Friday, October 5, 2012

Jack Welch Don't Know Stats

With his hasty conspiracy theory today "Unbelievable jobs numbers..these Chicago guys will do anything..can't debate so change numbers" (@jack_welch), Jack Welch proved to the world that while his former company may be one of the pioneers of utilizing statistics to improve business, he himself has no clue how stats work. 
  
From the BLS  "The coefficient of variation (CV) is 1.9 percent on national monthly estimates of employment level from the CPS, which translates into a change of 0.2 percentage point in the unemployment rate being significant at the 90-percent confidence level. Because the CPS has a much smaller sample than the CES, its margin of error on the measurement of month-to-month employment change is much larger. For a monthly change in CPS employment to be significant, it must be about plus or minus 436,000"

So, what does this mean?  It means that today's employment number is definitely positive, but could be as low as +437,000 jobs created.  But that assumes that last month's reported employment level was accurate as well.  Since last month, the Household Survey showed a decline in employment (at a time when the now revised Establishment Survey showed a gain of 142,000 jobs), we may be able to assume that last month's number wasn't exactly spot on.  What this means is that if last month showed job creation at the same level as the Establishment Survey (well within the margin of error at 90%), today's gain would only be 585,000 jobs and with the margin of error applied to that total, it would mean we are saying that between 149,000 and 1,021,000 jobs were created this month.  Notice that the lower level of created employment is pretty close to the Establishment Survey (and well within the Establishment Survey 90% confidence range). 

The fact is, we ARE seeing job creation and that job creation is bringing down the unemployment rate.  But, on a monthly basis the statistics are simply not precise enough to make a specific determination on exactly what happened in the measured month except for directional moves that fall outside the margin of error.

So, while as politically compelling as it may be to claim that there is some vast conspiracy between the BLS and the Obama campaign, a simple understanding of statistics can easily put those ideas to rest.