Tag: Social Security

Originally published at Forbes.com on June 23, 2020.

 

Elizabeth Warren supported it.

So did Bernie Sanders, and three other now-departed 2020 Democratic presidential candidates.

And so does Joe Biden, though you won’t find it on his website.

It’s a long-running proposal to make one particular change in Social Security: the implementation of the CPI-E, an alternate measure for the calculation of Social Security’s annual cost of living increase, a change which would increase the level of increase in benefits from year to year. And the experts at the Penn Wharton Budget Model, in their analysis of Biden’s plan for Social Security, report that he’s on record as supporting this change as well. (The folks at PWBH verified via e-mail correspondence that the Biden campaign had confirmed this detail with them, despite its absence from the Biden website.)

This question is newly relevant in light of a GAO study published last week addressing the question of whether the current approach to adjusting benefits is the right one. The study’s conclusions were, it turns out, not particularly conclusive — recommending further study on the issue — but the popularity of this change, touted as a win for the elderly because it will increase their benefits at a faster pace from year to year, means it’s past time for me to address it, and, with it, explain the basics of CPI calculations in the first place.

The basics

Here are the basics of how the CPI is calculated, in general, based (in part) on a detailed description at the Bureau of Labor Statistics website*:

“In the CPI, the urban portion of the United States is divided into 38 geographic areas called index areas, and the set of all goods and services purchased by consumers is divided into 211 categories called item strata. This results in 8,018 (38 x 211) item-area combinations.

“The CPI is calculated in two stages. The first stage is the calculation of basic indexes, which show the average price change of the items within each of the 8,018 CPI item-area combinations. For example, the electricity index for the Boston CPI area is a basic index. . . . At the second stage, aggregate indexes are produced by averaging across subsets of the 8,018 CPI item-area combinations.” This averaging is based on usage patterns based on consumer surveys, the “market basket of goods,” but the weights are kept constant from one month to the next, for purposes of calculating the index.

For example, in December 2014, the standard CPI measure used a weight of 16.011% for food and beverages, of which coffee had a weight of 0.15%; household furnishings had a weight of 3.455%, of which dishes and flatware had a weight of 0.03%.

Note that this is not a comprehensive calculation, and the math is not always what you’d expect. In particular:

The government only uses a sampling of items within a given category. For example, “a particular type of cheese item will be chosen, with its likelihood of being selected roughly proportional to its popularity. If, for example, cheddar cheese in 8 oz. packages makes up 70 percent of the sales of cheese, and the same cheese in 6 oz. packages accounts for 10 percent of all cheese sales, and the same cheese in 12 oz. packages accounts for 20 percent of all cheese sales, then the 8 oz. package will be 7 times as likely to be chosen as the 6 oz. package. After probabilities are assigned, one type, brand, and container size of cheese is chosen by an objective selection process based on the theory of random sampling. The particular kind of cheese that is selected will continue to be priced each month in the same outlet [store].” At another store, another type of cheese might be chosen, so that “the cheese sample (or the new vehicle sample, the television sample, etc.) contains a wide variety of styles and brands of cheese, vehicles, televisions, etc.” But this sampling is not guaranteed to produce the same types of changes as the “true” inflation rate for all goods and services in the entire country, simply by means of sampling error/random chance, and to the extent that this occurs, the CPI calculation will be off.

The index is limited in its ability to address improvements or declines in quality; it’s not that the experts ignore this aspect of price changes, but it’s difficult to do with precision. For example, automobiles have, over the years, improved in terms of the safety features, fuel efficiency, technology improvements to integrate cell phones, and the like. “Adjustments for quality change in the CPI new car index include structural and engineering changes that affect safety, reliability, performance, durability, fuel economy, carrying capacity, maneuverability, comfort , and convenience. Since 1999, quality adjustments have not been made for changes associated with pollution control mandates. . . . The adjustments exclude changes in style or appearance, such as chrome trim, unless these features have been offered as options and purchased by customers. Also, new technology sometimes results in better quality at the same or reduced cost. Usually, no satisfactory value can be developed for such a change. In such cases, the quality change is ignored, and prices are compared directly.”

Changes in housing-related costs are based on the change in rents, either actual (for rental units) or hypothetical, for owner-occupied homes, regardless of whether a homeowner actually experiences those cost increases. This is really hard to wrap your head around, but here’s a thought experiment to make sense of this: some homeowners, in the aftermath of the housing market crash in 2008, were unable to sell their homes when they moved away, and were left renting the house they had formerly lived in, and owning their new home. But imagine that the reverse were true — that they rented the home they occupied and at the same time, owned a home for investment purposes, for rental income and its appreciation value. (On a personal note, when I lived in Germany for a time, I learned this was not uncommon, because both renting, and small-time landlording, were both more common than here.) Is there any difference, in terms of practical outcome as far as cost increases are concerned, between the two scenarios?

Medical care CPI calculations are particularly complex. Only out-of-pocket spending and consumer-paid health insurance premiums are included in the index (that is, government and employer-paid costs are not included, because these are not costs borne by individuals). Quality improvements in medical care are also not reflected in the index, due to the impenetrability of data. Fees reflect negotiated rates for Medicare or private insurance but not any discounts offered uninsured patients after-the-fact.

The variations

There are also two basic CPI measures:

CPI-U is the CPI for All Urban Consumers, which includes 93% of the U.S. population, and CPI-W is based on the purchasing patterns of Wage Earners and Clerical Workers, in all, about 29% of the U.S. population. This latter index is what’s used for Social Security adjustments, but is not appreciably different from the CPI-U.

In addition, there are two “experimental” indices, the CPI-E, and the C-CPI, or chained CPI.

The CPI-E uses alternate weights for households headed by someone 62 or older; for example, medical care has a weight of 12% rather than 9% for the CPI-U. A 2010 comparison of general categories of expenditure shows that levels are fairly similar at the bird’s eye view, varying by as much as two percentage points, rarely more. 14.2% of the over-65’s expenditures were on transportation compared to 16.0% of the total population, for example, but 7.0% of the total was allocated to utilities rather than 5.5%. While the older group spent 13.2% vs. 6.6% on health care, they spent half as much, 5.1% vs. 11.2% on insurance and pensions.

Using the CPI-E rather than the CPI-W to increase Social Security benefits would produce a higher increase from year to year; the Senior Citizens’ League calculates that the average difference from 1983 to 2019 was 0.25% per year, which is small in any given year but compounds over time.

But there are two reasons why the existing method of calculating the index produces questionable accuracy. First, the CPI-E calculates a new set of weights but does not collect new data, with respect to shopping patterns within any individual category (do they purchase more or less organic food, for example, and are they more or less likely to shop at Whole Foods or Aldi?); and, second, the sample size that remains when only the over-62s are used, from the larger consumer survey, may be too small to be accurate, in terms of general statistical norms for accuracy of calculations.

While changing to the CPI-E would increase Social Security inflation adjustments, the chained CPI would reduce them, because it attempts to reflect the behavior of all of us, to choose lower-priced items when the item we’d usually purchase increases. We’re seeing this in action with the various meat shortages due to Covid-19; at least, I personally have compared the cost of a package of lamb to the steak I was planning to buy, and chosen the former, and compared the cost of chicken breast to ground beef and likewise revised the week’s meal plan. As the Brookings Institute explains, “the BLS calculates one measure of inflation that uses the first period basket and another measure of inflation that uses the second period basket (which might include different items because of price changes—more chicken and less beef, for example)—and then takes the average. It does this every month, and creates an index that “chains” these changes from month to month.

Critics of the Chained CPI with respect to Social Security say that, to quote, for instance, The American Prospect, “of course, no real substitute exists for the main drivers of elderly people’s budgets, namely housing, medical care, and prescription drugs, all of which typically rise faster than inflation. You can’t really substitute ham for arthritis medication.” But as we saw above, medical care has only a three percentage-point difference, and in other categories, older adults are famously (infamously?) more cost-conscious.

For the period from 2000 to 2017, the chained CPI was lower than the regular CPI by pretty much the same differential as the CPI-E was higher than the regular CPI — which, in the end, suggests that a chained elderly CPI (a C-CPI-E?) could end up at about where you started.

Finally, the BLS is considering a change to its method of calculating the cost of housing, determining, instead of the equivalent rental cost, “how much it costs to own and occupy the home (e.g., by including mortgage interest payments but not the purchase price of the home).” How would this impact the CPI-E? Since a larger proportion of the older adult population own their homes than the adult population in general (81% vs. 71%), and they are likewise more likely to own homes free and clear, this ought to reduce the CPI index if based specifically on costs for the elderly, to the extent that rents rise faster than inflation as a whole.

International Comparisons

Is the United States in line or out-of-line, in comparison with other countries?

By the GAO’s count of practices in the 36 OECD countries, 27 index state pensions/Social Security benefits in whole or part based on prices. Of those, only 4 use a special index for elderly-specific inflation. None of them use a chained-type formula. (10 countries calculate a special formula for elderly-specific inflation, and 5 calculate a chained CPI index, but except for the 4 above, the remainder don’t use it in their adjustments, only for research and analysis purposes.)

Which is all well and good — but it omits the crucial piece of information that more prevalent in other countries than either of these contemplated changes is a calculation that’s based, in whole or in part, on changes to average wages rather than prices. A cursory review of the OECD Pensions At A Glance Country Profiles shows that, to take a few examples, in the Netherlands, pensions increase to keep pace with minimum wage rates; in Germany, pensions increase with wages, adjusted for “sustainability” (relative changes in the dependency ratio); and in Australia, increases are the greater of a standard CPI or a retiree-specific index, but with a minimum adjustment relative to average wages. Should the U.S. do likewise? Intellectually, it’s appealing, but it hardly seems practical given the system’s sustainability worries.

The Bottom Line

Are you still with me? The concepts here are confusing even for me — and I’m an actuary!

Making a decision about a fine-tuned inflation rate is not simple, and nothing that I’ve written (or read) suggests that there’s a single “right” or “fair” answer. In fact, here’s a thought experiment: imagine that Social Security benefits were cut, now, in an actuarially-fair way to make up for future higher cost-of-living increases. Some people would win and some would lose, depending on whether the CPI increase was higher or lower than the inflation that they themselves personally experience based on their own personal “basket of goods.” What would be the “fair” policy to enact?

In fact, no matter what adjustment method is chosen, some people will benefit more than others, because spending patterns are different — own a house or rent? have high medical bills, or in good health, or get your bills covered by Medicaid anyway? drive a car frequently or take the bus? Even something as simple as the degree to which one is accustomed to eating meat will impact how your “personal” price index compares to the Consumer Price Index.

Is the Biden campaign supporting this change because it’s the Right Thing To Do? Or because it’s an easy way to promise spending more money in favor of a voting bloc? Since it’s touted as “free money” even though it comes out of the pocketbooks, collectively, of American taxpayers, I worry that it’s the latter — but, at the same time, perhaps the fact that they don’t widely promote this position is that they recognize it’s not as simple as it appears.

 

* Thanks to Laura Miller (twitter handle @curiouser_Alice) who provided some additional clarifications and review.

 

December 2024 Author’s note: the terms of my affiliation with Forbes enable me to republish materials on other sites, so I am updating my personal website by duplicating a selected portion of my Forbes writing here.