The issue of measurement error rests on the variables that affect measurement tolerances. There are three that I could come up with. One is called "residual votes", which is the difference between the number of people who show up on election day and the total number of votes actually reported for that precinct. This difference is due to disallowed ballots, which depends entirely on the technology being used and the zealousness of the people deciding which ballots get tossed. The second is the finite accuracy with which any voting technology can count votes (anybody who has had to count a jar of pennies knows that accuracy is finite). The third is outright fraud, which is very, very difficult to measure. I couldn't find any published research on the web that addresses the question of total accuracy in voting, either. Before throwing numbers around, though, we have to recognize that the electoral college magnifies the effect of error - if a populous state has a lousy system of vote counting, for whatever reason, it can easily overwhelm dozens of states with very good systems. Seven states (California, Texas, New York, Florida, Pennsylvania, Illinois and Ohio) have the greatest number of electoral votes, and together have 39% of all the electoral votes. Any one of these states alone is enough to swing the result of a very close election. This thread is addressing the question of whether this system is "fair" or "democratic". If you go by the standard of one person - one vote, then it clearly is neither fair nor democratic. It makes the job of swinging an election through vote fraud a whole hell of a lot easier. Back to vote count accuracy. The National Commission on Federal Election Reform (http://www.reformelections.org/data/reports/99_full_report.php) recommended that residual rates in vote counts should be reduced to 2%. This implies that current residuals are greater than 2%. And this is before you talk about simply miscounting the number of legal votes or vote fraud! So - two percent: In the 2000 election a grand total of 105,326,325 votes were cast for the president. If the residual rate was "only" 2%, then that means that 2,149,516 people went to the polls and voted only to have their votes thrown out. Who were they, what technology did they use to vote and for whom did they intend to vote? It seems to me that if the U.S. is going to erase at least 2% of the votes in any given election, then the margin of victory should be at least 2% (assuming that disallowed ballots are random, and not more likely to have favored one candidate over another). If the margin is less than 2%, then the only reasonable course is a runoff election. Oh, and get rid of the electoral college. It may be consitutional, but that doesn't make it right. Failing that, insert an amendment that if a state produces a result with less than 2% margin, then its electoral votes go unawarded. I would be willing to bet that the U.S. Presidential election system has as much as 5% error. I think that anything greater than 0.1% is totally unacceptable.