How a half-truth becomes a zombie stat

🏅12 Olympic records have been broken so far at the 2024 Paris Olympics
Issue #131: breaking down complicated AI concepts, propaganda x-ray vision, and unreliable facts
By Zulie @ Medium

Did you know that 60% of marketers create one piece of content per day? It must be true because it’s in a post about content marketing stats every marketer should know.

But wait: When copywriter Noya Lizor began to dig into the legitimacy of that stat, she discovered that the article, updated in 2021, was originally written in 2016, and included that statistic then. (I corroborated this by looking at the source code.) The blog post itself isn’t the original source of the statistic; it cites that 60% figure from a study conducted in 2012 and published in 2013. This means that when the article was updated in 2021, the fact was nine years out of date. Plus, when I tried to read the original study today, it led to a 404 page.

That’s a “zombie statistic” — a numerical half-truth that staggers around the internet without a clear source. Here’s how to spot a zombie stat:

• It’s several years out of date, like Lizor’s example above.
• It can lack a hyperlinked citation. For example, this Forbes post is full of improperly cited marketing statistics.
• Sometimes, it’s totally made up. For example, the 10,000-step myth was invented for marketing purposes by a Japanese pedometer company.

Those easily-repeated numbers are misinterpreted at best, or just plain wrong at worst. That doesn’t stop them from being plastered all over the interwebs.

Another kind is what I call the “telephone zombie stat” — one that loses context as it gets passed from headline to screenshot to group chat. I loved data scientist Elle O'Brien, Ph.D.’s takedown of one of these she found in the wild a few years ago: the shocking fact that 44% of Americans don’t want kids. Even though Today cited the number from the reputable Pew Research Center, it turned out that this 44% figure only included Americans aged 18–49 who did not yet have kids, not “all Americans.” If you ask the older child-free community if they’re planning on having children, the answer probably skews to “no” more than is truly representative of all Americans. Plus, 44% includes both those who said they’re “not at all likely” and those who are “not too likely” (so maybe somewhat likely!) to have kids.

In short, the study actually concludes that 23% of Americans aged 18–49 who are not already parents are not at all likely to become parents. That’s one heck of an asterisk.

One more story

Despite being fascinated by AI, many of the technological concepts behind it still elude my understanding. That’s why I loved Shreya Rao’s promise to “take complex-sounding concepts and make them fun and nbd by illustrating them,” in her deep and illustrated dive into neural networks. At its most basic, a neural network (modeled after our own brain neurons!) takes inputs, like how many bedrooms a house has and its location, and gives you an output, like how much that house should cost. You can then attach various weights to each factor, depending on how much it should influence the final outcome. For example, the number of bedrooms might matter more than location to determine the price. Each of the factors can influence the weight of the other factors, too. Small houses with too many rooms, Rao points out, could feel cramped, lowering the price. Neural networks can go deeper and deeper into conditionals like that, ultimately giving you a pretty accurate estimate of whatever it is you’re trying to predict.

Your daily dose of wisdom: become immune to propaganda

Apropos of trustworthy news sources, I enjoyed ethnographer Regie Stites, Ph.D.’s explanation of how he learned to see through propaganda. One important step is to “​​consider the feelings a political image or piece of news is intended to evoke.” Once you’ve identified the feeling (normally FOMO, fear of others, paranoia, or hope), you can combat it.

Learn something new every day with the Medium Newsletter. Sign up here.

Edited and produced by Scott Lamb, Harris Sockel, & Carly Rose Gillis

Questions, feedback, or story suggestions? Email us: tips@medium.com