Tuesday, October 13, 2009

Lurking Variables and Statistical Literacy

People are squealing about pork and the payrolls for federal employees. Fact: Federal employees make more on average than private-sector employees. Fact: They are not overpaid.
Economix explains why:
In 2008, only 14 percent of federal workers were on part-time schedules, compared to 26 percent in the private sector. Federal workers were far older on average: 55 percent were between the ages of 45 and 64, compared to 36 percent of private-sector workers. Furthermore, 45 percent of federal workers held a college degree or higher educational credential, compared to 29 percent of private-sector workers.
analysis of the impact of individual education and experience on earnings in the United States by the Harvard economist George Borjas showed that federal employees are paid considerably less than comparable private workers at the top end.
Basically, people looking at the simple averages are not accounting for several important ("lurking") variables that more than explain the average differentials between federal employees and the average private-sector employee.  Andrew Gelman calls this a lack of "statistical literacy"; some call it "lying with statistics". These statistical liars include Chris Edwards at Cato (who by the way should probably know better but is advancing a deliberate agenda), The Free Enterprise Nation, Ilana Mercer of World Net Daily, and Jeff Jacoby of The Boston Globe.


  1. Author is correct in saying that "people looking at simple averages are not accounting for several important variables..." Author gives data on plausible confounders (full-time, age and college degree) for public and private, but gives no data on the strength of the relationship between these plausible confounders and the outcome of interest: pay. The Borjas copmarison applies only to employees "on the top end". So, there is no strong evidence given for the author's claim that these "lurking" variables "more than explain" the average differentials. The author's claim may be true, but this essay does not provide adequate evidence.

  2. The fact remains: controlling for education, level of experience, and occupation, workers in the public sector jobs make less, not more, than workers in private sector jobs.

  3. Your claim may well be true, but stating that it is a fact is not evidence in an argument. While the Borjas paper says the standard deviation of wages is higher in the private sector than in the public, I see nothing in the abstract or the comment by Balls that mentions the size of the average wage gap for comparable jobs -- and nothing that takes into account the long-term security of having a tax-paid job.