Chain, Chain, Chain

September 30th, 2011 at 8:26 pm

A number of commenters have expressed interest (or worse) in my views on moving to a “chained” version of the consumer price index (CPI) for price adjustments in various programs, notably including Social Security and the tax brackets.

As briefly as I can, let me give the what’s and why’s of this.  I understand that this change cuts benefits for Social Security recipients, so I don’t go there lightly.  Retirement insecurity is a huge concern, and I join others in suggesting an idea below to help mitigate that effect.  But I endorse the switch because it’s a more accurate price measure, and, applied to the tax brackets, it raises significant revenue as well.

In order to capture the overall cost of living for a typical household, these price indices track the change in the average price of a “market basket” of goods purchased by a typical family, including food, clothing, shelter, fuels, and other goods and services.

What is it?: Consumer price indexes are built by putting together a “market basket” things consumers purchase.  The price of the market basket then depends both on the price of the items in the basket and their shares.  For example, according to the CPI, shelter costs make up about 40% of household spending, food about 15%, transportation 17%, and so on.

Over time, both prices and these relative shares of the goods in the market basket will change.  The difference between the CPI and the chained CPI is in the way these changes are measured.  The CPI only captures the changes in prices, not the changes in shares.  The chained index captures both.

For example, if the price of meat goes up, families might adjust by buying less meat and more pasta.  Or, it could go the other way.  If the price of steak falls relative to chicken, shoppers might buy more steak and less chicken.  The CPI misses these substitutions, but the chained index captures it, which is why it grows more slowly: it captures consumers’ shifts toward items that have become relatively less expensive.

Whereas the CPI uses only the consumption shares from period one in measuring the price changes in the market basket in period two, the chain-weighted index “chains” together (averages) the shares from period one and period two, thus allowing for the substitution engendered by relative price changes.  Since it captures these actual changes in consumer behavior, the chained index is considered a more accurate measure of changes in the cost of living.

What is its impact?: The BLS has been calculating a chained CPI since 2000, and the figure below plots it against the CPI—not that interesting to look at, I guess, but the fact that the chained version grows 0.3 percentage points per year less quickly than the current index is a big deal over time.  That old compounding thing, again.

Source: BLS

The figure below takes the annual Social Security benefit for a middle-income retiree in 2011 whose initial annual benefit is around $17,000.  It then plots the nominal values of that recipient’s benefits out to their 95th year.  At first, the difference is about $50, but should they live so long, by the end of the graph, the difference is in the thousands.

Source: Memo to Xavier Becerra from the Social Security Actuary, June 21, 2011 (hat tip, KR); Actuaries report, intermediate assumptions re inflation.

That’s why numerous budget types who have recommended this switch, like Bowles-Simpson and Rivlin-Domenici have suggested an upward adjustment to the Social Security benefits of the old elderly under this plan, an idea I’d agree is essential.  In fact, I wouldn’t make the switch without it.

The other important impact here is on the tax brackets.  As your nominal income grows you pass into higher tax brackets.  EG, in the Federal income tax schedule, a married couple going from $69K to $70K passes from the 15% to the 25% bracket.

Now, these brackets are adjusted by inflation, so if inflation grows by 3%, a bracket that starts at $100,000 (just keeping the math simple here) becomes $103,000.  But if we switch to a slower growing price index that has inflation growing by 2%, that bracket starts at $102,000.

Suppose your income grew that year from $100,000 to $102,500.  Under the switch to slower inflation, you pass into the higher bracket, and visa versa.

According to CBO, this impact of switching to the chained CPI raises almost $90 billion over 10 years.  The Social Security switch raises about $110 billion, but that doesn’t include the elderly bump up, which would have to be applied against those savings.

So, there you have it.  Chain weighting the price index improves its accuracy relative to the current price measure, which means benefits adjusted by a price index will grow less quickly.  Some of these benefits, like Soc Sec to the old elderly, would need to be adjusted to protect them from cuts they can’t afford.  It also raises significant revenues through the tax code.

One more thing: if anyone tries to apply to this change only to the benefits and not to the tax brackets, tap the secret OTE code into your calculator and I’ll be there in moments to put a stop to that nonsense.



Print Friendly, PDF & Email

13 comments in reply to "Chain, Chain, Chain"

  1. foosion says:

    Does regular CPI or chained CPI do a better job of tracking cost of living changes for the elderly?

    It may be the case that chained CPI is more accurate for the general population, but that regular is more accurate for the elderly.

    • davesnyd says:

      I think foosion is asking the right question. And I suspect the answer is more complicated and varies in a way that isn’t easy to capture.

      Here’s the deal: we’ve defined CPI as our indicator of prices. If chained CPI is more accurate, then we ought to revisit the definition of CPI.

      Furthermore, how do we account for the fact that some seniors own a home free and clear and have no inflation from housing? While some rent and are extraordinarily affected by housing cost changes?

      Are changes in the price of gas less relevant for seniors who have downsized from two cars to one and are walking everywhere (or riding electric golf carts in their retirement communities)? or are the more relevant because of seniors who have sold their home, bought a winnebago, and are now driving cross country?

      How many seniors buy new clothes at a rate approaching that of people in the active work force? How should that affect whatever CPI numbers are used? On the flip side, aren’t they much more affected by price increases in health care (at least, if they aren’t paying for “medigap” insurance)?

      We need to make sure that the COLAs in social security are accurate– but simply choosing a more slowly increasing CPI substitute may or may not be the right way to do that.

      I’m not saying don’t switch– I’m saying make sure there is a good reason for whatever you/we switch to.

      • Procopius Furioso says:

        davesnyd says more eloquently what I was intending to say. Because of Mr. Bernstein’s earlier post I discovered the BLS already has an index designed to cover the elderly. I can see how chaining makes the regular index more accurate (I’m most concerned about the CPI(W) because that’s the one used for Social Security COLA). If that experimental CPI(E) isn’t already chained I’m sure they could tweak that relatively easily, although that might raise problems with comparability with earlier results. Meanwhile, I don’t think any of this discussion matters because Peter G. Peterson, the Koch Brothers, the Republicans, and many “centrist” (read “extreme right-wing”) Democrats have already decided their goal is to reduce/destroy Social Security, and they’re going to use every tool possible to do that. And I’m sorry to say I think that President Obama wholeheartedly believes they are right.

  2. Cat says:

    “The CPI misses these substitutions, but the chained index captures it, which is why it grows more slowly: it captures consumers’ shifts toward items that have become relatively less expensive.”

    The CPI measures the change in the cost of the standard of living of the ‘middle class’ while the chained CPI measures the change in the standard of living of the “middle class” as they slide into the lower class.

    The chained CPI is one of those BS economic stats that measures the state of the economy as if everyone was just some cog in a machine rather then as people who deserve to be able to buy the same goods they could buy last year. They shouldn’t have to be told by people, who dont have to worry about the size of their SS checks, that they aren’t getting a COLA adjustment because they can just substitute in an inferior cheaper good.

    • pjr says:

      That was along the lines of my reaction, Cat, to the explanation offered here. When all of us can afford to eat only gruel, and travel by bicycle, and a lawn sprinkler instead of a weekend at the beach, the chained-CPI will show little inflation. Can’t wait. (As I say about my teenager all-too-often, what in the world is he thinking?!?!?)

    • Carol says:

      You hit the nail right smack on the head, Cat. And no pretty words from Washington can hide what they’re doing.

  3. urban legend says:

    And so when the recipient switches to cat food instead of real food, the chained CPI will incorporate this shift into its cost of living calculator. In other words, it is simply wrong to say chained CPI is “more accurate” because that is a loaded concept. Indeed, it seems inherently corrupt because it encourages and reinforces a constantly declining standard of living.

    Social Security from the very beginning has been designed to maintain benefits in relation to changes in standard of living. This would be a departure from that principle. Time to re-examine your assumptions here, Jared. Methinks you have been had by an insidious disease, the symptoms of which are an occasional reflexive need to appear “reasonable” (and “Serious”) by meeting the other side half way.

    The one and only correct answer to the Social Security retirement fund is that we don’t have a clue what will happen in 30 years — as evidenced by the fact that 14 years after 1997 the Doomsday Date when the Trust Fund surplus, that date is only seven years closer to the present than it is now — and will not have a clue until we have gone through an economic expansion again. Accordingly, we should do nothing whatsoever until we see what that expansion does to the projected exhaustion of the Trust Fund surplus. If we ever have a genuinely clear picture that there is likely to be a shortfall — which will mean that revenues going forward will not be sufficient to operate on a pay-as-you-go basis without changes, any changes should be strictly on the revenue side (i.e., higher payroll taxes). This country cannot afford any hits whatsoever, including reduced Social Security benefits, to present or future middle class incomes.

  4. Ken M says:

    There’s a big problem with using chained CPI for social security, as others have alluded to. As I’m sure you know, the Bureau of Labor Statistics has an experimental elderly-specific CPI, the CPI-E, which takes account of the typical elderly basket of goods — for example, a greater portion of income is spent on health care, which rises a lot faster than other costs. Anyhow, the CPI-E rises *faster* than the CPI currently used, which means that the elderly are already seeing their effective social security income decreasing year over year due to inflation. To go to the chained CPI, which grows even more slowly, is going to make this problem even worse — it will make the elderly slip further behind. There may be good reasons overall to replace CPI with chained CPI for many purposes, but for social security it’s taking a bad situation and making it worse. Why not advocate switching to the CPI-E for social security? Wouldn’t it be the right thing to do? And deal with the financial issues by extending payroll taxes to all sources and levels of income.

  5. Lars Olsson says:

    Why not CPI-E? That’s specifically targeted at the elderly (hence the “E”), as you know – and it more-accurately reflects the fact that seniors spend a greater percentage of their income on things like medical care/treatment/prescriptions, etc.

    I’ve never understood the preference for chained CPI over CPI-E.

    • Jared Bernstein says:

      The chaining is the important part–accounting for the substitution issue I note in the posts. No reason we couldn’t applying the chaining technique to the cpi-E–that would be preferable, I think.

    • Carol says:

      The chained cpi cuts benefits. The cpi-e gives too many benefits.

      • Procopius Furioso says:

        Ah, “The cpi-e gives too many benefits.” You do understand that most people living on Social Security get less than $20,000 a year, don’t you? And you’re offended because people are living large on that? And you don’t want them to live any larger? Even though it wouldn’t cost you any more than you already pay? I suspect you must have a large enough income to afford some sort of savings program, maybe a 401(k). Most of us don’t. But I’m sure Pete Peterson will be gratified to see you agree with him.

  6. Jim Z. says:

    Jared, OF COURSE Congress and the President will use the chained CPI for SS benefits and not tax brackets. Anyone who thinks otherwise lives in fantasyland. The pain caucus rules!