Wednesday, April 10, 2013

Lies, damned lies, and statistics

Without supporting analysis data is generally neither useful nor particularly enlightening.  For example a recent study of household wealth revealed that "on average, the wealthiest households in Europe are in Luxembourg, but Cyprus... was second."

The White House's recent push for the female vote equal pay (see here) is based on similarly absurd analysis (or lack thereof).  As Mark Perry writes:

...comparing average wages by gender is statistically meaningless because that comparison violates the most basic statistical principle that you can only compare two groups if you “hold everything else constant.” In the case of comparing wages by gender, you have to control for all of the other relevant variables that affect income: hours worked, education, years of continuous work experience without interruption, marital status, number of children, age, industry, workplace conditions, and occupational risk of death or injury, etc. To say that women on average make 23% less than men without controlling for any variable except gender is statistically meaningless, i.e. it’s an “apples to oranges” comparison.

In fact, analysis on the raw data provides evidence of a reverse pay gap in favor of women:

James Chung of Reach Advisors spent more than a year analyzing wage data from the Census Bureau and found that in America’s largest cities, single, childless women under the age of 30 earned 8% more on average than their male counterparts (see Time Magazine article here). In some cities like Atlanta, the wage premium in favor of women was as high as 21% 

Wonder when Obama and Pelosi will speak out against that?

No comments:

Post a Comment