The Prediction Graveyard

resolved · Earned

Silver's final 2012 model gave Barack Obama a 90.9% probability of winning the presidential election.

Source-backed summary; not presented as a verbatim quotation.

Nate Silver
FiveThirtyEight election forecaster · United States

Made: 2012-11-06Resolved: 2012-11-06
Original conceptual trajectory diagram illustrating No cause of death assigned
Reasoning and outcomes. Original illustration for The Prediction Graveyard. Aniruddh Jangid, 2026. Original digital diagram. CC BY 4.0. Changes: none. This is a conceptual diagram, not an archival image.

The world at the time.

Silver aggregated polls and evaluated paths through the Electoral College. His pre-election analysis explained that pooling surveys increases information while correlated polling errors can remain.

What happened.

Obama won the election and received 332 electoral votes against Romney's 206. The high-probability outcome occurred, although one election cannot by itself establish the calibration of a probability model.

Predicted vs actual.

MetricPredictedActualVariance / caveat
Probability of Obama victory90.9%Obama won, 332 electoral votesFavored outcome occurred; single-event calibration untestable

The counter-signal.

Gallup's final likely-voter survey put Romney at 49 percent and Obama at 48 percent nationally. This conflicting contemporary poll could support a different reading, but national popular-vote estimates were not equivalent to an Electoral College forecast.

Verdict.

Reasoning: sound · Outcome: right

The realized winner agrees with the model's favored outcome, not with a guarantee of victory. No error is computed by subtracting an observed binary outcome from a probability as though both were point estimates of the same quantity.

Exact quote null. Original FiveThirtyEight URL now redirects to ABC, but search snapshot retained contemporaneous article text; 90.9% corroborated by dated Slate report. Do not call 332 an expected-vote point forecast: contemporary reports confuse modal map and expectation.

Cause of death / explanatory lens.

No cause of death assigned

The lesson.

Are we judging a probability by one result or by calibration over many comparable forecasts?

Sources.

  1. Nov. 5: Late Poll Gains for Obama Leave Romney With Longer Oddsprimary · 2012-11-06 · model reasoning and pre-election direction; search archive retained
  2. Nate Silver's Election Day Projectioncontemporary report · 2012-11-06 · 90.9% forecast corroboration
  3. Romney 49%, Obama 48% in Gallup's Final Election Surveyprimary · 2012-11-05 · contemporary counter-signal
  4. 2012 Electoral College Resultsofficial · 2012 · outcome

Related predictions.