Naked Capitalism Ate Another Post, So I’ll Put My Comments Here:
<b>Big Data and Econometrics That Doesn’t Add Up</b>
Loved the “definition” of Big Data and the juxtaposition of that with the Bloomberg report on reality vs. econometrics.
Big Data looks like another high-tech money sink in the making. The problem, as the auothor of the definition so rightly lampoons, is that “data” implies statements that are true or have a high degree of reliability. In short, when we talk of “data” we imply that we actually know something. I think that reflects the long history when computers and computing time were so expensive that users were very careful not to waste their use on computations that had unreliable inputs; you had to be very sure about what you were doing before you were allowed to do it. Now, when computing is cheap, the attitude is “who cares?”
So, when you hear about “Big Data” just remember what you’re really talking about: A computer that holds and maipulates electrical signals that in turn represent numbers that in turn represent something of interest, which may or may not be real. In short, computers provide representations of representations of things that may be real or abstractions. The identification of “corrleations” or other “statistical” relationships in the “data” only adds more abstraction from which we have to <i>infer</i> something of human intellectual significance.
In short, Big Data looks a lot like sculpting with shit.
Which brings us to the Bloomberg piece. Funny how when the nice acounting of econmic theory can’t be found to add up in real life, we just shrug our shoulders and say “statistical discrepancy” and then keep on going. In most real sciences, that “discrepancy” would be the alarm that either the theory is wrong, the techniques for measurement are wrong, or both. In short, referring to my comments above, such statements can’t be called “data” because they’re clearly unreliable.
But of course, this being economics, we define the theory to be truth and reality as a “deviation”.