2021 OASDI Trustees Report

skip to main content
Table of Contents Previous Next Tables Figures Index

V. ASSUMPTIONS AND METHODS UNDERLYING
ACTUARIAL ESTIMATES
The future income and cost of the OASDI program will depend on many demographic, economic, and program-specific factors. Trust fund income will depend on how these factors affect the size and composition of the working population as well as the level and distribution of earnings. Similarly, program cost will depend on how these factors affect the size and composition of the beneficiary population as well as the general level of benefits.
The Trustees make basic assumptions for several of these factors based on analysis of historical trends, historical conditions, and expected future conditions. These factors include fertility, mortality, immigration, marriage, divorce, productivity, inflation, average earnings, unemployment, real interest rates, and disability incidence and termination. Other factors depend on these basic assumptions. These other, often interdependent, factors include total population, life expectancy, labor force participation, gross domestic product, and program-specific factors. Each year the Trustees reexamine these assumptions and methods in light of new information and make appropriate revisions. The assumptions for this report were selected by the middle of May 2021.
Future levels of these factors and their interrelationships are inherently uncertain. To address these uncertainties, this report uses three sets of assumptions, designated as intermediate (alternative II), low-cost (alternative I), and high-cost (alternative III). The intermediate set represents the Trustees’ best estimate of the future course of the population and the economy. With regard to the net effect on the actuarial status of the OASDI program, the low-cost set is more optimistic and the high-cost set is more pessimistic. The low-cost and high-cost sets of assumptions reflect significant potential changes in the interrelationships among factors, as well as changes in the values for individual factors.
While it is unlikely that all of the factors and interactions will differ in the specified directions from the intermediate values, many combinations of individual differences in the factors could have a similar overall effect. Outcomes with overall long-range cost as low as the low-cost scenario or as high as the high-cost scenario are very unlikely. This report also includes a section on sensitivity analysis, where factors are changed one at a time (see appendix D), and a section on stochastic projections, which provides a probability distribution of possible future outcomes, with most of the key factors being varied around the intermediate alternative (see appendix E).
Readers should interpret with care the estimates based on the three sets of alternative assumptions. These estimates are not specific predictions of the future financial status of the OASDI program. Rather, they provide a reasonable range of future income and cost.
All of the key demographic, economic, and program-specific assumptions reach their long-range ultimate values within the next 25 years. For extrapolations beyond the 75‑year long-range period, the ultimate levels or trends reached by the end of the 75‑year period remain unchanged. The assumed ultimate values represent average annual experience or growth rates. Actual future values will exhibit fluctuations or cyclical patterns, as in the past.
At this time, there is no consensus on what the lasting effects of the COVID-19 pandemic on long-term experience might be, if any. Given the unprecedented level of uncertainty, the Trustees have assumed that the pandemic will have no net effect on the individual long-range ultimate assumptions. However, pandemic effects are reflected in the short-range (first 10-year) period, as described in the remainder of this chapter.
The following sections briefly discuss the various assumptions and methods used in making the estimates of trust fund actuarial status, which are the focus of this report.1 There are, of course, many interrelationships among these factors that are important but are beyond the scope of this discussion.
A. DEMOGRAPHIC ASSUMPTIONS AND METHODS
This section of the report provides a brief overview of the demographic historical data and the assumptions used for the projections.
1. Fertility Assumptions
Birth rates by single year of age, for girls and women aged 14 to 49,2 are the basis for the fertility assumptions. These rates apply to the total number of women, across all marital statuses, in the midyear population at each age. Table V.A1 displays the historical and projected total fertility rates.3
Historically, birth rates in the United States have fluctuated widely. The total fertility rate decreased from 3.31 children per woman at the end of World War I (1918) to 2.15 during the Great Depression (1936). After 1936, the total fertility rate rose to 3.68 in 1957 and then fell to 1.74 by 1976. After 1976, the total fertility rate rose above 2.00 by 1990, where it generally remained through 2009, but dropped below 1.90 for 2011 and has been at relatively low levels since then. It is estimated to be 1.65 for 2020.
The Trustees assume that the COVID-19 pandemic will affect birth rates in years 2021 through 2026. Under the intermediate assumptions, the total fertility rate is projected to decrease to a level of 1.54 for 2021 and 1.62 for 2022. The total fertility rate for 2023 is projected to be 1.71 children per woman, which is the rate that would have been projected in the absence of the pandemic. The low birth rates for 2021 and 2022 are then projected to be fully made up with elevated rates in 2024 through 2026, before returning to the levels that would have been projected in the absence of the pandemic for 2027 and thereafter. In other words, the pandemic is expected to cause a number of women to defer childbearing from 2021-22 to 2024-26.
The variations in the historical total fertility rate resulted from changes in many factors, including social attitudes, economic conditions, birth-control practices, and the racial/ethnic composition of the population. Since the baby-boom era (1946-65), women have had higher educational attainment, higher labor force participation, an older average age at first marriage, a higher propensity to remain unmarried, and higher rates of divorce. All of these factors are consistent with continued lower total fertility rates than those experienced during the baby-boom era. Based on consideration of these factors, the Trustees expect the ultimate total fertility rate to be relatively close to the average level since the end of the baby-boom era. The assumed ultimate total fertility rates are 2.20, 2.00, and 1.70 children per woman for the low-cost, intermediate, and high-cost assumptions, respectively. The low-cost and intermediate assumed ultimate rates are 0.05 higher than those used in last year’s report and the same as those used in the 2019 report. For the high-cost assumption, the ultimate rate is 0.05 lower than the assumption used in last year’s report.
The 2021 Trustees Report uses a cohort-based projection approach rather than the period-based approach used in the 2020 and prior Trustees Reports, resulting in a much longer transition to ultimate birth rates from the current low birth rates. The cohort-based approach assumes a more gradual shift to older ages of childbearing for birth cohorts who have not yet reached fertility completion at age 49. For the intermediate assumptions, after the assumed pandemic effects (i.e., after 2026), the projected total fertility rate gradually increases through the year the ultimate value is reached (2056). While the ultimate total fertility rate for alternative II is higher in this year’s report than in the 2020 report, the total fertility rate is lower for years after the pandemic through 2036 due to the new cohort-based projection approach. The assumed low-cost and high-cost total fertility rates trend away from the intermediate path, reaching their ultimate levels in 2056.
2. Mortality Assumptions
Mortality projections are developed by assuming ultimate average annual percentage reductions in future mortality rates by age group and cause of death. For this year’s report, these ultimate rates of reduction were revised and a new cause of death, dementia, was added. The assumptions were then used to estimate future central death rates by age group, sex, and cause of death. Adjustments were made to the death rates in 2020, 2021, 2022, and 2023 to account for the effects of the COVID-19 pandemic. Preliminary data for 2020 show that while the effects of the pandemic led to significantly higher death rates for those aged 15 and older, the death rates for those under age 15 were significantly lower, though not quite to the same degree. Thus, for ages 15 and over, death rates are assumed to increase above what would have been projected in the absence of the pandemic by 16.4 percent for 2020, 15 percent for 2021, 4 percent for 2022, and 1 percent for 2023. For ages under 15, death rates are assumed to decrease by 14 percent for 2020, 10 percent for 2021, and 2 percent for 2022, with no adjustment for 2023. From the estimated central death rates, probabilities of death by single year of age and sex were calculated. Projected death rates for years after 2023 are unchanged from the levels that would have been projected in the absence of the pandemic, under the assumption that increased deaths from the residual effects of living through the pandemic (both physiological and psychological) will be roughly offset by decreased deaths that instead happened sooner (during the pandemic).
Historical death rates were calculated for years 1900 through 2019 for ages below 65 (and for all ages for years prior to 1968) using data from the National Center for Health Statistics (NCHS).4 For ages 65 and over, final Medicare data on deaths for years 1968 through 2017 and preliminary data for 2018 and 2019 were used.5 Death rates by cause of death were produced for all ages for years 1979‑2019 using data from the NCHS.
The total age-sex-adjusted death rate6 declined at an average annual rate of 1.02 percent between 1900 and 2019. Between 1979 and 2019, the period for which death rates were analyzed by cause, the total age-sex-adjusted death rate, for all causes combined, declined at an average rate of 0.85 percent per year.
Death rates have declined substantially in the U.S. since 1900, with rapid declines over some periods and slow or no improvement over the other periods. Many factors are responsible for historical reductions in death rates, including medical advances, increased availability of health-care services, and improvements in sanitation and nutrition. Historical death rates generally declined more slowly for older ages and more rapidly for children and infants than for the rest of the population. Between 1900 and 2019, the age-sex-adjusted death rate declined at an average rate of 0.78 percent per year for ages 65 and over, and 2.97 percent per year for ages under 15.
Mortality assumptions differ for the low-cost, intermediate, and high-cost scenarios. Throughout the projection period, the low-cost scenario contains annual percentage reductions that are smaller than those in the intermediate scenario, while those in the high-cost scenario are larger. The ultimate annual percentage reductions for each of the three alternatives were revised for this year’s report. For the intermediate assumptions, the resulting total ultimate rates of improvement are slightly lower in this year’s report than they were in last year’s report. The changes to the low-cost and high-cost assumptions widened the range of rates of mortality improvement.
The trends in the annual reductions in central death rates were calculated for the period from 2008 to 2019 for both the NCHS and Medicare data, by age group, sex, and cause of death.7 These trends are the starting reductions for alternative II. For alternatives I and III, 50 and 150 percent of the starting reductions are used, respectively. These annual reductions, by alternative, are assumed to transition rapidly from the starting reductions until they reach the ultimate annual percentage reductions assumed for 2045 and later.
Table V.A1 contains historical and projected age-sex-adjusted death rates for the total population (all ages), for ages under 65, and for ages 65 and over. Age-sex adjustment eliminates the effect of a changing distribution of population by age and sex, allowing the pure effects of changes in death rates to be observed. Under the intermediate assumptions, projected age-sex-adjusted death rates are higher than the rates in last year’s report for the age group under 65 and slightly higher than the rates in last year’s report for the age group 65 and over. These changes primarily result from incorporating more recent historical data, which continue to show low rates of improvement.
The projected average annual rate of decline for the total age-sex-adjusted death rate is about 0.46 percent, 0.93 percent, and 1.48 percent between 2020 and 2095 for alternatives I, II, and III, respectively. In keeping with the patterns observed in the historical data, the assumed future rates of decline are greater for younger ages than for older ages, but to a substantially lesser degree than in the past. Accordingly, the projected age-sex-adjusted death rates for ages 65 and over decline at average annual rates of about 0.45 percent, 0.88 percent, and 1.34 percent between 2020 and 2095 for alternatives I, II, and III, respectively. The projected age-sex-adjusted death rates for ages under 15 decline at average annual rates of about 0.32 percent, 1.32 percent, and 2.74 percent between 2020 and 2095 for alternatives I, II, and III, respectively. It should be noted that these average annual rates of decline between 2020 and 2095 are not comparable with the rates of decline between 2019 and 2094 in last year’s report. As mentioned above, to reflect the effects of the pandemic in this year’s report, death rates for 2020 were increased by 16.4 percent for ages 15 and older, and decreased by 14 percent for ages under 15. For all three alternatives, this short-term change in death rates for 2020 increases the total (all ages) average annual rate of decline between 2020 and 2095 by about 0.20 percentage point from what would have been projected in the absence of COVID-19.
Demographers express a wide range of views on the likely rate of future decline in death rates. For example, some believe that the long-standing historical tendency for mortality to decline more slowly at the oldest ages will cease in the future. Others believe that biological factors, social factors, and limitations on health care spending may slow future rates of decline in mortality.
.
Table V.A1.—Fertility and Mortality Assumptions,a
Calendar Years 1940-2095 
Total
fertility
rateb
Age-sex-adjusted death rate c
per 100,000
d803.4
d4,434.6
d792.3
d257.6
d4,357.1
e1.65
f923.7
f299.9
f5,083.2

a
This table contains basic assumptions along with key summary values that are derived from basic assumptions.

b
The total fertility rate for any year is the average number of children that would be born to a woman if she were to experience, at each age of her life, the birth rate observed in, or assumed for, the selected year, and if she were to survive the entire childbearing period.

c
Based on the enumerated total population as of April 1, 2010, if that population were to experience the death rates by age and sex observed in, or assumed for, the selected year.

d
Estimated using final data for ages below 65 and preliminary data for ages 65 and older.

e
Estimated using provisional data through the third quarter and the intermediate assumptions.

f
Estimated using the intermediate assumptions for all ages.

3. Immigration Assumptions
Projections of the total Social Security area population reflect assumptions for annual immigration flows. For this report, four categories of immigration flows are used:
Lawful permanent resident (LPR) immigration: Persons who enter the Social Security area and are granted LPR status, or who are already in the Social Security area and adjust their status to become LPRs.8
Net LPR immigration is the difference between LPR immigration and legal emigration. Net other-than-LPR immigration is the difference between other-than-LPR immigration and other-than-LPR emigration. Total net immigration refers to the sum of net LPR immigration and net other-than-LPR immigration.
Immigration assumptions differ for the low-cost, intermediate, and high-cost scenarios. The low-cost scenario includes higher annual net immigration and the high-cost scenario includes lower annual net immigration. Table V.A2 contains historical and projected levels of various immigration flows.
LPR immigration has increased significantly since World War II, due to various factors and legislative changes, including the Immigration Act of 1965 and the Immigration Act of 1990.
For all three alternatives, the Trustees assume that the COVID-19 pandemic will affect LPR immigration in years 2020 through 2025. The assumed LPR immigration levels for 2020 and 2021 are 390,000 lower than would have been assumed in the absence of the pandemic. LPR immigration levels for 2022 are assumed to be the same as would have been assumed in the absence of the pandemic. LPR immigration levels for 2023 through 2025 are assumed to be higher than would have been assumed in the absence of the pandemic, fully making up for the lower levels in 2020 and 2021.
For the intermediate alternative, the ultimate level of annual LPR immigration, which includes residents who adjust their status to become LPRs, is assumed to be 1,050,000 persons for 2026 and later. For alternative I, ultimate annual LPR immigration is assumed to be 1,250,000 persons for 2026 and later, and for alternative III, ultimate annual LPR immigration is assumed to be 850,000 persons for 2026 and later. The ultimate levels of LPR immigration are unchanged from last year’s report.
The assumed ratios of annual legal emigration to LPR immigration are 20, 25, and 30 percent for alternatives I, II, and III, respectively. This range is consistent with the limited historical data for legal emigration from the Social Security area. These ratios are unchanged from last year’s report. Under the intermediate alternative, by combining the ultimate annual LPR immigration and legal emigration assumptions, ultimate annual net LPR immigration is about 788,000 persons. For the low-cost and high-cost scenarios, ultimate annual net LPR immigration is 1,000,000 persons and 595,000 persons, respectively.
The estimated number of other-than-LPR immigrants residing in the Social Security area and the annual level of other-than-LPR immigration were affected significantly by the economic recession of 2007-09. Although other-than-LPR immigration was greatly reduced during the economic downturn and immediate years thereafter, it returned to higher levels for 2014 through 2018, and is assumed to remain at a higher level in 2019, reflecting a recovery from levels experienced during the recession.
For all three alternatives, the Trustees assume that the COVID-19 pandemic will affect other-than-LPR immigration in years 2020 through 2025. The assumed other-than-LPR immigration levels for 2020 are 780,000 lower for the low-cost and intermediate scenarios, and 720,000 lower for the high-cost scenario, than would have been assumed in the absence of the pandemic. The assumed other-than-LPR immigration levels for 2021 are 450,000 lower for the low-cost scenario, 420,000 lower for the intermediate scenario, and 360,000 lower for the high-cost scenario, than would have been assumed in the absence of the pandemic. Other-than-LPR immigration levels for 2022 are assumed to be the same as would have been assumed in the absence of the pandemic. Other-than-LPR immigration levels for 2023 through 2025 are assumed to be higher than would have been assumed in the absence of the pandemic, fully making up for the lower levels in 2020 and 2021.
The ultimate levels of other-than-LPR immigration are 1,350,000 persons for alternative II, 1,850,000 persons for alternative I, and 850,000 persons for alternative III. The ultimate levels are attained in year 2026 for all three alternatives. In last year’s report, the ultimate levels were 1,650,000 persons and 1,050,000 persons, for alternatives I and III, respectively. The ultimate level for alternative II is unchanged from last year.
Emigration from the other-than-LPR immigrant population includes those who leave the Social Security area and those who adjust their status to become LPRs. This other-than-LPR immigrant population is highly mobile and far more likely to leave the Social Security area than is the citizen or LPR population. However, as other-than-LPR immigrants stay in the country for longer periods of time, they generally become less likely to leave the country.
Under the intermediate assumptions, the total annual number of other-than-LPR immigrants who leave the Social Security area averages about 421,000 through the 75‑year projection period. The ultimate annual number of other-than-LPR immigrants who adjust status to become LPRs is assumed to be 450,000 for the intermediate assumptions, and is unchanged from last year’s report. For the low-cost and high-cost scenarios, the total annual number of other-than-LPR emigrants averages about 577,000 and 264,000, respectively, through the 75‑year projection period. The ultimate annual number of people adjusting status to LPR status is assumed to be 550,000 persons and 350,000 persons, for the low-cost and high-cost scenarios, respectively, and is unchanged from last year’s report.
Under the assumptions described above, the projected size of the other-than-LPR immigrant population grows substantially, from about 15.1 million by the end of 2021 to about 35.4 million by the end of 2095. This growth reflects the excess of annual immigration over the combined annual numbers of emigrants (including adjustments of status) and deaths that occur within the other-than-LPR immigrant population.
Under the intermediate assumptions, projected net other-than-LPR immigration gradually decreases over time. Because the projected number of other-than-LPR immigrants leaving the Social Security area is based on rates of departure, an increase in the number of other-than-LPR immigrants residing in the Social Security area results in an increase in the number who emigrate out of the area. All other components of net other-than-LPR immigration are assumed to be stable after 2025, and thus do not contribute toward any change in annual net other-than-LPR immigration. Under the intermediate assumptions, the projected average annual level of net other-than-LPR immigration over the 75-year projection period is about 488,000 persons. For the low-cost and high-cost assumptions, projected average annual net other-than-LPR immigration is about 732,000 persons and 245,000 persons, respectively.
The projected average annual level of total net immigration (LPR and other-than-LPR, combined) is about 1,280,000 persons per year during the 75-year projection period under the intermediate assumptions. For the low-cost and high-cost assumptions, projected average annual total net immigration is about 1,736,000 persons and 844,000 persons, respectively.
Demographers express a wide range of views about the future course of immigration for the United States. Some believe that net immigration will increase substantially in the future. Others believe that potential immigrants may be increasingly attracted to other countries, that the number of potential immigrants may be lower due to lower birth rates in many countries, or that changes in the law or enforcement of the law will reduce immigration.
Table V.A2.—Immigration Assumptions,a Calendar Years 1940-2095 
Other-than-LPR immigrationb
Adjustments
of statusc d
403
810
e901
e234
e264
e1,075
e566
e 258
e 466
e 774
f1,100
e 246
e466
f 389
f1,163
f 210
f165
f450
f495
f420
f253
f450
f -283
f212

a
This table contains basic assumptions along with key summary values that are derived from basic assumptions.

b
Historical other-than-LPR immigration and emigration estimates depend on a residual method. The Office of the Chief Actuary developed these estimates, as well as the resulting other-than-LPR January 1 stock estimates, for years through 2000. For years 2001 and later, the residual method uses stock estimates. For 2001 through 2004, the stock is set to values that linearly grade from the 2000 stock estimate to the DHS 2005 stock estimate. For 2005 through 2012, stock estimates come from DHS. For 2013 through 2019, stock estimates are developed by the Office of the Chief Actuary, based on the same methods used by DHS.

c
Estimates do not include persons who attained LPR status under the special one-time provisions of the Immigration Reform and Control Act of 1986.

d
Adjustments of status are a positive for net LPR immigration and a negative for net other-than-LPR immigration.

e
Estimated.

f
Estimated, intermediate alternative.
Note: Components may not sum to totals because of rounding.

4. Total Population Estimates
The starting Social Security area population for December 31, 2018, is derived from the Census Bureau’s estimate of the residents of the 50 States and D.C. and U.S. Armed Forces overseas. Adjustments are made to reflect mortality assumptions for the aged population since 2010 that are consistent with Medicare and Social Security data, net immigration assumptions for the aged population since 2010, estimates of the net undercount in the 2010 census, inclusion of U.S. citizens living abroad (including residents of U.S. territories), and inclusion of non-citizens living abroad who are insured for Social Security benefits. The Office of the Chief Actuary projects the population in the Social Security area by age, sex, and marital status for December 31 of each year from 2019 through 2095 by combining the assumptions for future fertility, mortality, and immigration with assumptions for marriage and divorce. Previous sections of this chapter present the assumptions for future fertility, mortality, and immigration. Assumptions for future rates of marriage and divorce reflect historical data from the National Center for Health Statistics, the Census Bureau, and selected individual States.
This report presents a July 1 (i.e., midyear) population for each year, which is derived from surrounding December populations. Table V.A3 shows the historical and projected population for July 1 by broad age group, for the three alternatives. It also shows the aged and total dependency ratios (see table footnotes for definitions).
Aged a
Total b
2018c
2019d