The reason for the discrepancy? Blatant statistical manipulation. We live in a period of inflation. Bill Fleckenstein explains. He takes on the unemployment statistics as well. Fleck's column is based on this interview with economist John Williams from a Wall Street firm Weeden's web site:
For those not familiar with "substitution," [Williams] explains the practice's evolution in the CPI calculations. The concept of substitution was a concoction of Alan Greenspan and Michael Boskin, who basically argued that if one item were too expensive, consumers would substitute that with a cheaper one. Williams' response: "The problem is that if you allow substitutions, you aren't measuring a constant standard of living. You're measuring the cost of survival. You can keep substituting down and have people buy dog food instead of hamburger. It happens. But that's not the original concept behind the CPI."
Williams says that the government's motive in all of this, if there is a motive (of the government collectively; don't picture a group of men cooking up something in a back room), is its desire to put a favorable spin on all the data.
Another motive? Transfer payments like Social Security are indexed to the CPI, and they would be far higher if the CPI were accurate. In fact, says Williams, if the "same CPI were used today as was used when Jimmy Carter was president, Social Security checks would be 70% higher." That's seven-zero.
Though Williams doesn't get much into hedonics, he does talk about the inflation-understating impact of geometric weighting versus arithmetic weighting in the CPI statistics: "Geometric weighting ... has the 'benefit' that if something goes up in price, it automatically gets a lower weight, and if it goes down in price, it automatically gets a higher weight."
Don't buy bonds. With a 4.7% yield before taxes, you're barely keeping up with the government's phony inflation statistics, and you're actually losing money using more reasonable inflation measures.