Jump to content
Sign in to follow this  
Adrian

How to Lie With Statistics

Recommended Posts

I wouldn't have bothered, but 67% of BGO posters wanted me to recover this thread.

 

 

How to Lie With Statistics

 

Darrell Huff's classic book on statistics was first published more than fifty years ago. In it he details how advertisers, politicians, newspapers, market researchers and the like manipulate numbers to give the answer they want. We all remember lines like:

 

- Top breeders recommend it

- Eight out of ten cats prefer Whiskas

- More dentists use Colgate

 

and this book explains what most of us knew already: how they are all lies.

 

One example is called by Huff "The One Dimensional Picture." Before and After pictures are used to show the increase in something. In this picture different sized money bags are used to illustrate that American workers earn twice as much as their Russian counterparts:

 

statPic.gif

 

This is done by doubling the height of the money bag. That's fair isn't it? Of course not. The width is doubled as well, showing a four-fold increase in area and thus a four-fold increase in the American wage. And bags are of course three dimensional, so the American's bag can hold eight times as many coins as the poor Russian's. All from a simple doubling.

 

But that was fifty years ago. Surely nobody would try such a ruse these days, would they? Check out the Greenmetropolis trees.

 

There are many more such concepts in this slim, easily understood book. Another one still used today is the Repeated Small Sample (ask ten people to try your shampoo against a "leading rival" and if you don't get the result you want, ask another ten, and another ten until you do).

 

It's not in the book but my favourite was a politician saying that crack cocaine usage had risen 167% from the previous year. I view any percentage higher than 100 with distrust bordering on contempt, and rightly so in this case. Crack cocaine usage had risen from 0.03% of the population to 0.08%, but she'd managed to turn that 0.05% increase into 167%. Quite impressive in a way, and arguably true in that 0.08 is one and two-thirds of 0.03. She got her headline, but the more people who see through such "lies" the better.

 

I consider How to Lie... a classic book, and if the President Roosevelt and Mickey Mantle references are dated, the subject matter certainly isn't. I'd like to see taught in schools as a way to use statistics in real life.

 

Hazel 25th July 2006 08:54 AM

That sounds like a really interesting book.

 

It drives me crazy when you see cosmetic/toiletry adverts on TV which say "84% of women who tried this saw an improvement" then at the bottom of the screen it says '243 women tested'. First of all, 243 is not a representative chunk of the the female audience, and to get such an odd number they clearly test the minimum number of women to get a high percentage (the target they set out to obtain). When one of these adverts says "33% of women...1 million tested" then I might be impressed with the product.

 

Adrian 25th July 2006 10:01 AM

Exactly Hazel. It's called a "statisically representative sample."I think the example he uses is a barrel full of Smarties and trying to find out how many of each colour there are. Reach in and grab a handful? In this case each colour was a layer in the barrel and so you'd get a wrong answer, but much worse never know it was wrong.

 

It all sounds obvious but I bet there are millions who fall for it. In fact I know there, I asked ten of them :rolleyes:

 

megustaleer 25th July 2006 11:46 AM

I view any percentage higher than 100 with distrust bordering on contempt, and rightly so in this case.

Aha! One of my pet peeves!! :grumble:

 

Hazel 25th July 2006 12:10 PM

They should point out that it is 88% of a specifically targeted group of 243 women, who were given a £5 Marks & Spencer voucher to agree.

Grammath 25th July 2006 01:13 PM

It should be common knowledge that 42.7% of all statistics are made up on the spot.

 

Mark Twain summed it up best when he said that "there are lies, damned lies, and statistics."

 

Cathy 28th July 2006 09:36 PM

Sounds worth a read. There was a programme on ITV a while ago about how the government make it sound like they're putting more extra money into something by including money already pledged or interest rates and stuff like that. Haven't trusted statistics since then!

 

Mungus 29th July 2006 10:34 AM

Statistics are a tricksy way of generating sales. It's like medical statistics that can use 'number needed to treat', 'risk per 100,000', 'mortality after 5 years' and so on depending on which presents the required picture. No-one seems to care that the case of mumps has increased from 94 in 1996 to 43,322 in 2005 but are fixated on this inaccurate belief that the MMR jab can cause autism just because a dodgy study by a dodgy doctor got so much press. Not that I have a bee in my bonnet about it.

Momo 30th July 2006 09:45 PM

This book sounds really interesting. I have put it on my wishlist!

Share this post


Link to post
Share on other sites

I'd also be interested in this and yes, like everyone else I fume at the way statistics are mishandled.

 

I could rattle on for hours (78.4% of which would have you reaching for the hemlock) about many areas of this, but my own pet area is education. League tables 'empowered' parents with statistics, yet they've been tinkered with ever since as each deliberate weakness in their compilation has been exposed. So, originally they took no account at all of the social circumstances of the school or of whether there had been a significant raising of pupils' performance from an already low base. More recently the huge increase in GCSE results per school had to be reassessed because it was finally accepted that this didn't take full enough account of the core subjects like English and maths. And so on, and so on.

 

Plus the old scientific truth that once you start to measure something you change it. Many schools now try to encourage pupils into subjects they are more likely to earn good grades in, rather than subjects that are necessarily in their best interests. Also, a large number of teachers increasingly teach not the subject, as such, but how to pass an exam in it, which is not the same thing at all. Who can blame them? There are huge pressures at every level of a school to perform well in the tables and thereby produce the best statistics. This as opposed to providing pupils with the best possible education. I also know of teachers who've turned a blind eye to coursework cheating because they feel under so much pressure to produce the best results for the tables.

 

The same is true in hospitals, where decisions are increasingly made not in the interests of patients but in terms of the Government's targets. So Blair can reel out all sorts of statistics that are not only misleading but actually mask the damage the rush for those statistics has created.

 

I'm off now to check the statistics for strokes suffered by irate ranters...

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×