Committee for a Responsible Federal Budget
Social Security_Card_Layered

Developing Social Security Disability (SSDI) Reform Demonstrations to Improve Opportunities and Outcomes Based on Lessons Learned

by Jason Fichtner and Jason Seligman

Read the full paper.

I. Introduction

There is both strong interest and strong skepticism in pilots and demonstration programs for the Social Security Disability Insurance (SSDI) program. The Government Accountability Office (GAO), Social Security Administration (SSA), and the research community generally agree the demonstration process has suffered from substandard design and execution. There are many questions surrounding future protocols, such as which agency or entity should decide what to implement, who should administer them; how to ensure adequate data are collected from participants as they engage and disengage at various points, who should perform the subsequent review and analysis, whether participation should be voluntary, and when pilots should be expanded or discontinued. Past demonstration projects and the program design literature both offer important lessons going forward on what works and what doesn’t work. We bring these lessons to bear on our own design of demonstration projects to help beneficiaries engage work. These policy recommendations target reforms for the SSDI program in the window of time currently granted by the 2015 Bipartisan Budget Act (BBA). Being pragmatic, we target current issues; being policy-process oriented researchers, it is our intent that the tenets we offer will be applicable to project design and execution well beyond the BBA window.

Primarily through a temporary payroll tax reallocation, the 2015 BBA avoided a pending 2016 trust fund depletion and has extended the exhaustion of the disability trust fund from an initial estimate of 2022 up to 2032 under the assumptions of the 2018 Social Security Trustees Report. That is good news; however, some continue to feel that the 2032 estimate is highly uncertain and thus possibly over optimistic. Whatever the case, the time from now until 2032 is still short when considered against a backdrop of two decades in which research and demonstration projects have failed to identify robust paths to work for SSDI program participants.

The 2015 BBA calls for demonstration projects that help beneficiaries engage in work. Currently, the SSDI program is a reasonably well-designed system of policies and procedures meant to support those who cannot work during a longer-term disability. However, changes in health, medical technology, and the workplace increasingly offer some potential applicants and some participants a better opportunity to support themselves through work than the current program does. It is somewhat tragic then that demonstrations of program components attempting to support engaging work have not lived up to promise of helping those who can do better, to do better. To be fair, reforming the SSDI program in this way takes a great deal of consideration. The program does a lot of things quite well and interacts with other programs in important ways that must be considered when designing opportunities to engage work. These include Medicare, Medicaid, the Social Security retirement program, and the more recent Americans with Disabilities Act (ADA) of 1990 and Affordable Care Act (ACA) of 2010. These last two programs are newer, and we are still learning how they may improve or otherwise alter work motives among SSDI applicants and awardees. There are many program interactions to consider when designing and testing reforms, which program modernization must address. Further, to best succeed, program changes will need to be field tested.

Research programs should be designed and evaluated properly; design should be informed based on the lessons learned from previous research and demonstrations. We believe these insights are immediately necessary, as SSA must begin fielding new demonstrations very soon. With limited time, SSA must consider what sorts of demonstrations can be fielded simultaneously (in parallel), which must be sequenced, and how to effectively design and administer such demonstrations. When thinking about time-related pressures, we appeal to minimizing project scopes in ways that improve efficiency without sacrificing the robustness of findings. Doing this requires pre-planning, via Power Analysis. Project designers need to think about what minimum impact a demonstration project needs to achieve in order to contribute a meaningful reform; this is what researchers will need to be able to detect.

Notably, program changes with large impacts require smaller samples to detect, whereas to detect slight changes or improvements, larger samples are required. Especially for voluntary research and demonstration projects, the time to engage a large sample is an important investment. Consider that some tested reforms will have larger and some smaller changes in participant behavior as a goal, based on differences in the minimum necessary contribution of a reform. For those where the largest goals are sought, it may make sense to test several variants of a program change at the same time with smaller samples, and then move forward with successes from the batch of variants. This may be somewhat counterintuitive because it means that program reform components requiring the largest behavioral changes in order to be meaningful can be tested in smallest populations. Because big changes for these components are most necessary, and because results are unsure, parallel testing of variants in these small batches further makes sense.

Based on our review of previous research and demonstrations and the program design literature, we define a suite of new projects and a related sequencing protocol to inform policymakers in the window of time allotted. We further work beyond this minimum and consider what projects might afford continued advancement of the program given the changing nature of work in our evolving society. We will address the role for a permanent program component devoted to research and development and offer principles for evolutionary improvements to the SSDI program at minimum resource cost.

The 2015 BBA acknowledged the poor financial state of the SSDI Trust Fund, which at the time was roughly one year from being depleted. Congress reallocated some portion of the payroll tax funds from the Old-Age and Survivors Insurance (OASI) Trust Fund to the Disability Insurance (DI) Trust Fund to allow the continued payment of full benefits under the current system through fiscal year 2022, as estimated at the time of the act’s passage. According to the most recent report of the Social Security Board of Trustees, the DI Trust Fund will now be depleted even later, in 2032, slightly over one decade from now. Acknowledging the financial and programmatic challenges of our nation’s current Disability Insurance design for both beneficiaries and taxpayers, Congress took the opportunity to offer SSA resources to test projects that promote workforce attachment. Specifically, sections 821–23 of the BBA granted authority to engage in demonstration projects under “Subtitle B—Promoting Opportunity for Disability Beneficiaries.”

In a series of earlier writings (Fichtner and Seligman 2014, 2015, 2016a, 2016b, 2018), the authors of this paper detailed tenets for disability insurance reforms that are designed to integrate well with (1) broader U.S. Social Security retirement program reforms – which are themselves necessary ahead of 2034 (the date currently estimated for the depletion of the OASI Trust Fund), (2) other developments in the U.S. social safety net since 1965 – specifically the Americans with Disabilities Act of 1990 and the Patient Protection and Affordable Care Act of 2010, (3) opportunities for reintegration of disability insurance beneficiaries into the workplace, and (4) lessons from other nations’ disability insurance reform efforts. In this paper, we consider additional lessons learned from demonstration programs in general as well as tenets for the design of demonstration projects that can test some of these ideas. We also consider whether in retrospect some of these studies could have been broken up into smaller components, or whether sample sizes could have been amended to yield results more efficiently.

Even given the extended timeline recently offered by Social Security Administration actuaries, the relatively short window of time from 2018 to 2032 suggests that the design and implementation of testing cannot be put off. Congress is depending on the data and experiences accumulated through earnest work to evaluate various reform proposals – as it has made clear by the resources it has allocated to program experimentation activities.

Proposing program reforms without testing them and exploring the associated nuances of practiced implementation can be a recipe for failure, not only in terms of budget savings but more importantly in terms of the social benefit the program yields to society. A suboptimal system harms those it fails to serve, be it before or after reform. Those the SSDI system is meant to serve are defined as disabled, and they deserve a program that supports their convalescence as well as their rehabilitation. A better-designed program will do both. Hence, a goal of this paper is to contribute to the delivery of more successful executable designs. There will, of course, be many important and detailed aspects of demonstration project design that this piece will not and cannot address because the context of specific field research conditions is not fully known to us at this time. Rather than dabbling in false precision, we will lay out principles for the design of a suite of demonstration projects. Here is a summary of these principles:

  1. Demonstration projects should test things that are uncertain. In other words, testing should be limited to changes or features that have not been demonstrated to succeed or fail in accomplishing their objectives in the past. This will make the best use of lessons learned from past work and of the time remaining.
  2. Demonstrations should limit the number of changes per project whenever possible and should otherwise be designed so that it is easier to determine the contribution of each change to the program and those it serves.
  3. Demonstrations should be run in parallel when possible. This will make the most of limited time and will hold baseline economic conditions as constant as possible for the sake of facilitating more controlled comparisons.
  4. When demonstrations cannot be run in parallel, they should be sequenced so that knowledge gained from testing prior features can be built upon. The order of operations is important to consider – sequence has consequence.
  5. Demonstrations will necessarily rely on volunteers, as prescribed in the law, which may make their results less generalizable. However, pure experimentation is not a requirement in order for results to be generalizable. Matched randomization of participation among the willing can help inform the debate and provide useful evidence for national program design and rollout. Notably, if estimated impacts are expected to be larger in a voluntary population, then smaller samples may be needed. Of course, the opposite is true as well. If voluntary groups are less likely to carry through with exercises in one or another reform pilot, secure in the knowledge they can revert to older practices, then estimated impacts may be lower, and sample sizes should be higher to detect smaller impacts in the testing phase. Regardless of the direction of impact from program to program, once reforms are broadly implemented, policymakers and administrators must be prudent regarding estimated improvements in service benefits and service costs, especially as reforms initially scale up.

We place these five basic principles for the organization of demonstration projects here, acknowledging that they are relatively straightforward and hoping that they are easy to keep in mind as one goes through the reforms we next describe. In our previous work, we emphasized three main proposed program reforms for Disability Insurance:

  1. Changing the structure of the program to include both temporary and partial disabilities.
  2. Integrating existing employer-based disability insurance programs into the national disability insurance system, and supporting the expansion of such programs.
  3. Integrating other federal social support programs into an overall system designed to support the disabled in various ways, including by promoting their recovery and rehabilitation to meaningful workforce participation. By “meaningful” we mean participation that enhances participants own welfare and that participants themselves understand as being to their benefit.

In our work to date we have often pointed out that these three reform channels can work as standalone reforms, or they can be integrated – and our previous research offers a nice context for dividing and conquering the work of testing program reform features over the limited time ahead, before the estimated DI Trust Fund depletion date. The rest of this paper proceeds as follows: In Section II we provide a review of the successes and failures of selected research and demonstration projects. Following the literature review, in Section III, we consider how lessons learned and the five broad principles we introduced above can come into play when designing and fielding projects targeting important aspects of our three proposals for reform between now and 2032. We then propose SSDI demonstration projects along these principles in Section IV. Section V summarizes what we have discussed and offers some public policy implications. Section VI concludes by emphasizing again that these suggestions are not meant to be a detailed blueprint but rather are meant to serve as an outline based on fundamental principles.

Read the full paper.