By Mark J. Warshawsky

In Part I, we discussed various measures of price changes and how they differ. In Part II, we will look at the data to precisely estimate these differences and discuss policy implications.

Below we see a graph of the monthly measures of annual inflation
from January 2001 through June 2022, using the four price measures described
earlier. In broad terms, they move closely together, but certain differences
are apparent. The CPI-W measure somewhat exaggerates the highs and lows of
inflation. Given that this extra volatility is not a property of the most
accurate measure from the C-CPI-U, it is undesirable for public policy use, as
it causes needless anxiety and concern about inflation measurement for benefit
purposes. In particular, deflation, which is particularly upsetting to the
Social Security population, is more likely with this measure. The inflation
measured from the C-CPI-U is a bit below the other measures, and there is no
discernable pattern in either direction from the R-CPI-E.

These observations can be confirmed more formally by examining
some statistics in the table below. The mean inflation rates based on the CPI-U
and CPI-W are nearly identical. The R-CPI-E rate is only 0.06 percentage points
higher than those rates—hardly noticeable to the intended recipients of the
proposed policy change, being a lower increase than commonly attributed to it,
of 0.2 or 0.3 percent, while still being a cost to the system and the taxpayer.
The C-CPI-U measure, however, is lower by 0.29 percentage points, which would
represent considerable savings over the long-run in a solvency reform for
Social Security; indeed the savings is estimated by the actuary at nearly a
fifth of the total financial shortfall of the program. By contrast, the CPI-W
has the greatest volatility as measured by standard deviation, at 1.92 percent.
The correlation of CPI-U with CPI-W and C-CPI-U is high though, at 0.995 and 0.993,
respectively, whereas with R-CPI-E it is only 0.976, indicating a source of
confusion and dissonance with current familiar measures. The relative pattern
of correlations with the most accurate measure, C-CPI-U, is similar.

The policy implications of this discussion and data analysis
are clear. The most accurate and low volatility measure to use for the Social
Security COLAs, while still being familiar and producing savings for Social Security
reform, is the C-CPI-U. If savings, however, are not desired from COLAs,
switching from the CPI-W to the CPI-U would be good policy to reduce volatility
and increase familiarity and transparency. On most of these policy metrics,
proposals to switch to the R-CPI-E fare poorly.

The post Inflation and the Social Security COLA, Part II: Data and Policy Implications appeared first on American Enterprise Institute – AEI.