BLOG

Book Review: How to Lie with Statistics, Part II

More Tools You Can Use to Create Convincing, and Arguably True, Untruths and How to Spot Such Shenanigans When Others Use Them on You

“Everyone is entitled to their own opinion,” wrote Daniel Patrick Moynihan, “but not their own facts.”

How to Lie with Statistics

The legendary U.S. Senator must have known this wasn’t exactly true, given his long political career—though the quote certainly has the ring of truthiness. As Darrell Huff’s book, How to Lie with Statistics, shows, with an aptitude for math and some sociopathic tendencies, almost everyone really can have their own facts.

This is part two in Chainsaw’s book review, which includes fresh examples of classic statistical manipulation; for part one, click here.

Why two parts? Because professional statisticulators in boardrooms and newsrooms all over the world have an amazing array of methods to collect their own facts and use them to prove wildly untrue points, and we needed two blog posts to even explain the tip of the iceberg. Enjoy.

The Gee Wizz Graph

You have probably been conditioned since grade school to think look at a graph and just assume that the x and y axes begin with zero, and rise from there in an even, proportional manner. If you look closely at graphs presented by political and business entities, however, you might notice that this isn’t always the case.

This one comes from FOX News, but liberal types shouldn’t be complacent about sources either. When there’s an axe to grind, axes get ground down.  Image courtesy of Media Matters

Huff’s book offers several excellent examples of graphs made more dramatic by simply “lopping off the bottom,” and “changing the proportion between the ordinate and abscissa.”

This technique is being used right now to misrepresent rising health care costs as a function of aging Baby Boomers. If you look more closely, however, you’ll see that those associated costs are slightly higher, but overall, almost negligible. Who benefits from this misrepresentation?

Another easy way to make a graph more dramatic is to use a three dimensional figure—say, a bag of money, or a drawing of the item being discussed—that increases in volume as well as height. So, if it rises 50% on the X axis, it increases in volume by 100%.

Your conscious mind might think this a cheap trick, but your unconscious registers it as much more dramatic growth than a simple bar graph would imply.

For more examples, read the book, or watch the news a bit more closely. “Some of this may be no more than sloppy draftsmanship,” admits Huff. “But it is rather like being shortchanged: When all the mistakes are in the cashier’s favor, you can’t help wondering.”

The Semi-Attached Figure

This is a figure that, while true, has little or nothing to do with the fact or product you need to shill. The classic example is the tagline, “Four out of five doctors smoke Camel cigarettes.”

The advertiser wants to give the impression that cigarettes are safe and healthy. Of course, they aren’t, and any relevant statistic, involving cancer, infant mortality, heart disease, and the like, would make that unpleasant fact clear.

But, by advertising a semi-attached figure—a number that suggests safety without actually proving it—you’ll have your customers justifying “just one more cigarette” in no time.

A common strategy is conducting an opinion poll. For example, your new client, Organic Corn Crunch cereal, wants an advertisement that suggests it’s safer to feed to your children than the competition cereal, which is made with GMO corn. Unfortunately, you can’t find scientific evidence that GMO food is unsafe, because there isn’t any.

The solution? Take a poll, in front of your neighborhood Wholefoods, asking if shoppers consider wholesome, organic, free-range corn safer than GMO corn—or how about frankencorn?—for their tiny, precious babies.

When 70% of those polled (12 out of 17 shoppers, you cleverly quit while you were ahead) choose organic corn, you’ve got your lede: “More than 70% of Americans Consider GMOs Unsafe for Children!” And deck: “Be A Better Parent By Feeding Your Child Organic Corn Crunch!”

Any statistics pertaining to finances, particularly from a biased source, are suspect. Even economically savvy people have a hard time parsing out numbers designed to mislead, and if you’re a layperson, you’re probably a mark.

Let’s say you’re thinking about investing with your friend’s cousin’s business, which is increasing its profit by 10% each year! You don’t have to fake any axes on that graph.

But, last year they had sales of $100,000, and made a profit of $10,000. This year, sales were $200,000 and income $11,000.  The profit margin has fallen from 10% down to 5.5% year over year. Which means the company is driving sales by discounting and margins are being squeezed, probably because it’s tanking.

How to Statisticulate
duck meme
 “Remember, not everyone is out to deceive you,” writes Huff. “Most people are incompetent when it comes to statistics. But, those who present statistical arguments on behalf of industry are seldom found, in my experience, giving labor or the customer a better break.”

Whether you’re a wizard with numbers or not, there are all sorts of shortcuts that amateur statisticulators can use to make their musings more believable, and they do.

Decimal Points
For a spurious air of precision that will lend all kinds of weight to the most disreputable statistic, consider the decimal. People are more likely to believe that U.S. Americans sleep an average of 6.38 hours per night than seven.

Adding Percentages
Quick quiz: Your restaurant’s rent has gone up by 10%, wages by 10%, and food by 10%. About how much have total costs have risen? Probably around 10%. But if you’re looking for a bigger budget, add it all together and you’ll get… 30%.

Conflate Percentages with Percentage Points
If your restaurant’s profits increased from 3% to 6%, you can say profits increased by a modest three percentage points, or by 100%. Your choice.

Imply that Correlation Is Causation
While two events may be statistically related, true causation is rare, while correlation is surprisingly common. Unless there’s scientific data suggesting a true causal relationship, it probably isn’t.

correlation graph

Graph courtesy of mathcoachblog

Change Your Base Years

Are prices going up, or down? That depends on what base year you use. To oversimplify: Let’s say this year, milk has gone from $2 to $4 per gallon, while bread prices have dropped from $2 to $1.  What would you like to prove? That the cost of living has risen, or fallen?

If you use last year as the base (when prices were 100%), bread prices have dropped 50% and milk prices doubled (200%). Since the average of 50 and 200 is 125, the cost of living has risen 25% in one year!

Try it again, using this year as the base period. Bread cost 200% more last year, while milk was selling for 50% of current prices. Last year’s prices were 125% of this year’s, which means the cost of living has dropped 25%!

How to Talk Back to A Statistic

The final chapter of Huff’s book offers some advice for folks who mistrust a statistic, but don’t have the math skills or lab results to disprove it outright.

WHO SAYS SO?

Does the source have a bias? Be honest with yourself—even if you don’t trust FOX or MSNBC, your online life is surrounded by a filter bubble that reinforces your opinions. If the source of the data has a conscious bias—or more insidiously, an unconscious bias—that’s a red flag.

If there’s an obvious Appeal to Authority, say, an article is cited as being from “Cornell University” when it was merely published in one of the school’s many journals, that’s another red flag.

HOW DOES S/HE  KNOW?

What was the methodology? What was the sample size? Who was being polled? Can you think of an experiment, graph, or style of reporting that might provide or present better data than the method used?

WHAT IS MISSING?

Is the average a mean or a median? What is the base year? Are numbers presented per capita, or total? Are they presented in a useful context? For example, if each employee holds 600 shares, how many shares are there total? What percentage of shares do employees hold?

If information key to understanding the statistic has been omitted, it’s a good guess that statisticulators have been hard at work.

DID SOMEBODY CHANGE THE SUBJECT?

Does the data really prove the conclusion? Just because 70% of respondents to your survey think GMOs are unsafe to feed to children doesn’t mean that GMOs have been proven unsafe.

DOES IT MAKE SENSE?

“Many a statistic is false on its face,” writes Huff. “It gets by only because the magic of numbers brings about a suspension of common sense.”

Does it seem to good to be true? Statistics say that red wine and dark chocolate are your keys to rapid weight loss, but you may want to try a gym membership and plenty of vegetables as well.

Statisticulation is a part of the modern information landscape. As Huff wrote: “In commercial practice, a statistician is about as unlikely to select and unfavorable method as a copywriter is to call his sponsor’s product flimsy and cheap when he might as well say light and economical.”

So give this slim little book a read, and we guarantee that you’ll be 75% better prepared to navigate the Misinformation Age.

2016-10-21T11:03:08-05:00 By |Categories: Communication, Psychology, Reviews|Tags: , |