Skip to content

Social Security Reform: Economic Effects of Planned Changes in the CPI


The Consumer Price Index (CPI) plays a crucial role in the Social Security program because CPI adjustments affect future benefits, which are indexed to inflation. The Bureau of Labor Statistics (BLS), which estimates the CPI on a monthly basis, has made several corrections in the construction of the CPI. In 1996, the Advisory Commission to Study the Consumer Price Index (Boskin Commission), created by the Senate finance Committee, estimated an upward bias of 1.1 percentage points per year in the CPI relative to a true cost of living index. The range of plausible values, according to the Boskin Commission, could be 0.8 to 1.6 percentage points per year. Conceptual changes by the BLS incorporated in 1998 and those proposed in 1999 would further reduce the growth in the CPI. Most of these changes were already under study or planned by the BLS before the Boskin Commission report was issued.

In the 1998 Report of the Board of Trustees of the OASDI Trust Funds, the long-run inflation assumption is a flat 3.5 percent after 2006, which is the same assumption as in its 1997 report and 0.5 percentage points below the 1996 assumption.The BLS made changes to the CPI in 1998 and plans more changes in 1999 that are expected to reduce measured inflation further by a combined 0.46 to 0.56 percentage points.

The CPI for All Urban Consumers (CPI-U) represents the spending habits of about 80 percent of the population of the United States. The CPI for Urban Wage Earners and Clerical Workers (CPI-W) covers a subset of the CPI-U population, and represents about 32 percent of the total U.S. population. Adjustments to Social Security benefits are currently based on the CPI-W, measured from the average of the third quarter of one year to the third quarter of the succeeding year. The changes made in the construction of the CPI are reflected in both CPI-U and CPI-W.

Biases and Changes to the CPI

The American economy is flexible and dynamic. New products replace old ones almost every day, and the relative prices of different goods and services change almost daily. Thus, it is difficult to construct an accurate cost-of-living index that can truly reflect the ever-changing basket of goods.

The Boskin Commission pointed out four biases that cause the CPI to overstate inflation relative to a true cost of living index.

  • Substitution bias occurs when the CPI's fixed basket of goods fails to reflect the fact that, as a result of changing relative prices, consumers substitute less expensive for more expensive goods. For example, if chicken becomes more expensive relative to fish, consumers may substitute fish for chicken. Yet the current CPI continues to use the same quantity of both fish and chicken as in the original basket. As a result, the CPI is overestimated because it does not reflect substitution that consumers make. The formula currently used by the BLS is based on a fixed basket of goods.
  • Outlet substitution bias occurs when shifts to discount stores by consumers are not adjusted for quality differences in the stores themselves, such as differences in quality of service and convenience of location. Current methodology does not account for quality differences in purchasing the same item from different outlets.
  • Quality bias occurs when improvements in quality of products are not accurately reflected in the price, overstating the CPI by as much as 0.6 percentage points, according to the Boskin Commission.
  • New product bias occurs when new products are introduced into the market basket too late to capture the full price changes that occur after these products are introduced. For example, there are 36 million cellular phones in use, but the CPI has no price index for cellular phones, which have been in widespread use for years. During that time, prices have come down substantially.

Since January 1997, the BLS has made three major qualitative changes:

  1. Revised the Hospital Service Index so that it now reports the pricing of bundles of services on patient bills rather than individual items like a room charge, lab-tests charge, or hospital drugs charge.
  2. Replaced the personal computer's single price index with a hedonic regression approach (in which each computer component is individually priced) to reflect the true price of personal computer and peripheral equipment.
  3. Revised and updated the market basket of consumer goods. Effective February 1998, the 1993-95 consumer spending patterns have replaced the 1982-84 spending patterns as the basis for weighting 200 items in the CPI.

The new weights are slightly lower for food and beverages and higher for new items like food-away-from-home and a new major group, Education and Communication. It is estimated that updating weights may reduce measured inflation by 0.1 to 0.2 percentage points per year.

The BLS has also improved the CPI by making other changes such as the revision of sample-rotations (where new sample observations systematically replace the old ones), the inclusion of new geographic areas in the sample, and the collection of price quotations from outlet stores as well as retail stores. To reduce the substitution bias in the construction of the CPI, the most important change forthcoming in 1999 is the use of a geometric mean index in place of the current methodology for estimating price changes in nearly two-thirds of consumer spending. Since this change is likely to reduce measured inflation more than other recent changes, it would be worthwhile to discuss the conceptual framework of the geometric mean index as compared with the current methodology.

The Geometric Mean Index

Effective January 1999, a geometric mean estimator will be used for 61 percent of the total consumer spending represented in the CPI for All Urban Consumers (CPI-U) and 64 percent of the consumer spending in the CPI for Urban Wage Earners and Clerical Workers (CPI-W). The remaining categories will continue to use the current methodology. Based upon BLS research, the adoption of the geometric means approach will reduce the annual rate of inflation in the CPI by approximately two-tenths to three-tenths of a percentage point per year. The use of the geometric mean was also recommended by the Boskin Commission to reduce the substitution bias.

The current method measures relative price change from the base period or beginning year (0) to the current period (t) by comparing the sum of the weighted prices, the weights being equal to the quantities of a good being purchased in the beginning (base) year. The geometric mean multiplies the relative prices raised to the power of the expenditure-shareof each item. The purpose is to eliminate from the formula the base year quantities, which generate biases, and replace them with expenditure shares which happen to be more stable than quantities.

The current formula also estimates prices each month for a fixed basket of goods and services. If the prices of all goods and services increase by the same amount, say five percent, and the basket does not change, then both the current and geometric mean approaches will show the index increasing by five percent. The two formulas will give different results, however, if the prices of items within the basket or even within an item category increase by different percentages and consumers make substitutions.

The geometric mean formula assumes that the basket of goods changes with the change in the relative prices and that the quantities of items adjust in such a manner that the consumer expenditure on meat (e.g., chicken and fish ) as a proportion of total consumer expenditure remains constant relative to all other items in the family budget. For example, if a family is spending 10 percent of its budget on meat and poultry -- which we are calling an expenditure share -- the share would remain constant regardless of how this family substitutes fish for chicken or vice versa. An intuitive explanation of the geometric mean index is that it holds the share of expenditure on an item constant, whereas the current formula holds the quantity of the item constant.

The choice of items that are included in the geometric mean index by the BLS is based on their cross-price elasticity of demand. Economists have shown that the geometric mean formula is more appropriate when consumers can easily substitute new or cheaper products for old or expensive products. The fixed-basket formula is appropriate when substitution is difficult. Usually, goods with elastic demand, such as food and beverages, are easily substitutable, while goods with inelastic demand, such as housing, utilities, and medical services, are more difficult to substitute.

Thus, BLS has excluded the following item categories from the geometric mean estimation, as their demands are inelastic, and consumers are least likely to make substitutions for these items.

  1. Shelter Services -rent of the primary residence, owner's rent equivalent of primary residence, and housing at school.
  2. Utilities and Government Services - electricity, natural gas, water and sewage; telephone and cable TV; and state and local registration fees, license fees, and motor vehicle property tax.
  3. Medical Care Services - physicians' services, dental services, hospital services, eye care, nursing home and day care, and other medical professional services.

The specific changes and their effects on the CPI are listed below:

The use of the geometric mean index, which will begin in January 1999, is likely to reduce measured inflation by 0.2 to 0.3 percentage points, according to the BLS. Hence, our simulation has assumed an average value of 0.25 percentage points due to this change, splitting the difference in the BLS estimates.

The total correction in the CPI in 1998 was -0.21 [(- 0.06) + (- 0.15)] percentage points, and in 1999 is planned to be -0.25 to -0.35 percentage points. The sum ranges from -0.46 to -0.56, which we model as a change of -0.51.

Since Social Security benefit payments are indexed to the changes in the CPI, it is important for the actuaries of the Social Security Administration to incorporate the proposed changes in their calculations of trust fund balances. The 1998 Annual Report of the Trustees, however, did lower the expected annual percentage change in the CPI from 3.2 percent to 1.4 percent for 1998, from 3.2 percent to 2.4 percent for 1999, and from 3.2 percent to 2.6 percent for 2000 in their "intermediate" cost assumptions.

The following economic analysis of the planned changes in the growth of the CPI is based on the 30-year long-run simulation module of the Macroeconomic Advisers (MA) model, which closely follows the CBO's long run assumptions. The present simulation uses the CBO's current assumption on long term interest rates and federal budget surpluses. The present simulation, however, assumes no change in the Federal Reserve's monetary policies.

Economic Impact

This section presents the economic impact of the planned changes in the CPI on some key economic variables like the federal deficit, the trust fund balance, Social Security benefit payments, and Gross Domestic Product (GDP) and its components.

Federal Budget

According to the MA model, the FY 1998 surplus of $80 billion is expected to increase to $113 billion in 1999, $124 billion in 2000, and reach a peak value of $241 billion in 2010 before turning back into a deficit again in 2022. Our simulation shows that the forthcoming changes in the CPI would increase the surplus to $269 billion in 2010, $190 billion in 2019, and surpluses would continue until 2024. (See Figure 1.)

Social Security Trust Fund and Benefits

Social Security benefits are adjusted for inflation according to the CPI-W. When implemented, forthcoming changes in the CPI would reduce current benefits by 1.9 percent in 2005, 3.7 percent in 2010, 7.3 percent in 2020, and 9.4 percent in 2026, saving as much as $153 billion in benefit payments by 2026. Consequently, the trust fund, instead of declining to $2,673 billion after reaching a peak value of $3,334 billion in 2020, would continue rising throughout our forecast period. By 2026, it would reach $4,437 billion, about $1,764 billion over the base scenario. (See Figure 2.)

However, the reduction in Social Security benefits would not be sufficient to bring them into balance with payroll contributions on an annual basis, and annual benefits would still exceed payroll taxes even in 2013. The interest income on the trust fund, as shown in Figure 3, is higher than the baseline and would accumulate $607 billion more in interest income than the baseline over a period of 1998 through 2025.

While interest income to the trust fund starts declining after 2020 in the base scenario, the proposed changes in the CPI cause interest income to keep rising (see Figure 3), reaching $227 billion by 2026. This amount is higher than the MA baseline of $121 billion and even higher than the Trustees' estimate of $207 billion. It should be noted that while the Trustees assume a fixed interest rate of 6.1 percent in estimating the interest income, the MA model calculates the interest income based on the long-term yield on 10-year government bonds. This long-term yield is determined endogenously (internally) in the MA model as an equilibrium rate of interest that equates savings to investment in the long run. The long-term yield on 10-year government bonds in the MA model reaches 8.45 percent by 2026.

Social Security Payments Per Beneficiary

According to the MA simulation, a lower CPI would also reduce Social Security benefits per beneficiary in current dollars. Average annual benefits per beneficiary are projected to decline by $263 by 2005, $616 by 2010, $1,605 by 2020, and $2,396 by 2026. (See Figure 4.)

Social Security benefit reductions would cumulate over time, and the total reductions would be greatest for those who lived longest. For example, the present value of the cumulative benefit loss, using a five percent discount rate, for a beneficiary receiving a monthly check of $800 in 1998 would be $926 by 2005, $2,923 by 2010, $9,062 by 2020, and $17,239 by 2030.

Investment and Gross Domestic Product

Rising surpluses in the federal budget and lower inflation would result in lower interest rates. The real federal funds rate would decline from 6.8 percent to 6.2 percent in 2026, and that would help increase gross private domestic investment by 0.7 percent in 2010, 1.8 percent in 2020, and 2.7 percent by 2025. Real GDP may initially decline due to the lowering of benefits (and falling demand). However, as investment activity picks up speed, real GDP would gradually rise by as much as $70 billion in the long run, approximately 0.6 percent over the baseline. (See Figure 5.)

A lower CPI would not only increase investment, it would also promote capital deepening (more capital per unit of labor) and capital widening (more capital plant and equipment) in the manufacturing sector. Wages and salaries would decline due to lower COLAs, but capital deepening would increase productivity per hour in the long run. A lower growth in CPI would also encourage exports favorably, with net exports declining from -$275 billion to -$205 billion by year 2026.

Corporate Profits, Personal Income, and Consumption

A lower COLA would not only lower wages, but it would also lower corporate profits-after-tax by 2.7 percent and Federal Reserve profits by 4.2 percent by 2026. A decline in federal transfer payments to persons (at current prices) by $10 billion in 2005, $29 billion in 2010, $104 billion in 2020, and $187 billion in 2026 would shrink both personal income and personal disposable income by 2.3 percent, and real disposable income and real consumption (at 1992 prices) by less than one percent by 2026. The effect of proposed changes in the CPI is most severe on federal transfer payments, reducing them as much as five percent by 2026. The percent change in corporate profits, personal income, and federal transfers is shown in Figure 6.


Lower benefits would affect gross private saving adversely. However, despite a decline of $71 billion in gross private saving by 2026, roughly 2.5 percent below the baseline, gross saving (defined as the sum of gross private saving and the combined budget surplus or deficit of federal and state governments) would still increase by over $195 billion, roughly 6.2 percent over the baseline by 2026. (See Figure 7.)

Gross saving as a ratio of GDP is not affected much by the changes in the CPI. This ratio is projected to decline any way from 17 percent in 1998 to 12 percent by 2026 in the base scenario; it would, however, decline to 13 percent under the proposed changes. Personal saving, which is a component of gross private saving, has turned negative in the last two quarters of 1998. The ratio of personal saving to personal disposable income, more commonly known as the saving rate, has declined from 6.2 percent in 1992 to - 0.2 percent in 1998. Unfortunately, this trend is likely to continue in the future unless significant changes occur in the stock market and individual savings increase in the form of bank deposits. In the short-run module of MA model, the saving rate barely increases to one percent by 2007.

The reason for negative savings is not that individuals are not saving at all; the real reason is that the Bureau of Economic Analysis (BEA) has changed the National Income and Product Account (NIPA)definition of personal income. As of July 1998, personal income excludes the capital gains distributions of mutual funds. In brief, when capital gains are distributed by the mutual funds industry, consumers view it as an increase in wealth and, therefore, increase their consumption. The new NIPA definition excludes the capital gains from personal income but these capital gains are registered in consumption. As a result, the ratio of consumption to income keeps rising, while that of saving to income continues to show a decline. If such gains in wealth were not excluded from personal income, the saving rate would only show a decline from 6.2 percent in 1992 to 5.4 percent instead of -0.2 percent in 1998. The picture presented by personal net worth (net additions to wealth in the form of corporate equities and capital gains distributions) shows households in a stronger position than one might expect based on personal saving rates. In the 1980s and 1990s, the additions to wealth surpassed personal saving by a substantial margin. The growth in net worth also surpassed the growth in personal income.

Other Effects

Lowering the growth in the CPI does not seem to influence the civilian unemployment rate or the total "man-hours" worked. The annual growth in the producer price index would decline by 0.1 percentage points in the long run, lowering the producer price index by 2.1 percent below the baseline by 2026. The cost of capital net of tax would decline by about 0.6 percentage points in the long run, similar to other interest rates.

The household net worth (holding) of corporate equities has been rising enormously at a rate of 3 to 4 percent per year since the eighties, and the proposed changes in the CPI will further increase the holding by $926 billion, about 3.4 percent over the baseline by 2026.


The BLS has made adjustments in its methodology for calculating the CPI, which it estimates will reduce the CPI by 0.31 percentage points effective January 1999. At that point, much of the reduction in the CPI advocated by the Boskin Commission will have been achieved. Many Social Security reform proposals have included reductions in the CPI as one measure to reduce the actuarial deficit in the Social Security trust fund, although they are often merely recognizing the changes already made by the BLS.

Our analysis shows that the forthcoming reduction in the growth of the CPI would increase federal budget surpluses; expand trust fund balances; and increase fixed investment in plants, structures, and equipment. An increase in investment, in turn, would help increase the national product in the long run, thus making the economic pie bigger for all. On the other hand, it would also cut Social Security benefits, welfare payments, and unemployment benefits. Because the CPI also governs the adjustments to the poverty line year by year, it would also influence the measured numbers of persons in poverty every year.


  1.  U.S. Senate. Advisory Commission To Study The Consumer Price Index, "Towards a More Accurate Measure of the Cost of Living," Final Report to the Senate Finance Committee, Washington, DC, December 1996.
  2. U.S. Social Security Administration. The 1998 Report of the Board of Trustees of the Old Age Survivors Insurance and Disability Insurance Trust Fund, Washington, DC: U.S. Government Printing Office, April 1998.
  3.  Stewart, Ken.,"Planned Changes in the Consumer Price Index Formula," Accessibility Information, April 16, 1998, Home Page, The Bureau of Labor Statistics (, Washington DC.
  4. see reference 1.
  5. Moulton, Brent R., "Biases in the Consumer Price Index: What is the Evidence?" Working Paper 294, U.S. Dept. of Labor, Bureau of Labor Statistics, Washington DC, October 1996.
        ______ , "Anatomy of Price Change. Basic Components of the CPI: Estimation of Price Changes." Monthly Labor Review, December 1993.
  6. The CPI-U is used to adjust poverty thresholds, income tax brackets, and many benefit programs.
  7. The CPI for Urban Wage Earners and Clerical Workers (CPI-W), which covers 32 percent of the population, is used to calculate the Cost of Living Adjustments (COLAs) for Social Security, SSI, and federal retirement programs.
  8. The current index, Sum PtiQ0i / Sum P0i0i , is based on fixed base year quantities, where Pti is the price of item i in year (t), P0i is its price in the base year (0), and Q0i is the quantity of item i in the base year (0). The numerator is the sum of consumer expenditures on all items in year (t) based on "base year quantities" Q0i s, and the denominator is the sum of the consumer expenditures in year (0) again based on "base year quantities." The geometric mean index is defined as ÏÏ (Pti/P0i)Si , where ÏÏ denotes the product of price ratios raised to the power Si, the expenditure share of item i (for example, (Pt1/P01)S1. (Pt2/P02S2. (Pt3/P03)S3... and so on). The expenditure share Si is the ratio of consumer expenditure on item i to total consumer expenditure on all items. This proportion usually does not change when substitutions are made within items.
  9. The cross-price elasticity of demand is defined as the percentage change in demand for one commodity in response to the percentage change in price of another commodity.
  10. AARP is a licensed user of The Washington University Macroeconomic Model, provided by Macroeconomic Advisers, LLC, St. Louis.
  11.  All dollar figures are current dollars.
  12. The discount rate is a proxy of the interest rate on 30-year long-term government bonds, which is often used as the discount rate in the long-run cost and benefit calculations. Since the discounted benefit loss is in present value terms, it needs no adjustment for inflation.
  13. Federal reserve profit is interest earned on the Fed's portfolio of government securities. The Fed's holdings of government securities are measured by the stock of non-borrowed reserves.
  14. Gross private saving comprises personal saving and business saving. The share of personal saving is about 3 percent in gross private saving.
  15. U.S. Bureau of Economic Analysis, National Income and Product Accounts, Survey of Current Business, August 1998.
  16. The implications of this exclusion are discussed at length in a separate Issue Paper, forthcoming.
  17. Browne, Lynn E. and Gleason, J.,"The Saving Mystery, or Where Did the Money Go?" New England Economic Review, September/October 1996.

Written by Satyendra Verma, Ph.D., Economics Team, AARP Public Policy Institute

April 1999
©1999 AARP
May be copied only for noncommercial purposes and with attribution; permission required for all other purposes.
Public Policy Institute, AARP, 601 E Street, NW, Washington, DC 20049

AARP In Your State

Visit the AARP state page for information about events, news and resources near you.