With all the debate about the recent Boskin commission recommendations, I thought I'd take a moment to summarize for the layman the economic issues surrounding the construction of the CPI and its biases. This is not meant to be an exhaustive summary, but rather an informative introduction to the subject. Construction of the CPI The CPI is constructed in order to measure changes in prices and living standards from year to year. It is a Laspeyres, or base weighted, or forward looking price index. The BLS assembles a "market basket" of goods purchased by a typical consumer in a year (so many eggs, so much milk, gas, housing, entertainment, etc.) and adds up the total cost of this basket. It then moves to the next year and adds up the cost of *that same basket* at the new prices. Dividing the second number by the first and multiplying by 100 (to get rid of that cumbersome decimal point) gives us the CPI, which purports to be a measure of the increase in the cost of living. If the CPI is 125, you would need 25% more income this year to buy the same bundle of goods that you bought last year. Biases Holding the bundle of goods constant from year to year is necessary in order to make meaningful comparisons, but is not without its problems. This is where most biases come from. The four most important sources of bias are substitution bias, quality improvement bias, new product bias and sale bias. Substitution Bias Substitution bias comes from the fact that relative prices change from year to year; a CPI of 125 does not mean that all prices have increased by 25%, but rather some have increased less, some have increased more, and the average, weighted by the quantity of the good in the market basket, is a 25% increase. It's perhaps the toughest to understand, but it probably also has the largest effect, so it's important to understand it. I'll start with a hypothetical example and work from there. Imagine that you're indifferent between Coke and Pepsi --- this is not to say that you can't tell the difference between them, only that you have no preference for either over the other. The Coke machine in your building charges $.50 per can, while the Pepsi machine (for whatever reason) charges $.60, so you buy Coke. You drank 200 cans of Coke last year, so those 200 cans are in your CPI "basket". Now imagine that the Coke machine, for whatever reason, changes its price to $1, while the price of Pepsi remains constant. The BLS will look at your "basket" of 200 cans of Coke and calculate that the price of those 200 cans has gone up from $100 to $200. According to the CPI, the inflation rate for soda is 100%, and your income must be raised by $100 in order to maintain your real standard of living. The way the CPI is calculated does not account for the possibility that if the price of Coke increases to $1, you'll switch your consumption to Pepsi. A true price index would in this case measure inflation at 20%, and you would need an income increase of only $20 to maintain your real standard of living. Now, this effect only happens to the extent that goods are substitutable for one another. The above case is one in which the goods are perfect substitutes, so the substitution effect is quite large. If goods are perfect complements (consumed in proportion to each other), we would not expect any substitution if the price of one good changes. The recent comment by the director of the BLS ridiculing the Boskin commission's findings, "If the price of washing machines goes up, you won't consume more dryers," is just such an example of perfect complements. However, this comment is somewhat disingenuous, since all goods are not perfect complements with all other goods, and there will be, generally speaking, some scope for substitution when relative prices change. Much hay has been made on sci.econ about substitution into "lower quality products" like substituting into chicken when the price of beef goes up or substituting into margarine when the price of butter goes up. The fact is, this doesn't really matter. If the price of butter goes up and you switch to margarine, you are better off than if you had stayed with butter at the new, higher butter price, simply because you could have stayed with butter but chose to switch to margarine instead. You must be better off with margarine, or you wouldn't have switched. The BLS assumes you'll stay with butter for purposes of CPI calculation, and thus ignores the possibility that you'll improve your position by substituting. Mind you, this is not to say that you are no worse off than you were before the price of butter changed, only that the CPI will tend to systematically overstate the degree to which you're worse off. Quality Improvement Bias Some goods improve in quality from year to year. You can buy a new Lexus this year for the same price as last year's model, but this year's model comes with more standard features. This year's crop of computers is faster and more powerful than last year's. When constructing its "basket", the BLS considers goods like cars and computers, but does not consider the possibility that you're getting more for your money. For example, you could buy a midline computer 2 years ago for about $2000. You can buy a midline computer today for about $2000, so the CPI would show no change in living standards. Except that 2 years ago, that $2000 computer was a 486DX2, and now a $2000 computer is a 200 Mhz Pentium. The faster computer is better and increases your enjoyment (and your real standard of living). New Product Bias New products come into being all the time. Since the CPI looks at the price of last year's bundle of goods at this year's prices, it can't account for the possibility of people increasing their standard of living by purchasing goods that weren't available last year. Sale Bias The employees of the BLS who collect the price data used for CPI calculation tend to go to the same stores at regular intervals. In doing so, the BLS ignores people's tendencies to adjust their buying habits in order to take advantage of sales, bulk discounts, rebates and the like. For example, you can buy audio CD's at many electronics stores for around $12 today, but the BLS is probably still going to record stores where the same CD is $15. In the past 20 years there has also been an explosion of discount stores in the USA (Wal-Mart, Home Depot, buyer's club stores, etc.) whose lower prices are systematically undercounted in CPI calculation. While it may be true that you get a lower level of service at Home Depot than you would at a neighborhood hardware store, the fact that many people shop at Home Depot indicates that they prefer it to the higher prices and better service of a hardware store. I hope this helped outline the issues surrounding CPI calculation. Don