En todos los análisis post-mortem de la campaña americana, uno de los argumentos más repetidos es que los analistas de datos (data scientists, o como toque llamarlos en cada momento) se equivocaron completamente en sus predicciones, empezando por Nate Silver y terminando por esa maravillosa gráfica en tiempo real que nos dejó el New York Times durante el recuento (aquí la tienen antes de la debacle, y aquí en todo su esplendor posterior).

En 2008 y 2012, la campaña de Obama se había diferenciado enormemente de la de sus contrincantes, entre otras cosas, en el uso que hacían de los datos que tenían a su disposición, en cómo intentaban generar nuevas mediciones allí donde encontraban huecos y en cómo se iban adaptando en función de lo que le iban diciendo los números. Aquí tienen un artículo que habla de ello, y aquí otro.

¿Y la campaña de Clinton? Politico tiene un extenso artículo sobre la derrota que habla de muchas cosas. Entre ellas, de incompetencia suma:

The only metric that people involved in the operations say they ever heard headquarters interested in was how many volunteer shifts had been signed up — though the volunteers were never given the now-standard handheld devices to input the responses they got in the field, and Brooklyn mandated that they not worry about data entry. Operatives watched packets of real-time voter information piled up in bins at the coordinated campaign headquarters. The sheets were updated only when they got ripped, or soaked with coffee. Existing packets with notes from the volunteers, including highlighting how much Trump inclination there was among some of the white male union members the Clinton campaign was sure would be with her, were tossed in the garbage.

Léanlo entero, que no tiene desperdicio. ¿Y Trump? Todo lo contrario.

Inside his campaign, Trump's analysts became convinced that even their own models didn't sufficiently account for the strength of these voters. "In the last week before the election, we undertook a big exercise to reweight all of our polling, because we thought that who [pollsters] were sampling from was the wrong idea of who the electorate was going to turn out to be this cycle," says Matt Oczkowski, the head of product at London firm Cambridge Analytica and team leader on Trump's campaign. "If he was going to win this election, it was going to be because of a Brexit-style mentality and a different demographic trend than other people were seeing.

Trump's team chose to focus on this electorate, partly because it was the only possible path for them. But after Comey, that movement of older, whiter voters became newly evident. It's what led Trump's campaign to broaden the electoral map in the final two weeks and send the candidate into states such as Pennsylvania, Wisconsin, and Michigan that no one else believed he could win (with the exception of liberal filmmaker Michael Moore, who deemed them "Brexit states"). Even on the eve of the election Trump's models predicted only a 30 percent likelihood of victory.

No tengo sistema de comentarios. Si quieres decirme algo sobre este artículo, puedes ponerte en contacto conmigo mediante e-mail o Mastodon.