Short-Range Actuarial Projections of the Old-Age, Survivors, and Disability Insurance Program, 2001
Actuarial Study No. 115
Chris Motsiopoulos and Tim Zayatz, A.S.A.
Expenditures from the OASI and DI Trust Funds include the following items:
Current-payment and retroactive benefits;
Payments for vocational rehabilitation services;
As of calendar year 2000, benefit payments account for roughly 98 percent of outgo for the combined OASI and DI Trust Funds. In general, future benefit payments are estimated by projecting the number of beneficiaries and the average benefit, by type of beneficiary. Following is a description of the projection of benefit payments. The projection of other items of outgo, as well as all items of income, are described in section IV.
Fully insured status is required of an aged worker for eligibility to primary and auxiliary benefits, and of a deceased worker for eligibility to survivor benefits. Exceptional categories are child survivors and parents of child survivors, who may alternatively be eligible if the deceased worker had currently insured status. Table III.A1 shows the total number of workers who are fully insured for OASI benefits.
Table III.A1 also shows the total number of workers who are permanently insured; that is, the worker has sufficient quarters of coverage not only to be fully insured at his or her current age, but also enough quarters of coverage to be fully insured upon attainment of retirement age. For most individuals, this is 40 quarters.
The number of workers who are fully or permanently insured is projected based on data from the Continuous Work History Sample (CWHS), which is a 1-percent sample of the Master Earnings File (MEF) combined with data on the same sample from the Master Beneficiary Record (MBR). The CWHS contains coverage information for workers, but has incomplete reporting of deaths, and thus overstates the number of fully insured workers. To account for the overstatement, mortality adjustments are made to the gross counts on both a historical and projected basis.
The following subsections provide detail of the projections of the number of beneficiaries for the OASDI program—section III.A along with tables III.A2-III.A33 present results for the DI program; section III.B along with tables III.B1-III.B84 present results for the OASI program.
Table III.A2 presents the Social Security area population under normal retirement age. The demographic section of the Long-Range Office of the Chief Actuary provides population projections based on assumptions such as fertility, mortality, and immigration. Data on quarters of coverage is used to extract the fully insured population from the general population, which is then related to data from the CWHS as described above. Table III.A3 shows the fully insured population under normal retirement age.
Disability insured status requires fully insured status as well as a recent connection to the labor force, as described in section II.C. Note, therefore, that a worker may be fully insured, but may not have the required connection to the labor force; conversely, a worker may have the required recent connection to the labor force, but may not have accrued a sufficient number of QCs to be deemed fully insured. A special disability insured test exists for younger workers, which provides an alternative to the "20/40" recency-of-work requirement. Table III.A5 shows the ratio of the disability insured to the fully insured.
The short-range model projects the number of workers who are disability insured by single year of age using the following time series regression:
The independent (regressor) variables are year—represented by X1, and labor force participation rate—represented by X2. The lag-1 autocorrelation coefficient—a measure of the amount of correlation between successive error terms in the regression—is represented by . Future response values are extrapolated using the time series model, and then transformed into their corresponding DIFI values.
Table III.A5 shows the DIFI values, which are applied to the projected fully insured population to create the projected disability insured population.
The time series model is an iterative autoregression. The residuals are examined each time and the procedure is repeated until the autocorrelation between successive error terms is sufficiently small. Finally, the regression results are subject to maximum and minimum values to keep the projected percentages within reasonable bounds. Labor force participation rates and the disability insured population are shown in tables III.A4 and III.A6, respectively.
The intermediate assumptions predict the total number of disability insured workers will increase steadily throughout the short-range period. The average annual growth rate over 2001-10 is roughly 1.4 percent. On a gender-specific basis, the annual growth rate is 1.1 percent for males, and 1.8 percent for females. Note that there are disability insured workers age 65 or older for the first time at the end of 2003. This is a consequence of the increase in the normal retirement age, which first affects individuals born in 1938. Those born in November and December of that year will still be considered eligible for disability benefits before they attain the normal retirement age of 65 years 2 months in 2004. Afterward, they are eligible only for old-age benefits.
Growth in the DI rolls is due to the difference between the number of new beneficiaries that have been added to the rolls and the number of beneficiaries whose benefits have been terminated, in a given year. The short-range model projects the number of new entrants, or awards, for disabled workers by single year of age and gender by applying disability incidence rates to the exposure1 of the disability insured population not already entitled to benefits—that is, not in force. Exposure is calculated as (i) the aggregate amount of time that insureds from a specific birth cohort are exposed to disability during a particular year, minus (ii) the aggregate amount of time contributed by those in force at the beginning of that year. Refer to appendix B. for details of incidence rate award calculations.
The model develops future incidence rates from a base year set of rates, by single year of age and gender. Although the rates are different for each age, the change in the rates from one particular year to the next is assumed to be constant across all ages of the insured population, for a particular gender. Also, the model does not separately identify increases or decreases in award rates by type of impairment. Occasionally, however, a specific category will need to be analyzed separately because the incidence of the impairment is concentrated within a specific age group or gender. For example, we currently split out HIV-related impairments because of its high concentration among the younger age groups.
The corresponding incidence rates are developed through a series of preliminary projections reflecting judgment on the expected number of awards. For continuity, the estimates developed for the short-range period phase into long-range estimates by the 10th projected year. In essence, the ultimate incidence rate is set in the long-range period, and the short-range model determines the path the DI award program will take to get to that rate. The following is a list of the major components and analyses performed by the short-range model in developing disabled worker incidence rates:
Compare actual data to prior estimates for most recently completed historical year, by age group and gender. This highlights areas where estimates may have "missed badly" over the previous year. We then try to identify one-time events versus permanent legislative changes or trends, and adjust the following year's forecast accordingly.
Perform a preliminary award projection assuming incidence rates remain unchanged from the most recently completed historical year. This purely demographic exercise shows the increase in awards due solely to changes in the underlying disability insured population. We then determine how incidence rates should be modified to account for various exogenous factors.
Examine the effect of changes in key economic variables. Estimates of labor force participation are already reflected in the disability insured projection. Increasing unemployment rates will often mean an increase in future disability claims, whereas steady or decreasing unemployment will result in the leveling off of claims. Tight labor markets, increases in real wages, and productivity gains provide work incentives, and may ultimately reduce the number of claims filed.
Examine 1-, 3-, 5-, and 10-year trends in incidence rates, as well as year-over-year growth in the number of awards over those periods. Short-term trends provide valuable insight into the first several projected years as we observe the effects of significant legislation and economic conditions. Long-term trends provide direction beyond the first several years as program dynamics tend to move toward historical average levels. However, even established dynamics can be reshaped by demographic factors that evolve slowly over time.
Monitor changes in incidence rates for ages 50-64. These are the ages which have the highest incidence of disability. This analysis has become increasingly important as the baby-boom generation—birth cohorts 1946-64, currently ages 36-54—begins to enter peak years of disability incidence.
Perform separate analysis on HIV-related impairments. Although not as significant as it once was, this impairment category is analyzed separately since award patterns deviate substantially from most other impairment categories.
Track the disposition of applications filed for disability claims. Since an increase in applications may or may not result in an increase in awards, claims data must be analyzed in relation to other factors. For example, increased filings during recessionary economic conditions have historically led to a significant increase in the number of awards—up to a point; more and more claims will eventually come from less severely disabled individuals, resulting in only a marginal increase in awards.
Track allowance rates at the initial and appeal stages of the disability determination process. Allowance rates measure the number of favorable determinations among disability claims. Historical levels have been consistent, ranging from 30-35 percent allowance at the initial stage to roughly 70-75 percent or more during the final appeal stages. Small positive deviations can result in a significant increase in the number of awards; whereas a decrease in the allowance rate—that is, an increase in denied claims—may result in an increase in appeals, which may cause a backlog of cases at the appeal stages. A large number of pending cases may in turn create a large number of deferred awards, and may even affect the processing of new claims.
Account for the effect of the increase in normal retirement age. Basic factors that affect the decision of older workers to apply for DI benefits as opposed to old-age benefits include: the amount of effort in getting medical evidence and "proving" the inability to work; longer processing times; a 5-month waiting period; high rate of benefit denial; difference in maximum family benefits payable—OASI usually allows a greater amount per household; and possible legal expenses. Despite the apparent hassles associated with getting a DI benefit, the alternative is to wait longer for full old-age benefits, or receive a greater actuarial reduction by taking early benefits. Consequently, the short-range model assumes an induced increase in disability awards among workers over the age of 60 beginning in 1999, as a result of the increase in normal retirement age.
The overall incidence rate for a particular year may be expressed as gross—awards for the stated year divided by exposure for that same year—or age-sex-adjusted. Gross rates are not directly comparable over time because of the year-to-year differences in the composition of the underlying insured population. To compare incidence rates for different years, we need to adjust the gross rate to factor out the effect of the changes in population in terms of both age and gender. In essence, the adjusted rates are "indexes" rather than direct measures and are standardized to the insured population for 1998. They show what the incidence rate would have been in any particular year given the age and gender distribution of the insured population of 1998. Refer to appendix C. for an example of age adjustment.
Gross and age-sex-adjusted disability incidence rates and the number of awards to disabled workers are shown in tables III.A8 and III.A9, respectively.
It is worth noting here that the projection of disability incidence rates is one aspect of the short-range forecast that continues to rely heavily on educated judgment in connection with a variety of factors affecting the DI program. While some factors can be modeled in a straightforward deterministic fashion, an all-encompassing structural model of disability incidence remains impractical because of the complicated nature and dynamics of the DI program. For example, traditional demographic models can account for the aging of the baby-boom generation and subsequent impact on the number of awards as more beneficiaries are awarded at older ages. At the same time, other factors exist that affect the number of disability awards at older ages, but cannot be readily modeled. An example of this would be the degree of use of vocational factors2.
Below is a list of the leading determinants which may have a significant impact on both the number of disability claims filed and the resulting number of awards3:
The impact of any one factor is difficult to gauge. As shown, they may be administrative, economic, or demographic in nature. Some factors evolve slowly over time with well-defined characteristics, and as such lend themselves to empirical modeling. Others are less predictable and must be considered within the scope of the entire program as it currently exists, and their interaction with other factors. Still others will have an obvious immediate effect, but the total future impact may not be readily apparent and will only become known over time. So we must rely on experience, judgment, and careful study of these individual factors to expand the basic mathematical projections.
Another closely monitored area of the DI program is the termination of benefits. Many variables affect the number and rate at which beneficiaries are terminated from the DI rolls, including:
The law contains several provisions to encourage individuals who wish to return to work, but continue to have a disabling impairment. Among these are the trial work period (TWP) and the extended period of eligibility (EPE). The TWP is a 9-month period during which an entitled beneficiary may work without affecting the right to benefits. Earnings during the 9 months are not counted toward substantial gainful activity4 (SGA), and benefits will continue as long as the beneficiary has not medically recovered. Individuals who continue to have a disabling impairment following the TWP, receive a 36-month EPE. Earnings during the EPE are counted toward SGA, and monthly benefits will not be paid when such earnings exceed the SGA level. If earnings fall below the SGA level anytime during the EPE, benefits are automatically reinstated.
The reasons for termination of DI worker benefits can be grouped into four main categories:
Conversion—The disability benefit is converted to an old-age benefit upon attainment of the normal retirement age (NRA);
All other reasons—This relatively minor category is mostly comprised of beneficiaries who switch to old-age benefits prior to NRA; but also includes withdrawal of application and erroneous entitlement.
The short-range model projects the number of terminations for disabled workers by reason, single year of age, and gender by applying disability termination rates to the exposure of the beneficiaries on the DI rolls. For termination purposes, exposure is calculated as (i) the aggregate amount of time that in-force beneficiaries from a specific birth cohort are exposed to termination during a particular year, plus (ii) the aggregate amount of time contributed by new awards in that year. Refer to appendix D. for details of termination calculations.
Most terminations occur as a result of either conversion or death—in calendar year 2000, roughly 46 percent of total terminations were by conversion; 37 percent were by death; 15 percent were by return-to-work or medical recovery; and 2 percent were by other reasons. In developing termination rates for the individual categories, we use several of the techniques described in the analysis for awards, namely: comparing actual data to prior estimates and explaining any large estimation errors; and examining 1-, 3-, 5-, and 10-year trends in termination rates and year-over-year growth in the number of terminations for the individual categories. However, further insight is needed in unique areas. The following discussion highlights special considerations in termination analysis for each category.
DI worker beneficiaries who do not die, recover, or terminate for some other reason are automatically converted to old-age beneficiaries on the last day of the month before the month they attain NRA. As a result of the scheduled increase in NRA5, conversions are expected to decline sharply in 2003 as some of the conversions are deferred to the following year. Specifically, 2 months of conversions from the 1938 birth cohort will occur in 2004; 4 months of conversions from the 1939 cohort will occur in 2005; 6 months of conversions from the 1940 cohort will occur in 2006; 8 months of conversions from the 1941 cohort will occur in 2007; 10 months of conversions from the 1942 cohort will occur in 2008; and finally all the conversions from the 1943 cohort will occur in 2009 when the NRA becomes integral age 66. This explains the sharp increase projected for that year. Conversions are shown in table III.A14.
Medical advancement has significantly changed life expectancies among the general population. However, the impact on a disabled life is less clear. Many outside variables can contribute to changes in disability mortality rates. Recently the DI program has experienced legislation eliminating drug and alcohol related impairments (Public Law 104-121); an increasing number of awards to older workers whose determinations are based on a set of vocational factors rather than a single severe disability; the diminishing impact of HIV-impairments and AIDS; and an increasing proportion of younger and physically healthier beneficiaries who are awarded as a result of mental impairments. From a modeling standpoint, these effects are quantified-in this case we predict how many fewer deaths will occur-and the results are blended with general long-range mortality improvements to predict death rates among disabled workers.
A statistic that provides insight into disability mortality trends is duration—a measure of the average amount of time current beneficiaries have been on the DI rolls. This measure shows that the "maturity" of the rolls has been lengthening in recent years, and across all ages. For example, as of the end of 1995 a 50-year-old male disabled worker in current-payment status had been on the DI rolls for an average of 8.0 years; at the end of 2000 the average duration increased to 9.1 years. Likewise, the average duration of a 50-year-old female disabled worker has increased steadily from 6.6 years in 1995 to 7.3 years in 2000. This indicates that a greater proportion of the DI rolls is made up of beneficiaries who have been current for some time, and proportionately fewer "new" beneficiaries. This condition tends to have a dampening affect on mortality rates. Refer to appendix E. for details.
Note that although disability mortality rates are expected to decline in the future, the DI program will experience a net increase in the number of deaths. This is purely due to demographic shifts in the general population. As the baby-boom continues to enter the DI rolls at progressively older ages, the program will naturally experience more deaths since the highest disability mortality occurs at ages 50 or older.
Disability recovery may occur when a beneficiary either provides notification of an improved disabling condition, demonstrates the ability to engage in SGA, or is judged to no longer meet the definition of disability based on a full medical continuing disability review (CDR).
Certain outside variables can exert significant influence on the overall number of recoveries, making this termination category difficult to predict. Trends in the nature of disabling conditions or changes in regulations that redefine the required severity of an impairment can directly affect the "natural" recovery rate. However, medical review of current beneficiaries is often the only way to distinguish those who no longer meet disability requirements. Therefore, factors that affect the level of CDR activity also have a big influence on recovery rates. These include budget restrictions, the number of new claims that need to be processed, caseload backlogs, and legislation.
Public Law 104-121 provides for the appropriation of special funds for conducting CDRs. Since its enactment in 1996, the CDR plan has been regularly evaluated by the Office of Disability to reflect emerging experience. As a result of this special funding, previously existing backlogs of scheduled CDRs have been eliminated, and maturing CDRs are being worked in a timely fashion. However, because of the recently enacted Ticket to Work and Work Incentives Improvement Act of 1999 (Public Law 106-170) certain restrictions have been placed on the scheduling of CDRs. In particular, CDRs may not be triggered solely as the result of work activity, and beneficiaries actively participating in the ticket-to-work program may not be scheduled for a CDR even though their review diary has matured. As a consequence of these two provisions, there will likely be a small decline in the number of scheduled CDRs and a deferral of the associated recovery terminations.
The short-range model projects recoveries under three individual categories: (i) return-to-work by those demonstrating SGA upon completion of an extended period of eligibility, (ii) terminations resulting from full medical CDRs, and (iii) "work alerts"—a minor category which has been virtually eliminated under the provisions of Public Law 106-170. As with death terminations, the rate of recovery among disabled workers can be significantly affected by the age at entitlement and the amount of time spent on the DI rolls. Refer to appendix E for details.
The main "other" reason for termination is the decision to receive old-age benefits in lieu of DI benefits, which a disabled worker may elect to do beginning at age 62. For the most part, this occurs automatically upon attaining NRA—roughly 85 percent of those who switch, do so at NRA. However, many factors influence the personal decision to convert prior to NRA. One of the most common reasons is eligibility for benefits outside the DI program. For example, workers compensation benefits may partially or totally offset a DI benefit, but would not affect an OASI benefit. Other factors include: the amount of actuarial reduction that would occur to the old-age benefit; the difference in maximum family benefits payable, which may be lower under the DI program than under the OASI program; and the beneficiary's own health assessment and outlook of life expectancy. The short-range model looks for any exogenous factors that may cause new trends to emerge.
As mentioned, conversion and death account for most of the terminations among disabled workers, and in general termination rates have been declining. Two significant trends have developed over the years which help explain the decline. The first is mortality improvements, as previously discussed. The second trend is a reduction in the average age of disabled workers. Data presented in appendix F indicates that over the period 1975-95, the average age of a disabled worker award has generally declined from 51.5 to 48.6 years for males; and from 52.1 to 48.5 years for females. Consequently, the average age of disabled workers in current-payment status has also declined over the same period from 53.0 to 49.8 for males; and from 53.9 to 49.9 for females. Also over the period, the fraction of 30-44 year olds on the DI rolls has roughly doubled from 14 percent to 27 percent for males; and has more than doubled from 11 percent to 26 percent for females. The increase in younger and physically healthier beneficiaries is largely attributable to the increasing proportion of new awards due to mental impairments. The result has been a smaller percentage of beneficiaries converting to old-age benefits each year, as well as lower death rates.
Since 1995, the average age of a disabled worker has trended back up, reaching 50.8 by the end of 2000. Over the next 10 years, beneficiaries from the baby-boom will enter the DI rolls at progressively older ages and the average age will continue to increase, reaching roughly 53.4 by the end of 2010. This trend will help stabilize overall termination rates beginning around 2002.
Disabled worker termination rates by reason are shown in tables III.A10, III.A11, III.A12 and III.A13, while the number of terminations are shown in tables III.A14, III.A15, III.A16 and III.A17.
Table III.A7 shows the number of disabled workers with benefits in force. In force refers to individuals who are eligible for benefits, including those actually receiving a payment (in current-pay) as well as those whose benefit has been suspended for one of several reasons. The in-force population is calculated as of December 31 of each year by adding the number of awards, less terminations, for the current year to the in-force population at the end of the previous year.
Disability benefits may be suspended for any number of reasons, including refusal to accept rehabilitation services, a pending determination of continuing disability, workers compensation offset, or extended period of eligibility requirements. Estimated future suspensions are developed by age and gender by applying withheld rates to the in-force population. Withheld rates and suspensions are shown in tables III.A18 and III.A19, respectively. The short-range model assumes that these rates do not vary much over time. However, legislative action will cause noticeable deviations from time-to-time. A significant jump occurred in 1995 as a result of changes in the treatment of drug addicts and alcoholics—by January 1997, these benefits were removed from suspense and permanently terminated. Most recently, a provision of Public Law 106-170 modified the rules governing benefits paid to prisoners. The result is an increase in suspensions among ages 25-49 in 2000 and later.
Table III.A20 shows the number of disabled workers with benefits in current-payment status. The current-payment population is calculated as of December 31 of each year by subtracting suspensions from the in-force population. Over the next 10 years, the model projects an average annual increase in the current-payment population of roughly 3.0 percent for males, and 4.6 percent for females. The relatively larger increase for females is attributable to the rapid growth in female labor force participation experienced during the 1990s. This has led to an increase in the number of females with disability insured status.
Refer to appendix G. for a comprehensive illustration of the disabled worker numbers projection by single year of age.
Awards to auxiliary beneficiaries of disabled workers are summarized in tables III.A21 and III.A22. The short-range model produces calendar year estimates for the following categories:
Young wife—benefits may be payable to a (divorced) wife of a disabled worker, where the wife is entitled because of a child in her care;
Young husband—benefits may be payable to a (divorced) husband of a disabled worker, where the husband is entitled because of a child in his care;
Aged wife—benefits may be payable to a (divorced) wife of a disabled worker, where the wife is entitled based solely on her age (having attained 62);
Aged husband—benefits may be payable to a (divorced) husband of a disabled worker, where the husband is entitled based solely on his age (having attained 62);
Minor child—benefits may be payable to a child of a disabled worker, where the child is under age 18;
Disabled child—benefits may be payable to a child of a disabled worker, where the "child" is a disabled person age 18 or older whose disability began before age 22; and
Student child—benefits may be payable to a child of a disabled worker, where the child is an elementary or secondary student between ages 18 and 19.
Awards to young wives and young husbands are based on trends in the historical ratio of these awards to minor child awards. Awards to aged wives are based on trends in the historical ratio of these awards to male disabled worker awards; awards to aged husbands are based on trends in the historical ratio of these awards to female disabled worker awards; and awards to children are based on trends in the historical ratio of these awards to total disabled worker awards. Finally, the model determines quarterly projections from annual awards using interpolation which takes into account historical seasonal patterns in such awards.
Projections for auxiliary benefit terminations are produced quarterly by applying assumed termination rates to the estimated exposure of the particular auxiliary category6. Projections for auxiliary in-force and current-pay populations, and suspensions are similar to those for workers.
Tables III.A23-III.A33 summarize quarterly data for disabled workers and auxiliary categories. The tables include the in-force population, awards, termination rates, terminations, withheld rates, suspensions, and the current-payment population.
Links to tables: III.A23 III.A24
III.A25 III.A26 III.A27
III.A28 III.A29 III.A30
III.A31 III.A32 III.A33
1 An actuarial quantity measuring the aggregate amount of time, or life-years, that a population is exposed to program forces such as incidence and termination. Aggregate life-years are calculated for each age during each year.
2 Many individuals are found to be disabled even though their impairments fail to meet the required level of severity. In these cases, an individual's medical condition is evaluated in conjunction with age, education, and job skills. These "vocational factors" are given increasing weight with the advancing age of the worker, and are particularly significant in the determination of disability among workers age 50 or older.
3 Borrowed from Actuarial Study No. 114: Social Security Disability Insurance Program Worker Experience (Tim Zayatz, June 1999). Consult this text for greater detail, and also The Social Security Disability Insurance Program-an Analysis (Department of Health and Human Services, December 1992).
4 One of the criteria for entitlement to DI benefits is the inability to engage in any substantial gainful activity by reason of a medically determinable impairment. Certain earnings criteria have been established as reasonable indications of whether an individual is engaging in such activity. As of 2001, an employee averaging over $740 per month will ordinarily demonstrate SGA; $740 or less per month will ordinarily demonstrate lack of SGA. Under certain circumstances, someone earning less than SGA may receive additional review of work activity. The SGA level is adjusted annually to reflect increases in the national average wage.
5 See Glossary definition of "Normal retirement age" for details on the scheduled increases.
6 Exposure is estimated as beginning-of-period in force, plus one-half of the awards for the period.
List of Studies
December 26, 2001