Election Math That Didn’t Compute
Shortly before settling in to absorb a bunch of voting returns on election night I ran across a website that made the whole thing simple. The author presented a vast array of charts and numbers, with razzle-dazzle analyses, that purported to predict accurately who would be the next President of the United States of America based on just a few early ballots. It was impressive, and no flaws in the presentation were apparent. If you wanted a quick answer, all the number crunching boiled it down to this:
If the race was close in Indiana, and Obama won Virginia, there was a 100 percent chance that Obama would be our next President. If those two things didn’t happen, we were in for a close race and a long night.
The polls closed very early in both states, so I figured I could learn what was going to happen in not more than an hour after dinner. Maybe I could even win a few last-minute bets (we lost a dinner bet when Obama eventually won). Early results showed a close race in Indiana. The trouble was those unpredictable voters in Virginia just kept standing in line for several hours. About the time CNN was willing to call Virginia for Obama, Obama clearly had won the entire election. Following the wizard’s advice didn’t help me learn a darn thing any quicker than anybody else did.
It wasn’t the first time a flaw in human performance caused frustration when I was trying to follow the guidance of a statistical genius at election time. When I worked in public relations for Allis-Chalmers in the 1960s, our supervisor “volunteered” himself and several of us to work for the Associated Press on election nights. Supposedly, we were doing a civic duty by helping tabulate incoming election returns. Actually, we were trying to butter up the news people at AP. We knew it, and they knew it, but they needed help and we were available.
At that time, the Associated Press was the principal news organization that took it upon itself to declare winners in elections. They used a fairly sophisticated statistical program that incorporated samples from key parts of a state to predict winners. In Wisconsin, where we helped out, the samples were mostly from the heavily populated areas in the southeast, but also included a county or two in the northern rural areas.
On the night of my frustration, we added up totals from report after report, but our leaders refused to declare a winner in the most important contest. Midnight passed; still no call was made. We finally had added up nearly 90 percent of the precincts in the state, and started clamoring for a call. We were told that the statistical formula absolutely required returns from one small county in the north, and it hadn’t reported.
About 2 a.m. somebody got somebody in the northern county out of bed and demanded to know what was going on. Our contact had merely forgotten about the AP commitment and retired for the night hours earlier!
This very unhappy camper got home around 4 a.m. that election night, just about in time to get ready to go to work at my day job. I should have learned right then not to get mixed up with political predictions, but Tuesday night I did it once again.