Eric, thank you for your interesting example (below). In the case you mention people's real income has clearly gone up, unlike that which has confronted most of us lately. How do we know this? The quantities, 10 of each good, which were purchased in Year 1 could have been purchased for $300*10 + $150*10 = $4500 in year 2, just the 125% increase implied by an increase of 225% in the fixed weight index. In fact people had $5100 to spend, and hence they were better off in Year 2. In fact, people could afford 11.33 units of each good in Year 2. In fact they purchased 14 units of good A and 6 units of good B in Year 2. Presumably they are better off with their actual purchases than by continuing with their previous pattern of purchasing equal quantities of each. The magnitude of the substitution effect is the amount by which they are better off with the new commodity bundle rather than the old. Since the old, the proportionate increase, is always available, the substitution effect can never be negative. Dave Richardson ---------- From: pen-l Subject: [PEN-L:3957] RE: the CPI Date: Thursday, April 25, 1996 11:48AM Forgive possible double posting, but my message from yesterday never appeared on pen-l. In any case, I've revised it to make it more clear. D Richardson wrote: > . . . the substitution bias is always positive. The answer is from > the micro theory textbook: as relative prices change people > substitute toward the now less expensive goods and away from > the more expensive . . . I disagree. The sign of the substitution bias is an empirical matter and not a "theoretical" matter. The sign of the substitution effect can be negative or positive. Example: Assume two goods (A and B) experience price increases between year 1 and 2. Assume good B increases in price faster than good A. Therefore, people shift to buying more of good A in year 2 compared to year 1. year 1 | year 2 good P rel P Q z P relP Q A 150 3 10 z 300 2 14 B 50 1 10 z 150 1 6 CPI: year 1 150x10+50x10 = 2000 year 2 300x10+150x10 = 4500 =225% increase using fixed weights (from year 1) But, expenditures in the two years are year 1 150x10+50x10 = 2000 year 2 300x14+150x6 = 5100 = 255% increase using actual buying patterns That is, in this case the CPI UNDERSTATES the rate of inflation experienced by consumers: CPI indicates 225% inflation rate while expenditures by consumers go up by 255%. Why? In this case the good that has its relative price falling between years 1 and 2 is the HIGHER priced good. A shift to a good whose relative prices has fallen need not be a shift to a lower priced good. The substitution bias only has a systematic downward impact on the CPI if goods that are LOW priced in the base year tend have their relative price FALL in later years. But there is no theoretical reason for this to occur. In fact, there are likely good reasons to suppose that goods that have their relative prices fall were HIGH priced goods to begin with. Conclusion: "theory" does not imply that the CPI necessarily overstates "true" inflation rates (based on the substitution effect). Eric .. Eric Nilsson Department of Economics California State University San Bernardino, CA 92407 [EMAIL PROTECTED]