U.S. President Barack Obama wants to trim the deficit by replacing the traditional Consumer Price Index with the "chained CPI." One appeal of this approach -- for the administration -- is that it simultaneously increases middle-class taxes and reduces future spending on programs such as Social Security and veterans benefits. According to the president's budget released today, the measure "will reduce deficits by at least $230 billion over the next 10 years," although my colleague Peter Orszag has noted that this might be overly optimistic.

Either way, adopting the chained CPI would be a mistake. It is a dishonest measure of the cost of living.

We have price indexes so that we can figure out how much money we would need if we wanted to buy a constant amount of real goods and services. Government benefits currently grow at the same rate as the price level because we don't want retirees and veterans to be penalized by inflation. Taxes are also indexed to prevent price increases from pushing people into higher brackets. The existing system isn't perfect: Retirees, for example, spend their money on different things than the average American represented in the CPI.

Switching to chained CPI, however, would make things marginally worse.

The problem is the way the new index treats what economists call the "substitution effect." Suppose that beef prices rise significantly while chicken prices increase by a smaller amount. An honest measure of inflation would count the increase in the price of beef based on how much beef people were eating prior to the price increase. Chained CPI, however, would underweight the increase in beef prices on the grounds that many people would react by switching to less expensive foods, such as chicken.

Chicken, however, isn't beef. And what happens when chicken becomes too expensive? According to the logic behind the chained CPI, cat food or even soylent green could be the next logical substitutes. That's not an honest way to measure the cost of living. The trouble began in 1996, when Michael Boskin and other economists argued that the U.S. Bureau of Labor Statistics was exaggerating the extent of inflation. According to them, the relative weights of the goods and services included in the CPI weren't adjusting frequently enough to account for Americans' changing tastes. Robert Gordon, who served on the Boskin Commission, wrote a helpful primer on the specific calculations involved.

In 1999, the BLS modified its main price indexes to "correct" some -- but not all -- of the so-called substitution bias identified by the Boskin Commission. The chained CPI was created in 2002 to "correct" completely the remaining bias, although it hasn’t been used by any government agency as an official price index. This makes a difference. Inflation looks slower when measured with the chained CPI than when measured with the regular CPI. The president claims that this would be an improvement:

"Beginning in 2015 the Budget would change the measure of inflation used by the Federal Government for most programs and for the Internal Revenue Code from the standard Consumer Price Index (CPI) to the alternative, more accurate chained CPI, which grows slightly more slowly. Unlike the standard CPI, the chained CPI fully accounts for a consumer’s ability to substitute between goods in response to changes in relative prices and also adjusts for small sample bias."

There is no single straightforward way to count changing tastes in a price index. For example, Americans now consume more wine and less beer than in the past, so it makes sense to adjust their relative weights in the price index. No honest observer would agree, however, that pink slime is a reasonable substitute for a grass-fed steak.

(Matthew C. Klein is a contributor to the Ticker. Follow him on Twitter.)