Princeton Election Consortium

A first draft of electoral history. Since 2004

Senate Seat History

This graph plots the average number of expected Democratic/Independent Senate seats, based on state-level polls available on the date of the analysis. The calculation is done using the same algorithm as the 2004-2012 Presidential Meta-Analysis: (1) use 3 or more recent polls* to calculate a median margin and estimated uncertainty between the two major candidates, (2) convert the margin and uncertainty to a win probability, and (3) calculate the exact distribution of all possible outcomes.

In 2014, the Princeton Election Consortium is tracking 19 out of 36 Senate races. The above graph shows the average outcome from all 2^19=524,288 possible outcomes. The other 17 races are in no doubt, and have virtually no polling data.

The gray band indicates the 1-sigma confidence interval, i.e. approximately the middle 68% of all outcomes. This confidence interval includes true uncertainty as well as pollster-to-pollster variability. If pollsters are, on average, good judges of the electorate, which they usually are, the real uncertainty is considerably smaller.

*For each state, the 3 most recent polls from different organizations are always taken. In addition, polls are taken for the last N weeks, where N=6 for July and decreases gradually to N=2 by October.