Yogi Berra supposedly said, “It ain’t over till it’s over.” The only official election count is the final, verified count of every single vote.
Everything else is just a projection – an estimate, a sample.
When elections results or projections don’t match the final count, well, people start spreading silly conspiracy theories. Sports help us understand that.
A Sports Analogy So, for the moment, let us think about elections like a sport. Suppose the New Paris Porcupines meet the Old Heathen Hedgehogs on the gridiron. With one minute to go in the fourth quarter, the Porcupines lead by 21 to 8. The Porcupines have 72% of the points, and the lowly Hedgehogs only have 28%. The Porcupines are running away with the game! The Hedgehogs don’t have a chance!
Then, disaster strikes! The Porcupines fumble the Hedgehogs’ onside kick. The Hedgehogs’ worst player scoops up the ball and runs for a touchdown! Add an extra point! The score is 21 to 15! Horrors! The Porcupines’ devoted fans cover their faces in terror!
 |
| High School Touchdown |
Ready for another onside kick, the Porcupines crowd up to the kicker, only to watch the ball sail over their heads to the 10-yard line. A Hedgehog scoops up the bouncing ball and, before you know it, the whistle blows, and the Hedgehogs win, 22 to 21.
Sports fans call that a come-from-behind triumph.
But what if the Porcupines’ fans are politicians? That changes
everything:
“We were winning with 72%, and then lost in only one minute! That proves bad scorekeeping! The Hedgehogs cheated! Throw the spiny fuzzballs in jail! Forfeit their season! Lock up the refs! It is impossible to lose when you lead by that much in the last minute! Statistics don’t lie! The refs should have stopped the game. It was all but over until the Hedgehogs cheated! Cheaters!”
Obviously, say the politicians, this must be football fraud. The numbers prove it!
That is nonsense. Sports fans know that teams can have a different score in every quarter. Newsreel touchdowns happen. Something similar is true in elections.
In fact, this is a perfectly fair comparison. Let us look at why.
Estimating the Final VoteLet’s suppose we are counting votes in Political City, USA. Political City has two precincts. Precinct #1 is a rich suburb, packed with Republicans and their SUVs. Precinct #2, however, is on the poor side of the tracks. Most of the voters are Democrats, many of whom hobble around in poor health and prefer to vote by mail. Political City has about 40,000 voters, but budget cutbacks left them with too few poll workers. Counting all the ballots will drag on and on. And on and on and …
All the same, everyone is excited to know whether Pat Psychopath or Chris Sociopath will win the mayor’s race. Long before the complete vote is tabulated, the local news wants to get a quick sample to find out who won. We can’t wait for days to hear the final count. I mean, seriously, think of the suspense! Nerve-racking!
That’s where statistics might help us. Statisticians take samples to estimate the final count. Does that work? Maybe, maybe not. It mostly depends on how you collect the sample.
Sampling Method #1: We collect all the ballots. But we don’t count them yet. Instead, we dump all the ballots into a huge, dry tub. We blindfold three honest poll workers, and they stir all the ballots with broomsticks for 30 minutes. The ballots are now completely mixed up. The rich Republicans, the poor Democrats, and voters who don’t fit into any category are all randomly mixed.
Now, still blindfolded, the three honest poll workers yank a thousand ballots out of the tub and those ballots are counted first. Do those thousand ballots predict whether Pat or Chris is going to be the next mayor? Well, with a certain margin for error, they probably do. It is remotely possible, of course, that by sheer bad luck the thousand ballots could be almost all Republicans. In that case, the thousand ballots would show that the Republican will win, even though the final count might say the Democrat wins. Still, most of the time – not all the time – but most of the time – a thousand ballots picked at random will closely but not exactly predict the final result. That’s because the thousand ballots were chosen fairly and randomly, and every voter had an equal chance to be included.
Anyway, let’s imagine that Sampling Method #1 correctly estimates that Democrat Chris Sociopath will be the next mayor.
Sampling Method #2: Most of Political City’s Republicans voted in person using electronic machines (which also create a paper ballot trail), and their votes are easy to count. The Democrats, however, mostly voted by mail. They are harder to count. Their ballots need to be opened, and the signatures must be verified. Then the ballots must feed into a scanner, and then the scanner’s information shuffles off into the main computer. There’s nothing wrong with that rigmarole, but it takes time. Now, since everyone is eager to know, the news reports the count two hours after the polls close. That count includes 20,000 votes! Half of the voters! A much bigger sample than we got with Method #1! A huge sample! The count for Method #2 mostly includes Republicans, since they were easier to count, and, when we look at the numbers, we think that Pat Psychopath will the election by a huge, huge margin. Maybe 72% to 28%! So, we all think that Pat will become mayor and ruin Political City.
However, once all the Democrats are counted days later, it turns out that Chris Sociopath won the election by 10 votes. We count and double check, since the election is so close, but it comes out the same every time. Chris won. Congratulations to Chris, and it will be Chris, and not Pat, who gets to make a mess of everything and ruin the town.
The difference between Sampling Method #1 and Sampling Method #2 is not the number of ballots in the sample. That matters, but less than you would think. No, the difference is how the sample was created. With Method #1, the sample is fair and every Republican and every Democrat, not to mention every independent and every confused voter, has an equal chance to be in the sample. However, with Method #2, although the sample is huge, it is not fair. Few mail-in ballots are included. Fewer Democrats have a chance to be included. Therefore, the sample from Method #2 is inaccurate, because Republicans are a bigger part of the sample from Method #2 than they are of the voting population.
So, if we project the election from Method #1, we will probably (not always, but probably) come close to the final result. If, however, we rely on Method #2, where the sample is unfair, the sample might not resemble the final tally.
Worse, all sampling introduces some error, probably 4% or 5% even using the best methods. The closer the election becomes, the more important sample errors become – since the sample will never exactly match the proportion of the entire population – and so it takes longer to produce a valid estimate. If the election is close, we might never have a valid estimate. That’s just how statistics work. The projections and the early reports are never the final total. That is why even the smartest news networks cannot calculate or even estimate accurate election results when the election is super-close.
“It ain’t over till it’s over.”
Remember – and this is critical – the only count – and I mean the only count – that means anything is the final tally that has been tabbed up, verified, double checked, and reported to the authorities. Everything you hear, either from the media or the voting officials, before we have a 100% verified count, is an estimate – an educated guess. That educated guess comes either from Sampling Method #1, which is often but not always right, or Sampling Method #2, which is typically wrong. An incomplete vote count is only a sample—and, worse, an incomplete count uses Sample Method #2. A count of 98% of the votes is still just Sampling Method #2. If the educated guess takes a long time, or if it is occasionally inaccurate, well, that doesn’t prove that anyone cheated. It only proves that samples do not always work, and that unfair samples often fail.
Down with the Conspiracy Theories
So, when Gripey Garrett gripes on social media, “My candidate was winning with 90% of the vote counted, but now she is losing,” all that means is that Gripey Garrett was using Sampling Method #2, and you should never trust Sampling Method #2.
Now, in the real 2024 election, Democrats tend to live in cities, and cities can take longer to count because they have more voters. How long does it take to count all the votes in a village of 500 people? How long does it take to count the votes of a million people? Also, Democrats often mail their ballots or drop them off in collection boxes. Those ballots need to be opened, verified, and fed into scanners. That is why it is not unusual for Democrats to need longer to count. When we see an early surge in Republican counts, followed by a surge in Democrats, well, that’s often just the process of counting. The early voting reports use Sampling Method #2. And you can never trust Sampling Method #2. Basic math.
So, if you want to prove that someone stole an election, you need to find evidence of the actual stealing. The fact that voting percentages change as the ballots are counted only means that you are using Sampling Method #2.
Earlier Post: Don't Expect Same Day Election Results Earlier Post: Were the Polls Wrong in 2020? And, if so, Why?
And, finally, congratulations to the Hedgehogs for their great victory! Hail to the Hedgehogs, good work. They played play hard to the end! The first, second, third quarters were not the whole game. The score after three quarters is just Sample #2. The fourth quarter still matters. The last 60 seconds still matter. We aren’t shocked when a team comes from behind to win. Likewise, we should expect voting percentages to change while the counters count. That’s just math.
Every eligible American citizen should vote in every single election. Never miss. Never forget. And always inform yourself before you vote.
Never trust Sampling Method #2. Thank you, and God Bless the United States of America.
by William D. Harpine
______________________________
P.S. How do athletic teams sometimes overcome late scoring deficits? Other than pure luck, the winning team might have more endurance. Maybe the losing team gets cocky and celebrates too soon. Each quarter is different—the score from quarters 1 through 3 is just Sampling Method #2! A come from behind victory isn’t cheating. It’s just sports. Only the final score counts.
 |
| Lever-Style Voting Machine |
P.P.S. Is it possible to cheat in an election? Of course. American elections were often (not always, but often) crooked as recently as my youth. Unsupervised partisan operatives sometimes managed the ballots. That was common practice here in Texas for decades. Elections in many states used paper ballots, and political operatives could erase the ballots, throw them away, or add more ballots if the results didn’t come out the way they wanted. Seriously. Have you never heard of the old (now defunct) Chicago political machine?
Likewise, mechanical voting machines, widely used for decades, were robotic fraud machines. I watched my vote get mixed up by a mechanical machine in 1980 in Virginia when I tried to vote for Independent John Anderson, and the officials just shoved me out the door. I should have protested, but, in Virginia, in 1980, who would have listened? Modern voting methods are, in contrast, stunningly secure. Is it still possible to cheat? Probably. But you need evidence before you call someone a cheater. Okay?
P.P.P.S. It's not easy to understand statistics. The math isn't necessarily hard, but the human brain is simply not wired to grasp probabilities. We think in terms of "yes" or "no," not "maybe." All the same, every school should require at least a brief stats course. It could save our nation so much grief...
I Voted Sticker, State of Texas, Nueces County
Image of lever-style voting machine by RadioFan, Creative Commons License, via Wikimedia Commons
Copyright © 2024 by William D. Harpine
.jpg "King Charles III")
.jpg){kind=link}
