For many years I've had a small book on my shelf called How to Lie with Statistics first published in 1954. I see that it is now in a 1993 paperback edition, with pictures.
Darrell Huff, How to Lie with Statistics, W. W. Norton & Company (September 1993).
I am reminded of Mr. Huff's book by this entry on the Crime & Consequences blog which I follow for interesting news and usable information. I'll quote and paraphrase extensively from the entry so you'll understand why I am reminded of Mr. Huff. Here is Kent Scheidegger's, The Incredible Growing Statistic.
Burgeoning gangs behind up to 80% of U.S. crime."
The story actually says, "Criminal gangs in the USA have swelled to an estimated 1 million
members responsible for up to 80% of crimes in communities across the
nation, according to a gang threat assessment compiled by federal
officials." That could be read the way the headline-writer read it, but maybe not.
The FBI's press release
for its 2009 National Gang Threat Assessment says, "Criminal gangs
commit as much as 80 percent of the crime in many
communities, according to law enforcement officials throughout the
nation." Ah, "as much as" and "many communities," meaning this is the
peak estimate for the most gang-infested communities, a far cry from
the national total implied by the USA Today headline.
Turning to the full document,
and skipping past the dumbed-down summary and key findings (which are
the same as the press release), what to our wondering eyes should
appear but this:
According to local law enforcement
information, gang members are responsible for as much as 80 percent of
the crime in some locations.5
————————————-
5.
Stated percentage is based solely on self-reporting by local law
enforcement. The figures given are not meant to represent an National
Drug Intelligence Center (NDIC) analysis.
only is the estimate for "some locations" and not nationwide, as the
USA Today's headline indicates, it is expressly disclaimed as a federal
government estimate in the very source the paper cites as a federal
government estimate.
The moral of the story is, don't believe anything you read about the
results of a study until you have gone into the text and checked it
out. With each step removed from the original, the
results grow simpler, more clear-cut, less nuanced, and more ominous.
As we consider the outrage and doom and gloom about crime in the military, we should consider Mr. Huff's homily and Mr. Scheidegger's exhortation to read the underlying reports — closely.