How To Lie With Statistics: Understanding Core Principles
October 23, 2019
Crafted by a non-statistician, in a contrived language and illustrated using humorous drawings, How To Lie with Statistics is relevant in the twenty-first century as when it was first published in 1954.
This book is still a bestseller, although some of its examples are outdated, such as the price of bananas and the earnings of Yale graduates. Besides, the tricks that Darrell Huff describes are still applicable today. These include misuse of averages, misleading charts, and more.
Like the author argues, statistics are often false on its face, and this continues to be the case because numbers have magic that tends to suspend common sense.
Indeed, lying with statistics is easy. Statistics are useful; they help us to explain and discover the way we live in every way. They provide perspective on the past while making the future predictable.
However, statistics can also be used to confuse, manipulate, sensationalize, and obfuscate. We will look at How to Lie with Statistics summary by looking at the different ways Huff highlighted and how that still applies today.
How To Lie With Statistics In 7 Ways And Get Away With It
1. Biased Sampling
By definition, samples are the incomplete picture of the whole. However, the question is, how much of the whole is it? When samples are big enough and chosen correctly, they tell us something.
Biased sampling refers to polling a group that doesn’t represent the entire picture.
For instance, a survey may show that “43% of commercial bank customers would opt for mobile banking if it was offered” becomes nonsensical after you realize that the survey only targeted individuals on their mobile phones.
2. Small Sample Sizes
Selecting a sample size is both an art and a science. To make sense, conclusions should be drawn from big sample sizes.
Thus, a statement, such as 15% of organizations, intends to adopt cloud storage this year” can quickly become suspect when you discover the sample size is only 20 companies. But as long as you quote 15% and omit the exact number of companies, no one will suspect anything.
The second example of such statistics was the “study” that HP carried out to discover that excessive use of email lowers someone’s IQ by 10 points.
3. Poorly-Selected Averages
This typically involves averaging values for non-uniform populations. For instance, I recently read an article that talked about a neighborhood as one of the city’s wealthiest areas.
The same article mentioned that the residents in this neighborhood had an average of $100,000 in annual income. However, the article failed to reveal the fact that a section of the neighborhood is wealthy, whereas the other section has income levels that are way below the national average.
Providing one average value for two distinct populations is not only misleading but also incorrect. The median value for the neighborhood’s income would be a better indicator of the residents’ income.
Another case in point is the story of the man who drowned in a pool with an average depth of one inch.
4. Results That Fall Within The Standard Error
There’s no perfect measuring or sampling technique; they all have a degree of error. In other words, surveys can only be as accurate as their standard errors.
The headline, “More men than women, prefer eBooks over paper” sounds fantastic until you discover that according to polling results, 52 percent of men preferred eBooks against 49 percent of men, and the survey had a standard error of + or – 5 percent.
5. Use Graphs To Create A Certain Impression
The two charts below have exactly the same data. Which chart shows an accurate rise in investment in mobile technologies between 2005 and 2007?
Only the scale is the difference between these graphs. As you can see, graphing data gives plenty of room for creating a false impression. The same applies to infographics and pictograms.
6. “Post-Hoc Fallacy”
This is where a researcher incorrectly asserts that two findings have a direct correlation between them. Compared to the other tactics mentioned, this can be tricky to figure catch.
For example, if a study says that vegetarians earn a higher income than people who eat meat, it may be absurd to start concluding that, to increase your income, you should abstain from meat. Surprisingly, that’s what some ‘researchers’ out there do.
7. “The Semi-Attached Figure”
This refers to a tactic where someone states one thing as proof of something different. For instance, if an ad claims that “20 percent of CEOs drive a Mercedes Benz more than any other car,” it may imply that CEOs are authorities on cars. This tactic is more common than you think and is one of the misleading statistics in the media.
Lying With Statistics Is Easy
Thus, when looking at statistics, you should always be skeptical. Remember, your professor knows all these tactics, so you should guard against them when writing your research papers.
Otherwise, you may end up receiving endless revisions or even low grades. To avoid losing marks due to questionable statistics in your assignments, consider hiring writing help.
Take a break from writing.
Top academic experts are here for you.