8/5/201317.58
8/6/201317.48
8/7/201317.38.1
8/8/201317.48.1
8/9/201317.48.2
8/10/201317.58.2
8/11/201317.58.2
8/12/201317.58.3
8/13/201317.68.4
8/14/201317.68.4
8/15/201317.58.4
8/16/201317.48.4
8/17/201317.68.6
8/18/201317.78.6
8/19/201317.88.8
8/20/201317.98.9


No, It's Not Seasonal

A close look at August 2013 vs. August in prior years shows no other strong seasonal spikes.

2010 did show a half percentage point rise August 1-20, but that was at a time when BLS seasonally-adjusted unemployment rates were steady or rising. 2011 sported a 0.3 percentage point rise August 1-20 and 2012 had an insignificant 0.1 percentage point rise August 1-20.

This upward trend is steady, stronger, more persistent, and with a unique divergence to BLS reported data as compared to prior years. Whatever is going on, it's not a typical seasonal pattern.

What About Sampling Error?

Even though Gallup states the "margin of error is ±1 percentage point", my interpretation is the dsclaimer pertains to the results in any one survey, not trends that strengthen in one direction for 20 straight days.

As Nate Silver proved in the last presidential election, repeated surveys and trends are important.

For details, please see my confident prediction 90% Chance of Obama Win; Three Things Romney Needs to Win; Election Night Coverage With Mish on National Syndicated Radio

In this case, there is only one pollster, but given no history of bias by Gallup, and given the steady, significant rise in polled unemployment, the trend looks ominous.

Monthly BLS Not Seasonally Adjusted Data 

Let's do a final check to see patterns in the BLS Civilian Unemployment Rate (non-seasonally adjusted).

DateBLS Not Seasonally Adjusted Rate
2010-07-019.7
2010-08-019.5
2010-09-019.2
2011-07-019.3
2011-08-019.1
2011-09-018.8
2012-07-018.6
2012-08-018.2
2012-09-017.6
2013-07-017.7


BLS vs. Gallup Comparison

In BLS not-seasonally adjusted data, in every year between 2010 and 2013, the unemployment rate declined from July to August and then again from August to September.

Why the seasonal decline? Teachers going back to school.

We do not yet have August or September 2013 numbers to compare, but if the BLS matches Gallup, the next couple of unemployment reports are going to look downright ugly.

Expect a Jump in BLS Reported Unemployment

I expect to see a strong jump in the BLS unemployment rate unless the Gallup trend reverses significantly, and soon.

Mike "Mish" Shedlock
http://globaleconomicanalysis.blogspot.com