Leaks from the fiscal cliff negotiations have been focusing a lot of attention on the proposal to cut future Social Security benefits, relative to their projected nominal growth rate under current policy, by basing their annual increase on "chained CPI," rather than the currently used version of the consumer price index. Chained CPI grows more slowly because it assumes that consumers respond to changing relative price levels by shifting their consumption from more expensive goods to cheaper ones. For example, if the price of beef rises relative to that for chicken, people respond by buying less beef and more chicken, so their expenditure levels don't rise as much as if they kept the proportions constant as prices rose differentially. Chained CPI takes this into account, in such a way as to produce lower annual rises in the CPI measure.
Though this might require a longer discussion than I have time or space for here, the underlying question has a lot to do with assumptions about people's utility that we cannot easily test. For example, the effect on my utility of the above relative price shift between beef and chicken depends on how I value each of them. In effect, what is my consumer surplus from each, per unit of consumption, at different price levels.
While this may sound abstruse, there are good reasons why we might wish we knew the answer. Suppose I am going to buy a fixed real-life annuity, but I can opt for using either regular CPI or chained CPI. Obviously, in an arm's length transaction with the annuity company in which I have $X to spend, the advantage to me of picking chained CPI is that I would get to start with a higher annual benefit.
Social Security in effect hands retirees a fixed real life annuity. The rationale for this is that people might otherwise under-save for retirement (or late retirement), so we give them a minimum "forced saving" level that is the same for all periods, reflecting that consumption often is separate between periods but has declining marginal utility in each period. (E.g., better one dinner tonight and one in a year, than two dinners on one of the two nights and zero on the other.)
So there is a constant-benefit-level comparison to be made, as between CPI and chained CPI real life annuities that have the same present value because the latter starts higher.
In the case of Social Security, then, there are two distinct questions to ask. The first is whether benefits should be cut relative to current law, while the second is whether a chained CPI methodology does better than the existing one in keeping the marginal value of the last dollar of consumption that you can fund through your Social Security benefits the same as between periods.
Hence, the question of whether the nominal growth rate of Social Security benefits should be pegged to chained CPI rather than regular CPI is distinct from that of whether benefits should be cut. For example, suppose chained CPI is the better measure but that benefit levels ought to be raised, all things considered. Then the correct response would be to switch to chained CPI and over-correct for the year one benefit increase that would make this budgetarily neutral. (As I think about this, the actual optimal policy problem is more complicated still - for example, one may want to distinguish between age cohorts in measuring what would be a present value-neutral change.)
An underlying problem, of course, is that, since the underlying issue pertains to utility, there really is no extremely robust right answer even to the pure analytical question of how CPI ought to be computed. For example, my CPI may differ from yours, since my market basket is different (even at the same income level) reflecting differences between our utility functions.
If you are trying to figure out how inflation distinctively affects seniors, from the point of view of optimal CPI design (distinguishing this from the question of how much their benefits should rise and fall), it is pretty obvious that healthcare is a large piece of the whole - larger than it generally is for people in younger age cohorts. With healthcare prices generally rising faster than the overall inflation rate, even (for seniors) counting out-of-pocket costs that Medicare and Medicaid don't cover, chained CPI presumably loses credibility as a "neutral" approach. This tentatively suggests to me that, even if we want to impose Social Security benefit cuts, this may not be a great way to do it.