Proofiness - Charles Seife [97]
Figure A3. Expected outcomes of polls of 1,024 people.
In the sixty-four-person poll, the curve has gotten thinner, so the margin of error has shrunk. Now, 95 percent of the time, the answer you get is somewhere between 61 percent and 39 percent for Bush—you’re within plus or minus eleven points of the true answer. The more people you ask, the smaller the margin of error. In the 1,024-person poll, the curve has gotten much thinner. The margin of error is now 3.1 percent; you’re 95 percent confident that your poll is within 3.1 percentage points of the truth. The thinner the curve, the less likely it is that the poll gives an answer far away from the truth, and the smaller the margin of error.
As your sample gets larger and larger, the margin of error gets smaller and smaller. It never disappears completely, but it can be brought down to a level that’s manageable—leaving the pollster to worry about other kinds of nonstatistical errors that can potentially render a poll meaningless.
Appendix B: Electronic Voting
Nowadays, no discussion of electoral fraud can be complete without at least a mention of electronic voting. A lot of critics, particularly on the left, are deeply concerned about the advent of electronic voting machines. They’re terrified of the prospect of digital machines spiriting our votes away into the digital ether, perhaps allowing politicians to steal elections with a flick of a switch. It’s true that electronic voting has the potential to cause electoral chaos. At the same time, though, it has great promise; it can make elections better, reducing errors and saving lots of time and money.
In the ideal—if the machines function perfectly—electronic voting has a lot of advantages. The most obvious is that people no longer have to count ballots; when a voter presses a button or a portion of the screen to indicate his preference, the vote is counted at the very same moment that it is cast. The act of voting is no longer separate from the act of tabulation. This alone saves a lot of time and money. Election officials no longer have to print (and store and transport) paper ballots, reducing costs even further. On top of that, machines are a boon to the disabled; they can prompt blind voters with audio signals, for example. It’s no wonder that election officials want to switch over to electronic voting machines as soon as possible.
However, the machines have a downside. Unlike systems that use paper ballots, there’s no physical record in a purely electronic voting machine. You press a button and you have to trust that the machine is recording your vote properly. You might never know if it gobbles up your vote or records it incorrectly. As a result, when things go wrong with the machines—as they occasionally do—they can completely obliterate people’s votes. In the 2000 presidential election, a programming error wiped out at least 678 votes in New Mexico. (Al Gore “won” the state by a mere 366 votes.) Reports of vote-eating machines have surfaced all around the globe; in France, for example, a computer scientist reported in 2008 that precincts with electronic machines seemed to be swallowing votes at an alarming rate.89 Also, there’s the possibility that the machines will create votes out of thin air: in 2003, an election in Boone County, Iowa, initially registered 144,000 votes on electronic machines even though there were only 19,000 eligible voters. Sometimes machines rob one candidate and give votes to another. In 2000, one Florida county’s electronic machines registered some negative 16,000 votes for Gore, while Bush received roughly 2,800 votes—all in a county with fewer than 600 registered voters. Many of the