Contents

• Introduction
• Expenditure Patterns of the Elderly
• The Experimental CPI for the Elderly
• Policy Considerations

Summary

The federal government, in an effort to protect the purchasing power of Social Security beneficiaries, indexes benefits to increases in the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W). Concern has periodically been expressed that the CPI-W may understate the impact of inflation on the elderly population and that it therefore may not be the most appropriate measure of inflation's impact on the elderly.

At the behest of Congress, the U.S. Bureau of Labor Statistics (BLS) developed an experimental price index to track changes in the cost of living for the population aged 62 and older. In most years since 1982, the start of the experimental consumer price index (CPI-E) for the elderly, the annual rate of change in the CPI-E has exceeded that of the CPI-W and CPI-U. But, methodological limitations in the experimental index may have contributed to this pattern. Were BLS to construct an index that is more representative of the elderly population than the CPI-E, there is no guarantee that the relationship between the new index and the CPI-W would be the same.

Interest in the CPI-E most recently emerged in response to deficit-reduction plans issued in 2010 and 2011 that recommend inflation-indexed provisions in federal law be based on the Chained Consumer Price Index for All Urban Consumers (C-CPI-U). Because the C-CPI-U has typically risen more slowly than the CPI-W, this proposal raised concern at the time among those Social Security recipients who already believe they have not been fully compensated for increases in their cost of living. Bills were then introduced to switch for purposes of Social Security indexation from the CPI-W to a CPI for those aged 62 and older (H.R. 456, H.R. 539, H.R. 776, H.R. 798, and S. 1876). As suggested by an amendment in the nature of a substitute to the FY2013 budget resolution in the House, interest has lingered into 2012 among some Members to switch to the C-CPI-U as a means of curbing the rate of growth in the budget deficit.

Introduction

In an effort to protect the purchasing power of Social Security recipients, Congress in the early 1970s indexed benefit increases to the only consumer price index available at the time. The index to which Social Security benefits are linked became known as the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W) after the U.S. Bureau of Labor Statistics (BLS) began publishing the Consumer Price Index for All Urban Consumers (CPI-U) in 1978.¹

Concern has periodically been expressed that the CPI-W may not accurately reflect the inflation experience of the elderly, who make up the majority of Social Security beneficiaries. It has been argued that because annual cost-of-living adjustments (COLAs) in Social Security benefits have not kept pace with increases in the prices of goods and services more often purchased by the elderly (e.g., health care), some index other than the CPI-W might be more appropriate (e.g., an index for the elderly that Congress directed BLS to develop in 1987). This is not the context within which reconsidering the index upon which Social Security benefits are based has most recently been raised, however.

Suggestions to change the index for inflation-adjusting Social Security among other programs and provisions in federal law most recently have arisen in connection with deficit reduction. Several plans to curb the growth in the U.S. budget deficit, which were put forth in 2010 and 2011, recommend that inflation-indexing be based on the Chained Consumer Price Index for All Urban Consumers (C-CPI-U) rather than on the CPI-W or CPI-U.² Because the C-CPI-U has typically risen more slowly than the two indexes, this proposal raised concern at that time among those Social Security recipients who believe they already are being insufficiently compensated for increases in their cost of living. A bipartisan amendment in the nature of a substitute to the FY2013 budget resolution in the House, which was introduced but not approved in March 2012, suggests that interest remains in slowing the growth in the deficit by changing the price index on which Social Security COLAs are based.

This report opens with an explanation of whom and what the CPI-W represents before examining how the spending pattern of the average elderly household differs from that of all households. It then focuses on BLS' experimental consumer price index (CPI-E) for the elderly and analyzes rates of change over time in the CPI-E, CPI-W, and C-CPI-U. The report closes with a brief discussion of policy considerations.

Expenditure Patterns of the Elderly

The CPI-W is designed to measure changes in the price of a market basket of goods and services purchased by those who earn at least half of their income by having worked in clerical, blue-collar, or service occupations for at least 37 weeks in the previous year. In other words, the CPI-W only tracks the buying habits of the employed. This particular group of employed persons has accounted for a dwindling share of the U.S. population over time.³ Today, it reflects changes in the cost of living of about 32% of the population.

Price changes may affect the average retiree's cost of living differently from that of the average CPI-W household to the extent that their purchasing patterns differ from one another. BLS collects data through the Consumer Expenditure Survey (CES) on how households spend their money in order to assign weights to each of the goods and services in the market basket. The weight reflects an item's relative importance in the market basket and determines how much a change in an item's price will affect the overall CPI.

As shown in Table 1, elderly households allocate their spending differently from the rest of the population across the major categories of goods and services in the CPI. The largest difference in spending patterns between the elderly and the general population is found in the shares of expenditures accounted for by health care.⁴ In 2010, the latest year for which data are available, those aged 65 and older spent twice as large a share of their total outlays on health care compared with the overall population. With respect to the population aged 75 and older, the share of their spending devoted to health care was two-and-one-third times the share of the total population.

Health care costs have consistently risen more rapidly than the average price level.⁵ Because the elderly consume a greater than average share of a good whose price has generally risen faster than overall prices, the CPI-W may understate the inflation experience of the average elderly household.

As noted above, the argument is often made that the CPI does not represent the average inflation experience of the elderly population. But, just as the inflation experience of the elderly population may differ from that of the population at large, so too are there differences within the elderly population itself.

No summary inflation measure for a large population group will exactly account for the experience of each member of that group. Differences in spending patterns, in combination with different rates of price change for all of the various goods and services included in the CPI, mean that individual inflation rate experiences may range considerably above or below the measured average. If there is a great deal of variation in both the general population and within subgroups such as the elderly, a small difference in average inflation rates between groups may not be significant.⁶

Suppose the average inflation rate of the elderly population is slightly higher than the rate for the overall population, but that the distribution of individual inflation rates among the elderly is widely dispersed. In this case all of the elderly would be better off if their benefits were indexed to an inflation measure based on the average elderly household. Within the elderly population, however, there would be several different consequences. First, there would be some elderly whose inflation rates would be understated by the overall rate, but exaggerated by the elderly inflation measure. Second, there would be those elderly whose inflation rates were higher than either the overall measure or one based on elderly consumption patterns. Finally, there would be a number of elderly whose actual inflation rates would be lower than either the overall measure or one based on the elderly.

One study of the distribution of inflation rates across the population found that differences in inflation rates among demographic groups were small in comparison with the variation within those groups. Further, it was found that differences among groups tended not to be stable over time. This study argued that no one group suffered disproportionately from inflation.⁷ If the variation in consumption patterns is great among the elderly and if the average inflation rate of the elderly is not dramatically different from the average rate of the overall population, then arguments for a separate index for the elderly population may be less compelling.

The Experimental CPI for the Elderly

In 1987, Congress amended the Older Americans Act of 1965 to direct BLS to develop an experimental price index to track inflation in the population aged 62 and older. BLS has calculated estimates of such an index, commonly called the CPI-E, dating back to December 1982.

In all but 5 of the 29 years between December 1982 and December 2011, the experimental index rose as or more rapidly than the CPI-W and CPI-U. (See Table 2.) In only three years was the increase in the CPI-E closer to that of the CPI-W than the CPI-U. The increase in the CPI-E has usually been closer to that of the CPI-U partly because a larger weight is given to health care outlays in the market basket of the CPI-U than the CPI-W. (Recall that unlike the CPI-W, the CPI-U covers persons not in the labor force including retirees.)

The difference in the annual rates of change of the CPI-E compared with both the CPI-W and CPI-U has generally decreased since 1993, largely because the gap between health care inflation and overall inflation has narrowed as has the gap between shelter inflation and overall inflation.⁸ The 0.5-0.7 percentage point gap in the annual rates of change between the CPI-E compared with the CPI-W and CPI-U that often was the case from 1982 to 1993, shrank to 0.3 percentage points or less in most years thereafter. (See Table 2.)

Although the differences in the three indexes usually have been in the expected direction, the relationships between the three might not be the same if BLS developed an official rather than experimental index for the elderly.

Optimally, when constructing a CPI for older Americans, a sample of geographic areas would be drawn for that specific population. In addition, surveys would be designed to collect expenditure weights for that specific population, a point-of-purchase survey designed for that population would be used to construct the outlet frame, and the distribution of items sampled would be representative of older Americans. Such an index would be costly to construct, however, and Congress has not appropriated the necessary funds to do so.⁹

For example, the number of households in the CES on which market baskets are based is much smaller for the CPI-E than for the CPI-W and CPI-U. The CPI-E therefore is subject to greater sampling error than the official indexes because of BLS' limited resources. The CPI-E also uses the CPI-U's sample of retail outlets to gather prices, but the outlets may not accurately reflect those at which the elderly shop and the prices may not be representative of those paid by the elderly. These methodological limitations may have contributed to the differences in the experimental compared with official measures of inflation and are the reasons for it being "classified as an experimental index."¹⁰

Policy Considerations

If the primary purpose of developing a separate index for the elderly is to inflation-adjust Social Security benefits, it should be kept in mind that not all Social Security recipients are elderly. Some receive benefits under the program because they have disabilities; others, because they are the spouse or young child of a deceased worker covered by the program. Thus, some would argue that the market basket of the elderly population is not the most appropriate one on which to base adjustments to Social Security benefits.

Having a separate price index for the elderly may introduce complications in other areas. For example, the income thresholds that define tax brackets currently are adjusted annually by the CPI-U. If it is appropriate to base Social Security benefit adjustments on a price index for the elderly, should it also be used to adjust income tax brackets of elderly taxpayers?

Finally, as stated at the outset, recent interest in the indexes used to inflation-adjust federal programs and individual income tax provisions has been prompted by the desire among policymakers to curb the growth rate of the budget deficit. Switching from the CPI-W or CPI-U to the Chained CPI-U may reduce government outlays and raise revenue in future years because the Chained CPI-U has risen more slowly than the two official indexes—and therefore, more slowly than the CPI-E.¹¹ Leaving aside whether the Chained CPI-U is a more accurate measure of inflation than the CPI-W and CPI-U,¹² it would not appear to achieve the purpose of those who have proposed changing to the Chained CPI-U to instead switch to the CPI-E for calculation of Social Security benefits as some have suggested.¹³ Were BLS to replace the experimental index with a newly developed index more representative of the elderly population, there also is no guarantee it would bear the same relationship to the CPI-W and Chained CPI-U as that of the CPI-E.¹⁴

Footnotes