Wednesday, December 17, 2008
STATISTICS VS LIES
FROM THE SOVEREIGN SOCIETY A-LETTER
EXERPT:
There are Lies, Damned Lies,
and then there are Statistics
Today's comment is a special lies report from A-Letter Editor Matt Collins, as a complement to Chairman John Pugsley's full lies report
Are you ready for the best thing you'll hear all day?
You can ignore the numbers. That's right, you heard me. The charts, the graphs the statistics, just forget about them. If they don't make sense to you, and even if they do, you should be skeptical about what they're saying and whether they're telling the truth.
But that old saying is popping into your head, "numbers don't lie." Well that statement's an even bigger lie than the numbers, and today you'll find out why. Besides there's another saying. And being that this one comes from Mark Twain, I'm more inclined to believe it.
On the topic of lying, Mark Twain paraphrased Benjamin Disraeli. He said "There are three kinds of lies: lies, damned lies, and statistics." No statement on the topic could ever be truer.
Busting the Myth of Statistics
Let's start with a simple example. We'll say that I'm doing a study, and in this study I've made a basic assumption that there's a relation between blue jeans and car crashes. Sounds strange but stick with me.
So I do a study across a wide range of people and places. I complement this study with a graph that shows two lines - for blue jeans and car crashes - and as the sales figures for blue jeans increases, so does the occurrence of car crashes.
But what if the two have nothing to do with each other? At what point does my graph answer that question? Well, it doesn't. It could be that one person is buying an enormous amount of blue jeans (more and more each day) and that the occurrence of car crashes is going up for other reasons entirely.
Now I know this is a strange example, but it proves at least one thing. It proves that our two variables - blue jeans and car crashes - may be totally unrelated. Something that the graph (the numbers) wouldn't even lead you to consider. Instead, the numbers, the graph, they make my weak argument - that blue jeans cause car crashes - seem more legitimate. After all, the numbers can't lie, right? Wrong.
A Frighteningly Real Example
Now we're going to move onto something a little bit scarier. And that's how our government calculates inflation.
What's scary here is that the inflation statistics that are circulated by the government, are acknowledged to be some of the most legitimate statistics out there, and they're used by major businesses to make market-shaping decisions.
But what we'll find here is a situation that makes the previous example look like child's play. Indeed, these numbers are subject to some of the most sophisticated tinkering in the financial world.
A process called ‘hedonic pricing' is used in determining statistics like inflation and the Consumer Price Index. Hedonic pricing is a way to adjust values for the progression of time. Sound confusing? Well it is, and hedonic pricing is something that can be directly manipulated to produce the desired statistics.
For example, I bought a TV five years ago. And it was cheap. And good. It was a wonderful TV and I still use it today. But this past year I decided to treat myself, and I bought a fancy new HDTV. This new TV cost me US$400 more than my old TV. But - thanks to hedonic pricing - it'll actually be treated as though it was cheaper than my old TV.
Why? Well, since the new TV has a better picture and better sound, I get greater enjoyment out of it (at least as far as hedonic pricing is concerned) and therefore it's of greater value. At least that's what hedonic pricing says. They can use hedonics and attach an arbitrary value to that added enjoyment - let's say they go with US$450.
Since I get US$450 more value from my new TV, it's actually written into the Consumer price index as being cheaper than the old TV. That's how they can use hedonics to make consumer products seem cheaper and inflation lower.
So they can use hedonic pricing to a very real extent in order to shape the results they get. Still think numbers can't lie?
What You See Depends on where you Stand
We'll go from that example to a much bigger problem with Statistics. And that's the fact that they're made by people.
Don't get me wrong, I think people are great. But I also realize the fact that everyone sees the world in his or her own personal way. And how they see the world depends largely on where they stand.
For example, Martin Hennecke, Senior Manager of private clients at Tyche, doubts that inflation is as low as the government reports. He cites the fact that if you used the 1983 formula to calculate inflation, then the result is currently between 10-11% (twice as much as the official numbers.)
But Mike Shedlock, a Global Economic Analyst from Sitka Pacific Capital, disagrees completely. He cites the fact that the government uses a formula known as Owners' Equivalent Rent (OER) as the housing component in the CPI and their calculations of inflation. We won't get into OER here, but let's just say that it's as easy to manipulate as hedonics.
Shedlock tries instead to calculate inflation using the Case Shiller Index, a statistic he sees as more honest than the OER. By his calculation, we've currently got a negative rate of inflation...that is, deflation.
So What do You Do?
First, you should relax.
Knowing that most statistics are meaningless should free you from worry, not introduce more. What you know now will make you less likely to be swayed by a weak argument sporting a handful of pretty pictures. You'll be more likely to ask questions about statistics, and not just accept them.
You should decide whether they make sense to you or not, and then just worry about why. Remember that every graph, chart and statistic is not an objective presentation of the facts. Rather, someone is always trying to prove a point or support an argument. Don't ever be lured in by weak arguments that use complex math to prove a point that fundamentally doesn't make sense to you.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment