## Tuesday, June 10, 2003

STATISTICS

Statistics are essential for dealing with most political issues nowadays. What are the effects of a tax cut? What would ratifying the Kyoto Accord cost? What would the benefits be? Are we in a recession? How much do the 'poor' pay in taxes? The answer to all these questions lie in statistics.

That partisans spin statistics like a top, when they aren't lying about them or outright making them up, is well established. I'm not interested in rehashing any of that, though a few days ago I found a good example in Newsweek. What I'm interested in exploring now is the extent that even solid figures, used appropriately in the correct context and by a fair and honest person, can STILL be misleading and wrong.

There is a concept in mathematics, the name of which unfortunately escapes me right now, in which the answer to a particular problem can only be taken to the decimal place of the least exact term in the equation. For example, in multiplying 102*10.5*2.05*7.055794654661, your answer should not contain ANY numbers to the right of the decimal point, because one of the figures being multiplied didn't have any. (By the way, if the 102 was given as 102.00, you would be justified in answering in tenths (the degree of precision now being limited by the 10.5)) Giving an answer out to tenths, or hundredths, or thousandths would give the illusion that your result is that precise, but that would not truly be the case. This rule really didn't really mean anything to me as a student, it was just one more thing to learn to keep from tripping up on a test. It wasn't until much later that I realized the wisdom of it.

Back to statistics, I wish there was a way to find a similar easy way to clue people in on the precision of any particular statistic, but I have never been able to come up with anything remotely feasable. Unfortunately, the very rule used in mathematics above actually acts to give users of various statistics a false sense of reliability.

Do I have an example? Sure, tons. Let's use a biggie - GDP. How is GDP calculated? I would bet that most people don't have a clue, perhaps even a majority of those who use GDP figures to buttress their own arguments or tear down someone else's. Here is a decent primer for the various methods used to calculate it. The most common (it has the benefit of being the easiest to understand and use) is the expenditure model:

GDP=C+I+G+NT

Which means that GDP is made up of C (consumer spending) plus I (investment) plus G (government purchases) plus NT (net trade, export minus import).

Remember that the growth of this figure is constantly being reported, and is always given - annualized - to the tenth of a percent. What does this indicate to someone with some vague recollection of their Jr. High math class? First off, it's given out to tenths. Hey, it must be pretty solid. Then remember that they are extrapolating an annual number from monthly results, and realize that they really seem to be claiming accuracy into the hundredths.

But how accurate are GDP figures really? According to the Bureau of Economic Analysis, GDP increased (in chained 1996 dollars (ie. adjusted for inflation)) from 8508.9 in 1998 to 9439.9 in 2002. But let's think about what we were buying in 1998, and for how much. In 1998, you could buy a brand new Pentium II, for the bargain price of about \$1500. Contrast that with 2002, when I bought one of the then new Pentium IVs for about 1500. Today you can get Pentium IVs for quite a bit less than 1000. (I bet you could easily buy a computer today for less than \$500 that would run circles around a 'top of the line' 1998 Pentium II.)

Let's try a quick test now: How much would selling 10 million Pentium IIs at 1500 in 1998 have added to the calculated national GDP? And how much would selling 15 million Pentium IVs at 1000 this year add to our GDP? If you came up with 15 billion in each case, have a beer on me. But how can this be? How can increasing our computer production by 50%, and the quality of our product by several magnitudes possibly leave our Gross Domestic Product figures... unchanged? The answer is, even though it appears to be a solid, precise figure, GDP is actually a rather blunt instrument, and the longer the period you are looking at, the worse it does. The next time you see someone trying to compare GDP growth between decades, remember this page, it's implications on the GDP comparisions, and try not to laugh in their faces.

THIS phone, weighing 'just 28 ounces' with a battery life 'for 30 minutes of talk time and eight hours of standby', was selling for 3995.00 in the 1980s!

Another problem inherent is the inclusion of government spending. The examples of government overspending have been with us for my entire life, with the \$100 hammer, and the \$200 toilet seat cover being particularly ubiquitious. But even those examples don't come close to illustrating the waste that the G component includes. For an example of what I'm talking about, let's say the federal government builds a bridge. Let's further assume, to be more than fair, that it is built on schedule and on budget, and is built with the proper materials and within code specifications - a beautiful 25 million dollar bridge. This is what GDP is supposed to measure, right? Now for the other shoe to drop: it is named the Byrd Bridge, and built in a location in WV that not more than 50 people a day use. Will we really get 25 million dollars of utility out of that bridge? Of course not. I'm certainly not saying that there isn't waste in any of the other figures, but all governments are prodigiously wasteful, they always have been, and almost certainly always will be. The bigger the 'G' in the equation, the more the GDP figure overstates the 'true' GDP in my opinion.

Keep this in mind whenever you see comparisons between various country's GDPs. How much of each consists of individual decisions, and how much that of such and such a country's government.

And how much should military spending count towards GDP? I understand why it is counted in full now, but, especially when you consider how much of the US military spending is actually being devoted to protecting our allies, I wonder if this component could be better if treated a little differently.

I could go on, but it's getting late, and I think I've made my point.

Statistics are indispensible. It's impossible to make an informed decision about so many things without the use of various statistics. However, you always need to keep in mind the limitations, both real and potential, inherent in each one.