How is the COLA for Social Security Computed?

It’s important to recognize how big a deal this is.  Social Security benefits automatically increase with inflation.  Almost no other product or pension available to the public does that.   This is huge.  If you were getting a fixed dollar amount every month forever, when prices go up, your fixed dollar amount would buy you less and less over time.  Social Security, while not perfect, goes a long way to making sure your benefits go up along with the prices of things you buy, so over time, the purchasing power of your Social Security benefits does not decrease.

There are many possible measures of inflation.  And every individual has his or her own actual rate of inflation — because everybody buys a different collection of goods and services, and the changes in prices for these things isn’t uniform.  If you buy more of a good which has gone up a lot in price and less of another good which hasn’t, your personal rate of inflation may be higher than the broader average.  Or vice-versa.

Moreover, the rate of growth of wages is different from the rate of growth of goods and services.

Nevertheless, once one has started collecting Social Security benefits, the annual inflation adjustment — the COLA, or Cost Of Living Adjustment — is based on the annual change in the CPI-W — the Consumer Price Index for Urban Wage Earners and Clerical Workers.  The CPI-W is updated and published monthly by the BLS — the Bureau of Labor Statistics.

The CPI-W itself is not a measure of inflation in the sense that the CPI-W at any point in time doesn’t tell you about the changes in prices.  It’s simply an average price level for a “basket of goods”.  To convert a price into a rate of change in prices, one needs to measure the price level at two points in time and compare them.   Inflation is the change in prices over a given period of time.

What gets confusing is that the monthly number published CPI-W by itself is, again, just a price level — for June 2022, the CPI-W is 292.542.  That doesn’t tell you anything about inflation by itself.  You need to compare it to the CPI-W at another point in time.  The May CPI-W was 288.022, for a monthly change of (292.542 – 288.022 == 4.52).  In one month, prices rose by (4.52/288.022 == 1.57%).  That’s a pretty substantial change in a single month.   If prices went up by 1.57% every month for 12 months, over the course of a year, prices would have risen by more then 20%!  That’s *not* what has actually happened over the course of a year, though. The price level changes are quite variable and they move around more, relatively, on a month-to-month basis then they do over longer periods of time like year-to-year.

Even though the CPI-W is recomputed monthly — the headlines — and how we actually think about things — is on an annual basis.  From June 2021 to June 2022, the CPI-W went from 266.412 up to 292.542.  This change represented a total of a little less then 10% year-to-year inflation.

So how does this all factor into the COLA you can expect?

The COLA is based on the annual rate of change in the CPI-W.  And an annual measure has to be constructed by looking at price levels a year apart.   So, over time, any given month could be chosen, an annual rate computed, and so long as the same month is used year after year, the long-term average change is going to be the same — every single month of change is factored in.

But as you can see from the June example above, month-to-month volatility in the CPI-W is substantially higher than year-to-year volatility.  In order to smooth that out, rather than simply picking a single month and looking at it from one year to the next, instead, Social Security takes three months and averages the year-to-year change for all three.  The months they use are the third quarter of the calendar year — July, August, and September.   They average the CPI-W for those three months last year, they average the CPI-W for those three months this year, then they compare the two averages.   That smooths things out a little.

But the important thing to recognize is that (a) it’s a year-to-year computation.  As such, every single month’s inflation is factored in.  Just because they use the July, August, and September CPI-W measurements doesn’t mean they are ignoring the inflation which took place in January or April.  They are comparing the CPI-W from last July to the CPI-W of this July — which means that every month between those two Julys is included.   And (b) because they use the same three months every year, every single month, over time, is included, but by using three months, they are rolling forward and smoothing it out a little bit.  The long term average has to be the same — each month is eventually included exactly the same number of times.  But in any single given year, certain months are included three times in the average (ie. When computing the 2022 COLA, they compared prices from 2020 to 2021 and so, April 2021 was included three times in computing that average) while other months blend previous and current years (again, when computing the 2022 COLA, they used July 2020-July-2021, August 2020-August 2021, and September 2020-September 2021 — so each of July, August, and September get either two from 2020 and one from 2021 or two from 2021 and one from 2020).  Over time, each month is included precisely the same number of times.  It’s just that those measurement months get blended year-to-year.

The up-shot of all this is that while it’s important to look at the July, August, and September annual changes when computing this coming year’s COLA — over the long run, it doesn’t matter because if the choice of months led to a slightly higher COLA this year, that’ll be offset by a slightly lower one in some combination of the previous and subsequent year.  And similarly, if the July, Aug, and Sept numbers lead to a slightly lower COLA this coming year (say, perhaps, inflation dropped a lot in Aug and Sep) — that’ll be offset by higher COLAs in prior or subsequent years as, again, every single month factors in over time in precisely the same proportions.

Sample Computation — How was the 2022 COLA computed?

In July, August, and September of 2020, the CPI-W was, respectively, 252.636, 253.597, and 254.004.  The Average of the CPI-W for 3Q2020 was therefore 253.413.

In July, August, and September of 2021, the CPI-W was, respectively, 267.789, 268.387, and 269.086.  The Average of the CPI-W for 3Q2021 was therefore 268.421.

Thus, the year-to-year change from the 3Q2020 average to the 3Q2021 average was (268.421-253.413)/253.413 == 0.05922 or 5.9%

No surprise, then, that when the SSA announced the 2022 COLA (announced on 10/13/2021, to take effect when 2022 started) — the COLA was 5.9%

https://www.ssa.gov/news/press/releases/2021/#10-2021-2

We don’t know what the COLA will be for 2023 yet.  The July CPI-W was just announced on 8/10/2022 and it came in at 296.276, an 8.5% increase over the July 2021 CPI-W.  Since the inflation in July, Aug, and Sept 2021 was pretty moderate, most of the annual increase reflected in the July 2021 to July 2022 number will show up again in the August 2021 to August 2022 and September 2021 to September 2022 numbers and there’s every reason to expect, therefore, that the 2023 COLA will be pretty close to the 8.5% which we saw from July 2021 to July 2022.  It won’t be exactly that.  But unless there’s some substantial disinflation or even less likely, deflation in the next two months, it’s reasonable to expect around an 8.5% increase in Social Security benefits for 2023.

Social Security Cost of Living Adjustments have been law since 1974, and the first COLA applied using the formula from that law took place for benefits in 1975.  Prior to that, any annual change in benefits required new legislation.  Since then, it’s been fully automatic.  (They’ve modified the formula a little bit.  The current “average third quarter change” formula was adopted in 1983.  Before that, it was first quarter-to-first quarter.  Third quarter makes it a lot easier to compute it several months before it takes effect.)

If you’d like to see the historical CPI-W levels, Social Security has conveniently gathered them, monthly, going all the way back to 1974, on this page:

https://www.ssa.gov/oact/STATS/cpiw.html