One of the very first classes, during the first year of my bachelor’s degree, was on accounting. I had to take this course in order to gain more credits for a module, which I had decided to complete asap. Now, looking back and having listened to dozens and dozens of lectures in many degrees and universities all over the world, I say that this first lecture on accounting, was one of the most boring classes, I have ever attended in my life. The idea of autistic number shifting from one table to another other, in a very mechanical fashion (we get to this later!), in accordance to ‘rules’ which are different from industry to industry and country to country, was just not appealing to me. Even more so, after I had realized that the results of the work of an accountant are occasionally checked by accounting firms, which hire more accountants in order to proceed with the same number shifting again in order to find this one missing digit, the needle in the haystack. Needless to say that it was the last course on the topic of accounting I ever sat in.
And needless to say how over the moon I felt, when I realised that this dull work can be done machines, 100%, now! Once a firm has digitalized all of its balance sheets, one mouse click is enough, (…) (…) (…), and DONE! Don’t get me wrong, I do believe that the work of checking debits and credits is crucial for the sustainability of every company. In particular, because it is a loophole where a lot of tax fiddling takes place. Tax fraud is a billion dollar business. And even if, one day, accounting processes all around the world will be automated, sneaky tax evaders will find new ways to make the numbers match in their favour. But this should be an even greater imperative for putting the responsibility in the hand of automated processes. Data-based forensic accounting could be a new profession for number crunchers .
Here is one simple example of how data algorithms can detect a fraud in the numbers, by applying Benford’s law, a rule of frequency distribution, which has been first formalized in the 1880s. Benford’s law states, simply put that in a naturally generated collection of numbers, the leading significant digit is likely to be small. For example, in a balance sheet the number 1 is supposed to be the most frequent leading digit, followed by the number 2, followed by the number 3, and so on, until the number 9 is the least frequently occurring leading digit in the body of numbers.
If this simple analysis of a company statement would reveal that, let’s say, the number 7 is the most frequently occurring leading digit, then something is fishy. Interestingly, this law not only applies to collections of numbers generated by humans but also to all physical constants in science. Beautiful.