It is well known that consumers tend to overestimate the rate of inflation. They see in particular the prices of gasoline and food rising, and given this "frequency bias," perceived inflation rates are far higher than measured inflation. Indeed, alternative price indices which include exclusively such daily consumer expenditures, such as the American Institute for Economic Research's "Everyday Price Index," reveal inflation rates are consistently above the CPI, PCE, GDP deflator, or other broader estimates.
But it's not just all perception. The CPI considers prices for the average urban consumer according to fixed weights, and this weighting system does not accurately reflect costs of living for what you and I would recognize to be discrete demographic groups.
One such group is the elderly. Given their relatively high expenditure on medical care as a percentage of their incomes and the increases in the relative price of medical care, using the CPI to measure the cost of living for an elderly person just does not make sense. That's why the Bureau of Labor Statistics has developed a CPI-E -- the "E" is for elderly, of course.
And yet, the federal government does not use the CPI-E for cost-of-living adjustments to either Medicare or Social Security. Medicare uses the CPI-U, which is the normal CPI index for the average urban consumer; Social Security uses the CPI-W, which is slightly different in that expenditure weights reflect expenses for only urban wage and clerical workers, a narrower group of consumers than in the CPI-U. Quite frankly, neither does an adequate job as a measurement of costs of living for Medicare or Social Security recipients.
(In case you are, like me, interested in these sorts of esoteric details which have huge and largely unrecognized policy implications, check out this link which explains the difference between CPI-U and CPI-W.)
Another group for which the CPI does not represent costs of living is the poor. There have been attempts in the U.S. to develop a "CPI-P" -- I found this BLS paper from 1996 -- but the work is so forgotten that the Bureau of the Census continues to adjust the poverty thresholds for inflation based on the CPI-U. This makes absolutely no sense at all, and when every dollar of income counts, there are very real human consequences to mis-measuring costs of living. Previously, studies cited in the BLS paper had found that the basket of goods and services purchased by the poor did not see relatively higher inflation; that finding for the last decade of prices data appears strained.
In this blog post, I want to revive the CPI-P. And as I normally do when I create datasets, it will soon be added to the "Data" page available on the righthand column for future reference, using FRED to automatically update the data. You can also get it here.
Like the CPI, the CPI-P will use a weighting scheme, derived from the Bureau of Labor Statistics' "Consumer Expenditure Survey," which measures how consumers spend their income on various goods and services. From this survey comes expenditures data on the bottom income quintile, which I'm using as the representative consumer for the CPI-P.
What I found was that the cost of living for the poor has increased significantly more than as is measured by the CPI. Relative to 1967, the earliest I can go back with my data, the CPI-U undercounts living costs for the poor by 12 percent of CPI-U. That implies that any safety-net program which indexes payments to the CPI-U has cut those payments on a real basis by 12 percent, all else equal.It may help to imagine it this way: there are two baskets of goods, one for the average urban consumer and one for the average person in the bottom income quintile. In 1967, it would cost $100 to purchase either basket of goods. As of June 2012, the conventional basket would cost $694.90, whereas the basket for the poor would cost $784.24.On a month-to-month basis, the drift of the CPI-P from the CPI-U is small, usually no more than a fifth of a percent. But it is consistently positive. And over a half a century, the consistent positive differential adds up to a meaningful difference in the cost of living as measured versus what it actually has become for the poor.