Clustering

Theres this guy I know - works for Merrill Lynch (or is that BoA now?  Whatever).  I'm really not quite sure what his group does, but it involves plenty of financial wackiness.

So, this guy - Lets call him John Puddingface - has a pretty good year last year.  Mind you, by 'good year', I mean his group ended up making around 2%, whilst all around chaos unfolded, the dark wings of night cloaked the fallen, and Merrill Lynch ended up screwing the pooch and getting bought by BoA.

John Puddingface makes a bonus of roughly (get this) $2M.  And hes *pissed*.  Righteously, royally, religiously *pissed*.  Its only half of what he was expecting, he has a family, hes got obligations, how is he going to pay his mortgage, blah, blah, blah.

Interlude

Let us say, for arguments sake, that John Puddingface is actually a brilliant financial analyst, and his group effectively made 40% (the market went down 38, he went up 2, etc.).  The question is, if a company is going in the toilet, should individual performers be rewarded?  

I know, we want to retain talent, etc., etc.  That said, think of how easy it is to game this system if there is no concept of malus (the opposite of bonus).  

You *know* the company is doing poorly, and is going to have a bad year.

So, you get together with some of your colleagues, and 'borrow' some of their action, i.e., instead of each of them going down 30 points, they go down 35.  You, on the other hand, get the upside on these transactions, and end up 2 points. Yay!  Bonus time for Bonzo! And, of course, you discreetly share the wealth with your colleagues.

If enough people game the system this way, you end up shelling out $18B in bonuses, and nobody is the wiser.

If you think this doesn't actually happen, you've got some growing up to do.  In fact, lets just extend this a wee bit - instead of colleagues, think companies.  

Enron did this for years, shuffling losses around  their subsidiary companies whilst people made huge amounts in bonuses. 

End Interlude

Turns out, John Puddinghead is probably not a genius.  In statistics, there is something called clustering, where random events end up forming patterns - at least, patterns as far as we puny humans are concerned.  Humans, as it turns out, are really, *really* good at detecting patterns in white noise.  In a world where survival is predicated upon differentiating (successfully!) between shadows and tiger stripes, this was a pretty useful trait.  Nowadays - not so much.

Anyhow, clustering is what makes people thing they are on a Hot Streak when they win 5 hands of BlackJack in a row (just before they Lose It All).  Clustering happens everywhere, including in the stock market, and in companies.  It is almost certainly the case that John Puddinghead was as much a victim, albeit in a good way, of Clustering. 

Interlude

Ever wonder why the Rookie of the Year rarely has a great Year 2, and in fact, they tend 

to suck? Or about the SI jinx?

Its all Clustering. There are a lot of Rookies. A *lot*. Just based on random chance, 

*one* of them is going to have a great year. If it turns out that it *was* random chance, then

the odds of random chance hitting two years in a row are really really small. The player is

probably going to regress to the mean. Compared to the great year they had before, the mean

is pretty sucky. Hence the jinx.

End Interlude

Oh, by the way, turns out that while John Puddinghead's group did OK, his Larger Group

(i.e., his division) lost Vast Sums of Money.