|
COPYRIGHT 2001 MIT Press Journals
I. Introduction
THE disability determination process used by the Social Security Administration (SSA) acts as the gatekeeper for several public programs: Disability Insurance (DI), Supplemental Security Income (SSI) and--for those judged disabled--Medicare and, in most states, Medicaid. Between 1985 and 1997, annual benefits under the two cash programs, DI and SSI, grew from $26 billion to $68 billion--increasing more than 150% in current dollars. Factoring in Medicare and Medicaid, total benefits to the disabled have grown to well over $100 billion per year. Entitlement spending on this scale argues not only for more analysis of applications decisions, but for an expanded focus on the complex procedure for determining eligibility. The eligibility criteria implicit in the screening process reflect basic normative decisions that define the population the program is intended to serve. Hence, it is important to understand the structure of the screening process and how it affects the size and characteristics of the targeted population.
Unfortunately, analysts studying disability program growth face a daunting information gap. Administrative data, tied to day-to-day operations, have no information on those who have not applied and little socioeconomic information that is useful for understanding application behavior. In contrast, household surveys of the general population provide information on nonapplicants and on a range of socioeconomic details; however, they contain little information on an individual's actual interactions with the administering agency--information that is needed to relate survey information to the agency's eligibility decisions.
SSA addressed this issue by implementing three major surveys: the 1966 Social Security Survey of the Disabled, the 1972/1974 Survey of Disabled and Non-Disabled Adults, and the 1978 Survey of Disability and Work. These surveys asked questions on health impairments and socioeconomic status that were designed for analysis of the disability programs. The resulting data have been utilized in a number of important studies of disability enrollment.(1) However, little effort has been devoted to modeling the structure of the disability determination process.
In this paper, we model the outcome of SSA disability determinations using survey responses on health, functional limitations, demographic traits, and work behavior for adult applicants. To do this, we exact-match data from the 1990 Survey of Income and Program Participation (SIPP) to SSA records on disability determination. We adopt a structural approach, using a multistage logit model that mimics the sequential process by explicitly modeling decisions made under distinct criteria. For example, SSA's decisions at some steps are based solely on medical evidence and health-related activity limitations, whereas decisions at other steps also take into account age, education, and past work. The matched data, combined with the structural representation of the process, enable more-appropriate use of survey data in explaining these differential eligibility outcomes.
The paper is organized as follows. Section II outlines the sequential disability determination process, by step. Section III introduces the four-stage sequential logit model. Section IV describes the data used in the study. Section V presents empirical findings, and section VI reports split-sample analysis results. Finally, section VII summarizes our conclusions.
II. Disability Determination
The two disability programs administered by SSA use the same disability determination process, but have different, albeit complementary, objectives. The DI program is part of the comprehensive social insurance program--Old Age, Survivors, and Disability Insurance (OASDI). It is funded through payroll tax contributions. The SSI program, by contrast, is means-tested and guarantees a minimal level of income to the poorest of the aged, blind, or disabled. SSI is funded from general revenues. Compared to eligibility decisions for other programs, disability determination is complex and not widely understood. Yet, in recent years, total applications for the two programs have reached as high as 2.5 million per year, so the budgetary and welfare impact of the process are undeniable. The process modeled in this study is that presently used by the state Disability Determination Service (DDS) agencies. It is used in initial determinations or DDS reconsiderations and it accounted for more than 80% of all allowances in fiscal year 1990. Other decisions made under the multilayered appeals process--by administrative law judges (ALJs), for example--are not modeled in this analysis.(2) As shown in figure 1, disability determination involves a sequence of five steps.
[ILLUSTRATION OMITTED]
A. Step 1: An Earnings Screen
SSA field offices screen out applicants who work. Applicants are denied if they engage in activities that are both substantial and gainful. Specifically, applicants earning more than the substantial gainful activity (SGA) amount are denied immediately; otherwise, the application is referred to the DDS.(3) The maximum SGA amount was $500 per month in 1993. Because this step is not based on medical criteria, it is not modeled here.
B. Step 2: A Medical Screen to Deny the Least Severely Impaired
An applicant is denied at step 2 if his or her impairment (or combination of impairments) is judged not severe. This judgment is based on a conceptual threshold: an impairment is considered nonsevere if it does not significantly limit the physical or mental abilities to perform basic work-related activities. Basic work activities include physical functions (such as walking, standing, or lifting), sensory functions (such as seeing, hearing, or speaking), and routine mental functions (such as understanding simple instructions, responding to supervision, or adapting to changes in the work environment). Applicants are also denied if their impairments fail the duration test, that is, if the impairment is not expected to result in death, and if the impairment has neither lasted twelve months nor is expected to last twelve months.
C. Step 3: A Medical Screen to Allow the Most Severely Impaired
Medical evidence on an applicant's impairment is assessed under step 3 using codified clinical criteria relating to both the nature and severity of the impairment. These codified criteria, including more than 100 impairments, are called the Listings of Impairments. Applicants with impairments that "meet" the listings are allowed immediately, solely on the basis of medical criteria. Moreover, if an applicant has an impairment that is not included in the listings but is considered medically equivalent to a listed impairment, the impairment is said to "equal the listings" and the applicant is allowed. Applicants neither denied at step 2 nor allowed at step 3 have impairments that, although severe, are not severe enough to consider them disabled purely on medical grounds. Such applicants are evaluated further in steps 4 and 5. During 1990-1993, more than two-thirds of all initial allowances for adult applicants were based on the listings.
D. Step 4: Can Severely Impaired Applicants Work in Past Jobs?
In this step, an applicant's residual functional capacity (RFC) to meet the requirements of major past jobs is considered. The evaluation of residual functional capacity determines to what extent the applicant can perform basic work-related activities despite a severe impairment. This analysis takes into account, for example, whether the applicant can walk, lift objects, follow instructions, and tolerate environmental conditions found in the workplace. The work-related functions considered are the same ones considered in step 2, but they are evaluated differently. In step 2, the presence of such limitations is taken as evidence of a severe limitation, but in step 4 the extent of the limitations is considered, by comparing the applicant's residual capacity with the demands of past jobs. Hence, step 4 also involves determining the requirements of jobs held in the fifteen years prior to application, if they were held long enough for the applicant to learn the requisite skills. For example, if two applicants have similar leg impairments, the applicant who held a desk job is more likely to be judged able to do sedentary work and hence denied, whereas the applicant who held a physically demanding job would probably be evaluated further in the final step. In general, applicants judged able to perform past work are denied; the remaining applicants, including those with no recent work experience, are passed on to step 5.
E. Step 5: Can Severely Impaired Applicants Do Any Work in the Economy?
Under step 5, the applicant's residual functional capacity is considered, along with vocational factors, to determine whether the applicant can work in jobs other than those previously held. The analysis of residual capacity from step 4 indicates whether the applicant can do sedentary, light, medium, heavy, or very heavy work.(4) Under step 5, the applicant's age, education, and work experience are used to determine whether the applicant can work in employment consistent with his or her residual capacity. This determination is based on a table called the vocational grid. The grid, like the medical listings, lends objectivity to the determination process, facilitating uniform administration. To illustrate the use of the grid, suppose an applicant's impairment permits only sedentary work. If the applicant is of advanced age, low education (six years or less), and has held only arduous, unskilled jobs, then the applicant is judged disabled using the vocational grid.
III. The Structural Model
The intent of the sequential process is to provide a cost-effective procedure for determining disability that can be administered uniformly throughout the nation. Intuitively, decisions made at early steps are more straightforward to adjudicate, because they require only development of medical information, whereas medical-vocational decisions--made at steps 4 and 5--require both medical and vocational information. Hence, the process is made more efficient not only by breaking the determination into separate steps, each with distinct decision criteria, but by ordering the steps in terms of their informational requirements.(5)
We model the last four steps of the determination process, which include four decision nodes and five final outcomes, [d.sub.2], [a.sub.3], [d.sub.4], [a.sub.5], and [d.sub.5]. The decision nodes correspond to the four steps modeled and, for convenience, will be denoted k, l, tn, and n, respectively. As shown in figure 2, [d.sub.2] = denial at the second step based on severity of impairments; [a.sub.3] = allowance at the third step based on listed impairments; [d.sub.4] = denial at the fourth step based on residual capacity for past work; [a.sub.5] and...
Read the full article for free courtesy of your local library.
|
Bookmark this article
Print this article
Link to this article
Email this article
Digg It!
Add to del.icio.us
RSS
