The consumer price index is used to measure the rate at which prices increase. It is calculated by taking a basked of common goods that people buy all the time (groceries, appliances, etc.) and looking at how their average prices change over time.
Suppose we are calculating the CPI using a basket of goods that contains only cinema tickets. Last year a cinema ticket cost \$15, but now it costs \$30. The CPI would then have doubled to reflect that fact that the prices in the basket of goods have doubled.
Some argue that this isn't a very good way to measure consumer price inflation. Here's why: if the price of cinema tickets doubles then people will buy fewer cinema tickets and buy, for example, more Netflix subscriptions instead. So although the price of cinema tickets doubled, people aren't spending twice as much on cinema tickets (because they aren't buying cinema tickets any more!) So, in some respects, the fact that the CPI has doubled doesn't tell us very much.
There are two ways to deal with this:
- the approach used when computing the CPI is to say that since people aren't buying cinema tickets we need to update the basket of goods we use to calculate inflation (remember the basket is supposed to be representative of peoples' consumption habits so if they buy fewer cinema tickets it should contain fewer cinema tickets). The problem here is that the basket is only updated infrequently so it takes time for the CPI to "catch up" with consumption habits.
the other approach (used in chained CPI) is to statistically adjust the standard CPI measure to account for the fact that we know people will substitute to different goods when the goods they were buying become more expensive.
To do this, the statistical authority gathers data on expenditure for each category in the basket of goods then then weights that category's contribution to the index by its share in expenditure. In the US, this is done on a monthly basis, so the chained CPI uses a more up to date measure of substitution between product categories.