Why a National Demonstration to Mitigate the Cash Cliff in One Safety Net Program Didn’t Increase Employment or Earnings

Why a National Demonstration to Mitigate the Cash Cliff in One Safety Net Program Didn’t Increase Employment or Earnings

Jun 29, 2022
On this episode of On the Evidence, Mathematica's Diane Beaver and David Wittenburg join the Social Security Administration's John Jones to discuss lessons from the Promoting Opportunity Demonstration that could help address cash cliff effects that disincentivize work in safety net programs.

On this episode of On the Evidence, Mathematica's Diane Beaver and David Wittenburg join the Social Security Administration's John Jones to discuss lessons from the Promoting Opportunity Demonstration that could help address cash cliff effects that disincentivize work in safety net programs.

Across many safety net programs, workers with low income and their families face the threat of a sudden and unexpected loss of benefits if their earnings increase too much, sometimes resulting in a net decrease in overall income. Policymakers have long worried that the phenomenon, often described as the cash or benefits cliff, discourages work and reinforces dependence on public assistance. Over the past decade, the Social Security Administration has launched two national demonstrations intended to mitigate this so-called cash cliff effect in the Social Security Disability Insurance (SSDI) program.

The latest episode of On the Evidence explores the results of those demonstrations and what they mean for future efforts to address program cliffs in the SSDI program and many other safety net programs. Our guests for this episode are John Jones, David Wittenburg, and Diane Beaver.

Jones is an economist at the Social Security Administration in the Office of Research, Demonstration, and Employment Support who has overseen several large-scale randomized controlled trials testing potential changes to the SSDI program, including two discussed on this episode: the Promoting Opportunity Demonstration (POD) and the Benefit Offset National Demonstration (BOND).

Wittenburg is a senior fellow at Mathematica whose research on interventions to promote employment for people with disabilities includes evaluations of POD and BOND.

Beaver is an advisory services analyst at Mathematica who spent more than a decade at a community nonprofit counseling people on how work and other entitlements would affect their Social Security benefits. In that role, she played a part in implementing POD and has firsthand knowledge of what beneficiaries experience as they navigate the patchwork of program rules that govern the amount of government aid people can receive for housing, food, child care, health care, and other needs as their work status changes.

During the conversation, Jones, Wittenburg, and Beaver discuss why the novel benefit offset tried in POD and BOND did not lead to increased employment or earnings. In particular, they discuss the challenges of timely reporting of monthly income, the cost to administer the benefit offset, and the implications of beneficiaries’ general confusion over program rules. Lastly, they discuss the policy implications of the demonstrations for the SSDI program and other safety net programs.

Listen to the full episode below.

A version of the full episode with closed captioning is also available on Mathematica’s YouTube channel here.

View transcript


I’m J.B. Wogan from Mathematica and welcome back to On the Evidence, a show that examines what we know about today’s most urgent challenges and how we can make progress in addressing them. On this episode, we’re going to talk about a common phenomenon in safety net programs called the cash cliff and what can be learned from a recent five-year, federally funded national policy demonstration intended to remedy unintended consequences of a cash cliff in the Social Security Disability Insurance program.

My guests for this episode are John Jones, Dave Wittenburg, and Diane Beaver.

John is an economist at the Social Security Administration in the Office of Research, Demonstrations and Employment Support who has overseen several large-scale randomized control trials testing potential changes to the Social Security Disability Insurance program, including two we discuss on this episode.

Dave is senior fellow at Mathematica whose research on interventions to promote employment for people with disabilities includes evaluations of demonstrations for the Social Security Administration that we discuss on this episode.

And Diane is an advisory services analyst at Mathematica who previously spent more than a decade at a community nonprofit counseling people on how work and other entitlements would affect their Social Security benefits. In her prior role, she played a part in implementing one of the demonstrations we discuss on the episode and she has firsthand knowledge of what beneficiaries experience as they navigate the patchwork of program rules that govern the amount of public aid people can receive for housing, food, child care, health care, and other needs as their work status changes.

In early June, I spoke with John, Dave, and Diane about the cash cliff in the Social Security Disability Insurance program because Mathematica and the Social Security Administration recently released findings from an evaluation of something called the Promoting Opportunity Demonstration, or POD for short. In a moment, we’ll get more into the details of POD and the evaluation findings, but I want to start by briefly explaining why policymakers worry about cash cliffs in safety net programs, including the Social Security Disability Insurance program.

A cash cliff, which can also be referred to as a benefits cliff, is a sudden and often unexpected decrease in public benefits that can occur with even a small increase in earnings. Such a cliff exists in many safety net programs, including Medicaid, the Supplemental Nutrition Assistance Program, the Temporary Assistance for Needy Families program, and the Social Security Disability Insurance program. Wage increases can result in a net loss of income or only a small overall increase, which can function as a disincentive to work more hours or earn higher pay. Policymakers who have long sought to encourage work and reduce dependence on government assistance programs worry that the cash cliff undermines both goals.

One possible solution to this so-called cliff effect is to adjust the eligibility rules for a program so that the cliff becomes an off-ramp with a gradual decrease in benefits that coincide with a gradual increase in earnings. In broad brush strokes, that’s what POD was designed to do in the Social Security Disability Insurance Program. Once earnings reached a certain point, every additional two dollars in earnings would result in a one-dollar reduction in benefits. The hope was that POD would promote work among beneficiaries, resulting in increased employment, increased earnings, and an overall reduction in benefits paid to people in the program as they became less and less dependent on the disability insurance.

The demonstration involved about 10,000 volunteers in 10 states, but many more people would be affected if the rule changes tested under POD were expanded and made permanent by Congress. In 2020, more than 9.5 million Americans received benefits through the Social Security Disability Insurance Program, which is the largest federal program that provides cash benefits for qualifying people with disabilities. A small share of beneficiaries already work and could earn more, but they don’t make enough to exceed the income eligibility thresholds of the program.

Unfortunately, the evaluation by Mathematica, which used a randomized controlled trial design, did not find that POD increased employment or earnings, nor did it find a reduction in the total amount of benefits paid. Those findings are largely consistent with the results of another five-year demonstration completed in 2018 called BOND, evaluated by Abt Associates and Mathematica, that also tried to mitigate the worst effects of the cash cliff in the Social Security Disability Insurance program. While neither demonstration yielded the impacts that policymakers had hoped for, they surfaced important insights about why the demonstrations didn’t have a bigger effect on employment and earnings, which could inform future efforts to address the cliff effect in the Social Security Disability Insurance program as well as other safety net programs.

Together, the evaluations raise questions about the administrative complexity and cost of replacing the cash cliff with a ramp. They shed light on the particular challenge of adjusting the amount of benefits someone receives based on changes in self-reported income each month. The research also suggests that beneficiaries’ confusion about program rules might make them less sensitive to changes to the current program’s work incentives than previously believed.

A full transcript of the episode is available on our blog at Mathematica.org. I hope you find the conversation useful.


I’ll put this question out to everyone. But I’d love to start with John. We’re going to spend a fair amount of time today talking about the problem of the cash cliff in disability insurance. And I know that other public assistance programs also have cash cliffs, sometimes referred to as benefit cliffs or the cliff effect. Would you mind defining what a cash or benefit cliff is and why it’s an area of concern specifically in the Social Security Disability Insurance program?


A cash cliff is often referred to a situation in which an individual receives some kind of financial support, and after receiving the support, they cross a barrier or a threshold that causes them to lose that benefit payment. So, they go from a state of having a benefit payment or a financial payment of some sort to a point where they no longer have it because they cross a threshold that caused them to lose eligibility to that benefit. So the Social Security Administration has a cash cliff in itself. It’s actually called substantial gainful activity. SGA for short. That threshold is actually a monthly earnings amount.

Now, SGA is actually a way that Social Security Administration judges whether a person has overcome the disability and is able to work. But most commonly, it’s a dollar amount of monthly earnings. So SGA is around $1,300 now. If you’re blind, it’s a little bit higher, around $2,000. But once you cross that threshold of SGA and you have sustained earnings above that threshold, you could lose eligibility to your benefit payments, because the DI program is receiving a lot of individuals who are disabled and then no longer can work, so we have this inflow coming into the program that’s much greater than the outflow; that is, people recovering from their disability and being able to get back to work. And so what happens when you have more people coming in than are leaving, the program gets larger, and that threatens the DI trust fund, which is the funding used to provide the payments to those beneficiaries. So, if you can’t regulate the flow of people coming into the program and leaving, then you stand a chance of making the program financially unstable, and so it jeopardizes the program itself.


Diane and Dave, is there anything you would add about defining this cliff or defining sort of the concerns around the cliff?


It is very difficult for not only the beneficiary to understand where they are relative to the cliff, but also for the agency to track exactly where a person might be relative to their trial work period or their extended period of eligibility. And SSA often has to do a lot of due diligence to understand exactly where a person is in regards to this.

The concern set of rules have, as John described, a lot of well-intentioned provisions to allow people to try work, but they’re so hard to follow that the implementation takes a benefits counselor to know exactly where people are relative to the cash work, which creates a lot of confusion, both with the beneficiary, but also at SSA, and can create unfortunate situations like overpayments. So, for beneficiaries that are currently working, they absolutely require the expertise of somebody like Diane.


Yeah, and I would just echo on that, that when you’re thinking about the cash cliff, that feeling of all or nothing is really prevalent for the beneficiary; right? It’s very scary to think about losing your SSDI check. And for some of us, we’re thinking, all right, what’s the average SSDI check? It’s like, what a little over $1,000 a month. But if you’re only earning about that much a month, or let’s say you’re earning $1,400 a month and you’re going to be over that SGA limit, you’re talking about 50 percent of your income disappearing all at once.

The concern is that when you’re a person with a budget and you’re looking at the amount that I have coming in at a monthly basis, you’re really trying to plan around that amount. And to tell somebody that you either can earn $1,349 and keep your $1,000 for SSDI, or whatever that amount is, or you have to almost double the amount that you’re earning right now, that you’re working, to make up that income. Those become really challenging choices for people and that’s, I think, the issue with the cash cliff. Someone can say, well, great, I can work 20 hours a week, and that’s really where I feel I am with my health to be able to earn, but I can’t work 35 hours, I can’t work 40 hours.

How do you create a system that can accommodate people for that? Because everyone’s situation is different. Even every state’s minimum wage is different. So the level of effort involved in getting to SGA is going to be really different depending on what the state’s minimum wage is, or the area’s minimum wage is. That’s part of the challenge too, in creating a national policy for Social Security is trying to accommodate what the whole landscape of our very diverse nation is like.


A couple of insights, Diane. thank you. I was reading a little bit about the benefits cliff in general, and there was some imaginary scenarios where someone could be considering going back to school or considering a job promotion and it actually would be worse for them to advance in their careers, because if they exceeded the threshold, they would actually have less overall resources when you combine their income and benefits, the concern that was being raised in one article from the federal reserve was that you could be not only decentivizing work but also career advancement.


As someone who is providing benefit counseling services to people, we evaluated those situations quite a bit. And part of what we did as providing benefits counseling, it’s not just looking at how your Social Security check is impacted but all the other benefits that you are on as well. Because, while Social Security, that lump sum is definitely usually the largest benefit that people are concerned about, they’re also concerned about their health insurance, they’re concerned about what their Supplemental Nutrition Assistance Program benefits are. People are going to be concerned about their housing benefits, which may or may not be income based.

So, it creates a large domino effect, where it’s not just one benefit that’s impacted, and, oftentimes, we could show people that when you start to work, you know, not limiting yourself to a life of poverty, which is what you’re doing if you try to stay under SGA. You’re going to limit yourself to be below the poverty line; that that risk, if you use your work incentive correctly, outweighs the guarantees of being on SSDI and a low-income part-time job; however, that’s really scary, and depending on if people have gone through all their work incentives and maybe didn’t know that they have, that cash cliff to be really present for them, and they could be making a decision today of, great, I’m going to take a job where I’m earning $1,800 a month, but that’s less than SGA, that’s $1,300, plus my Social Security Disability Insurance check, which was $900. So, why am I working 30-some hours a week to have less money than I would if I stayed at 15 hours a week? So, we played with all those sorts of equations, and that’s part of the personalized benefit counseling service is to help people run through scenarios so they could at least see at the end of the day what the money in their pocket is going to look like and make decisions that makes sense for them.


It sounds like a complicated task, and I can understand why individuals would need a benefits counselor to help them navigate the range of scenarios and choices.

John, I want to turn back to you and give our listeners a little context around some of the research that we’ll be discussing today. Historically, what has driven the Social Security Administration to test different program rules and test different benefit offsets? And what do policymakers within the agency and in Congress home hope to learn through these demonstrations?


For POD, the most recent demonstration, we were under the 234 authority. It’s a legislation that authorizes the Social Security Administration to change the rules of the DI program and test them. 234 authority is disappearing, we no longer have that authority. There is this kind of sunset clause in the legislation, and so after POD is finished, we no longer have the 234 authority to test policy changes within the DI program.

Now, Congress originated POD or the demonstration test. They had requirements that we had to follow. Like, for instance, we had to have a volunteer group of beneficiaries volunteer to participate. It requires a wet signature. They have to physically sign a piece of paper, acknowledging that they were aware of the experiment and all the rules that were involved. They also mandated that the program or the test would last no more than five years. So, that’s a very short period of time to get the experiment up and running, because there are legal requirements that we have to fulfill that take at least a year to fulfill, and then we also have to have time to recruit beneficiaries into the program, because it’s a voluntary program now that they have to volunteer into, so that takes quite a bit of effort.

And as far as the amount of time that we had to study the experiment, we only had, really, two observable years, because the last year is the year we have to start informing the beneficiaries that the demonstration is ending and transitioning them off the program, out of the demonstration, into the current program. And so as a result, that short window of time only allowed really, two years of clean observable information.


I want to talk about some of the specific ways that the Social Security Administration has tested different program rules to address the cash cliff in the past. What was the theory of change specifically in POD, and how does it build upon lessons from the previous Benefit Offset National Demonstration that some people may know by the shorthand BOND?


Sure. So POD stands for Promoting Opportunity Demonstration. And POD essentially was a test of a two-for-one benefit offset. There was other requirements in it, but that was the chief incentive. The two-for-one benefit offset works this way: It’s if you work and you earn a certain threshold level, that will trigger the offset; that is, a you earn $2 above that threshold, you would lose $1 of your benefit payment. So, for every $2 you earn above that threshold, you lose $1 of the benefit payment for each time you earn $2 above.

So, essentially what was happening in that demonstration is that we were replacing the cash cliff with a ramp. That ramp essentially is a 50 percent marginal tax on their benefit. So the cash cliff is a hundred-percent tax on the benefit for as little as $1 above the threshold. So, essentially what we were doing was replacing 100 percent tax on benefits for working with a 50 percent marginal tax. The reason why that was put in place, the theory behind it is that by reducing the severity of the exit barrier; that is, removing the 100 percent tax and putting a 50 percent tax in its place, it provides incentive for beneficiaries to try to work; that is, going back to work is not going to be so detrimental that they lose the entire benefit payments. What happens is, as they earn, their benefit payments will reduce slowly over time, and if they earn enough, they could actually zero out their benefit payments. We call that full offset. So that’s what POD was testing.


Dave, POD didn’t appear to have the intended effect on work and public benefits receipt, but the report that Mathematica put out has some important findings about why it didn’t work that could be useful going forward. What did you learn about beneficiaries’ understanding of the program rules, and why might that make a difference in terms of promoting work and increasing people’s earnings?


Yeah, J.B., thanks for the question. And the issue on understanding program rules does play directly into kind of the POD theory of change that John just mentioned, which is POD is a test of much more frequent but predictable earnings reporting versus the current system, with less demands on reports. That’s the current cash cliff. But the outcomes are less predictable, and the consequences of the cliff are severe; namely, under POD your benefit adjusts in the same manner for the same level of earnings each month, with a gradual reduction, whereas under current rules, your benefit adjustment can change depending on whether you are in the trial work period, as John had mentioned, or there’s a cash cliff where you lose your benefits.

So, what did we learn about the understanding of program rules? This gets to a fundamental problem of both current rules, as well as POD rules. Unfortunately, there is a lack of understanding of program rules even with very proactive benefits counseling. Under POD, less than half of the people understood that their benefits were reduced at any time that their monthly earnings were above the SSA threshold set for POD. The caveat is that this understanding was much higher among those who, not surprisingly, used the offset, and so there is a practical issue for those who used the program and understood it more clearly.

And to the other point of POD being a five-year demonstration, we saw understanding increase over time. And so, when people first got into the demonstration, they were fairly confused, and it took some time to adapt to the rules. But, later, in the period when both benefits counselors, as well as POD treatment group members, began to understand and become acclimated with the rules, there was much more timely earnings reporting.

Everything I might have just said says, oh, wow, that’s really bad that POD participants took so long to understand the program rules. But you have to remember, we’re also tracking what’s going on in the control group, and there, the understanding of the rules was even more severe. So, generally, less than half of the control members understood variations in current rules, including only 28 percent understanding there was a trial work period where their benefits remained unchanged regardless of earnings. So there is mass confusion out there, understanding what happens when I start earning money to my program benefits under current rules, as well as these new rules.


Can I ask one other question? I think that when a demonstration like POD doesn’t have the intended effect on work and earnings, the tendency is for people to equate no impact with failure. I was just curious what this group makes of that. How do you interpret a no-impact finding here?


People have to be very careful of interpreting no impact as failure, because if you’re afraid to test anything, you’ll never make any progress. And in the case of POD, without a doubt, the intended outcomes, particularly on employment and program benefit amounts, were not moved; however, there was not a negative impact on those outcomes. The real failure would have been not testing the intervention before going forward. If policymakers just decided, hey, the cash cliff is a bad idea, what we’re going do instead is a benefit offset and scaled that program, what would have happened was they would have spent billions and billions of dollars and they would have lost a lot of money, and there would have been a lot of burden on beneficiaries, as well as administrators, to shift those systems.

So even though the demonstration certainly didn’t meet the outcomes that I’m sure policymakers had hoped, and particularly around employment and earnings, the caveat is that, that doesn’t mean that you shouldn’t test these things. Google runs millions of experiments, and most of them fail. But the ones that succeed are the ones that push policy and innovation, and I think we have to consider that, particularly with public programs, and understanding the outcomes of POD.

The other thing that I would also push with POD, which is, if you think about POD as an overall test of what happens when you have more frequent reporting requirements with gradual offset reductions, I think it becomes more interesting to interpret, in the broader landscape, the questions you were asking at the top of the show, J.B., of like, you know, how should we think about offsetting people’s benefits, because what you’re doing, essentially, in tests like POD, is you’re increasing the monitoring of people’s work outcomes on a monthly basis, and what we found in POD is, not surprisingly, that puts burden on participants to report those earnings, but it also puts a burden on administrators to collect that information, and, as a result, there’s a cost to doing this.

So, it raises the question of, what is the rise-size way of essentially monitoring earnings? And there’s always this push-pull between program integrity intentions of, oh, we’ve got to monitor earnings to ensure that people are not overearning and somehow taking advantage of the system versus trying to support people going back to work and earning enough so that they can have a sustainable living. And I think that is truly one of the more interesting parts of POD, because it’s not just SSA who is struggling with this issue, it’s sort of right sizing what is the right reporting metric, because what we’re talking about is people having to report their earnings not only to the IRS but to SSA, and at what periodicity does that matter? Should that be monthly? Should that be quarterly? Should that be annually? Should that be on your IRS forms? I think that those things are really, really important questions to raise out of the context of POD.


John, is there anything you’d like to add, or that kind of get the gist of it?


Dave has said it quite well. I know from my own personal experience, reporting the results of a demonstration, that, oftentimes, you hear or even read about people saying, oh, that’s a waste of money. But they don’t understand the broader picture, because the reason why we were testing it is so Congress would know whether it was a viable policy to put into law for the entire nation. So, had they put these policies in place, we could tell them that it’s going to cost much more than the current program. And we don’t really see a lot of movement within that short period of time that we’re allowed to study it to justify actually implementing such a policy that doesn’t cover the cost and doesn’t seem to have a really large treatment effect.

And so what we’re essentially doing is exactly what Congress wants done; that is, test this thing to see if we should implement it. And what we’ve done is we showed them that, hey, if you do, the costs will exceed the benefit. So, it’s not a failure; it’s a success. So, oftentimes, people mischaracterize it by saying, oh, it’s a failure because it didn’t give Congress what they wanted to see. But in reality, we should recognize that tests that fail are oftentimes more valuable than tests that succeed.


That’s helpful framing. Thank you for that, John. The Mathematica report highlights some of the unexpected challenges of collecting monthly data on people’s income in order to adjust people’s benefits as their earnings change. Diane, I’m hoping you can speak from your prior experience as a benefits counselor, if you could speak to the experience of the beneficiaries here, what is that experience like of reporting data to the Social Security Administration? Under current program rules, what are the income reporting requirements for a Social Security Disability Insurance Program beneficiary? How often do you have to report? How do you submit information? How cumbersome is it? Give the listeners a sense of what it’s like from the beneficiary perspective, if you can.


Yeah, not a problem. So I will say that I think both beneficiaries and benefit counselors find the guidance around wage reporting to be unclear. A lot of it really depends on what your local office says they want. And so in the community that I worked in, I worked with two different Social Security offices, and they had two different kind of timeframes that they wanted beneficiaries to be reporting if they were on SSDI.

The guidance is generally that SSDI recipients should be reporting when there is a change in income that would change the receipt of their benefits. And just take a moment to think about that, because that requires, then, a person to have a comprehensive understanding of the benefits rules, as Dave kind of identified, which not a lot of people do, unfortunately. So, technically, what we always told folks is that people need to report when they start and stop jobs, and then when their wages are going to change from the below trial work period to above trial work period, or vice versa, or when they go above SGA or they go below SGA. So, it was always very easy for beneficiaries to grasp, I need to report to Social Security when I start and stop a job. That’s a very discrete, concrete moment to communicate with Social Security.

It was a lot harder to help beneficiaries to know when their earnings were going to go above trial work level and if that was relevant to where they were in their benefits experience, or if they’re going to be above or below SGA. One thing to think about is a lot of beneficiaries are working hourly-wage jobs, where their hours may not be consistent week to week, month to month, versus a lot of us, who are making these decisions are salaried. We know exactly what our income is going to look like. A lot of the beneficiaries I was working with, their schedules were going to be determined the week before, and that created a lot of challenges with trying to help people to understand what their income was going to look like.

What was also another challenge is that, as a benefits counselor, I was also advising people to use the work incentives that Social Security has in place, which some of which we haven’t even gone into, like an impairment-related work expense, which is a way to reduce the countable income that Social Security says you’re earning. And when you start to put in all those sorts of complications, the math is no longer as easy as, oh, I’m going to tell Social Security I started a job or I’m ending a job. There’s a lot more considerations that need to go into place. So, my complicated answer is, it’s not straightforward when to report or how to report.

Social Security has given people a number of ways that they can report. People can report online if they can create a mysocialsecurity.gov account, and you can FAX some documentation, you can mail in documentation, or you could always go in person, prior to the pandemic. That might sound easy, like, oh, I can call, I can just go online, I can mail, I can FAX. That sounds really easy. But a lot of times, it’s a lot more difficult in practice.

When someone does start working, you have to contact Social Security so that a person at Social Security manually adds that particular job to your account. In order to do that, the Social Security office needs to know what that employer’s tax ID number is. And they ask the beneficiary for the tax ID number. And I’m going to be honest, a lot the beneficiaries I looked at had a no idea what a tax ID number was. So, I’m the one that’s looking at EINs trying to make sure that this matches the business that this person is employed at. So, as the benefits counselor, I’m able to help facilitate that step.

But I imagine a lot of beneficiaries who don’t have a benefits counselor on their team wouldn’t really know how to go about that step to even get it into Social Security system so that then they could report that they’re working, and then be able to use the systems and processes that Social Security has in place to facilitate monthly reporting or quarterly reporting, or whatever it is that the Social Security office says they want from this particular beneficiary.

A further challenge is that if you use any of the fancy work incentives that Social Security has in place, like the impairment-related work expense, or you’re using a subsidy or a special condition, you can’t use any of the automated systems that are in place at Social Security. You have to go and actually talk to a person, or FAX or mail those in. And so, it makes it really cumbersome if you’re trying to use all of the incentives that are in place to communicate with Social Security. I can’t tell you how many hours I’ve spent on hold waiting to talk to a representative at Social Security in order to get some of these processes started for my beneficiaries.

So, you know, the other thing that I also got told is that a lot of people were told to just FAX or mail your pay stubs in. You know, do this frequently, let Social Security know how much you’re earning. And I was told by my multiple points of contact at Social Security that when you FAX or mail your pay stubs, they are sitting in a stack somewhere that a person at Social Security does not have time to go through and input each of these pay stubs in a timely manner. And I think that that’s another big piece of this, is that when we’re asking beneficiaries to report their wages, it’s difficult on the beneficiary end, and it’s difficult on Social Security’s end too. So it doesn’t work well for anybody. There’s no clean process right now that makes that happen.

A lot of the beneficiaries that I worked with would look at me and ask me why are we bothering to do this? What’s the point of all this effort to report wages? And, you know, my words to them were to prevent overpayments, to make sure that you are doing what you’re supposed to be doing. You were given the red book when you signed up for your benefits, let’s make sure we do what we’re supposed to do. But it makes it really tough, and I think that even when we were doing everything that we were supposed to do, quote, unquote, it was still difficult for Social Security to respond to those prompts in a timely manner, just given the volume that Social Security has to deal with.

And the other thing that I think about sometimes is that the DI program is such a small percentage of Social Security’s problems overall; right? They’re also supporting all of the retirement community. They’re supporting the SSI program. There’s a lot of other administrative work that Social Security is doing, and it’s just the process is always really challenge and frustrating, I think, for everybody involved.


Diane, you mentioned overpayments, and I just want to ask, beyond the initial frustration involved of reporting or collecting income data every month, I imagine overpayments are part of this. But what were the costs to the people and to the program? In POD, didn’t they discover that there were both over- and underpayments that they had to correct after the fact, that was somewhat of a byproduct of income data being collected or reporting initially not being correct on a monthly basis?


Yeah. So in POD, in comparison to current rules, in current rules, you report directly to Social Security. In POD, you got to report to your benefits counselor, and the benefits counselor had unique access to the Social Security system to help with the reporting to Social Security. So, a lot of times, the action was quicker on POD than you would see in normal rules. Even so, there’s always a lag with reporting wages; right? Think about it this way; in order for Social Security to have sufficient time to change your check, they have to set a deadline of when you need to give that information by, so that gives the beneficiary a really short window.

So let’s say, you know, it’s January, and I am reporting my income for January. I’m trying to figure out where my January pay stubs are and I need to communicate that to a benefits counselor, who then needs to send that on to Social Security to have that adjustment in my check. Social Security is always going to be a little bit behind, because you’re using retrospective information to determine payment based on those.

Generally speaking, outside of POD, let’s say I’m reporting a year’s worth of earnings at once and Social Security is compiling all that information and trying to figure out if they paid me the right amount. There’s a possibility of an overpayment at a certain point if you went over SGA and you were in your period of eligibility, or went over trial work month, and we can get into all that. In POD it was a lot quicker at finding those things. But there was a lot of challenges with reporting wages in general, because there’s certain things that Social Security can’t anticipate about people’s earnings; right?

We have moved into the gig economy, and many of the beneficiaries are working gig jobs that walk this weird line between wage employment and self-employment, and self-employment rules for Social Security are a whole other layer of complexity of trying to determine what countable earnings are for that program. And we would get questions from beneficiaries like, I’m walking dogs on WAG!. You know, here’s what I got deposited into my bank account, does this count? And that is a lot of conversation back and forth with Social Security on what do you want us to call this. How do we report this? How do we account for this? I think that there is a lot of labor involved on the beneficiary’s end, as well as on the benefits counselor’s end, as well as on Social Security’s end, three people really trying hard to make sense of what this income means and how it should be regarded and counted.

I also think that there is a challenge with asking people who have disabling conditions to bear the onus of reporting wages and pay stubs. Again, it sounds really easy, and it sounds like it is the fairest way to try to evaluate what someone’s earnings are. But we also found that a lot of our beneficiaries may not have had access to the internet at home to be able to access their pay stubs online. Even if they did have access to the internet, a lot of times people didn’t know how to access their pay stubs through the online portals, which I have a lot of empathy for, because I can’t tell you how many times I’ve forgotten my passwords to certain accounts; right? Even then, there’s maybe coaching someone on how to e-mail it. If they don’t feel comfortable e-mailing their pay stubs to their benefits counselor, then trying to coach someone to print it out, it’s just a lot of little logistical things that you don’t think about unless you’re the person trying to help somebody to get that information out of the computer and into Social Security’s hands. And a lot of that is really time consuming, hours to try to coach people on how to do that.

I think one thing we saw, like Dave mentioned, as we look at the POD data, people got better at it over time; right? People got used to the rhythm of reporting every single month, and it gets easier. But, again, we get into that whole time idea of, it takes time for someone to adjust to these new rules and this new expectation, and then you throw COVID in there and you have a whole other layer of complexity with POD that no one anticipated.


My understanding is that overpayment is something, really neither side wants, because if you get an overpayment based on bad data, out-of-date data or incomplete data that the administration is then going to have to go back and try and claw back that money.




And the person that’s received the benefits has to give up benefits, and these are individuals who don’t make a lot of money already, so there’s probably a meaningful loss of income.


Oh, for sure. Again, I think the thing to think about is that POD minimized those overpayments, which is awesome, because as the benefits counselor in normal rules, it wasn’t uncommon for me to see the beneficiary walking into my office with a letter saying that they owe Social Security $15,000, $40,000, very, very large amounts of money, where you look at this and you think, how’s the person ever going to pay this back? And these aren’t people we were helping along the way. These are people that come to us once there’s a problem because they hadn’t reported to Social Security in six years about their earnings from work.

In POD, you’re talking about people having a $100 overpayments. You’re talking about people having $60 overpayments. That is so much easier to manage and deal with, and that is certainly a positive. But, you know, going back to Dave’s point, what’s the administrative cost and benefit of having someone then be so on top of it to keep these overpayments so minimal, that’s what research is for; right? It’s to help us figure out what the best path forward is, because, obviously, the $15,000 overpayment is something we want to help Social Security avoid, and beneficiaries. But do we need to put the work in to monitor $30 bucks? It’s a good question.


Diane brings up an important fact that came out in the impact findings on overpayments in POD; right, which is the size of the overpayment in POD for the treatment group was much less substantial than it was for the control group. Why is that? Because people weren’t falling off of a cliff. But they were skidding down a ramp, to Diane’s point, which is there was a lot of these overpayments. So there was actually a higher prevalence of the number of overpayments in POD, even the amounts were overall lower, and I think that really gets to the point of policy when you’re trying to land a policy.

One of the fundamental challenges with an offset ramp is that you’ve got to get the earnings just right. You’ve got to get the adjustment just right or there will be an overpayment. Put differently, you might get a scrape on the elbow, and you have to compare that with a cash cliff, where you could fall off completely.


This is a two-part question to the group. I’d love to hear each of your answers here. Basically, where do we go from here, given recent evidence, including POD and BOND, where do we go from here with the Social Security Disability Insurance program? Are there ideas for addressing the cash cliff in SSDI that are currently being tested or that should be tested in the future? And then where do we go from here in terms of addressing the benefits or cash cliff across programs? Are there lessons from this research, this body of research that can inform public assistance programs taking a program elsewhere?


The Social Security administration has plans for demonstrations. One that seems to show promise, based on our past experience in BOND and POD, is a program that we call the STEPS program. It’s essentially an offset program, just like the Benefit Offset National Demonstration or the Promoting Opportunity Demonstration, there’s offsets. But the difference is, those offsets were continuous; that the beneficiary, their benefit payment changed for every dollar that they earned, and that’s difficult to navigate, because you have to figure out where you are on that ramp. So, if you’re fortunate to be in a job, working and drawing income, then you happen to get a weekend where you’re working longer hours, as a result, your earnings would go up but that automatically imposes a larger tax on the benefits.

So, a step process would be offering instead of a continuous ramp a step ramp, where there’s earning intervals, where your benefit payments would be constant over that interval. And once you go past the end of that interval, you’ll drop down to a lower step. It’s kind of like the way income taxes work. You have to figure out what bracket of earnings you have. Except in that demonstration I’m talking about, the proposal, is that the brackets would be a bit larger so that the beneficiaries would have certainty of where they are and wiggle room where they don’t have to worry about their benefits changing over that period of time.

So, what we’ve learned from BOND is that we can reduce uncertainty by simplifying the program, which we tried in POD, and the STEPS program would be further simplification of the rules and reducing the uncertainty for the beneficiary, which we think would have promise if Congress has the desire to see that demonstration, and also willing to let it run long enough so that we can see the equilibrium state.


Anything that we’ve learned from this body of work in SSDI that could inform taking up programs and methods of cliff problem more broadly?


To me, I always think broadly, and that the problem is, going back to the entry barrier to the program, carrying on into the post-program; that is, you enter a program through that entry barrier -- in our case, is it was SGA -- and then once you get into the program, you still face that barrier. So, welfare programs or safety net programs could learn from POD and BOND is that attach the post-experience from the pre-experience. In other words, if there’s a barrier at entry, it doesn’t necessarily have to be the same barrier after you’re in the program. So if you can change that barrier so it’s not just an entry barrier, as well as an exit barrier, if you can soften that, you can encourage beneficiaries to go back to work. And in both BOND and POD, we did see beneficiaries go back to work. It’s just they didn’t go into work in mass droves. So it could be that making that ramp less steep could be an answer, or like I said, it could be that we break it into intervals, earning intervals rather than a continuous steep sloping ramp.


Diane, where do we go from here?


Well, something that I would really want to see, especially as technology is improving, and the ability for government, like, entities to communicate with one another, is to figure out ways to remove the onus of wage reporting from the beneficiary onto other sources, especially because the information is now so much more accessible. When I look at the amount of effort that it takes for a beneficiary to report, it makes me wonder if there are other systems that can be in place that would make it easier, so it’s not cumbersome on the beneficiary, as well as on the benefits counselor, as well as on Social Security, especially because information sharing is happening. The Medicaid office knows when you start a job, so how do we get those systems to talk in the right way, so Social Security can make the right decisions about how much someone should be getting in a timely manner?

The other thing that really stuck out to me about the POD experiment, which, I think, applies to other programs, is that having a system navigator and having counselors available to people really matters. It makes a big difference when someone has somebody that they can call that can answer their questions that’s on their team that’s going to help them navigate through these challenging programs. And I know that that’s one of the biggest challenges that the beneficiaries that I’ve worked with would tell me before they were working with me, is that, I’d call this office and I don’t know if I’m getting ahold of anyone, or I don’t know when they’re going to call me back. I don’t understand this. And I think easing people’s anxiety about what is going to happen, what the rules are, having that be accessible, helps people to feel calmer about the decisions they’re making, and to know that when things change, there’s going to be someone that’s going to help them through.

We in the community-based world really saw work as an opportunity for recovery. And we saw people get better because they were working, because they were able to kind of reengage in their communities, and, you know, we wanted to see any ways that work could be incentivized rather than feel like you’re adding an extra step is really important to beneficiaries, and I think that Social Security is really mindful of that too, because they want people to recover just as much as the practitioners in the community, just as much as the beneficiaries do. So, I think it’s thinking about how do we frame work as a step towards recovery and work being recovery for people. And I think if we do that, we’re going to keep on creating better policies going forward.


Dave, I’ll give you the final word here. Where do we go from here?


I think what Diane said about onus is really important. Three quick points on this; the first is around where the burden lies for things like overpayments. Currently, beneficiaries, when they don’t report on time, experience a benefit adjustment, so the onus is on them. Pamela Herd recently talked about potential shifting administrative burden, and, particularly in a complex program, you can think of things with things like Social Security, shifting the burden of recording of earnings when somebody hits their trial work period, SSA sending direct notices to the beneficiary, hey, you’re at the end of trial work period, now this is what’s going to happen to your earnings, as opposed to a beneficiary sort of walking into this. So, point one is really reinforcing this point on onus, of how much should fall to the beneficiary versus how much should fall to the administration; that could fundamentally improve trust.

Point two, and this plays into point one, is building trust among beneficiaries. That’s one of the key things when people are not fully understanding the rules, they’re often not invested because they’re very complicated and thinking through options of highlighting different rules, and their presentations to beneficiaries, of whether they would opt into them. And this type of exploratory work doesn’t necessarily need a full demonstration, but looking in smaller samples to assess the foundations for how beneficiaries might respond to incentives when offered at a larger scale.

And the third idea of learning from this is, one of the findings we keep talking about is, it takes time for people to adjust to the new rules. And so thinking about the population receiving services, not being exposed to the existing rules, so really thinking about, say, new awardees. So, one way to improve beneficiary understanding of modified rules and future demonstration is to offer the demonstration supports to cohorts of new awardees who aren’t exposed to existing rules, and so that you can really test whether or not the existing rules are causing harm to beneficiaries through their lack of knowledge. And I think that that could be a fundamentally important way of rethinking about how to do demonstrations, not only within SSA but any sort of support program that is serving long-term beneficiaries, because there is always a period to adjust to new rules.

And as my son once told me, you know, like whenever we’re thinking about our own rules in the house, like, once it becomes too much work, people immediately will stop. But if you make it much more transparent and easier and the rules more readily accessible, your potential for having much larger impact is higher.

And one of the demonstrations that I think would be very interesting to test is SSA’s Ultimate Demonstration, which says, hey, we’re going to temporary restrict all adjustments to earnings. So imagine a demonstration where you don’t have to report any earnings to SSA, and then looking at what has to the employment response. And a critique of that demo would be, well that’s not generalizable. But it is a demonstration that would tell you how many people might actually respond to a new set of incentives.

And from there, you might be able to revise incentives a very productive manner, as opposed to sort of always waiting from the cash cliff.


What Dave’s talking about, the Ultimate Demo, what it would do is serve as a way of figuring out what is the upper bound that financial institutions could have in getting beneficiaries back to work. Because if we do that ultimate demo and it shows that the upper bound is still not cost effective, then that would suggest that maybe beneficiaries are too disabled to go back to work, and all of this effort is a waste of time. So the Ultimate Demo, if we were allowed to do that, would provide tremendous amount of information that could inform Congress about policy-making.


That’s helpful, yes. And, Dave, I identify with your son’s wise comments about cognitive overload and how that can paralyze people from trying something new. I just want to say thank you to all three of you for being so generous with your time in talking with me today.


Thank you, J.B.


Thanks for having me.


Thank you.


Thanks to my guests, John Jones, Diane Beaver, and Dave Wittenburg. In the episode show notes, I’ll include links to any research we discussed on the episode. A full transcript is available on a blog associated with this episode at Mathematica.org. There are a few ways you can keep up to date on future episodes. You can subscribe through Goodpods, Apple Podcasts, Spotify, Stitcher, Google Podcasts, or wherever else you listen to podcasts. You can also follow us on Twitter. I’m at JBWogan. Mathematica is at MathematicaNow.

Show notes

Read the final evaluation report on POD.

Read the final evaluation report on BOND.

Find a summary of lessons learned from several decades of demonstrations by the Social Security Administration to test policy ideas in the Social Security Disability Insurance and Supplemental Security Income programs.

Learn more about the potential Ultimate Demonstration.

Watch a recorded discussion between Pamela Herd and Sebastian Jilke, professors at the McCourt School of Public Policy at Georgetown University, about administrative burden in the Social Security Disability Insurance program.

About the Author

J.B. Wogan

J.B. Wogan

Senior Strategic Communications Specialist
View More by this Author