miércoles, 29 de junio de 2011

Jonah Lehrer sobre sabermetrics o la obsesión por medir

El artículo va de la aplicación de las matemáticas a los deportes
Buying a car is a hard decision. There are just so many variables to think about. We've got to inspect the interior and analyze the engine, and research the reliability of the brand. And then, once we've amassed all these facts, we've got to compare different models.
How do we sift through this excess of information? When consumers are debating car alternatives, studies show that they tend to focus on variables they can quantify, such as horsepower and fuel economy. (Psychologists refer to this as the "anchoring effect," since we anchor our decision to a number.) We do this for predictable reasons. The amount of horsepower directly reflects the output of the engine, and the engine seems like something that should matter. (Nobody wants an underpowered car.) We also don't want to spend all our money at the gas station, which is why we get obsessed with very slight differences in miles per gallon ratings.
Furthermore, these numerical attributes are easy to compare across cars: All we have to do is glance at the digits and see which model performs the best. And so a difficult choice becomes a simple math problem.
Unfortunately, this obsession with horsepower and fuel economy turns out to be a big mistake. The explanation is simple: The variables don't matter nearly as much as we think. Just look at horsepower: When a team of economists analyzed the features that are closely related to lifetime car satisfaction, the power of the engine was near the bottom of the list. (Fuel economy was only slightly higher.) That's because the typical driver rarely requires 300 horses or a turbocharged V-8. Although we like to imagine ourselves as Steve McQueen, accelerating into the curves, we actually spend most of our driving time stuck in traffic, idling at an intersection on the way to the supermarket. This is why, according to surveys of car owners, the factors that are most important turn out to be things like the soundness of the car frame, the comfort of the front seats and the aesthetics of the dashboard. These variables are harder to quantify, of course. But that doesn't mean they don't matter.
A mí, la aplicación a los deportes me aburre un poco. Pero me parece de interés la introducción que he transcrito.
Comentarios:
1. el “número” fácil de comparar puede ser valioso, no por lo que indica directamente, sino por lo que indica indirectamente: por ejemplo, un motor potente suele ir acompañado de otras características propias de un coche de calidad. Un fabricante no mete un gran motor en un coche malo.
2. Usamos los números cuando no podemos realizar la comparación directamente (si nos gusta la línea del coche o el salpicadero o las llantas lo decidimos comparando directamente unos modelos con otros. Pero no podemos comparar directamente la eficiencia o la potencia del motor).
3. Lo que tiene de razonable el argumento es que, a menudo, las variables que podemos medir son muy malos proxys de las que queremos medir, de manera que los modelos que utilizamos son desastrosamente ineficaces para guiar cualquier medida de política jurídica o económica. O sea que, en ocasiones, puede ser peor medir que no medir, realizar una valoración cuantitativa que una meramente cualitativa por la falsa seguridad que crea la valoración cuantitativa. Detrás de la enorme asunción de riesgos en el sistema financiero que condujo a la crisis está, en una medida importante, la falsa seguridad que daban los modelos económicos para calcular el value at risk. Y ahí ganamos los juristas a los economistas. Nosotros hacemos ponderaciones cualitativas y, ellos, cuantitativas.
Y sobre medir la eficiencia de los maestros y los incentivos perversos que se pueden generar, miren lo que dice Matt Yglesias
My own take is that talk of incentives is massively overrated. Baseball teams don’t pay a premium to guys who hit lots of home runs in order to create “incentives” for people to hit home runs. If that worked, we’d all be major league sluggers! Baseball teams pay a premium to guys who lots of home runs because home run hitters are valuable contributors to baseball teams. Hitters who perform worse are less valuable. And hitters who perform very poorly are drummed out of MLB. That’s not really about incentives; it’s about attracting and retaining high performers to your organization.

No hay comentarios:

Archivo del blog