Electoral Scoreboard Explained

• The statistical method is very simple, but also very powerful. We simply take the median margin of victory by either candidate in all available polls in the last month. We also list wins-losses-ties for the candidate winning in the median in each state. By definition, the candidate with the most wins will have the higher median result. In that sense, the method essentially becomes a scoreboard of poll wins.
  
  
• The median is formed by sorting the list of all polls, then taking the middle value. For example, if we had these polls for a given month in a particular state:
  • Romney +5, Romney +3, Romney +5, Romney +3, Obama +1, Obama +1, Romney +1, Obama +1, Romney +2, Romney +8, Obama +1, Tie, Romney +6, Romney +1, Romney +1
  ... the sorted list would be this:
  • Obama +1, Obama +1, Obama +1, Obama +1, Tie, Romney +1, Romney +1, Romney +1, Romney +2, Romney +3, Romney +3, Romney +5, Romney +5, Romney +6, Romney +8
  There are 15 entries in the list, so the middle entry is the 8^th, which is Romney +1, so the median result is Romney +1, and as such, Romney is reported as the winner.
  
  
• We also report Wins-Losses-Ties. One may look at the 1-point median lead for Romney, and say that the state was essentially a toss-up in September. However, Romney's record was 10-4-1 against Obama, which is a healthy winning record.
  
  
• The median is powerful, because it throws away discrepant points automatically, without bias. Though it is slightly slower to converge to the true value than is the sample mean (by a factor of sqrt[2*pi] for a normal distribution), it avoids the sample mean's high susceptibility to outliers. As such, a particular poll that may be biased, have systematic errors, or occur during a temporary "bump" for either candidate, introduces far less error in the median than in the sample mean.
  
  
• An illustrative example of the power of the median may be found in this manuscript in which the median is applied to estimate the long and hotly contested value of the Hubble Constant (H₀):
  
  
  • Gott, J. R., Vogeley, M. S., Podariu, S., Ratra, B., 2002, Median Statistics, H₀, and the Accelerating Universe, Astrophysical Journal, 549: 1
  
  The authors of this paper note an amusing, but highly relevant quote by one of the greatest luminaries in astrophysics:

  ... Ya. B. Zeldovich commented on [the median]. He noted that [in his home country] some watches are not made very well, so when three friends meet they compare the times on their watches "one says `it's 1 o'clock,' the second says, `it's 5 minutes after 1,' the third says `it's 5 o'clock.' Take the median!" Perhaps no one has ever stated the benefits of the median over the mean better than Zeldovich.