Possibility of Recounts Would Be Reduced By a National Popular Vote
There were 23 recounts among the 7,645 statewide elections conducted in the 26 years between 1980 and 2006 ― that is, one recount in 332 elections. The average change in the margin of victory was a mere 274 votes in these 23 recounts. Moreover, the probability of a recount drops as the size of the jurisdiction increases, and only two of the 23 recounts were in big states. The original outcome remained unchanged in 91% of the 23 recounts.
Under the current state-by-state system of electing the President, there are 51 separate opportunities for a recount in every presidential election. The reason why there have been so many serious disputes in our nation’s presidential elections (5 disputes in 55 presidential elections) is that these 55 elections entailed 2,084 separate statewide counts. In contrast, if the President were to be elected using a single nationwide pool of votes, one would expect a recount once in 332 presidential elections ― that is, once in 1,328 years.
Indeed, the topic of recounts only comes to mind in connection with presidential elections because the current state-by-state system repeatedly produces artificial crises in which the vote is not at all close on a nationwide basis, but extremely close in a particular state.
The 2000 presidential election was an artificial crisis created because of Bush’s lead of 537 popular votes in Florida. Gore’s nationwide lead was 537,179 popular votes (1,000 times larger). Given the miniscule number of votes that are ever changed in recounts (averaging only 274 votes), there would have been no recount in Florida (or any other state) in 2000 if the national popular vote had controlled the outcome. Indeed, no one (except almanac writers) would have cared that Bush had a 537-vote margin in Florida.
Samuel Tilden’s 3% lead in 1876 was a solid victory in terms of the national popular vote (similar to Bush’s solid percentage lead in the 2004 election). However, an artificial crisis was created because of the razor-thin margins of 889 votes in South Carolina, 922 in Florida, and 4,807 in Louisiana.
There was a recount, court case, and reversal of the original outcome in Hawaii in 1960. Kennedy ended up with a 115-vote margin in Hawaii in an election in which his nationwide margin was 118,574.
For those genuinely concerned about recounts, a single national pool of votes is the best way to drastically reduce the likelihood of recounts and eliminate the artificial crises repeatedly produced by the current system. No one has ever suggested that the possibility of a recount constitutes a reason why state governors and other officials should not be elected by a popular vote.

The state-by-state winner-take-all system is not a firewall, but instead causes unnecessary fires.
Under the current system, there are 51 separate vote pools in every presidential election . Thus, our nation's 56 presidential elections have really been 2,135 separate elections. This is the reason why there have been five seriously disputed counts in the nation's 56 presidential elections. The 51 separate pools regularly create artificial crises in elections in which the vote is not at all close on a nationwide basis, but close in particular states.
A recount is not an unimaginable horror or logistical impossibility. A recount is a recognized contingency that is occasionally required (about once in 332 elections). All states routinely make arrangements for a recount in advance of every election. The personnel and resources necessary to conduct a recount are indigenous to each state. A state's ability to conduct a recount inside its own borders is unrelated to whether or not a recount may be occurring in another state.
If anyone is genuinely concerned about the possibility of recounts, then a single national pool of votes is the way to drastically reduce the likelihood of recounts and eliminate the artificial crises produced by the current system.