"The average commute from Discovery Bay (eastern part of the San Francisco Bay Area) is 50.6 minutes." So says an article in the May 19, 2002 Contra Costa Times. What it doesn't say is where the commute is to. Later on it does quote Dave Ginter of Brentwood who commutes 68 miles to Mountain View which takes "80 minutes on a good day and more than two hours on a bad one ... and even worse when there's an accident enroute."
So let's see. 80 minutes on a good day. Say 140 minutes on a bad day. That's an average of 110 minutes one way! But that's only if Dave travels only two days a week. Now Mondays are always a bear, so that would be closer to 150 minutes. Fridays are great going in, but worse than Mondays coming back. Tuesday and Wednesday are about average, and Thursday just before a long weekend can be murder. Then, the California Highway Patrol reports that there's "an average of three accidents a day on Route 4." Throw all this into your computer, and we can say that Dave spends an average of four hours a day commuting to and from work ... unless there's also an accident on 242, 680, 580, 92, and 101.
All of this began with the statement that the "census confirms commutes are longer." Duh. Ask anyone who has lived more than 10 miles away from work in a metropolitan area for a couple of years, and they'd tell you. And they wouldn't say - as the article does - that it's averaging them 4% more than it did two years ago.
There are two things in this scenario that drive me nuts: statistics and averages, both of which can be used or misused to make a point. In 1954, Darrell Huff wrote a book, How to Lie With Statistics, which should be must reading for any of you who use even the most minor of stats on a day-to-day basis ... which would be all of us since even a 20%-off sale can be misleading. Oh, by the way, the book is so good that almost 60 years later it's still; in print: W. W. Norton & Company, New York.
How about, "The market was off 1.2% today." That could mean that the market lost 120 points or 150 points ... depending on where the market was yesterday. And tomorrow, if it goes up, the news starts with, "The market was up 245 points today [for no apparent reason]" and they leave off the percentages. There's a built-in editorial bias, just as headlines are meant to get your attention and not inform. Raise your hands: how many of you have seen, "XYZ.com's stock was up 175% today to close at 38¢?" [Now let's not always see the same hands.] I have no idea what it tells me, but I don't like the presentation. There's something misleading and slightly disingenuous about it.
None of that affects you? I'm the only one who reacts? Okay, try this: "Dear Mr. Winning, in an attempt to keep your insurance rates down, Blue Cross [read two more pages] will not be raising your premiums for the next half-hour. Rather, we're increasing your deductible and the co-pay for all prescription medications." That after a 31% increase in premiums just six months ago. And what this amounts to is a $200 a month increase in out-of-pocket costs to me. My insurance is up 100% over a three-year period. Now the statistics and percentages mean something.
"Small companies with three to 40 employees had an increase of 11.5 percent, compared with 8.9 percent for firms with 200 or more workers." What does that tell you? Actually, nothing. Since increases are based on location, ages of employees, the company's size and (I have a strong hunch) how many employees were severely ill or used more than "the average" set by the insurance industry, the company could easily be forced into passing the costs on to employees or getting rid of health insurance entirely. Further, those percentages were for one company. We know that several national firms raised their premiums between 40 and 60% over the last 16 months.
Is it any wonder than 34% of all Americans, including many working for companies which can no longer afford the premiums, are without health insurance?
Edward Liddy, CEO of Allstate, stated recently that the increases in homeowners and automobile insurance (18.5 and 8.9%) "are dramatic" but says that "the industry is adjusting out of necessity after years of offering relatively low-priced policies." Blaming much of the increases on pharmaceutical and hospital costs, what he should have said is that Allstate, as almost every company, made some lousy investments in the 90s and now everyone has to pay.
I use the same type of statements in surveys, always mindful of what the mean means. It's a little easier for me since I don't have an agenda to push. If the number of respondents to our job satisfaction survey doubles in the negative in terms of "Company Stable," that means that there is a trend toward the demise of companies. It doesn't mean that twice as many companies are folding or will be closing because our sample is too small and, perhaps even more important, because those responding are not directly involved in the finances of a company and may not know any more than what they were told, i.e., "We stand a 50/50 chance of making it."
So What? There are a couple of I want to get across to the reader. First, don't put much if any stock in averages. What you're really after is the median. In a discussion of salaries, the median is that point at which half the sample earns more and half earns less. The average could include Bill Gates and someone packing Windows Whatever. On average, the two of them make $2 billion a year. Or you could be taken in by the fact that it is not uncommon for very senior execs to have a lower salary than bonus, yet only the salary is thrown into the mix.
Second, make sure the sample is large enough to be meaningful. If nine out ten dentists prefer Crust, make sure that the same was greater than 10 dentists. What about all these drugs you see on TV? "A small percentage of users may suffer from diarrhea, nausea, cramps, migraines, chest pains, and maybe a little death." What does "small percentage" mean? What are they trying to hide. And why is there so much small print that there's no time to read it even if you could?
Third, watch out for graphs. It's like the stock that went up 300% to 38¢ a share. A large graph which extends from one cent to three dollars may look impressive with its line going straight up over a one-day period, but I'd rather see a graph from one cent to $100 over a 6-month period. Look for flat-lining.
Fourth, be mindful of the "group" or sample being used. If an article reads that "Dot-commers earned an average of $160,000 a year between 1995 and 1999," make certain that the "dot-commer" is defined. That could be anyone in a dot-com company. And as we've already seen, that $160,00 could include bonuses, salary, stock, and all the Pepsi they could drink. The same goes for any article talking about the "family." The family of 1900 is certainly different than the one in 2000.
Last (although there are many more), don't believe in national maps. The fact that there are only 22 people in Wyoming and 16 in Montana (all park rangers and Attorneys General) skews the impression that such maps give. Rhode Island has greater clout than either of those states.
I don't know why, but I just got done counting the number of times a percentage was quoted in the first two pages of last Friday's Wall Street Journal: 15. My favorite is the following: "Some 550,000 U.S. households owned rats or mice as pets in 2000, up 49% from 1990, according to the American Pet Products Manufacturers Association, which says that most of these pets are rats."
This "newsletter" reaches approximately 6,000 people. How many of you have been called by the APPMA and been asked whether or not you have a rat - a real rat - living with you? How do they know this stuff?
All Rights Reserved. Copyright 2002. E. A. Winning Associates, Inc.