The scene: a warm mid-April afternoon in New York in a brightly lit meeting room. The protagonists: a small group of repentant pollsters and data analysts − mostly male, bearded, white and wearing square glasses. The glasses, the angular table, the curt, edgy statements, forensic lights and multiple giant screens bring a touch of the TV drama CSI to the gathering.
We have all been asked to examine a victim: the wounded American polling industry. The invitation suggested an afternoon of soul-searching: Why had the risk of Brexit been underestimated by pollsters? How had Hillary Clinton’s chances been over-estimated in the American election?
The opening salvo from the head of a giant US polling company was humble: We are at sea, we need to revise how we do things, we need to ask some fundamental questions about our models.
