Sunday, June 21, 2009

Great minds think alike
And the pounds took care of themselves

Via Alex at the Fistful of Euros, I learn that John Graham-Cumming, Slayer of Spam, has had similar ideas regarding Benford's law. He has looked at the figures in the forms submitted by the chancellor, the prime minister, and Harriet Harman. The PM and HH expenses follow the distribution, but the chancellor's do not. There are rather more 3's and 4's than you would expect by chance. He has identified the source of these: the chancellor claims exactly £300 for food every month, and reqularly claims £45 for his telephone bill.
He also observes that Hazel Blears has submitted claims for whole number of pounds. Hmm.

Were I to fabricate my expenses, I -or at least, whichever of my sons was working for me full-time whilst also a full-time student  - should be sure to add an appropriate and possibly random number of pennies to each of my more creative claims, if only to add an air of vermislitude to an otherwise bald and unconvincing narrative. [Benford's law probably wouldn't be able to pick up divergences south of the decimal point, as the distribution tends to uniform as the order of the digit increases]
On the other hand, you are allowed to round down in (some of) the boxes on your self-assesment, so it might well be legitimate to do so on an expenses form. 

In any case, I think the best test is reciepted expenses vs. unreciepted expenses. As soon as we see some tabulation at the Guardian project, it'll be time to break out the Benford's.

No comments: