Historical Election Analysis: Uncounted and Switched votes

http://www.geocities.com/electionmodel/Tru...

The Democrats lose votes in every election. This analysis determines the effect of uncounted and switched votes on presidential elections going back to 1948 – as well as on the 2006 midterms. The landslide elections are not included: Eisenhower (1952-56); Johnson (1964); Nixon (1972) and Reagan (1984). T

The adjusted vote shares show that Gore and Clinton won by higher margins than the recorded votes indicate. The 1988 Bush-Dukakis election is especially suspicious.

The True 2004 vote (Kerry 66.1- Bush 58.38mm) was determined by calculating the approximate number of returning Gore, Bush and Nader voters (2000 recorded vote less 3.5% % mortality, assuming 95% 2000 voter turnout). The 12:22am National Exit poll vote shares were the most-likely base case scenario. According to the U.S. Census, 3.4mm of 125.74mm votes were uncounted. Assuming that 75% of the uncounted votes were for Kerry, then 7.3% (4.5mm) of the total votes cast for Kerry must have been switched. The switched vote rate was based on the following equation:

True Vote = Recorded +Uncounted + Switched.

In 1988, Dukakis led Bush I in the preliminary exit polls, just as Kerry led in 2004. Assuming a 3% uncounted vote rate and that 7% of Dukakis votes were switched to Bush I (the same percentage switched from Kerry to Bush II), then Dukakis won the election by 50.3-49.7%.

For 1980, a 3.0% switched vote rate is assumed (Bush I was running as VP). In 1968 Nixon won by 500,000 votes; we assume a 1% switched-vote rate. For Carter in 1980, JFK in 1960 and Truman in 1948, the switched vote rate is assumed to be zero.

A sensitivity analysis table for each election displays the effects of changes in the Democratic share of uncounted votes and percentage of switched Democratic votes.