The myth of the Bradley effect

Private, daily tracking polls showed that, with a retooled campaign, Mr. Deukmejian methodically closed the gap. On the Sunday night before the day of the election — usually the last day of tracking polls the campaign will pay for — Mr. Deukmejian had closed to less than two percentage points. The campaign polled Monday night, too. It showed Mr. Deukmejian less than 1% behind. Private pollster Lawrence Research predicted to the campaign a razor-thin victory — exactly what happened.

The public polls stopped polling too soon, missing the Deukmejian surge. Most important, they ignored the absentee ballot. Mr. Deukmejian’s polling asked if people had voted absentee; other polls, including the exit polls, did not.