Until June 2024, Social Security adjudicators were denying disability benefits by telling claimants they could still work as pneumatic tube operators or nut sorters. These weren't outlier cases. The Washington Post covered it, Congress looked at it, federal courts kept questioning it, and the practice continued because the federal catalog of occupations Social Security uses to decide whether disabled people can do "some other job" was last updated in 1991 — before widespread internet, before email was common, before the modern economy existed in any recognizable form. In 2024, the agency issued two emergency messages. The first removed 114 obscure jobs nobody was actually citing, like astronomer and oral surgeon. The second imposed a documentation requirement on the 13 jobs that mattered — the ones courts had been complaining about — without prohibiting them. The catalog itself wasn't updated.

This episode is a stress test rather than a new doctrine. The first half of the season built a model of how administrative adjudication is supposed to work — due process, the inquisitorial role of the ALJ, credibility findings, the structural mechanisms that actually steer outcomes. This episode asks what happens when you run that model 2.5 million times a year. The catalog isn't really the problem — it's the symptom of a system that has to decide that many cases and can't afford to look too closely at any one of them.

Social Security receives roughly 2.5 million disability applications a year — the population of Chicago, applying for federal benefits, every twelve months. Around 60-70% get denied at the initial paper review stage by state disability examiners the claimant will never meet. Most reconsiderations affirm those denials. Only then does the case reach an ALJ, where approval rates rise dramatically — a signal that the front-end paper process is producing results the hearing keeps correcting. The backlog at hearing offices sat at 330,000 cases as of January 2026, with average wait times around nine months, down from 595 days in fiscal year 2018. People die waiting. That's not the system failing in some exotic way — it's the system working as designed at volume.

The doctrinal heart of disability adjudication is the five-step sequential evaluation. Step one filters out people earning above $1,690 per month. Step two filters out people with truly minor conditions. Step three approves the most clearly disabled — about 10-15% of approvals — based on medical listings. Step four asks whether the claimant can return to their past work. Step five — where most cases are won and lost — asks whether any other job exists that they could do. This is where the 1991 catalog operates, and this is where the medical-vocational guidelines (the "grids") take over. The grids are literal rows and columns: four inputs (residual functional capacity, age, education, work experience) produce one output (disabled or not). The table decides the case for an enormous category of step-five claims.

The Supreme Court blessed this in Heckler v. Campbell (1983), holding that the existence of jobs in the national economy isn't unique to each claimant and can be resolved through rulemaking rather than individualized testimony — because requiring otherwise "would hinder needlessly an already overburdened agency." That phrase is the doctrinal admission that mass adjudication requires trading individualized judgment for rule-based output. The grids are administrative law's clearest example. The cost lands on the people who don't fit the table cleanly, and the brittleness shows up most sharply at age boundaries. A claimant at 54 and a claimant at 55 with identical files can get opposite outcomes because the grids treat the move from "closely approaching advanced age" to "advanced age" as a structural break in the labor market. A regulatory rule exists to soften that cliff in borderline cases, but the agency reads "borderline" narrowly — the Fifth Circuit affirmed a denial in 2025 where the claimant was five months shy of 55.

The constitutional layer makes the picture more uncomfortable, not less. The 2.5 million initial applicants don't have a property interest in benefits they haven't been awarded, so the Constitution requires no particular process — what they get is whatever Congress chose to provide. The Mathews v. Eldridge balancing test governs only the post-award population: continuing reviews, terminations, overpayment recoupments — itself a multi-million-case-per-year system. Mathews built administrative burden into the test as the third factor, and accepted scale as a legitimate counterweight from the start. The doctrinal question isn't whether the test is being misapplied. It's whether a test that counts administrative burden as a factor at all can ever require more process once the system is big enough that the burden factor dominates the math. The test has no floor.

The episode closes by pointing forward. Social Security is the largest administrative adjudication system in the country, but it's not the only large one. The next episode covers immigration adjudication — specifically asylum — where you might expect stronger procedural protections given that the question isn't income but whether someone gets to stay in the country. The protections are weaker. The volume problems are worse.

Listen Now

Spotify | Apple Podcasts | Listen on our site

What We Cover

• Why Social Security was denying disability claims by citing pneumatic tube operator and nut sorter as jobs claimants could still do — and what the 2024 "fix" actually changed
• Why the underlying federal catalog of occupations dates from 1991 and still governs step-five denials despite courts and Congress raising concerns
• The 2.5-million-application-per-year scale of Social Security disability, and why 60-70% are denied on paper before any judge sees the file
• Why approval rates jump sharply once a case reaches a hearing, and what that signal means about the front-end paper review process
• The five-step sequential evaluation framework, walked through with a hypothetical 54-year-old warehouse supervisor with degenerative disc disease
• How the substantial gainful activity threshold ($1,690 per month in 2026) functions as a categorical filter at step one
• Why the medical listings at step three approve only about 10-15% of cases, and what the residual functional capacity finding at step four actually involves
• The medical-vocational guidelines ("the grids") at step five, and how four inputs produce one categorical output that the table decides
• Heckler v. Campbell (1983) and the Supreme Court's blessing of rule-based adjudication when individualized determinations would "needlessly hinder an already overburdened agency"
• How the age categories produce cliff outcomes — same file at 54 vs. 55 can flip the result — and why the regulatory rule meant to soften that cliff has narrow edges of its own
• Why the Mathews v. Eldridge balancing test governs only post-award cases, and how scale can dominate the third factor at the volumes Social Security operates
• Why initial applicants get whatever process Congress chose to provide rather than constitutionally required process
• How the structural mechanisms covered earlier this season — variation, credibility, agency steering — all flow through credibility findings and RFC determinations to produce step-five outcomes

Full Rough Transcript

Episode 16 — Social Security Disability: The System at Scale

Accurate Transcript by Speaker

[Standard intro]

Gwen: Hello, and welcome to Administrative Remedies, because you can't fix what you don't understand. Brought to you in part by the University of Tulsa College of Law. I'm Gwendolyn Savitz, an associate professor here at TU and the associate dean of research and intellectual life.

Marc: And I'm Marc Roark. I'm the dean of the College of Law.

Gwen: We'll be breaking down complex doctrines with real-life analogies and examples to demystify the world of administrative law for everyone trying to understand how government actually works.

Marc: Agencies are the main way the federal government gets things done. It's not through Congress, for reasons we'll be addressing over the course of this series.

Gwen: All right, Marc, picture a job — pneumatic tube operator. You receive cylinders that arrive through a network of tubes running through a building. You unload whatever's inside, paperwork or small items. You route them to the right department and you send the tubes back.

Marc: Sure, like what they use at pharmacies and banks at the drive-thru.

Gwen: Yeah, they still have a specific purpose, but what they don't really have are people dedicated solely to operating them all day long. In large department stores, hospitals, and office buildings, that used to be an actual job, doing that all day. All right. Let's talk about another one now — nut sorter. You sit at a conveyor belt and you separate the good nuts from the bad nuts as they go by, by hand.

Marc: Yeah, the trained squirrels in Willy Wonka, right? So not really a thriving profession.

Gwen: Right. Except until June of 2024, the Social Security Administration was using both of these jobs and others like them as reasons to deny people disability benefits.

Marc: Okay, what do you mean, using them?

Gwen: The last step of the disability evaluation — and we will get to this in detail — the agency has to show that even if you can't do your old job, there's some other job you could do. And when it's looking for that other job, it reaches for a federal catalog of occupations. And that catalog's last edition came out in 1991.

Marc: 1991.

Gwen: Right. And for years, if you applied for disability and you couldn't go back to your old job, the agency might tell you, well, you could be a pneumatic tube operator or a nut sorter, and deny your claim on that basis.

Marc: So real people have actually been denied disability because a federal adjudicator said that they could potentially be a pneumatic tube operator.

Gwen: Thousands of times. The Washington Post did a piece on it. Congress looked at it. Courts have been complaining about this for years.

Marc: And in 2024, they fixed it, right?

Gwen: Well, they did something. They issued two emergency messages on the same day. The first one identified 114 occupations that it called isolated. These were jobs that only existed in very limited numbers, like astronomer or oral and maxillofacial surgeon or director of a state historical society. Adjudicators can't cite to any of these to deny a claim anymore.

Marc: Those don't sound like the jobs we were just talking about, though.

Gwen: No, they're not. And that's the point. So most of those on the first list weren't jobs that people were citing in the first place. Nobody was saying, we understand you can't go back to your old job, but that's fine, you can just become an astronomer.

Marc: Yeah. So what about the pneumatic tube operator or the nut sorter or the document preparer or the addresser?

Gwen: All right. Those are on the second list. Those are 13 occupations that had been showing up in denials and that federal courts had been questioning.

Marc: So the second list is where the reform actually bites.

Gwen: Right. It doesn't prohibit citing those jobs, but it requires the ALJ to get additional evidence from the vocational expert that the job, as it is currently performed, is consistent with the claimant's functional capacity and exists in significant numbers.

Marc: Okay, hold on. That's not a fix. That's a paperwork requirement. The same VE who was citing pneumatic tube operator in 2023 can now cite it in 2026 as long as they say the magic words. Hundreds of cases per year per judge — how much friction does adding a sentence to the testimony actually create?

Gwen: Well, potentially not that much, but that's the question of the whole episode. Notice what they didn't do in either message. They didn't update the catalog. They didn't revise any of the job descriptions. They didn't fix how the agency understood modern work. They pulled a few entries off one list and they added a documentation step to another. Although certainly if someone were denied because it was claimed that there were more than a thousand local pneumatic tube operating jobs, that would be something they could challenge.

Marc: Okay, I feel like I should be angrier about this than I am. Why am I not that angry?

Gwen: It might be because you're not applying it to the 2.5 million disability applications that Social Security receives every year.

Marc: 2.5 million?

Gwen: We've talked about it before, but really we're talking about the population of Chicago applying for disability every year. And every one of those applications has to get decided. Someone has to look at the medical evidence, figure out whether this person can still work, and say yes or no.

Marc: Okay, I think I see where you're going. The stuff we've been talking about this season with due process, the inquisitorial model, the ALJ developing the record, credibility findings, internal appeals — those are designed for the individual case. We were talking about one claimant, one record, one hearing, one decision.

Gwen: And today we're asking what happens to all of that when you run it two and a half million times a year?

Marc: And the 1991 catalog is what it looks like.

Gwen: Right. The catalog isn't the problem. The catalog is the symptom. The problem is a system that has to decide all these cases and can't afford to look too closely at any one of them, including the question of whether the job the agency says you could still do exists in the form it's described.

Marc: So this isn't a new doctrine episode.

Gwen: Right. It is a stress test episode. We've spent the season building a model of how adjudication is supposed to work. And Social Security is what happens when you run it at industrial volume. And this 1991 catalog is one of the places where the strain shows. Plus, it's also just generally interesting.

Gwen: So there are roughly two and a half million applications every year, and most of them get decided on paper at the initial stage. And most of those initial decisions are denials.

Marc: So how many denials of that two and a half million are we talking about roughly?

Gwen: So they historically run 60 to 70 percent. So out of two and a half million applicants —

Marc: A million and a half-ish are being told no in the first round, before any hearing, before any judge, just based on the paper review by a state disability examiner working with the medical file. And then what happens to those denials?

Gwen: They could ask for a reconsideration, which is another paper review by a different examiner in the same state agency. Most reconsiderations affirm the denial. Only then do you get to ask for your hearing in front of the administrative law judge.

Marc: So by the time you actually see a judge, you've already been told no twice.

Gwen: Right. And statistically, the people who make it to the hearing level are much more likely to win than people at the earlier stages, which is kind of telling on its own.

Marc: That's a strange signal. The initial determinations are getting things wrong often enough that a meaningful percentage of them flip when someone actually listens.

Gwen: That would certainly be one reading of it. The agency would tell you that the cases look different by the hearing stage because by that point, more medical evidence has been developed. Both things are probably true, but yes, the paper review process at the front end is doing something the hearing keeps correcting.

Marc: And how many hearings is that now?

Gwen: Hundreds of thousands a year. And these are conducted out of the more than 160 hearing offices scattered across the country. And their own performance dashboard put the backlog at about 330,000 hearings as of January 2026, with an average wait time of about nine months, which is actually a big improvement. In 2023, the average was more like 15 months. And it was worse in fiscal year 2018 when it was 595 days.

Marc: Wow. People wait over a year, sometimes years, multiple years for their hearing.

Gwen: Right. And so before we go further, nothing we're about to talk about is an accusation of misconduct. The ALJs, the disability examiners, the support staff — they're doing their jobs. The system itself is what's under stress.

Marc: So when I hear a judge is deciding hundreds of these cases a year...

Gwen: This is not a criticism of the judge. That's their job. We talked about this in the inquisitorial model episode. The ALJ isn't a neutral referee listening to two sides. The ALJ is supposed to be developing the record, questioning witnesses, and deciding the case.

Marc: So the judge is theoretically the prosecutor's job, the defense attorney's job, and the judge's job all at once.

Gwen: Right. And they get to do that several hundred times a year. So you can ask how deep the record realistically is going to be developed in any single case.

Marc: The structural problem isn't that anyone is lazy or careless. It's that the model assumes an amount of attention per case that's just mathematically impossible at this volume.

Gwen: Yeah, that's part of what we're talking about this episode. All of this is in one way or another a response to that math.

Marc: Okay, how do you actually decide in two and a half million cases? What's the framework?

Gwen: This is the five-step sequential evaluation. It is the doctrinal heart of how disability gets decided, and we haven't really talked about it before. Let's look at 20 C.F.R. § 404.1520.

Marc [reading]: The sequential evaluation process is a series of five steps that we follow in set order. If we can find you are disabled or not disabled at a step, we can make our determination or decision, and we don't go to the next step. If we cannot find that you are disabled or not disabled at a step, we go on to the next step.

Gwen: All right. So there was a previous episode where we talked about Kathleen. She was 54, living in Oklahoma. She'd worked 22 years as a warehouse supervisor. She had degenerative disc disease and she'd gotten denied at the initial stage by a state DDS analyst she'd never met. Then she got denied again at reconsideration by a different analyst at the same agency. And we sat with her in her hearing. But what we weren't walking through is exactly what the determination was based on, what those five steps ask. That's what we do today.

Marc: Okay, step one.

Gwen: Substantial gainful activity. Are you working right now above a certain income threshold? If yes, you're automatically denied. It doesn't matter how sick you are. What amount do you think that is?

Marc: Oh, I would say it's got to be above the poverty rate. I'm going to say $40,000.

Gwen: That's per month.

Marc: Oh, per month. So per month, 40,000 divided by 12 would be 3,000. So I'm going to say $3,500 a month.

Gwen: All right. It is actually $1,690 per month.

Marc: Wow. So if you make more than $1,690 per month, you're automatically denied.

Gwen: Yes. So here Kathleen lost her job. She used it for Family Medical Leave Act. She couldn't perform the essential functions anymore. So she's not working. She's earning nothing. So she passes step one.

Marc: Okay, what's step two?

Gwen: Step two, we ask, is there a severe impairment? Does she have a medically determinable impairment that significantly limits basic work activities? Degenerative disc disease with imaging to back it up? Sure. That counts as severe. She passes step two. This isn't a really hard bar to pass. And step two is mostly a filter for people with really minor conditions.

Marc: Right. So mild allergy, a sprained ankle, these won't make it past step two.

Gwen: But most serious medical conditions do. Step three is where things start to get interesting.

Marc: Okay.

Gwen: This is where we say, does your condition meet or equal one of the listings? The listings are Social Security's catalog of conditions that are presumptively disabling. If you can document that your impairment matches the specific medical criteria of one of these listed impairments, you're approved without having to go any further. About 10 to 15 percent of approvals happen at step three. This would be like you've been diagnosed with Huntington's disease or ALS. They're not going to need a lot of time figuring out exactly what you can do.

Marc: Okay. So most people don't clear step three.

Gwen: Right. Most people don't get approved at step three. We still have to keep going. So here for Kathleen, her degenerative disc disease is real, but it probably doesn't meet the specific listing criteria for spinal disorders. They require things like specific neurological findings, particular MRI evidence, a lot of really detailed specific stuff. She's not going to get approved at step three, but she's also not going to get denied, so we move on.

Marc: Okay, what about step four?

Gwen: Can you do your past relevant work? Here, the agency is making what it calls a residual functional capacity finding, an RFC. What can you still do given your limitations? Can you lift stuff? How much can you lift? How long can you stand? Can you bend and reach?

Marc: Okay, what about for Kathleen?

Gwen: This can find that she can do light work, like maybe lifting up to 20 pounds occasionally, 10 pounds frequently, standing and walking about six hours a day. She clearly cannot go back to warehouse supervision at that level. That job required serious lifting, prolonged standing, climbing. So she passes step four.

Marc: And now we're at step five, which I'm guessing is the one that actually matters in most cases.

Gwen: Right. So this is where most claims are won or lost. Given this person's residual functional capacity, plus their age, their education, and their work history, is there other work in significant numbers in the national economy that they could do?

Marc: Any other work at all?

Gwen: Right. In the entire national economy. This is where the vocational expert comes in. They testify at the hearing, and the ALJ will describe a hypothetical claimant — a 54-year-old with limited education and a residual functional capacity for light work — and ask the vocational expert what jobs someone like them could do.

Marc: And the VE says, what exactly?

Gwen: So in a lot of these cases, they might say something like surveillance system monitor or document preparer or addresser.

Marc: But surveillance system monitor, like watching security cameras?

Gwen: Yes.

Marc: Where are these jobs coming from? Like, who actually has this list?

Gwen: This is the Dictionary of Occupational Titles, the 1991 catalog.

Marc: The one we opened with.

Gwen: Yep. Now you can see where it actually operates in the framework. It is step five. The vocational expert at Kathleen's 2026 hearing can look at her file and say she could be a document preparer. Document preparer is in the catalog. The description is from 1991. This was before widespread internet, before most people had email.

Marc: And the ALJ handling hundreds of cases generally isn't going to go behind the VE's testimony and litigate whether the document preparer is performed the same way it was in 1991.

Gwen: Right. Not at this volume. So the vocational expert cites the job, the ALJ accepts it, and the claim is denied at step five.

Marc: Even after the 2024 cleanup.

Gwen: Right. The cleanup wouldn't reach this case. All it did was remove jobs that nobody was citing anyway. There's not really a lot of National Historical Society operators at the national level. The list that actually mattered was the 13 occupation one. And for those, the agency didn't even prohibit citing it. It just added a documentation requirement. At the volume we're talking about, a documentation requirement can get satisfied.

Marc: And so the VE says yes, as currently performed. And here's my supplementary explanation.

Gwen: And the ALJ accepts it and writes the denial. So the underlying problem, the catalog that dates from the year AOL was born and years before most of our students were born. So this is being used to tell present day claimants what jobs they could do. It's still there. The cleanup changed the words that get spoken at the hearing, but it didn't change a lot of the outcome. And that is because of the grids. They are officially called the medical vocational guidelines, but everyone calls them the grids.

Marc: And what are the grids?

Gwen: These are literal rows and columns. They have four inputs — your residual functional capacity level, your age, your education, and your work experience. You plug those four values in, you read across the row, and the table tells you whether you're disabled or not.

Marc: So the table decides the case.

Gwen: For a huge category of step five cases, yes, the table decides.

Marc: So that's a core tension from Goldberg, isn't it? The whole point of an individualized hearing is that the human being considers your particular circumstances. But what the grid does is it basically matrixizes your disability. It puts a matrix in the middle of that.

Gwen: Right. This is before we've had any sort of property interest attached. We're deciding whether somebody gets something. The due process protections don't happen. And this mattered when it was challenged. In Heckler v. Campbell, this is 1983, Carmen Campbell had a bad back and hypertension. She couldn't do her job as a hotel maid anymore. The ALJ applied the grids and denied her claim. She argued the grids violated her right to an individualized determination.

Marc: Okay, and so what happened?

Gwen: The Supreme Court upheld the grids here.

Marc: Ah, that's shocking.

Gwen: I know.

Marc [reading]: The inquiry requires the Secretary to determine an issue that is not unique to each claimant — the types and numbers of jobs that exist in the national economy. This type of general factual issue may be resolved as fairly through rulemaking as by introducing the testimony of vocational experts at each disability hearing. As the Secretary has argued, the use of published guidelines brings with it a uniformity that previously had been perceived as lacking. To require the Secretary to relitigate the existence of jobs in the national economy at each hearing would hinder needlessly an already overburdened agency.

Gwen: A lot of people will have really standard residual functional capacities, RFCs. And so the grid works well for those. It's when you're deviating from what they define it. That's when we need the vocational experts.

Marc: All right. So that last line there, that it would hinder needlessly an already overburdened agency. That was 1983. So the court is openly saying, we're letting you do this because doing it the other way would crush the system.

Gwen: Yes. These grids are a deliberate trade. Congress and the agency looked at the scale of disability claims and decided that some part of the individualized determination had to give way to a rule-based table or the system wouldn't function.

Marc: So the grids are admin law's clearest example of agencies replacing individualized judgment with systemized rules for the explicit reason of keeping up with volume.

Gwen: Yes. And Campbell gave that trade a constitutional blessing.

Marc: Who gets hurt by that trade-off?

Gwen: It's the people who don't fit into the table cleanly. And the cleanest example of fitting is probably age. Kathleen is 54. There is a reason for that. That puts her at what the grids call closely approaching advanced age, the 50 to 54 bracket. If she were 55, she'd be in advanced age. So on her exact facts — limited education, semi-skilled past work with limited transferable skills, light work residual functional capacity — the grids direct different outcomes for the two brackets.

Marc: How is that different?

Gwen: So at 55, the grids direct a finding of disabled. At 54, it directs a finding of not disabled.

Marc: So one year of age is the difference between a yes and a no.

Gwen: Yes. Same medical condition, same residual functional capacity, same education, same work history. The grid treats the move from 54 to 55 as a structural break in the labor market. The working theory is someone closer to retirement has a meaningful, harder time learning new work. And the regulations bake that theory in as a categorical line.

Marc: That's the grids doing what they were designed to do. Predictable, uniform, and quite honestly, brutal.

Gwen: Yes. The whole point is to remove judgment from this question. The cost is a cliff at the category boundary, which the regulations themselves noticed eventually and created a rule for it. Here, this is 20 C.F.R. § 404.1563.

Marc [reading]: We will not apply the age categories mechanically in a borderline situation. If you are within a few days to a few months of reaching an older age category, and using the older age category would result in a determination or decision that you are disabled, we will consider whether to use the older age category after evaluating the overall impact of all the factors in your case.

Marc: So if Kathleen is 54 and 10 months, the ALJ can then just round up her age to 55.

Gwen: The ALJ can consider rounding up. This rule is permissive. It is not automatic. And the agency reads "a few days to a few months" really narrowly. HALLEX guidance treats six months as the outer edge. And in 2025, the Fifth Circuit affirmed a denial where the claimant was five months shy at 55.

Marc: Five months shy of 55 at his onset date.

Gwen: Five months wasn't close enough.

Marc: Five months? Wow.

Gwen: The court held the ALJ wasn't required to apply the borderline rule. So the rule exists, but the agency's working definition of borderline is tight enough that a substantial number of claimants who are close to the next category still get their current chronological age outcome.

Marc: So, Kathleen, at 54, how close to 55 does she need to be to get pushed over the edge?

Gwen: All right. 54 and 11 months, she should be inside the borderline window. Any lawyer would argue for the older category and probably get it. 54 and 7 months? Well, we're really at the outer edge here. Some ALJs would apply it. Plenty wouldn't. The circuits are split on whether the failure to address this issue is itself reversible error. If she's 54 and 2 months, the borderline rule isn't going to help her. She has to win on the merits at light work residual functional capacity in front of a vocational expert who's about to cite document preparer.

Marc: So the grid's uniformity has a softening rule built into it. And the softening rule has its own categorical line. And Kathleen is on the wrong side of that one, too.

Gwen: Right. It's worth noting that somebody can also partially win. The ALJ might say they are disabled as of their 55th birthday. But this is how the system handles its own brittleness. It adds a rule. The rule has its own edges. The cliff doesn't go away. It just moves.

Marc: Okay. We spent three episodes on how the structural picture works. How does it actually land here? I mean, what does this trilogy look like when you run it through two and a half million cases per year?

Gwen: So you get the numbers we were looking at in that first episode. The ALJ approval rates can range from below 20% to above 80%. Same agency, same regulations, same five steps, same grid — which means that two claimants with essentially identical files can get different outcomes based on which judge they happen to draw.

Marc: All right, we covered that as a structural claim, but what does it look like for Kathleen?

Gwen: Last week, we talked about Judge Williams. This was the approval rate around the office median. Disc disease claim that she had approved and the quality review memo three months later that taught her something about what the system considers adequate. So if we put Kathleen in front of her in 2026, after a decade of those memos accumulating, whereas if we put her in 2010 before any of them — same medical record, same testimony, this would even be the same ALJ — but we could have a different system around her.

Marc: And the outcome then also moves.

Gwen: Right. And it can move pretty dramatically. That's what we were talking about last week. Because Judge Williams in 2026 has absorbed a decade of guidance, training, and quality review feedback that the Judge Williams of 2010 hadn't seen yet. Same judge, same evidence, but the system around the judge tilted. Now let's hold the year steady and run the other axis. She's in 2026 with two different judges. She testifies she can only stand for about 20 minutes before the pain gets severe and that she has to lie down a couple of times during the day and her pain medication makes her foggy. The objective medical evidence, the MRI, is the same in front of both judges.

Marc: But?

Gwen: But Judge A finds her testimony credible. The residual functional capacity ends up more restrictive. With her age and education and limited transferable skills from the warehouse job, she can land in a category where the —

Marc: Grids direct to find disabled and she gets approved.

Gwen: Judge B might find the same testimony only partially credible. And so we'd come out as a residual functional capacity of light work with some qualifications. The grids say not disabled and she's denied.

Marc: Okay, so the same woman with the same medical record, but with different outcomes because of different years.

Gwen: So that was different outcomes because of different judges.

Marc: Okay, same woman with the same medical record, but different outcomes because of different judges.

Gwen: Right. And that judge who denied her wouldn't have inherently done anything wrong. They might have written a decision citing the medical evidence and explained why they found the testimony only partially credible and then apply the grids. On review, that credibility finding gets substantial deference.

Marc: So all of the structural features we talked about across the trilogy — the horizontal variation, the weak internal review, the ex-ante mechanisms — they all run through the credibility finding and the RFC, which determine what the grids then output.

Gwen: Right. This is why Social Security outcomes look the way they do. The structural picture isn't abstract. We have this five-step framework, and the grids give it a specific funnel at the end. Credibility determines what evidence counts. That leads to what you have as a residual functional capacity, and that determines where we land on the grids. So whatever is shaping the ALJ's approach to credibility and the residual functional capacity, like their experience in the local office culture and quality review environment, those shape the outcomes.

Marc: And the data on claimants who don't make it to the end?

Gwen: So the GAO and Social Security's own inspector general have both reported on claimants dying while waiting for a hearing, which makes sense. This is a process people only start when they're not 100%. So at current processing times with a nine-month average and some offices running longer than that, the number of people who die during the pendency of their claim is going to be meaningful every year. And that's not the system failing in some exotic way. That's the system working as described at volume with a better wait time than it used to be.

Marc: Okay. Everything we built this season shows up in this one system. Let's walk through how it does.

Gwen: What a great idea. Here's the doctrinal point anchoring all of it. Once disability benefits are awarded, they are a statutory entitlement that the Constitution is going to protect. That was Goldberg, Roth, and then Mathews itself. If you have benefits and the government wants to take them away, due process attaches. You're entitled to notice. You're entitled to be heard. The Constitution has things to say about how that has to happen.

Marc: And once it's awarded, that's the termination cases?

Gwen: Right. So Mathews v. Eldridge was a termination case. He was already receiving benefits when Social Security moved to cut them off. That's when the three-factor balancing test got applied. What the court was deciding was how much process was constitutionally required before the government could take away something Eldridge already had.

Marc: Okay, and the initial application is different.

Gwen: Constitutionally different. So the first-time applicant doesn't have a property interest. There's no entitlement to a benefit you haven't been awarded. You have a statutory right to whatever process Congress put in the statute. Paper review, reconsideration, the ALJ hearing, Appeals Council, eventually federal court. But those are procedural rights Congress chose to give. The Constitution isn't requiring them. It's not requiring much of anything at the initial stage.

Marc: So the two and a half million initial applications aren't getting a constitutional hearing. They're getting a statutory one.

Gwen: Yes. And it's going to be thinner. And that's a design choice that is driven in part by volume. So if we run Mathews over it, not for those initial applications, but for what happens when someone's in the system and the agency wants to deny continuing benefits or try and terminate them or recoup overpayments, those are the cases where the three-factor test actually does work.

Marc: So Mathews balances the private interest, the risk of erroneous deprivation with additional procedure, and the government's administrative burden.

Gwen: Right. We plug the numbers in. The private interest — pretty big. These are people losing benefits they've been living on. I would say enormous, but the court has said it is inherently not enormous. The risk of erroneous deprivation — it's demonstrably real. The GAO has documented that it's real. And the administrative burden of more individualized process — this is where the burden can get staggering. Continuing disability reviews alone run somewhere between one and a half and two million a year. So overpayment cases, there's hundreds of thousands more of those. The post-award population that Mathews actually governs is its own multi-million case per year system.

Marc: And when you balance those three things honestly at that volume, you tend to get this.

Gwen: Yes, you get these truncated termination procedures. You get CDR processes that mostly run on paper, overpayment recoupment that starts before the claimant has meaningfully been heard, and an appellate path that can take years. The Constitution is in the room applying Mathews, and Mathews is doing what it does, which is balance. And what it produces at this volume is a process that's much thinner than you might imagine.

Marc: Mathews itself was a Social Security disability case. The court knew it was setting the rules for federal benefits program processing hundreds of thousands of terminations a year. It wasn't sitting there imagining one isolated termination case.

Gwen: Right. The court built the administrative burden into the test as the third factor precisely because it was thinking about these high volume benefit systems. Scale isn't an accident in Mathews. It is part of the intentional design.

Marc: And so the third factor was designed to then count the volume.

Gwen: Yes. And it was designed to do that from the start. But what has changed since 1976 is the ratio. The disability adjudication system today is an order of magnitude bigger than the one the court was looking at. The initial applications, the CDRs, the terminations, the overpayments — same test, but a much bigger denominator.

Marc: So the doctrinal question isn't whether Mathews was misapplied. It's whether a test that counts administrative burden as a factor at all can ever produce a result that requires more process once the system is big enough that the burden dominates the math.

Gwen: Yeah. If one of those factors is going to be the administrative burden, then eventually at a sufficient scale, that factor will tend to outweigh or almost outweigh everything on the other side. Not because the test is being misapplied, but because this is the way it was built. The court accepted scale as a legitimate counterweight from the beginning.

Marc: And the test doesn't have a built-in floor.

Gwen: Right. It tells you you need to balance. It doesn't tell you that some minimum amount of process is required regardless of what the balance produces. There's no real doctrinal limit on how thin process can become if the system is big enough to make the burden factor dominant.

Marc: And let's be clear, what we've been talking about so far this episode is the largest administrative adjudication system in this country.

Gwen: It is. It is big, but it is not the only big one. So next episode, the last official episode of the season until Slaughter comes out, is the other major deep dive we're doing — on immigration adjudication. And specifically, we're talking about asylum. It's a different system, a different structure, and the constitutional stakes are higher. Social Security is about income. Immigration is about whether you get to stay in the country. You might expect the procedural protections would be stronger there, but —

Marc: Surprise, they are not. They are weaker, and the volume problems get worse, not better.

Marc: So that does it for today's episode on Administrative Remedies. Thank you for joining us today. Please, if you enjoy this podcast and enjoy this episode, give us a like on Spotify, iTunes, or whatever platform you're listening on. And be sure to tune in next time where we'll continue to dive into the contours of administrative law, because remember — you can't fix what you don't understand.

Related Guides

• How Agencies Steer Outcomes Without Reversing a Single Decision — The structural mechanisms covered in the consistency trilogy all land here, where they flow through credibility findings and RFC determinations into step-five outcomes the grids then control.
• The Mathews Balancing Test Explained — The doctrinal test that governs post-award terminations and continuing reviews, and that this episode argues has no floor once administrative burden dominates the math.
• The Lifecycle of an Administrative Case — The journey of a single disability claim through the system, which this episode shows is one of 2.5 million journeys happening in parallel each year.