Yesterday our local paper, along with many prominent national publications, ran an article about the cost of state-government corruption. It was based on an academic paper. The paper estimated that in the worst states (including mine, Illinois) corruption was increasing the cost of state government by about 5%. Although there's a certain faux-precision there, given, among other things, the fuzzy choice of what to include in that category, that doesn't sound like a crazy figure. What is crazy, however, is the sentence picked up by the press:
During that time [1997-2008], the 10 most corrupt states could have reduced their total annual expenditure by an average of $1,308 per capita—5.2 percent of the mean per capita state expenditure—if corruption had been at the average level of the states.
Before heading below the squiggle, try to figure out for yourself what's crazy about that sentence.
Right: If $1,308 = 5.2% of per capita state spending, then per capita state spending is $25,154. With per capita U.S. income running about $43k, that would mean that states are spending about 60% of our total income! And that's not counting the feds, local government, etc. Holy shit, no wonder we're in trouble. (Here in Illinois total state spending runs about 10% of income, not 60%.)
I wrote the authors last night asking them to correct this goof. I'm guessing that they just didn't mean to include the word "annual", and thus ended up off by about a factor of up to 10 on that dollar amount. (Most of the corrupt states spend less per capita than Illinois.) They haven't yet responded (not surprising on a weekend), but the error was so obvious and so prominent that it seemed worth publicizing the correction right away regardless of their response.
Meanwhile, it's not just my local paper that ran the crazy scary error. The $1308/year figures prominently in stories in Fortune, an official Indiana U. publication, various business journals (e.g. Dayton), etc. Google "Indiana Hong Kong state corruption cost" for a huge list of places to read the same nonsense figure.
This little accidental test is a reminder that the media are uniformly unable to handle even the simplest quantitative problems, at least if the errors fit one of their standard story lines.
BTW- Yes, it would be nice to see a reduction in corruption. It would make a difference, but nothing even remotely close to the huge difference suggested here.