Lie detector: how to critically interpret published statistics
Statistics are numbers, but because it seems to us cold and hard. There is a sense that it reflects the facts that gave us the nature itself, and our job is just to find them. But it is important to remember that collecting statistics people, and no one else. People choose what to count, how to count, what results to share, what words to describe them and how to interpret the numbers. Statistics are not facts, it’s interpretation. And your interpretation can be as good (and as bad) as the one that offers you a different person.
The number is not always true, and for starters, the easiest way to quickly check their credibility. Even if they passed the test, you may have questions of three types: how data were collected, how they were interpreted, and as presented graphically. The answers will help you formulate the right conclusions.
You can check (in most cases this is possible), most likely the fact in the mind or on the reverse side of the envelope. Don’t take anything for granted, try to understand.
When we conduct a similar test, the accuracy of the figures is not very important, as paradoxical as it may sound. Simply common sense: if Bert says that the crystal glass fell from the table onto the carpet and crashed, it seems plausible. If Ernie says that the glass fell from a height of 40‑storey building onto the pavement without crashing, it will be implausible. Here you will knowledge about how the world works, as well as basic life experience. Similarly, if someone says he is 200 years, or that he constantly wins at roulette in Las Vegas, or that he can run 40 miles per hour,— all this will be unlikely and very improbable.
What can you say about the following statement?
35 years after California ceased to operate the law on marijuana the number of marijuana users is doubling every year.
Sounds plausible? Let’s see, but where do you start? Suppose that 35 years ago in California there was only one smoker of marijuana — certainly a very low estimate (in 1982 across the country were half a million arrests for Smoking marijuana). If every year you double that number for 35 years, will receive 17 billion people— more than the population of the entire globe. (Try to calculate yourself and you will see that the annual doubling in for 21 years takes you to number more than a million: 1; 2; 4; 8; 16; 32; 64; 128; 256; 512; 1 024; 2 048; 4 096; 8 192; 16 384; 32 768; 65 536; 131 072; 262 144; 524 288; 1 048 576.) Thus, this statement is not that implausible— it’s impossible. Unfortunately, it is not at all impossible to think clearly when it comes to numbers: many of them are just afraid. But, as you can see, such calculations will be enough school-level arithmetic, plus a healthy skepticism.
Here is another example. You just took the sales Department on the phone and you need to call some unsuspecting (and, no doubt, angry) potential customers. Your boss is trying to motivate you, says:
Our best specialist have sold thousands of products per day.
Is this plausible? Try to dial a phone number — the least you need five seconds. Plus another five seconds to get through. Now let’s assume that each call actually ends the sale is, of course, not very realistic, but let’s imagine the ideal option to see what happens. Add ten seconds: you say an offer to sell, a potential client would accept it. Then another 40 seconds to find his address and write the credit card number. It gives one call per minute (5 + 5 + 10 + 40 = 60 seconds), or 60 sales per hour, or 480 sales for a very busy eight-hour day, without breaks. So a thousand items sold per day is unrealistic, even in the most optimistic scenario.
Some statements are harder to estimate. For example, a headline from Time magazine in 2013:
People with mobile phones more than those who have a toilet.
And what about this statement? Nahum come, on the one hand, people in developing countries without water, and on the other hand, those numerous residents of prosperous countries who have more than one mobile phone. It seems that the statement is quite plausible, it does not mean, however, that we must accept it. Rather, we cannot reject it simply because it is ridiculous. We will need other techniques for its evaluation, but the test for credibility it passed.
Sometimes it is impossible to assess whether the statement is true, without conducting their own research. Yes, of course, Newspapers and websites would be doing it for you, but not always — that’s when statistics goes bad. A few years ago was very common that such a statement is based on statistics:
Every year in the U.S. die from anorexia 150 thousand girls and young women.
Well, let’s see if this is in fact plausible. According to the us Centers for control and prevention of diseases, the annual number of deaths of girls and young women aged 15 to 25 from all kinds of diseases — 8500. Add to this women from 25 to 45, the figure will still only reach 55 thousand. The number of deaths from anorexia for a year cannot exceed three times the number of all deaths.
In his article for the journal Science Louis Pollack and Hans Weiss reported that since the formation of the Communication Satellite Corp
…the cost of telephone calls declined by 12 thousand percent.
If costs are reduced by 100%, they fall to zero (no matter what they were originally). If costs fall by 200%, it means that someone pays you the same amount once you pay him to get his product. A reduction of 100% occurs very seldom, a decline by 12 thousand percent seems altogether unlikely. In an article in the professional Journal of Management Development asserted that the number of customer complaints decreased by 200% as a result of the transition to the new policy, customer support
Writer Dan Keppel even called his book Get What You Pay For: Save 200% on Stocks, Mutual Funds, Every Financial Need (“Get what you pay for: save 200% on the stock exchange, investment funds of open type, for any financial need”). Have Cappela have a MBA degree. He should better understand the issue. Of course, to carefully compare the interest rates, they need to take from the same baseline. You cannot go back to the original salary level, reduced by 50%, increasing to 50% of your new, lower salary.
Interest seem simple and logical, but sometimes they can be confusing. If the interest rate increases from 3 to 4 percent, it increases by 1 percentage point, or 33% (increase of 1 percentage point measured from a baseline of 3; this increase is 1/3= 0,33 3). If the interest rate falls from 4 to 3 percent, it will decrease by 1 percentage point. However, it will not diminish 33%, as in the previous case, a 25% decrease of 1 percentage point measured from baseline to 4 (1 is 1/4, or 25%, from 4). Researchers and journalists were not always scrupulous in this matter and do not see sometimes the difference between percentage points and percentages, but you don’t have to be confused.
The New York Times announced the closure of a textile factory in Connecticut and move to Virginia. The reason for this decision were increased costs to the employees. According to the newspaper, “the payroll, all types of compensation, and unemployment benefits in Connecticut is 20 times higher than in Virginia.” Is this plausible? If that were so, you would probably expect a mass Exodus to Virginia — all companies and not just this factory, wanted to move, willow would know about it. Actually it is not true, and the Times had to publish a refutation. How could this happen? The fact that the journalist simply misread the report. Single indicator — unemployment – actually cost the company 20 times more expensive in Connecticut than Virginia, but given the other rates in Connecticut all costs for staff were higher than 1.3 times and not 20 times. The author of the article had no education in business-administration — and we have no right to expect it. To track this kind of error, you just need to calmly think things through. It is possible for everyone (and the journalist and her editors just had to do it).
In new Jersey approved a new legislative initiative, whereby mothers who are on social security received no additional benefits. Some members of the legislature felt that women in new Jersey, specifically children, to increase the monthly allowance received from the state. Two months later, lawmakers said that they were able to resolve this issue, as the birth rate decreased by 16%. Here’s what he wrote the New York Times:
Just two months later, the state released data that the number of newborn children of women already on welfare, decreased by 16%. The authorities congratulate themselves, with stunning results they have achieved in such a short time.
Please note that it was considered not pregnant, and the number of births. What’s wrong with this picture? Since pregnancy lasts nine months, no changes over the past two months cannot be associated directly with the law. Most likely, this is the role played by normal fluctuations in the birth rate (it is known that the birth rate — it is seasonal). In this regard, and other inaccuracies that cannot be detected by a simple test of credibility:
…over time these 16% reduced to 10%. The fact that the state became aware of the birth, which have not been previously reported. It turned out that many mothers did not consider it necessary to announce their newborn children, as their social grants for the period of child care did not increase.
Here’s an example of what problems can be encountered when collecting statistics: it turns out we take into account not all people, but I think that covered everything.