Introduction

The Social Security Board of Trustees projects the combined trust funds' assets of the program to be depleted in 2034. After this, the program, barring congressional action, would operate as a strict pay-as-you go system than can pay out in benefits only what it receives in revenue. Under current law and the trustees' best estimates, continuing program tax revenue will be sufficient to pay about 80% of scheduled benefits after the asset reserves held in the trust funds are depleted.

As discussed later in this report, the trustees' past 12 annual reports have projected the combined trust funds to be depleted at some point during the 2033-2035 time period. The annual projections rely on a wide range of economic, demographic, and program-specific factors. Economic factors include productivity, price inflation, unemployment, and growth in gross domestic product (GDP); demographic factors include fertility, mortality, and immigration; and program-specific factors include covered and taxable earnings, revenues from the taxation of benefits, and average benefits indexed to growth in average national wages. Assumptions across this wide range of factors are used in models to project the program's future financial status for the short-range and long-range time periods. The most recent 2023 annual report projects the program to fail both the short-range and long-range tests for financial adequacy.

Congress has demonstrated that Social Security's financial status is of high concern. Social Security is commonly referred to as the federal government's largest program in terms of both the number of workers covered (about 183 million in 2023) and the number of beneficiaries (over 66 million in June 2023).¹ Monthly Social Security benefits constitute a substantial portion of income for a large segment of recipients.² Recognizing this, past Congresses have introduced a variety of bills to address the program's future financial status.

This report first provides background on the program before it presents Social Security's short-range and long-range actuarial estimates. Next, the report discusses uncertainty in those projections and alternative scenarios provided by the trustees to illustrate the uncertainty of future conditions. Lastly, the report analyzes the accuracy of past projections of the future financial status of the combined trust funds.

Background

Social Security is a self-financing program that provides monthly cash benefits to retired or disabled workers and their family members and to the family members of deceased workers.³ As of June 2023, there were approximately 66.6 million Social Security beneficiaries. Of those, 52.1 million (78.2%) were retired workers and family members, 8.7 million (13.1%) were disabled workers and family members, and 5.8 million (8.8%) were survivors of deceased workers.⁴

Social Security is financed primarily by payroll taxes paid by covered workers and their employers. Employers and employees each pay 6.2% of covered earnings, up to an annual limit, and self-employed individuals pay 12.4% of net self-employment income, up to an annual limit. The annual limit on taxable earnings is $160,200 in 2023.⁵ Social Security is also credited with tax revenues from the federal income taxes paid by some beneficiaries on a portion of their benefits. In addition, Social Security receives interest income from Social Security trust fund investments. Social Security income and cost are accounted for in two separate trust funds authorized under Title II of the Social Security Act: the Federal Old-Age and Survivors Insurance (OASI) Trust Fund and the Federal Disability Insurance (DI) Trust Fund.⁶ This report refers to the separate OASI and DI trust funds on a combined basis as the Social Security trust funds.⁷ In 2022, the combined Social Security trust funds (OASDI) had total receipts of $1.22 trillion, total expenditures of $1.24 trillion, and accumulated holdings (assets) of $2.83 trillion.⁸

Annual Reports and Projection Periods

The OASI trust fund and the DI trust fund are overseen by a board of trustees. The board is composed of the Secretary of the Treasury (the managing trustee), the Secretary of Labor, the Secretary of Health and Human Services, the commissioner of Social Security, and two public trustees.⁹

The trustees are required by law to report to Congress annually on the financial status of the Social Security trust funds. Among other duties, Section 201(c)(2) of the Social Security Act requires the trustees to

[r]eport to the Congress not later than the first day of April of each year on the operation and status of the Trust Funds during the preceding fiscal year and on their expected operation and status during the next ensuing five fiscal years.¹⁰

While the Social Security Act requires the trustees to report on the projected operations of the Social Security trust funds for a five-year period, the trustees currently report 10-year (short-range) and 75-year (long-range) projections of the financial status of the trust funds. Projections covering 75-year periods first appeared in the 1965 Social Security trustees report.¹¹

Although current law does not direct the trustees to perform long-range projections for 75-year periods, all annual reports have used 75 years since the practice was first adopted in 1965.¹² Additionally, statute requires the trustees to include a statement of actuarial status and determine if the trust funds are in close actuarial balance (as defined by the trustees and discussed in the sections below on short-range and long-range actuarial estimates).¹³ Income rates and cost rates for a valuation period are used to calculate the actuarial balance (discussed below). For this calculation, the trustees define the long-term valuation period to be 75 years, which covers approximately the maximum remaining lifetime for virtually all current Social Security participants (i.e., covered workers and beneficiaries).¹⁴

Assumptions¹⁵

In each annual report, the trustees present three alternative sets of assumptions for demographic, economic, and program-specific factors.¹⁶ The low-cost set of assumptions represents a future experience that is the most advantageous to the program's financial status. The high-cost set of assumptions represents a future experience that is the least advantageous to the program's financial status. As the trustees state: "These alternatives are not intended to suggest that all parameters would be likely to differ from the intermediate values in the specified directions, but are intended to illustrate the effect of clearly defined scenarios that are, on balance, very favorable or unfavorable for the program's financial status."¹⁷ In actual experience, it is unlikely that all demographic, economic, and program-specific factors move in a manner that is either favorable or unfavorable to the program's financial status. Thus, the trustees use the intermediate set of assumptions to illustrate their best guess as to the future experience. This report, therefore, focuses on the intermediate set of assumptions.

The Trust Funds

In the context of federal program accounting, a trust fund is an accounting mechanism that allows a program to track revenues and expenses. Additionally, a trust fund provides a means for a program (e.g., Social Security) to hold any accumulated assets—that is, money not immediately needed to pay benefits—for the payment of future benefits.¹⁸

Under current law, there are two separate trust funds for the Social Security program: (1) the OASI trust fund and (2) the DI trust fund. Monies credited to each trust fund cannot be lent or transferred to the other trust fund without authorization from lawmakers. OASI benefits can be paid only from the OASI trust fund, and DI benefits can be paid only from the DI trust fund.

For the purposes of this report, the trust funds—and other program information—will be considered on a combined, hypothetical basis. Essentially, this suggests that data and measures presented illustrate a weighted averaged of the two separate funds. Naturally, this implies that most data and measures will be more weighted to OASI, as this program is larger in terms of annual costs, annual income, and asset reserves held in its trust fund. At the end of 2022, the combined value of the trust funds was $2.83 trillion. The OASI trust fund accounted for 95.8% of the combined OASDI trust fund value.¹⁹

The Office of the Chief Actuary (OCACT)

In practice, the trustees do not perform the modelling functions required to make projections for the status of the trust funds. Rather, that function is performed by SSA's Office of the Chief Actuary.²⁰ OCACT's mission—including its support for the trustees—is outlined on SSA's organization structure website:

The Office of the Chief Actuary (OCACT) plans and directs a program of actuarial estimates and analyses pertaining to the SSA-administered retirement, survivors and disability insurance programs and supplemental security income program and to projected changes in these programs. Evaluates operations of the Federal Old-Age and Survivors Insurance Trust Fund and the Federal Disability Insurance Trust Fund; estimates future operations of the trust funds; conducts studies of program financing; performs actuarial and demographic research on social insurance and related program issues; and estimates future workloads. Provides technical and consultative services to the Commissioner, the Board of Trustees of those two Trust Funds, and, as requested, congressional committees. Appears before congressional committees to provide expert testimony on the actuarial aspects of Social Security issues.²¹

Specifically, SSA's organizational manual specifies that OCACT's Office of Long-Range Actuarial Estimates is responsible for planning, directing, and coordinating long-range cost estimates for the retirement, survivors, and disability program (i.e., Social Security) both under current provisions and proposed changes in law or regulation.²² OCACT's website states that its Office of Long-Range Actuarial Estimates is responsible for estimates for up to 75 years in the future.²³

Short-Range Actuarial Estimates

For the short-range (10-year) period, the trustees measure the program's financial adequacy using the trust fund ratio—that is, the trust fund's asset reserves at the beginning of a year expressed as a percentage of the projected total cost for the year. For instance, a trust fund ratio of 200% indicates that two years of projected benefits could be paid with asset reserves absent any additional income.²⁴ The trustees state that maintaining a trust fund ratio of at least 100% is a good indication that the trust funds can cover most short-range contingencies.²⁵ A test of the program's short-range financial adequacy is satisfied if (1) the projected trust fund ratio is at least 100% at the beginning of the 10-year period and remains so for the 10-year period or (2) the ratio is below 100% at the beginning of the 10-year period but is projected to reach at least 100% within five years and remain at least 100% for the remainder of the 10-year period.²⁶

Figure 1 shows the most recent projections for the trust fund ratio on a combined basis under the intermediate assumptions. The Social Security program does not meet the trustees' criteria for short-range financial adequacy for the current 10-year period. The trust fund ratio is projected to fall below 100% by 2029 and remain below 100% for the remainder of the current 10-year period.²⁷

Figure 1. Social Security Trust Fund Ratio, 1975-2035 On a Combined Basis Under the 2023 Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Table IV.A3, pp. 49-50, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B4, https://www.ssa.gov/OACT/TR/2023/lr4b4.html. Notes: For comparison purposes, Figure A-1 includes the Congressional Budget Office's 2023 Long-Term Projection for Social Security's estimates for combined trust fund ratios.

Long-Range Actuarial Estimates

For the long-range (75 year) period, the trustees assess the program's financial adequacy using three measures: (1) annual cash-flow measures (i.e., income rates, cost rates, balances); (2) trust fund ratios; and (3) summary measures (i.e., actuarial balances). Cash-flow measures and summary measures are commonly expressed as a percentage of taxable payroll or GDP.²⁸ A test of the program's long-range close actuarial balance is satisfied if (1) the trust fund meets the definition of short-range financial adequacy and (2) the trust fund ratio is projected to remain above zero throughout the long-range (75-year) period. As shown in Figure 1, the first condition of the test is not satisfied (i.e., the trust fund ratio falls below 100%), nor is the second condition (i.e., the trust fund ratio is projected to remain at 0% throughout the 75-year projection period). The trust fund ratio is projected to fall below 100% in 2029 and fall below 0% (i.e., exhaustion of assets held in the trust funds) in 2034.²⁹

Annual Cash-Flow Measures

Annual cash-flow measures are one set of measures the trustees use to assess the program's long-range financial adequacy. Figure 2 presents the historical and projected income rates, cost rates, and annual balances as a percentage of taxable payroll under the trustees' intermediate projections.³⁰ As shown in Figure 2, the projected income rate is relatively stable, whereas the projected cost rate is expected to increase until the late 2070s before a relatively small decrease. Although the cost rate is projected to stabilize, and partially decline, for the remainder of the long-range period, it is expected to remain larger than the income rate.

As observed in Figure 2, the difference between income and cost rates—or balance—is largely driven by increases in projected costs. The projected income rates, expressed as a percentage of taxable payroll, are relatively more stable than the projected cost rates. Since the income rates are expressed as a percentage of taxable payroll, and the combined Social Security payroll tax is fixed under current law at 12.4% of taxable (i.e., covered) earnings, any remaining variation is caused by assumptions in taxation of benefits. The taxation of benefits is a relatively small portion of income.³¹

Figure 2. Social Security Cash-Flow Measures as a Percentage of Taxable Payroll, 1975-2097 On a Combined Basis Under the 2023 Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 2023, Table IV.B1, pp. 56-57, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B1, https://www.ssa.gov/OACT/TR/2023/lr4b1.html. Note: For comparison purposes, Figure A-2 includes the Congressional Budget Office's 2023 Long-Term Projection for Social Security's estimates for combined income rates, cost rates, and balances.

The trust fund ratio is expected to fall below 0% in 2034. Depletion of the trust fund reserves means that the trust funds could no longer augment continuing income in the payment of scheduled benefits. As shown in Figure 2, at the time of projected trust fund depletion, the program's project cost rate exceeds that of its projected income rate. Thus, in 2034, the program is not projected to have sufficient income to support the payment of full scheduled costs (i.e., monthly benefit payments). Under the intermediate assumptions, the trustees project that program income will be sufficient to cover about 80% of scheduled benefits in 2034. (This percentage of payable benefits would fall to 74% by 2097.³²)

The trustees attribute the projected rising costs to demographic factors: "Under the intermediate assumptions, demographic factors by themselves cause the projected cost rate to rise rapidly for the next two decades."³³ The historical and projected cost rates are, in effect, representative of the age distribution of the Social Security population. As the trustees further state, "The cost rate is essentially the product of the number of beneficiaries and their average benefit, divided by the product of the number of covered workers and their average taxable earnings."³⁴ To illustrate this relationship, the trustees present the number of covered workers per beneficiary (i.e., the ratio of the number of people paying into Social Security to the number of people getting paid by Social Security).

Figure 3 shows the historical and projected number of covered workers to Social Security beneficiaries under the intermediate assumptions. As can be seen, the decline in this ratio reflects an inverse relationship to the cost rate presented in Figure 2. Said differently, the projected increase in the cost rate reflects a projected decrease in the number of covered workers to beneficiaries.

Figure 3. Social Security Covered Workers to Beneficiaries, 1975-2097 On a Combined Basis Under the 2023 Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Table IV.B3, pp. 64-65, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B3, https://www.ssa.gov/OACT/TR/2023/lr4b3.html. Note: For comparison purposes, Figure A-3 includes the Congressional Budget Office's 2023 Long-Term Projection for Social Security's estimates for the ratio of Social Security–covered workers to beneficiaries.

Summary Measures

The trustees use two summary measures to assess the program's long-range financial adequacy. The first measure is the actuarial balance. The actuarial balance is the difference between the summarized income rate and the summarized cost rate, expressed as a percentage of taxable payroll.³⁵ In this context, the summarized income rate is the ratio of the present value of non-interest income to the present value of taxable payroll for the 75-year period.³⁶ Similarly, the summarized cost rate is the ratio of the present value of cost to the present value of taxable payroll for the 75-year period. Alternatively, the actuarial balance can also be thought of as the summarized balance. That is, the actuarial balance is the sum of the differences between the projected income and projected cost in Figure 2, discounted to present values. Or, more simply, it is the difference between the summarized income and cost rates.

Under the 2023 intermediate assumptions, the trustees estimate the actuarial balance to be -3.61% of taxable payroll (see Figure 4).³⁷ This measure represents the change in income or cost that would be required to achieve program balance over the 75-year projection period and to achieve a trust fund reserve equal to one year's projected cost by the end of the period (i.e., a trust fund ratio of 100%). For instance, the program being in balance would require program costs to decrease by 3.61% of taxable payroll or program revenues to increase by 3.61% of taxable payroll.

Figure 4. Social Security Actuarial Balance as a Percentage of Taxable Payroll, 1982-2023 On a Combined Basis Under the 2023 Intermediate Assumptions

Source: Sharon Chu and Kyle Burkhalter, Disaggregation of Changes in the Long-Range Actuarial Balance for the Old-Age, Survivors, and Disability Insurance (OASDI) Program Since 1983, Social Security Administration, Office of the Chief Actuary, March 2023, https://www.ssa.gov/OACT/NOTES/ran8/an2023-8.pdf.

The second measure the trustees use to assess the program's long-range financial adequacy is the cumulative present value of the projected financial shortfall (i.e., program income less cost). A positive measure for this statistic represents a positive trust fund balance, whereas a negative measure for this statistic represents an unfunded obligation.³⁸ As discussed earlier, the value of the trust fund is projected to remain positive through 2033. Under the intermediate assumptions, the trustees project the trust funds to be exhausted sometime in 2034. The trustees state: "Through the end of 2097, the combined funds have a present value unfunded obligation of $22.4 trillion."³⁹

Figure 5. Cumulative Present Value of the Social Security Projected Financial Shortfall, 2022-2097 Projected on a Combined Basis Under the 2023 Intermediate Assumptions and in Trillions of Dollars

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Figure II.D5, p. 19, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and https://www.ssa.gov/OACT/TR/2023/LD_figIID5.html. Notes: Present values as of January 1, 2023.

Uncertainty in Projections

Taken together, Figure 2, Figure 4, and Figure 5 show that under the trustees' latest intermediate assumptions, the program's financial adequacy is projected to decline under current law. All projections come with a degree of uncertainty. To help illustrate the uncertainty, the trustees use three different sets of assumptions in their annual reports. For example, Figure 6 shows the historical trust fund ratio and projected trust fund ratios under the low-cost, intermediate, and high-cost alternative scenarios. As discussed, the intermediate projections represent the trustees' best estimate, while the low-cost and high-cost scenarios help to present a range of possible outcomes. To accomplish this range, assumptions are presumed to all be either advantageous or disadvantageous to the financial position. As the trustees state

The low-cost alternative includes a higher ultimate total fertility rate, slower improvement in mortality, higher real wage growth, a higher ultimate real interest rate, a higher ultimate annual change in the CPI [consumer price index], and a lower unemployment rate. The high-cost alternative, in contrast, includes a lower ultimate total fertility rate, more rapid improvement in mortality, lower real wage growth, a lower ultimate real interest rate, a lower ultimate annual change in the CPI, and a higher unemployment rate.⁴⁰

Figure 6. Social Security Trust Fund Ratios Under Alternative Scenarios, 1990-2100 On a Combined Basis Under the 2023 Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Figure II.D6, p. 20, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and https://www.ssa.gov/OACT/TR/2023/LD_figIID6.html. Note: The above figured is replicated on an individual trust fund basis in Figure A-4.

While the three sets of long-range projections highlight the uncertainty in these estimates, this approach does not provide any indication of the probability that the future financial status of the trust fund is within or outside the range of these estimates. That is, this deterministic modelling approach does not indicate the likelihood for each set of assumptions.

In order to help assign probabilities to possible outcomes, annual reports include outcomes from 5,000 stochastic simulations. The stochastic approach is built on thousands of independent simulations where the values of the assumptions are allowed to vary. The distribution of these simulation outcomes is then used to determine the probability of solvency occurring within a range of years or the probability of key trust fund indicators falling within a particular numerical range. The stochastic results for the combined trust fund ratios are presented in Figure 7. The results suggest that, with 95% confidence, the combined trust funds would be exhausted between 2031 and 2040.⁴¹ Furthermore, they suggest that the low-cost and high-cost scenarios are "very unlikely."⁴² Figure 7 shows lines for the 2.5^th, 50.0^th, and 97.5^th percentiles of the estimated annual combined trust fund ratios. For example, the line representing the 2.5^th percentile reaches zero in 2031, which indicates that 97.5% of the 5,000 simulations resulted in combined trust fund ratios that remained positive through at least 2031, whereas 2.5% of the 5,000 simulations result in trust fund ratios that reached zero before 2031. Alternatively, the line representing the 97.5^th percentile reaches zero in 2040, which indicates that 2.5% of the 5,000 simulations resulted in combined trust fund ratios that remained positive beyond 2040, while 97.5% of the 5,000 simulations resulted in combined trust fund ratios that reached zero before 2040. The line representing the 50^th percentile indicates the median combined reserve depletion date (2033). For comparison purposes, the intermediate projection for the combined trust fund ratio is added to Figure 7. Under the intermediate assumptions, the projected date for combined trust fund depletion is "mid-2034."⁴³

Figure 7. Projected Social Security Trust Fund Ratios, 2023-2045 Using Stochastic Simulations

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Figure II.D7, p. 22, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and https://www.ssa.gov/OACT/TR/2023/LD_figVIE2.html.

Accuracy of Projections

The accuracy of future projections of Social Security's financial status is unknown. However, the accuracy of some past projections is calculated and presented in this section. For instance, the actual trust fund ratio for 2030 is unknown at this point. This value will be published in the eventual 2031 annual report. The actual trust fund ratio for 2020 is known, and (intermediate) projections from prior annual reports can be used to determine how accurately the 2020 trust fund ratio was projected. Moreover, using prior annual reports from 1983 through 2023 allow the projections to be assessed over time. For example, in the 1990 annual report, the intermediate projection for the 2020 trust fund ratio was 30 years in the future, in the 2000 annual report it was 20 years in the future, and in the 2010 annual report it was 10 years in the future. In general, uncertainty increases for projections further in the future. That is, it could be expected that the projections for the 2020 trust fund ratio was more accurate in 2010, less accurate in 2000, and even less accurate in 1990. This report assesses the projection accuracy for selected measures of Social Security's financial status starting with the 1983 annual report.⁴⁴

Possible Causes of Inaccuracy

This section discusses several, albeit non-exhaustive, possible causes for projection inaccuracy. The annual reports reflect the trustees' understanding at the beginning of a calendar year. For instance, the 2020 annual report was published on April 22, 2020, and the intermediate assumptions reflected the trustees' best guess as to the future experience starting at the beginning of the year. Therefore, the 2020 annual report did not reflect the potential changes from the COVID-19 pandemic (e.g., the pandemic-induced recession).⁴⁵ Said differently, the projections reflect a best guess at a specific date. The spectrum of events that may have significant effects on the Social Security program after the specific date of said best guess is unknown. For example, the projections in the 2000 annual report could not account for the 2007-2009 recession, the pandemic-induced recession in 2020, or legislative responses to each of those events. Given this, the uncertainty for projections further into the future is higher.

Although time may be the most obvious source of projection inaccuracy, it is not the only reason. One major source of projection inaccuracy is changes in legislation or regulations. Furthermore, these laws need not change the specific laws and regulations governing Social Security. For instance, the 1991 annual report credits the Immigration Act of 1990 (P.L. 101-649)—a law modifying the general immigration policy—with having a "significant effect" on Social Security.⁴⁶ However, there have been changes in law that directly affect Social Security. For instance, as part of the Omnibus Budget Reconciliation Act of 1986 (P.L. 99-509), lawmakers changed the cost-of-living adjustment (COLA) rules to effectively allow COLAs smaller than 3% to be payable. Since then, the change has accounted for 21 annual COLAs that would not have otherwise increased benefits (i.e., program costs).⁴⁷ Another source of projection inaccuracy is changes in methods and data collection. For instance, the 2021 annual report highlighted several changes in methods and data that contributed to a change in projected financial status. First, a new fertility model resulted in the ultimate fertility rate being reached 25 years later than projected in the prior year's report. Second, updates to the civilian labor force model to reflect new economic data lowered projected labor force participation rates. Third, updates to the methodology for projecting average benefit levels and Treasury tax information indicated future lower levels of revenue from the taxation of Social Security benefits. In this case, all three changes resulted in a worsening financial status.⁴⁸

Congressional Budget Office and Projection Accuracy

Since 2005, the Congressional Budget Office (CBO) has published its Long-Term Projections for Social Security. CBO's estimates use internally developed assumptions and utilize the CBO Long-Term Model,⁴⁹ which is used to produce projections for the economy and federal budget that extend beyond CBO's usual 10-year projection period.⁵⁰ CBO's assumptions and modeling process generally result in cost projections that are higher than the trustees' annual reports (see Figure A-2).⁵¹ This characteristic, among others, typically results in CBO projecting fewer years until depletion of the asset reserves held in the trust funds (see Figure A-1).

The structure of CBO's Long-Term Projections for Social Security has changed over time. For instance, many of the measures used to assess short-term and long-term actuarial status of the trust funds (e.g., income and cost rates as a percentage of taxable payroll) did not appear in the annual outlooks until 2015.⁵² Given this, CBO's projections are not analyzed in this report. However, in early 2023, CBO evaluated its projections for Social Security outlays and attributed differences between forecasted and experienced costs to "misestimates of future COLAs."⁵³ Similar to the trustees' estimates, CBO noted that misestimates were generally higher for relatively longer-term projections than for shorter-term projections. Also, the CBO analysis found that their projections were more likely to overestimate than underestimate Social Security outlays.

Measures of Projection Accuracy

For each intermediate projection of income rate, cost rate, worker-to-benefit ratio, and trust fund ratio—the past projections can be measured against historical (actual) values—this report presents and examines three measures of forecast accuracy. The measures are the mean absolute deviation (MAD), mean squared error (MSE), and mean absolute percentage error (MAPE). Although there are many measures of prediction (i.e., projection) error, these three were chosen for their relative simplicity in interpretation.

The MAD is the sum of the absolute differences between the actual value and the projected value, divided by the number of observations. This statistic measures how the projected values have deviated from actual values on average. The lower the MAD, the better the projection accuracy. However, one drawback to using the MAD to assess forecast accuracy is that possible large errors—actual value less projected value—may be averaged out over a larger number of observations.

The MSE helps to add context for this issue. The MSE is the sum of squared differences between actual value and projected value, divided by the number of observations. Because this calculation involves squared error terms, larger differences between actual and projected values have a greater impact than do smaller differences. Similarly, the lower the MSE, the better the projection accuracy. Additionally, the closer the MSE is to the MAD, the less likely relatively large errors were experienced. Conversely, a large difference between MSE and MAD suggests that large errors played a role.

Lastly, the MAPE is calculated as the average of absolute percentage differences between the actual value and projected value. Similarly, a lower MAPE reflects more accurate projections, whereas a higher MAPE reflected less accurate projections. Expressing an error in percentage terms also accounts for the relative size of the projected measure. The formulas for each statistical measure of projection accuracy can be found in Table 1.

Results

Table 1 shows the three statistical measures—MAD, MSE, and MAPE—for selected intermediate projections for the Social Security financial status. The table includes statistical measures for those where past projections can be measured against actual values: income rate, cost rate, worker-to-beneficiary ratio, and trust fund ratio. Additionally, the statistical measures of projection accuracy are shown by year. That is, accuracy is assessed for projections one year in the future, two years in the future, three years in the future, and so on.

As shown in Table 1, the number of observations (i.e., projections) available for analysis decrease further out in the future. This first annual report analyzed is the 1983 annual report, which presented a projected combined trust fund ratio for 2060. However, only that report's projection through 2022 can be assessed with a (now) historical value.⁵⁴ The table shows fewer projections for more long-range projections. Also, not all annual reports presented single-year projections. For instance, the 1985 annual report projected a combined trust fund ratio for the next 25-year period and then every five years until projected trust fund exhaustion, whereas the 1986 annual report projected a combined trust fund ratio for the next 10-year period and then every five years until projected trust fund exhaustion. The following discussion will focus on projection accuracy for those projections less than 15 years in the future. (As discussed, the 2023 annual report has projected the combined trust fund depletion in 2034.)

The first set of columns in Table 1 shows the statistics for the income rate. This report analyzed 40 predictions for income rates one year in the future (e.g., the 1983 annual report's projections for 1984, the 1984 annual report's projections for 1985, etc.). Over these 40 observations, the MAD was 0.10 and the MSE was 0.02. This suggests that the income rate projections one year in the future were relatively close to the actual values and relatively consistent. Since the MSE uses squared error terms, an MSE less than the MAD implies that most of the error terms were less than one. The MAPE for one-year income rate projections was 0.80%. This report analyzed 15 predictions for income rates 14 years in the future (e.g., the 1983 annual report's projections for 1997, the 1984 annual report's projections for 1998, etc.). As one might suspect, the forecast accuracy for the income rate declines the further into the future. Over 15 instances of projections for 14 years into the future, the MAD and MSE were 0.26 and 0.11, respectively. This suggests that income rate projections for 15 years in the future were also relatively close to actual values and relatively consistent (i.e., no large errors). The MAPE for 15-year income rate projections was 2.05%.

The second set of columns in Table 1 shows the statistics for the cost rate. This report analyzed 40 predictions for cost rates one year in the future. Over these 40 observations, the MAD was 0.17 and the MSE was 0.05. This suggests that the cost rate projections one year in the future were relatively close to the actual values and relatively consistent. The MAPE for one-year cost rate projections was 1.37%. Although this MAPE is relatively higher than the income rate MAPE, it does suggest a relatively high projection accuracy for cost rates one year in the future. As suspected, the forecast accuracy for cost rates declines the further into the future. Over 14 instances of projections for 15 years into the future, the MAD and MSE were 0.63 and 0.49, respectively. This suggests that cost rate projections for 15 years in the future were also relatively close to actual values and relatively consistent (i.e., no large errors). The MAPE for 15-year income rate projections was 4.86%.

Table 1 also shows that the projection errors for the worker-to-beneficiary ratio are relatively small and relatively consistent over time. For instance, over 40 observations for projections one year in the future, the MAD and MSE were 0.03 and 0.00, respectively. As stated previously, a lower MSE than MAD suggests no large errors. The MAPE for projections one year in the future is also relatively low. As suspected, the projection accuracy is lower for projections 15 years in the future. Over 11 observations for worker-to-beneficiary ratios 15 years in the future, the MAD was 0.12 and the MSE was 0.02. The MAPE for the worker-to-beneficiary ratio 15 years in the future was 4.09%.

Lastly, Table 1 displays the projection accuracy statistics for the trust fund ratio. Over 40 observations for the ratio one year in the future, the MAD is 3.43 and the MSE is 22.38. For all projection periods, the MSE is larger than the MAD. For the 14 projections for the trust fund ratio 15 years in the future, the MAD is 105.57 and the MSE is 13,402.85. This suggests that for the trust fund ratio projections, the accuracy has been affected by larger errors. Given this, it is not surprising that the MAPE for one year and 15 years in the future is 2.50% and 39.59%, respectively.

Conclusion

The Social Security system is facing a projected financial shortfall. Under the 2023 intermediate assumptions, the program's financial status fails the trustees' tests for both short-range and long-range financial adequacy. This failure is largely reflective of the program's projected rising costs and the related imbalance between covered workers (i.e., those paying into the system) and beneficiaries (i.e., those receiving payments from the system).

Although uncertainty exists in the magnitude of the projected shortfall, past trustees' projections are shown in this report to be more accurate in the short-run than in the long-run. That is, as the program approaches depletion of the asset reserves held in the trust funds, the projections of the magnitude of the shortfall become more accurate. In the trustees' 2023 annual report, the depletion of trust fund assets results in a de facto benefit reduction of about 20%. Additionally, as the program approaches the depletion date, the projections of the date itself become more accurate. As shown in Figure 8, the projected year of combined trust fund depletion has become more concentrated on the 2033-2035 window. In the past 10 annual reports, most have projected 2034 as the year for asset depletion.

Figure 8. Projected Years Until Combined Social Security Trust Fund Depletion Under the Trustees' Intermediate Assumptions, 1984-2023

Source: Congressional Research Service (CRS).

Appendix. Supporting Information

Formulas for Statistical Measures of Projection Accuracy

Mean Absolute Deviation MAD=|Actual-Projected|n

Mean Squared Error MSE=(Actual-Projected)2n

Mean Absolute Percentage Error MAPE=Actual-ProjectedProjectedn×100

Figure A-1. Historical and Projected Social Security Trust Fund Ratios, 1975-2035 The 2023 Congressional Budget Office (CBO) and Social Security Trustees Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Table IV.A3, pp. 49-50, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B4, https://www.ssa.gov/OACT/TR/2023/lr4b4.html; and CBO, CBO's 2023 Long-Term Projections for Social Security, June 29, 2023, supplemental Information, table 8, https://www.cbo.gov/publication/59184.

Figure A-2. Social Security Cash Flow Measures as a Percent of Taxable Payroll, 1975-2097 The 2023 Congressional Budget Office (CBO) and Social Security Trustees Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Table IV.B1, pp. 56-57, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B1, https://www.ssa.gov/OACT/TR/2023/lr4b1.html; and CBO, CBO's 2023 Long-Term Projections for Social Security, June 29, 2023, supplemental Information, table 8, https://www.cbo.gov/publication/59184.

Figure A-3. Social Security Covered Workers to Beneficiaries, 1975-2097 The 2023 Congressional Budget Office (CBO) and Social Security Trustees Intermediate Assumptions

Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Table IV.B3, pp. 64-65, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf and supplemental single-year Table IV.B3, https://www.ssa.gov/OACT/TR/2023/lr4b3.html; and CBO, CBO's 2023 Long-Term Projections for Social Security, June 29, 2023, Projections Underlying Social Security Estimates, https://www.cbo.gov/publication/59184.

Figure A-4. Individual Social Security Trust Fund Ratios Under Alternative Scenarios, 1990-2100 On a Combined Basis Under the 2023 Intermediate Assumptions	Source: 2023 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds, March 31, 2023, Figure II.D7, p. 21, https://www.ssa.gov/OACT/TR/2023/tr2023.pdf.