And the answer is…

What was the approximate vote total difference across three key states which decided the Presidential election?

A) 10,000

B) 80,000

C) 3,000,000

If you guessed 3,000,000, don’t feel bad. That’s the approximate number by which Clinton won the popular vote. So it was a number “in the news” that many would recognize. The correct answer is B, 80,000, the total number of votes which tilted the electoral college in Pennsylvania, Wisconsin, and Michigan into the Trump column. That’s right. 0.06% of the vote – and 0.0012% of the American people – decided who gets to be “the leader of the free world.”

If you got it right, I tip my hat to you. You’re on your way to being “a numbers guy (or gal!)”. 

Pollsters, the people who hire them, and many of rest of us who suffer through the endless news reports on who’s up, who’s down, and who’s sideways, will be smarting over the vast discrepancy between the forecasts and the outcome of the recent presidential election.  But this tiny difference highlights what most people miss about virtually all numerical results used and abused by policy makers, academics, business associates, and even your friends and colleagues: The uncertainty and error around numerical results are as important as the numbers themselves.

Pollsters in this case were trying to predict an outcome of an election which turned out to be exceedingly close, and very different from recent presidential elections. No one can predict the future, so forecasts are only informed guesses based on past data and experience.

Some polling turns out to be wrong. But some polling is more than that. It is nefarious. It’s used to drive voter sentiment, rather than indicate it. Campaigns, and often surrogates and advocates which are less than transparent, manipulate the news cycle using their own polling. Conduct a poll, show your candidate is winning, get the results reported, and try to reinforce the inevitability of your guy or gal. Create momentum.

In Painting By Numbers: How to Sharpen Your BS Detector and Smoke Out the Experts, I give you twelve layman’s commandments for evaluating any numerical result and then demonstrating how you can apply the commandments to your daily life. When it comes to polling, pollsters typically violate many of these commandments. They don’t acknowledge error except statistical error (commandment one), they rarely identify the assumptions used in their models (commandment two), they are trying to gauge sentiment not measure a physical quantity (commandment seven), and they often start with the answer (commandment eleven) and work backwards.

Want to learn more? I’m giving away the first five commandments from Painting By Numbers for free when you sign up for my mailing list. This will start you down the path to breaking down numerical analysis and smoking out the quant experts.

All you have to do is tell me where to send the PDF.