05/16/2013 02:56 pm ET Updated Jul 16, 2013

Orszag Underestimates Chained CPI Cut

President Obama's first budget director, Peter Orszag, chimed into the debate on the chained CPI last month with an analysis in Bloomberg View suggesting that the chained CPI COLA cut -- as well as the resultant cost savings -- might turn out to be less than projected by the actuaries of the Social Security Administration (SSA). Ezra Klein picked up on Orszag's claim in his Wonkblog, noting "... if Peter Orszag ... is right, perhaps there's less for liberals to be upset about than meets the eye," citing Orszag's argument that "[w]hat neither side seems to have noticed ... is that the difference between the chained CPI and the standard CPI has been diminishing."

Before considering Orszag's argument, some background is required. Social Security's cost of living adjustments are based on the government's original Consumer Price Index, the CPI-W. Switching to the chained CPI, which has been tracked by the Bureau of Labor Statistics (BLS) since 2000, would downwardly adjust the COLA when, in response to inflation, consumers substitute for their preferred items other goods whose prices have risen less. The current index, the CPI-W, already does this for similar items, like buying a different kind of apple or buying the same product on sale; the chained CPI would also downwardly adjust for substitution across unlike items, like buying extra blankets instead of heating oil; or spending less on health care and more on food when prices in the former domain rise relative to prices in the latter.

SSA's team of actuaries estimate that over the long-term, on average, the chained CPI will grow 0.3 percentage points slower per year than the current CPI, and hence that adopting the chained CPI would cut the COLA by that amount each year. The Congressional Budget Office (CBO) doesn't make long-term projections of this differential; but for its 10-year projections, as we recover from the Great Recession, it estimates the chained CPI will grow 0.25 percentage points slower. Both of these estimates are based on analysis of empirical trends over the past 10-20 years, as well as on econometric assumptions about the future.

Orszag observes, correctly, that from the period 2000-2003, to 2003-2006, to 2006-2009, the gap between the chained CPI and the current index has become steadily smaller. But from this decline from the beginning to the end of the 2000s, Orszag speculatively extrapolates that in the coming decades, the differential between the chained CPI and the current index could remain much smaller than 0.3 basis points.

The problem with this analysis is that it speculates about long-term trends based on a very short period of time -- the collapse of the housing bubble and the onset of the Great Recession. Projecting from such a short time period is methodologically unsound in and of itself. Even more problematic, the short time period from which Orszag extrapolates is aberrational.

The period from 2006-2009 was a period of abnormally low inflation -- the lowest inflation since the 1950s. In periods of low inflation, there is less scope for price dispersion, and hence less scope for substitution. With less substitution, there is less downward adjustment of the chained CPI vis-à-vis the current index. For example, during the years since the chained CPI has been officially tracked, i.e. since 2000, the years characterized by above-average inflation have evinced an average annual differential between the chained CPI and the CPI-W of 0.6 percentage points, while the years of below-average inflation have evinced, on average, no differential whatsoever.

Thus it is no surprise that during the Great Recession, a period of abnormally low inflation, the differential between the chained CPI and the CPI-W was lower than normal. When inflation is at historically more typical levels, the differential between the chained CPI and the CPI-W tends to be about 0.3 percentage points. BLS simulations found that during the 1990s, a simulated chained CPI lagged the CPI-U (an index that closely tracks the CPI-W) by an average of 0.32 percentage points per year. And since 2000, the chained CPI has, on average, grown 0.29 percentage points slower each year than the CPI-W.

Moreover, contrary to Peter Orszag's claim, there is absolutely no evidence of a long-term decline in the differential between the chained CPI and the CPI-W. In those BLS simulations, that differential was not in steady decline; rather, it was rising throughout the second half of the 1990s, going from 0.24 percentage points from 1991-1995, to 0.4 from 1996-2000. It declined at the end of 2000s, but has been increasing again where Orszag's analysis left off, i.e. 2009-12 (although the chained CPI values for 2011 and 2012 are not yet final).

Since historically, the differential between the chained CPI and the CPI-W closely tracks inflation (and hence price dispersion and scope for substitution), Orszag's speculation that the long-term differential between the chained CPI and the CPI-W will be as low as it has been in recent years is tantamount to a projection that inflation will remain at aberrationally low levels over the long-term. Yet he provides no evidence for this. In fact, the five-year moving average of inflation was increasing from the mid-1990s through the eve of the onset of the Great Recession in 2007. While it dipped during the Great Recession, the independent, bipartisan Social Security Advisory Board's Technical Panel on Assumptions and Methods projects that it will average 2.8 percent over the next 75 years, "consistent with the historical evidence from the earlier two decades." And the historical data indicate that at that level of inflation, we can expect the chained CPI cut to be about 0.3 percentage points, as SSA predicts.

In sum, there is every reason to believe that over the long term, the difference between the chained CPI and the CPI-W will be around 0.3 percentage points, and hence that SSA's estimate of the magnitude of the chained CPI benefit cut is sound. And while a cut of less than one percent per year may sound small, COLA cuts compound, so that by age 95 the chained CPI would cut the Social Security benefits of the average worker by 9.2 percent, and cumulatively by $28,015.

This is ultimately not a debate about percentage points, but a fight for a benefit Americans have earned through a lifetime of contributions. Social Security does not contribute one penny to the deficit, and has no place in discussions over grand bargains aimed at reducing our nation's debt.