17 June 2014

Driven to Data Distraction, Part The Second

There wasn't intended to be a Part The Second, but in reviewing (after hitting "Save", naturally) the text, the punch line and trail of breadcrumbs to same is less than obvious. Hindsight, and all that.

Since the point of part the first was: The quants took a beating on this one, a clearer road to that end might be warranted.

Let's start with Nate. While I acknowledge that Nate is a saint, the title of his piece, "Eric Cantor's Loss Was Like an Earthquake", and the substance of it, are so far up his ass that it's hard to overstate his error. Nate makes the signature mistake of quants: analogizing from natural processes to human processes. Natural processes obey external (God or Nature or the Stay Puft Man as one prefers) rules which such processes can't alter, while human processes are always subject to the dominant actor changing the rules to suit. While natural processes exhibit black swan events, to the instant human eye, they're not. They're just the result of insufficient data. Weather forecasting is the prime example. At one time, weather events could be predicted only grossly, and mostly if one could communicate with others upstream of the weather highway. With more data, and very big computers, weather and climate models turn black swans into Daffy Ducks. To the creatures of the time, the Yucatan asteroid was a black swan event. Today, not at all. We have the data. We might even be able to avoid the collision with available tech.

In the human sphere data, in the sense of not being anecdotal information (which quants sneer at, of course), will never trump knowing what the puppet masters are up to.

As the local reporter from Carr's piece says:
Jim McConnell kind of saw it coming. As a staff reporter at The Chesterfield Observer, a large weekly that serves the suburbs of Richmond, he wrote several articles in the spring suggesting Mr. Cantor was in for a fight. And on June 4, he suggested that "Brat's campaign is gathering steam as it hurtles toward the finish line."

Those quants who look only at historical election data, polls, and current public polling will never see that the rules had changed. Change the rules, change the outcome. The presence of Ingraham and the other astroturfers in the field was the "data" that mattered. Carr is right that national media didn't especially notice (none has put his hand and said, "I told you so", that I've read; I can wait). The over vote was what mattered. How to have known that before election day? Probably not, but the media following the vote totals as precincts reported (assuming that they reported serially, which I don't know), would know something was up.

The notion that yesterday looked like the day before, today looks like yesterday, so tomorrow will like today (mostly) is the Achilles heel of time series analysis of human processes. There's no guarantee that the rules controlling the data are stable. Fact is, I'd assert that the rules will be changed by the dominant actor once their pain reaches some Δ.

Now that wage arbitrage has been played out to its end (too little labor left in most production to make it worth the effort), we see corporations running like the bulls of Pamplona to tax arbitrage. They have found their Δ. Alito, et al, have declared corporations people, but Leona made it clear, "only little people pay taxes". Eventually, they'll kill off the middle class golden goose consumer across the planet. Good on them.

No comments: