In light of the enormous discrepancy between the Household Survey and the Establishment Survey in last Friday’s jobs report, I thought it might be worthwhile to revisit the discrepancy between Gallup and the BLS that I have talked about before.
Recall that last Friday the Establishment survey showed a net gain in jobs of 288,000 while the household survey showed a net loss in employment of 73,000.
Those are seasonally adjusted numbers. For details please see Nonfarm Payrolls +288,000, Unemployment Rate Drops to 6.3%; Household Survey Employment -73,000, Labor Force -806,000,
Seasonally Adjusted Numbers
- Nonfarm Payroll: +288,000 – Establishment Survey
- Employment: -73,000 – Household Survey
- Unemployment: -733,000 – Household Survey
- Involuntary Part-Time Work: +54,000 – Household Survey
- Voluntary Part-Time Work: -330,000 – Household Survey
- Baseline Unemployment Rate: -0.4 at 6.3% – Household Survey
- U-6 unemployment: -0.4 to 12.3% – Household Survey
- Civilian Non-institutional Population: +181,000
- Civilian Labor Force: -806,000 – Household Survey
- Not in Labor Force: +988,000 – Household Survey
- Participation Rate: +0.2 at 62.8 – Household Survey
Gallup Unemployment Rate
The weekly Gallup Employment Poll shows the percentage of unemployed is 6.9% and the percentage of underemployed is 16.4%.
Gallup explains “Weekly results reflect 30-day rolling averages ending each Sunday, based on telephone interviews with approximately 30,000 adults. Because results are not seasonally adjusted, they are not directly comparable to numbers reported by the U.S. Bureau of Labor Statistics.“
Actually, results are directly comparable if Gallup would simply use the BLS non-adjusted data. Here are the comparable charts.
Gallup Non-Adjusted Unemployment Rate
BLS Non-Adjusted Unemployment Rate
The scale on the Gallup chart is skewed by inclusion of the payroll-to-population statistics, otherwise it would be easy to spot cyclical month-to-month variations on the Gallup chart as well.
The valid comparison in this case is 6.9% vs. 5.9%. One percentage point is quite a difference.
Using Gallup data from the closest week I could find each month going back to 2010 the differences are rather startling.
Gallup vs. BLS Not Seasonally-Adjusted Unemployment Rate
|Gallup Date||BLS Date||Gallup||BLS||Difference|
|2014 April 28-May 4||2014 April||6.9||5.9||1|
|2013 April 29-May 5||2013 April||7.5||7.1||0.4|
|2012 April 30-May 6||2012 April||8.3||7.7||0.6|
|2011 April 29-May 1||2011 April||9.4||8.7||0.7|
|2010 April 26-May 2||2010 April||9.6||9.5||0.1|
In 2010, the BLS and Gallup were essentially in agreement. A large discrepancy developed in the 2011. That discrepancy narrowed in 2012 and 2013, then exploded higher in 2014.
The BLS surveys about 60,000 and Gallup 30,000 but both should be sufficient to determine a relatively accurate answer.
Does the BLS try harder than Gallup to prove someone is not in the labor force and therefore not unemployed? Perhaps.
And this past month certainly provided bell-ringer wake-up call from the BLS. As noted above, 806,000 people dropped out of the labor force on a seasonally-adjusted basis causing the unemployment rate to plunge in spite of the fact household employment fell!
On a not seasonally adjusted basis, the labor force fell from 155,627,000 to 154,845,000. That is a decline of 782,000. Either way, it’s a bell-ringer.
I suspect that for whatever reason, the BLS picked up a huge labor-force decline but Gallup did not. If so, that explains the jump in the difference between the two surveys from 0.4 percentage points to 1.0 percentage point.
Even if that thesis is true, it does not explain “why”. Nor does it explain which survey is more accurate. Nor does it account for large differences that started in 2011.
Regardless of numerous unanswered questions, the Gallup number seems closer to reality given the overall struggling nature of this economy, the meager rise in median income, and the historically poor recovery in GDP.
And if we added back in those who dropped out of the labor force by fraudulent disability claims, an unemployment rate closer to 9 would seem more accurate still.
Mike “Mish” Shedlock