Oral argument — Biestek v. Berryhill

We'll hear argument first this morning in Case 17-1184, Biestek v. Berryhill. Mr. Bhabha. Thank you, Mr. Chief Justice, and may it please the Court. When a vocational expert testifies about the existence of a specific number of jobs in a specific location at a specific time, that testimony can only be based on statistical, data-driven sources. And when the expert refuses upon request to provide those sources, the expert's testimony standing alone cannot constitute substantial evidence for three reasons. First, this Court's definition of the term substantial evidence and its application of that term in reviewing the decisions of administrative tribunals. Second, because the government's arguments as to why the expert's testimony standing alone is sufficient are unpersuasive. And third, because this is the rule that has worked since 2002 in the Seventh Circuit and, indeed, is consistent with the government's own policy as encapulated in the Social Security Administration's Vocational Expert Handbook. Ginsburg. The counsel for the claimant asked for the source material, but then didn't engage in any cross-examination of the witness, of the expert witness. Your Honor, as a factual matter, after being denied the material, he did engage in some cross-examination that appears in the record. But I think critical, Your Honor, is that without the material itself, any meaningful cross-examination regarding the expert's methodology, the provenance of the expert's labor market surveys was impossible. And I would note that it would be a rare case when you would be asked to cross-examine a statistical expert who is opining upon specific numbers that the expert has then modified through calculations without actually seeing the data sources itself. I think this Court's decision in Florida Power and Light is an important precedent in this respect, because there the Court noted that the testimony, the well-founded testimony of an expert may be enough if firsthand information is unavailable. And here, particularly because there were two sources the expert relied upon, the Bureau of Labor Statistics data, which was public, but then also her private labor market surveys, which the ALJ never saw and we never saw, that made any form of cross-examination or meaningful inquiry into the basis for these numbers impossible. Alito, what I don't understand about your argument is how it connects with the substantial evidence question. Substantial evidence refers to a quantum of proof, and I — it's hard for me to see why that inquiry is different depending on whether the underlying information in question was requested or not. The argument that you're making sounds like a procedural question, a due process question, whether it was unfair not to require the expert to produce the underlying data. Can you just explain how the procedural question that seems to be at the core of what you're arguing fits in with the substantial evidence test? Of course, Your Honor. So I think we have two answers to that question. First and foremost, we are not asking for the document-on-demand procedural rule that the government characterizes us. The government, who bears the burden at step 5, is choosing in these cases to only rely on the testimony of a vocational expert. Now, the government has added availability to enter other evidence into the record as well. If the government chose to submit its own labor market surveys, we would not say, and I'm not saying that there's a stand-alone constitutional violation because the vocational expert didn't give her surveys. Well, I know you're not saying that, but you have a quantum of evidence. It's substantial or it's not substantial. Explain, if you can, why the question of the issue of substantiality depends on whether the evidence was asked for or not. Your Honor, I think it flows from the basic intuition that when you question an expert about data, if the expert cannot then back up their testimony, that creates doubt. And although I don't think it's an exact match, I think a useful analogy are the adverse inference cases, which this Court has recognized when a party or a witness says, I am presenting evidence to the Court, but there is better evidence somewhere else that substantiates what I am saying. Alito, I see that argument. Now, in this case, was it the attorney for — who was it who said that the — the expert was not required to produce the evidence? Was there opposition on — by your adversary, or was it the ALJ? It was the ALJ. And so, Your Honor, what I think is critical here is that when the government says the record is exactly the same, it is not. If, for example, the expert had said when questioned for the data, I have a source, but I'm not going to identify it, I'm not going to show it to you, I think that would create real doubt. And here — Alito, I see that. But when the expert says, this is my opinion, and you say, produce the underlying evidence, and the expert doesn't say, no, I'm not going to do that, I won't do that, somebody — the judge intervenes and says, no, that's not required, why does that create doubt about the validity of the expert's testimony? The expert hasn't refused to produce that evidence. It's been the intervention by the ALJ. So, Your Honor, if I can just clarify, it was both here. The expert said, that data is in my confidential client files, and therefore, I don't want to produce it. And the ALJ said, I'm not going to require that. And I would note that the government doesn't defend in this Court the confidentiality rationale. And I think the reason for that, Your Honor, is that as the claimant's lawyer said below, it would be easy to redact or black out the names of the clients. Indeed, Federal courts and administrative agencies deal with highly sensitive information about national security or intellectual property every day. But here's the thing I don't — sure, maybe he made a — you're right in this case, maybe. But, I mean, even if we were in court, experts rely on all kinds of things. And if you — the other side makes a case, says, what are you relying on for your testimony? Ptolemy. Not Copernicus? No, Ptolemy. Well, that might be a good ground for going into it. And so, whether there's a good ground for going into it or not depends on the case. And why would it be any different here, where, in fact, it's not even a court, and you have a law which says you don't even have to use court rules of evidence? Do you see my point? Maybe you're right. Do I repeat it? I understand your point, Your Honor. But with respect, I disagree with it. And for two reasons. I'm not — I'm asking a question. I'm saying, why is it different from even a trial where the standard is less, and in a trial, my understanding is, is that what you disagree with? Or here, certainly, it would depend on the case. So, Your Honor, I have a practical answer to your question and a legal answer. And I'd like to give both, if I might. As a practical matter, vocational expert testimony has been widely criticized in the courts of appeals, with courts noting that the methodology is at times preposterous, leading to numbers that are likely fabrications. And as the Noshka Amoskov brief demonstrates, for the exact same jobs that our vocational expert in this case opined, there were 120,000 jobs. Experts for that exact same job in almost identical time periods have given numbers from the hundreds up to 480,000. And the reason for this huge range in the answers experts give is it's not simply like an expert in these cases is Googling the number of appellate lawyers in Washington, D.C. They have to go first to the Dictionary of Occupational Titles, a book that was written in 1977 and was last updated in 1991, and involves also — should — they should have gone into it in this case, because it was really rotten. Well, you have a pretty good bar, and you would think that the bar there would find a case, maybe it's yours, and go to the court of appeals and say, look, you should have looked into this one. And then when they look into that one, if they disagree with you, you would have said just what you're saying now. But I don't understand, you see, and you might have won. And maybe you should win. What I don't understand is having an absolute rule that every expert who's vocational, regardless of what he relied on or how much trouble it would be, would have to, without cross-examination or a strong basis in the law — in the facts of the case — why he'd have to produce all this stuff. Maybe it is confidential. Maybe sometimes he should. Maybe sometimes he shouldn't. Did you see? That's my problem. I understand your problem, Justice Breyer, and let me see if I can address it. The government itself, I would note, as a policy matter, tells experts, you should be prepared to explain, cite, and furnish the sources upon which you rely. In the handbook that the Social Security Administration gives to experts, it says that five times. Sotomayor, that was after this case. That's correct, Your Honor. So that wasn't in place at the time. Can I go back to Justice Breyer's initial question, and perhaps Justice Alito's? You say very explicitly in your reply brief, you're not asking for an absolute rule that an expert must, before testifying, produce these materials. You make an exception for when it's asked for. But you also concede in your reply brief that there might be situations where it's not necessary, where the expert opines, doesn't produce their materials, but there's other independent evidence that's reliable and could be relied upon to constitute substantial evidence. And so there's no absolute rule according to you, but what you're asking us now to hold, I think, is that as a matter of law, an expert who opines on something and refuses to provide the sources is sufficiently unreliable that it doesn't constitute substantial evidence. Is that what you're trying to get us to? Not exactly, Justice Sotomayor, if I can be very clear on the rule that the Seventh Circuit has held and that we are arguing for in this Court. In this context, when an expert is providing statistical data and then cites to statistical sources they rely upon and say these are the sources and these are not public sources, this expert relied on two sources, one of which was private, we had never seen it, the ALJ never saw it, in that situation the Seventh Circuit has held that if you ask the expert, just show me the sources that you yourself are relying upon so that the agency can make a determination and so that we can conduct a meaningful cross-examination, in that situation the say-so is not enough. Sotomayor, you're defining meaningful, but let's, let me pose a hypothetical. Expert says what this expert says, and you get up and say, how many of these people had the same conditions as my client has, and the expert says virtually every one of them. How many people did you do research on with respect to this issue? I contacted about 15 businesses. Have you placed these kinds of individuals in the kinds of jobs you've talked about? Yes, a hundred of them. Whatever do you believe that in that circumstance you could stand here and say that there wasn't substantial evidence from which the ALJ could rely upon, even if you had not seen the underlying records? Your Honor, I think the best case for you is for her to say, well, I've never really placed anyone with that, with those conditions. I only checked one employer, but I'm extrapolating from that some sort of methodology that really could be questionable. Those are the two extremes we have, isn't it? That's right, Your Honor. All right. The potential extremes we have. But no one asked any of those questions below for us to make a judgment about whether the ALJ's ruling was reasonable or not. So, Your Honor, if I can answer your question in two ways. First, I think ours is the easier case, because here the expert didn't say it was my experience. She said labor market surveys. And so the expert explicitly didn't say, I have placed a number of people. But taking Your Honor's hypothetical head-on, in a situation in which a vocational expert says, I've placed a number of people over my 10-year career in this position or in these sets of positions, I think it is likely that would not be substantial evidence, and here's why, Your Honor. Section 423d2a, the statutory provision here, requires that there be a significant number of jobs in the national economy. Now, that number has what significant means is the subject of some debates in the courts of appeals, and it appears to be a multifactored test. But what is critical is that that's at least in the hundreds and likely in the thousands. And these types of jobs that are getting placed are very specific jobs with people with very specific limitations. So even if an expert said, I've placed 15 people over the last 10 years in these jobs with similar limitations to your client, that doesn't provide a basis to know that there are hundreds or thousands of those jobs, and I think it's for that reason that vocational experts rely on surveys and not only their own personal experience in propounding their testimony. Sotomayor, but I think you're missing the point, which is you wouldn't have needed the surveys to make the argument you just made. You could have questioned the expert and shown the lack of a sufficient basis for their conclusion, and then made that argument to the ALJ. What I'm trying to get at is I understand the need in some situations to actually see the surveys, but don't you have to lay at least some predicate ground for why that's necessary in your particular case? I understand your question, Your Honor. I think initially, I will just say as a factual matter, and I think the record bears this out, as soon as my client's or claimant's representative asked for the surveys, the ALJ made clear she wasn't going to give them. So I don't think there was any opportunity even to proffer reasons. But even beyond that, Your Honor, it is the government's burden at this stage. And given the nature of this type of statistical testimony, even across examination when, as in our case, an expert says, I relied on public data and I relied on private data, it is hard for me to conceive of what kind of a meaningful cross you could do of the private data, the expert then provides her answers about this is what the data said, this is what I did to the data, but there's no way of verifying any of those kinds of answers. Roberts. One way to be to look at the publicly available data, right, and there was no questioning about that, was there? I mean, to the extent the Bureau of Labor Statistics information shows, you say 8,000 jobs in southwestern Michigan, if that's where it was, and the Bureau of Labor Statistics shows a different number. Why did you choose 8? In other words, there were fields, I'm not saying there were ample fields, but there were fields for fertile cross-examination that weren't explored, I think. Mr. Chief Justice, I think it is certainly a different case if an expert only relies on public data, then I absolutely agree you have exactly what the expert relied upon. In a case like this, however, and I think in many cases like this, and this is why vocational expert testimony has been the subject of criticism, the public data alone is not only not enough, but it is often of an entirely different character, because the taxonomy in the public data, in the BLS data, uses a far larger definition of jobs than what the DOT, the Dictionary of Occupational Titles codes, which is what the ALDVE is required to identify. So what happens is, in our case, for example, the SOC, the Standard Occupational Classification Codes in the census encompass, in some cases, hundreds of DOT codes. And so when you look at the overall number that comes from the census, that can be in the thousands or in the hundred thousands, but the critical question, and this is what the vocational expert is required to do, is to say how many jobs with somebody with this level of education, this skill level, these kind of disabilities, how many jobs are available for them? So while I agree that BLS source was available, it was a very partial source at best, and I think the critical question, and that's why the experts in this situation didn't only rely on the BLS data, was because there needed to be a significant winnowing of those numbers to get to the answers to the ALJ's hypo. Roberts. It strikes me that that's a good argument to have made before the ALJ, at which point the ALJ may have said, well, I see that now, and you should have access to the private data, and you can just redact it. But that wasn't done here. Well, Your Honor, the one thing that was done, I agree, that wasn't laid out to the ALJ. What was laid out, the first time that we asked for the data and were refused, immediately counsel said, look, it's a substantial evidence standard. These vague conclusions are not enough. So I think it was certainly put in issue to the ALJ. We need something more. And I would just note that this is a very specific type of testimony. This is testimony where a witness is giving statistical answers, identifying statistical sources, and I think it would be rare in administrative procedure in that situation for an agency to say, yes, there are these data sources. They're the sole basis for the testimony, but no regulated party or no party before the agency, you can't see those sources in order to meaningfully challenge the conclusions of the expert. And I think, Your Honors, Perales is an interesting case in this respect. Now, we agree with the government. For sure, Perales discusses procedural due process. But equally crucial in Perales, Perales came under section 405G of the Social Security Act, the very same provision we have at issue here. And the question presented by the government in Perales to the Court was whether or not the medical expert testimony in that context could be substantial evidence. And I think what's important about Perales is that in Part V of the opinion, while certainly also discussing procedural due process issues, the Court in Perales gives indicia of why the testimony there had probative value and why it was reliable. And I think it's very important to look at that case, because there, the medical experts didn't just give conclusions. What the Court noted specifically in Perales was that they laid out the tests that were conducted, the results of those tests, the types of surgeries that were conducted, and the results after surgery. Alitoso, your reliance on that case raises the interesting question, and a question I think is interesting, which is whether there would be any basis for limiting the rule that you're asking us to adopt to the specific situation here, which you have stressed where its testimony, its statistical testimony by a vocational expert in a Social Security disability hearing. Why wouldn't the rule that you're asking us to adopt apply whenever there is the question if a determination by an agency is supported by substantial evidence? Well, Your Honor, I think for sure the term substantial evidence is one that applies throughout administrative law. But what this Court has made clear in a number of cases is that, of course, substantial evidence looks at the record and, thus, inherent in the question, in the answer as to what substantial evidence means is what is the question that the agency is trying to address. In a situation like this, where you're talking about specific numbers that the agency is relying on as the sole basis to deny my client benefits for the applicable period, there I think when the expert points to data sources that she has modified in order to come up with these numbers, there substantial evidence requires more. But in other situations, Your Honor, for example, qualitative testimony, even testimony such as these are the kind of jobs in my experience I believe somebody with these sorts of limitations can do, i.e., the first part of what vocational experts testify, I think that's entirely different. But I do think in this situation it would be very normal when an administrative agency comes up with specific numbers and bases a determination on specific numerical conclusions. It is the norm to then make the agency or make the expert show their work. Kagan. Why wouldn't your argument be the same if there had been no request at all? Your Honor, certainly we would win if that – if the rule was that vocational experts have to give over their testimony regardless of a request or not, we would win. I'm just asking why is that part of your proposed rule? Your Honor, no court of appeals has ever held that there is a requirement, but I think I'm sympathetic to the rationale behind that rule, but I think that's Kagan. It seems to me that your rationale suggests that the on-demand part of your test is irrelevant. Your Honor, I don't think it's irrelevant, and as I said, I don't think it's irrelevant for the reason that when you question a witness and the witness then doesn't give you the test, the very data sources they rely upon, that introduces doubt. And I think that is where the question and the answer play into the test, and I think that's the rationale of the Seventh Circuit. They haven't elucidated it exactly in those terms, but that is what I take those decisions to say, which is that when the testimony is given alone, if it's not challenged, then maybe it's enough. But when you ask a witness, what is your data, and the witness doesn't give it over, and again, the government isn't defending here the confidentiality rationale, that does create doubt. And I think the adverse inference cases show that intuition in the law. Kagan. I guess I'm just not getting it, because either it's enough or it's not enough. It doesn't have anything to do with whether somebody demands the information. Well, Your Honor, I think the demand part of the test is something about have you forfeited the right to complain about it. Well, Your Honor, the Seventh Circuit and the Ninth Circuit, which has also applied a similar standard, have talked about this in terms of forfeiture or waiver, but I don't think from a legal perspective the term forfeiture is exactly correct here, because you cannot waive the substantial evidence standard. That is a reviewing standard that the Court must apply in any circumstance. So the way I read those cases is to essentially say if you question somebody, they cannot back up the data. And I think it would be analogous, Your Honor, if you asked a witness in exactly the same case, where did you get your data sources from, and the witness said either I'm not going to tell you, or they said, well, they occurred to me one day, I looked up something, but I'm not going to identify what it is on the Internet for you, that does create doubt in a record, Your Honor, just as here the failure to look at those numbers when requested, even when the claimant's representative said I don't need your client's names, you can redact those client's names, that creates doubt in the record, and that is why, particularly in a context such as this, when you have multiple outdated data sources which vocational experts have to bring together, when vocational experts come up with numbers that vary enormously for the same jobs, and all the more so when this is the sole basis upon which disability benefits applicants are being provided, then the statements of the expert are what is required. Can you draw an adverse inference against the expert if the expert sincerely believes, perhaps mistakenly, but sincerely believes that what is requested is confidential? Your Honor, I think the adverse inference case law, again, it's not directly controlling here for sure, but I think you can use the same intuition from those cases, because there's a very easy method, which again the government doesn't dispute here, to redact that information and provide only the crucial information. And I think the reason, again, particularly in this context of vocational experts, why this data is so important and why the failure to provide it is so practically harmful, in addition, we think, to being in conflict with the substantial evidence standard, is because these experts are bringing together very old definitions of jobs with entirely different taxonomy of job numbers, putting the numbers together and coming up with a result, and then, in certain circumstances, reducing that result further based on further hypotheticals posed by the ALJ. So it is a data-intensive process. The vocational expert identified as much in her testimony. And not even having the numbers that the vocational expert is admitting she relied upon hampers not only the claimant's representative's ability to cross-examination, but also, crucially, the ALJ's ability to actually determine whether there are real numbers here. Mr. Chief Justice, if I may reserve the remainder of my time. Thank you, counsel. Mr. Yang. Mr. Chief Justice, and may it please the Court, I'd like to address two general subjects. First, how Social Security hearings work and how vocational experts decide whether there are significant numbers of relevant jobs in the national economy. Excuse me. Second, I'd like to address the substantial evidence theory, which is evolving and it is built on a procedural predicate that is incorrect. Each year there are about 2.6 million initial disability claims that are filed with SSA. And at the third level of review, the SSA conducts 670,000 hearings. That's about 2,500 a day. Over 1 million people are waiting for just a response for their hearing, and they wait, on average, about 605 days. There is no adjudicatory process on a scale comparable to this, and Congress has properly vested SSA with broad authority to determine the right process to develop records. The question, then, here is the question, then, doesn't that solve this case? Because your agency now has in its handbook that every expert has to come prepared to cite the sources that they relied upon and to produce the materials. So if they've made that judgment, why shouldn't we make the same judgment? That a failure to do so goes to the heart of the expert's unreliability. The vocational expert handbook that we're talking about, the informal guidance, really is referring to vocational resource materials that they use. There are not individual case client files. And if I can explain, there's two levels of analysis that a vocational expert can do. Sotomayor, you want to show me where in the handbook that's set so I can follow you? The citations are in our briefing op at page 9, we discuss it, and then on page 18. I don't have the handbook in front of me, unfortunately. The language that the handbook says is cite, explain, furnish any sources you rely on to support your testimony. But the reason we think this is talking about resource, vocational resource materials, is because an expert just doesn't know the questions and the granularity of the questions that the expert's going to be asked. So, for example, there's two stages of question. Sotomayor, are you picking numbers from the air? No. I wouldn't say from the air. I'd say from experience, which can be educated by the jobs that they do for individual disability clients. If I can explain, there's basically two levels. Aren't they given the file before they come to testify on the individual claimant? No. They are given the relevant vocational background, but they're nothing about the disabilities of these claimants. In fact, ALJs are prohibited from communicating with vocational experts prior to the brought by an impartial agency to adjudicate a case. They are then You're worrying me that this is, in fact, what all of the critics are saying, that these are numbers pulled out of a hat as a person sits there. If I can explain, there are two levels, and I think this will help to address your concern. At the first level of inquiry, the vocational expert is determining a category of jobs based on generally things in the dictionary of occupational titles that discuss exertional limitations and such for those jobs. So, for instance, sedentary jobs that do not require any kind of advanced or education, unskilled jobs. Then the expert relies on public sources, statistical sources, like the BLS, that provide numbers in the national and local economies. And the expert, they don't map exactly, so the expert is going to have to use some judgment to extrapolate. Those are questions that can be explored fully at hearing, and, in fact, are the guidance suggests that they should provide these. These are things that you predict in advance. But there's a second level. At the hearing, there are very specific types of impairments that are not addressed by the grids. The grids take into account these high-level impairments. But there are things like non-exertional impairments, depression, ability to concentrate, things like unable to lift above your head or more than 5 pounds. These are things that you just don't know until you come into the hearing. And in this case, at page 10, for instance, of the reply brief, Petitioner acknowledges that the testimony of the vocational expert, that if her numbers were to be reduced by about 20 to 30 percent, if the Petitioner could not lift above his head or lift more than 5 pounds, that's the second level of inquiry. And that is what's educated by the vocational expert's experience reflected in her own surveys done for individual clients. And remember, these are the things that are not addressed by the grids. Ginsburg. And can you explain what is the confidentiality here? She says she relies on the Bureau of Labor Statistics and her own independent research. What is your own independent research? I can't tell you, because that would be part of client files. I think it would help to look at page 118 and 119a of the Petition Appendix, because that's the actual testimony. The first time this comes up is in the context of where the expert got her on-task percentage, that is, you have to be on-task about 80 percent. And so she says she gets it from her experience doing job analyses. A few lines down, can you provide the job analyses? The expert says, I cannot. It doesn't say that. She just says, observe. Those would be part of people's, that is, individual's private confidential files. Now, remember, this expert, her resume is in the record, and she was certified as a vocational expert. Ginsburg-Gillian What are the people? I don't have a grasp on whose confidentiality is at stake. These are individuals that have disabilities. This vocational expert works as a rehabilitation consultant to find jobs for people. That's her job. So she is very well situated to take a person, have hypothetical questions about what the person can do, and answer questions. And this is precisely what she's doing. She's saying, on my experience, you need to be able to focus, you'd be on-task 80% of the time. That's the type of information that you find. Kagan Without the data, how is somebody to cross-examine her on that? And how is the ALJ to verify that conclusion? Well, you cross-examine by simply asking basic questions. Well, when you say doing job analyses, what does that mean? Where are you getting this information? When you talk about your confidential files, what is the information in the confidential file? Where do you develop that information? There's a whole string of questions about methodologies, sources that can be explored. And if there are you can get to a point, there are case law, we cite a case, at least one case in our brief, where if you go down the cross-examination path and the expert simply is not able to provide cogent responses, that will undermine the testimony and maybe render it not substantial evidence. But here, there was no cross-examination. The judge intervenes. The judge says, I'm not going to require that. If it's in individual confidential files, individual people, I'm not going to require it. It's not – there's no adverse inference. The judge is – or the expert is simply explaining that this subset of data, not the high level, but the second level of data, is in confidential files. Then she goes on to the job numbers. Now, remember, the job numbers are, again, on two different levels. There's the high-level job numbers. That is 360,000 jobs between bench assemblers and sorters. And then there's an additional one when you're given the hypothetical about lifting. That's going to be reduced 20 to 30 percent. And counsel asks – Ginsburg. Can we go back to what you said? You dismissed the handbook, but I think this is an accurate quote from it. At all hearings, you, the vocational expert, should be prepared to cite, explain, and furnish any sources you rely on to support your testimony. I – that is probably an exact quote, but what I will say is that there's no way – a vocational expert cannot know. They're going to ask, can you lift 5 pounds? Can you lift 7 pounds? This person has this specific type of depression that requires X, Y, and Z. You don't have public sources for that. The expert's relying on expertise that's built and may be reflected in confidential client files. But you can't bring your entire source of your files through your professional experience to the hearing. That's impossible. So the only – we think the only reasonable way to read this is we're talking about the upper order, the first-order things, BLS data, Dictionary of Occupational Data, that kind of stuff can be reasonably expected to be brought. But this other stuff, it would be impossible. So any doesn't mean any? Like some agency pronouncements, they're not – We shouldn't defer to the agency's handbook on any. Well, the non-binding guidance is written for non-lawyers. It's not intended to be a statute. And I think it has to be – It's written for the experts, right? It's written for the experts. But again, if you're a vocational expert and you know how the game is played, you can't just proceed. Well, there's no way you could bring all your files. Just a slightly different line of question. If we were in Federal district court, and I know we're not, and forget about the rules of evidence, but if on the key issue in the case, the evidence depended upon the testimony of an expert, and the expert said, ah, I'm not going to give you my underlying data, it's secret, I don't think we would hesitate to find that no rational in favor of the party propounding that expert. Why isn't the same true here? If that were all that's standing alone, that may well be true. Okay. So you admit the principle. But I wouldn't concede that, though, because it depends on the entire record. Sure it does. But my hypothetical is that that's the key point. And on that key point, the only piece of evidence is an expert who says, I have secret data. You'd agree that we would reverse. Likely reverse. However, however, I would like to point out some key differences. There's factual differences, which is, one, this expert's not refusing. There's a ruling by the adjudicator that says the expert's not refusing. Fine. You can modify my hypothetical. That doesn't make any difference, does it? Well, no. I actually not. If the district court said I'm not going to make the expert turn over his secret evidence, we'd still reverse because no rational jury could find. Actually, I'm not sure that's right. This Court has held that, for instance, there's no process problem with admitting allegedly totally unreliable evidence so long as you have the ability to contest it. Not a process problem. Not a process problem. And sufficient evidence for a rational jury to reach the conclusion. But if this is a qualified expert, there's no contest on the qualifications. The expert testifies to a fact. Now, there may be additional things that undermine that, but there's otherwise no other evidence. There's no evidence that contradicts the expert. I think that's a tough call. I think that's a tough call. Breyer. How does that work? I mean, I actually don't know. But think of any kind of an expert, a house painter or a doctor, and the plaintiff has a certain kind of injury to himself or his house, and the doctor says, well, a person who coughs like that and a person who has that kind of lifting problem, I've looked up all the treatises, and they suggest there might be X or Y, and in my experience, I can refine that further because I've had thousands of clients, and when they cough like that, it means da, da, da. Okay? And that's all there is. And if that's all there is, does that expert or the house painter says the same thing about a rotten board? A rotten board, in my experience, means termites if it's like this, but not if it is like that. And I've talked to many doctors or many house painters, and I've seen a lot myself, and that's my experience. Now, is that reversible error? No. Well, then why would it be reversible error here if the vocational expert says exactly the same thing? You're going to say it isn't, okay. The one thing that I'm holding out is this standard looks to the whole record. And there are things that could be evidence that can be admitted into the record that can undermine the bottom line of the conclusion. So if, for instance, and there are cases like this, the vocational expert gives a bottom line number, and then there's cross-examination, and the vocational expert cannot answer in any credible way, how the expert came about doing this? Well, you know, I'm a vocational expert. Well, that doesn't make any sense, and I'm asking you how. And you probe, and you probe, and it ends up under so undermining that testimony that the record evidence shows that not a reasonable person, a reasonable person would not have relied upon that. Sotomayor, that's a little different. Roberts, no matter how much of an expert person is, what she basically said is, trust me. I've done this for a while, and I think, and it's not just trust me, I think, in general. Trust me, I think it's 20, you know, 20 percent. It does have a sense of being pulled out of the air. There are two points that I think are important. One, there are many issues in a Social Security hearing, just like in other cases. This is one of them. And oftentimes, issues are not contested. And when they're not contested, you don't have to develop much of a record on it. The reason is, if you had to develop a record on uncontested things, it would be an unmanageable process. It happens in court litigation. It happens here. Secondly, what we're talking about is the question of what the expert didn't say no, but the expert, the question to the expert was, can you provide those surveys? No. They're confidential. So the expert said no, and then the ALJ backs her up. I'm not seeing that in the record, Justice Ginsburg. I see, can you provide the job analysis, answer, they would be part of people's confidential files. Judge, the attorney says, well, you can black this out. At that point, ALJ comes in and says, I'm not going to require that. Second time it comes in, can you provide your own, it would again be the same answer. Is there part of the confidential files? ALJ says yes. Sotomayor, does it matter who? Mr. Yang, you seem to, in answering Justice Gorsuch, believe that because the ALJ was the one who said, I'm not going to order that, that that somehow elevates the prior answer into being reliable? I mean, no expert is the judge in a case, correct? The judge has to accept the testimony. All right? So how can an ALJ accept testimony for which it's blocked an answer about how the expert came to their conclusion? That's his basic argument, which is this testimony is unreliable because I was blocked from being able to show the basis. Yang, I understand that process argument. Well, my point is, and I think as we've explored already, that that's a process argument, not a substantial evidence argument, because the substantial evidence depends on what actually comes in. I can get on the stand and say anything I want, and if the ALJ stops the other side from giving the basis for that, the record is devoid of a basis for that answer and it's unreliable for that reason. That, I think, is the core of their argument. But here, the ALJ, which is an impartial ALJ or not, excuse me, an impartial expert whose qualifications are determined and not objected to, there's no reason to think that she's the expert. Well, that's fascinating because there was a whole exchange at the beginning of this hearing where the ALJ asked the attorney to stipulate to the expertise of the expert, and the attorney refused to do so. And he said he thought that meant they could not dispute anything the expert said, which the judge says no, no, no, and the attorney then clarifies, well, I don't object to the testimony then. So there's a very different But not to the expertise, and certainly not to the fact that the expertise matches the disabilities at issue in this case. Let me step back a second, and I think this might help a little. There's a good reason why counsel doesn't probe numbers like this. The numbers, the testimony here were based on two categories of jobs that totaled about 360,000 jobs in the national economy. The Sixth Circuit, there's no magic number about what constitutes the standard of a significant number of jobs, but the Sixth Circuit, which is the governing circuit here, has held, consistent with other courts, that 6,000 is enough. So we're talking about numbers, testimony at 360,000, and you've got to get it so far, it's got to be so off that it can't even be ballparked to 6,000. And so it doesn't matter whether the job number is 100,000 or 250 or 360, all of it is far beyond what would matter. And remember, this individual is the expert. I'm going to give you an example of the numbers that these experts give. Mr. Bhabha says that, when talking about nutsorters, and I guess I want to know why everybody talks about nutsorters, too, but it varies from 260,000 to 470,000, and that's a huge variance, and it makes you think, where is that coming from and what are we to do when somebody says one of these numbers. Hoffman. I totally understand. I totally understand your point. I've got two basic responses. Petitioner had their own expert that made very specific objections to the expert testimony presented. They did not contest the job numbers, and I think for good reason. Now, what Petitioner cites, too, is an amicus brief that provides for a list of various numbers, right? Now, if you look at the cases that are cited, at the stage one of the analysis for the vocational expert estimates, they have to do a few things, one of which is you have to determine the category of jobs. Now, in this case, the category was sorter. These are sedentary positions, but that encompasses a whole number of jobs that were defined in the dictionary occupational titles. Now, different vocational experts are going to focus on different sets. So some of them may have a broader set of sorters, some of them have a lesser set. And you don't list, unless it's cross-examined, all of them. You're asked initially to provide an illustrative DOT number. The DOT number provided in this case is nut sorter. If you look it up, it's an agricultural nut sorter. We are not suggesting that there are 125 agricultural nut sorters or 125,000 agricultural nut sorters in the country. It's illustrative of the type of sorter position, sedentary sorter positions. So when you look at these cases, what you find are the courts simply saying there are so many either. In one case, it's agricultural sorter. In one case, there was sorter. In another case, it was sorter. But you're saying they're all answering different questions. Different questions. You cannot tell from this. And the right way to examine that is to cross-examine it. Cross-examination, as the Court has explained, is the time-tested way of discerning truth and accuracy and testimony. And there's no reason to exclude that here. When you usually, when you're having someone testify to data and numbers, the way you cross-examine is to ask what she relied on and then see if that testimony lines up. Yes. And so, for instance. Well, yes, but here she said, I can't give you the data on which I relied. Well, no. This is – I'll give you a few questions I think would have clarified this case considerably. The – with respect to the job numbers, the expert testified that she relied on BLS numbers. Now, remember, BLS provides both national and regional numbers for various types of jobs. And she additionally relied on her individual market surveys. But remember, in this context, she is an expert who helps individuals find jobs. So her labor market surveys, one would assume, are for individuals finding jobs. Well, they can redact the names, right? They could redact the names. But if you would ask, well, what was – when you say you rely on the BLS and then you say you rely on the individual market surveys, what do you rely on the individual market surveys? And if the expert said, as I think is probably the case and as the ALJs probably assume, given the course of testimony, it was, well, there are about 20 to 30 percent of these sedentary jobs that are excluded when you have to lift above your head, right? Then we would know. Okay. Well, what if you think that's wrong? Well, then you can probe further. You can say, well, how did you – like, what is the basis? How many people are we talking about? How far did you survey out? All of these are things that go to weight. But again, to find no substantial evidence, you'd have to find that no rational decision-maker could have relied upon that. So there are all kinds of cross-examination questions that would have clarified this. And all we're saying – and we're not saying that when you cross-examine an expert, if for some reason this impartial expert who's already been certified as an expert can't respond. Sotomayor, isn't it your burden to prove that those jobs exist? Why is it their burden to show the basis for your expert's opinion? It's me getting up there and just saying, this is the number, believe me. This is a public right, and claimants bear the burden of proving disability. The way the Social Security Administration has administered the process is it has relieved them of the burden of production with respect to the Step 5 inquiry, by providing a vocational expert such that the vocational expert provides something that the claimant can respond to. But that doesn't mean that the claimant can simply say, well, I want a report. And if you don't provide a report from your private clients, there's not substantial evidence. You're at least going to have to require cross-examination to be able to discern what was this – how relevant is this report and how does it affect the bottom line. For instance, the job numbers that ultimately became relevant did not include a 20 to 30 percent reduction because the ALJ did not find that the claimant couldn't lift above his head or lift more than 5 pounds. It may be completely irrelevant, but we don't know that because it's not in the record. Although we suspect that it is relevant, we can't show that. So there are all kinds of reasons. Now, going to Petitioner's position, Petitioner, from the third petition to their opening brief, now let's shift a little bit in the reply brief, have asked for a categorical rule. It's substantial evidence and you can rely upon it if there's no demand, but if there is a demand and you fail to respond to it, for instance, because the ALJ says I'm not going to require it, and it's in your office, which you have to travel to, then it's not substantial evidence. That ignores what's in the evidence, as several questions pointed to. It's not a substantial evidence question. It's a procedural question. It's also not coherent, because if it's true, as Petitioner's argued – fact in dispute, and it really is in all these cases, right? I mean, that's why you have a vocational expert, is how many jobs are there going to be that this person could do. If it's the key thing in dispute, and the expert has said I want to keep my evidence on which I'm relying secret, it's not as in Justice Breyer's example saying I'm basing it on my experience over 30 years in the industry, and then you could maybe ask about that, and you could have your own expert with 30 years in the industry who could opine on that. And, in fact, the Federal Rules of Evidence, of course, treat that kind of expert very differently than an expert who relies on data, as this one did. This one says I've got secret data, all right? And it's the key question in the case. Well, then why doesn't that create an inference that they're – an adverse inference that that witness is hiding something? And why doesn't that undermine substantial evidence? There's two questions. The adverse inference line of doctrine concerns permissible inferences, right? Whether you can permissibly infer. There's a second question. Why would that be a compelling inference? I understand it's an analogy. We're not – we're not – Well, but it's meaningful when you're talking about review of a jury verdict or review here under substantial evidence, where you have to find no reasonable adjudicator would have gone – concluded that. So it's a very different inquiry. Having bracketed that, I will admit there are some answers that can undermine a bottom-line response. Now, we can quibble about where, you know, how far down that road you have to go to. But you'd agree that – okay, so we have some common ground, that an expert could say something or withhold something in a way that – on a key question that would be sufficient to undermine substantial evidence, would raise enough doubt about – Certainly. I think that's right. But again, look at what we're talking about here. This is on pages 118 and 119. There's nothing like that in this case. There's no follow-up. There was questioning, but not on the relevant issues. We do object. There's obviously some kind of a problem. I mean, that's apparent from a lot of the briefs and so forth. And suppose if I were writing a concurrence or something, I put in a paragraph which said if there really is a problem here, it's not – it may or may not be dealt with in gross. I'm not sure. But why doesn't they – why don't you find a test case or suggest to the bar, find a test case where you do probe, and indeed it does turn out to be resting on nothing, and you would either win or you would lose. And if you lose, you appeal it. And if you win, you have a model of how to proceed for others in the future. And they – I mean, is there anything wrong with doing that? And could that help solve the problem? I think it would depend on how that's written, but I would like to just focus on if it is contested significantly with cross-examination, which is not in this case, then I would be very hesitant, though, to require things like personal confidential files that's going to require significant delays in the process. No, you wouldn't have to do that. I mean, the Fed, you know, gets a lot of information on the basis of surveys that they certainly don't want to reveal. And that can happen in various agencies. Though sometimes you might have to reveal it. Well, there are cases – there are cases, I think, that highlight exactly what you're talking about, where there's cross-examination, and we cite, again, at least one or two of these in our brief, where the – on review, it's like, you know, this cross-examination is so undermined, the basic predicate of – or the basic testimony that it can no longer be substantial evidence. We're not quibbling with that. What we're saying is that the categorical document-on-demand rule, which they – which, frankly, is not a coherent view of substantial evidence, would be imposed on the agency. And they've mentioned the Seventh Circuit. The Seventh Circuit does not apply this rule. The Seventh Circuit has never required information from personal files. There's two Seventh Circuit cases that are central, Donohue, Dicta, McKinney. There was cross-examination by counsel, and the answer about – from the vocational expert about how the vocational expert got the numbers was basically, well, it's based on my knowledge as a V.E. And these followed up, and there was still nothing substantial after that. And so in that context, that's not a big problem. The two district court cases that Petitioners cite at page 40 of their brief and then in reply on 24, both involve cross-examination. Powell reversed on other grounds, didn't reverse on the question that we're talking here, and simply encouraged revisiting the issue on demand. And in Reynolds, there were multiple problems with the V.E. testimony that was revealed on cross-examination. There's no tradition in the Seventh Circuit or anywhere requiring this categorical rule that this Court granted cert on. And it would substantially impair the operations of the social administration. How? How would it substantially impair? Remember, these cases are – hearings are tightly packed. The volume of the cases is immense. People wait on – No, I don't understand that as how that's going to substantially impair. Oh, the reason is because you're not going to be able to predict in advance what documents might be relevant beyond the high-level documents that we're talking about, and you'd have to get them from your files. And that's going to require continuances. And having a continuance in the situation where claimants are already waiting on average, a million people – think about this – a million people are waiting 605 days on average just to get a response to their hearing request. What is the experience in the Seventh Circuit that has – Now, there's some dicta that suggests it in the Donohue opinion, but the Court has since stepped back from it. And if the requirement – How is the Government – oh, sorry. Is the new – are the new requirements in the handbook going to slow things up? No, because we don't understand that to mean bring all your personal files. We understand that to mean bring things that you're likely to be reciting from the BLS or other – DOT, these publicly available sources that you can provide those upon request. Is the Government's argument that its failure to provide timely hearings should be an excuse not to comply with other requirements? Not at all. But I will say that there is a significant undertone of fairness and due process here. And what the Court did in Perales, which I think is undisputably a due process case, is it applied before Matthews v. Eldridge, but it applied the same basic framework as what the Court solidified in Matthews v. Eldridge. And one of the things to consider is the cost of the additional process. The cost of the additional process here would be significant. Thank you, counsel. Thank you. Five minutes. Mr. Babak. Thank you, Mr. Chief Justice, and may it please the Court. I'd just like to respond to three things the Government said. First and foremost, the Government's lawyer represented that the vocational expert's handbook simply requires them to bring public information they might perhaps rely on. I'm sorry, but if you look at page 3 of the handbook, it explicitly says, at all hearings, you should be prepared to cite, explain, and furnish any sources that you rely on to support your testimony. And what I take Mr. Babak, you're right about that, but instead, it's just, well, that's the best practices rule. But the best practices rule is not necessarily a legal requirement. I agree with you, Justice Kagan, but what I take from the Government's own handbook is two things. Firstly, their administrability arguments are inconsistent with the policy the Social Security Administration itself is telling vocational experts. And on the specific question of workability in the Seventh Circuit, I disagree with my friend from the SG's office that this isn't the rule that's being applied. There is binding Seventh Circuit case law. We cite cases where this has been the basis for reversal. And my understanding from practitioners in the Seventh Circuit is that this is applied and that when data is not handed over and a case gets appealed to the appeals counsel, in some situations the Social Security Administration lawyers agree to a remand so the data can then be provided. And it's not always provided at the agency hasn't been applying it in the Seventh Circuit, right? Not voluntarily, Justice Kavanaugh. That's totally right. They haven't issued an acquiescence ruling. But the way it works in practice is that if this sort of a request is made and the specific data is not there, it is not uncommon for the record to be held open for 14 days and the data is then sent to the LJ and to the claimant's representative. Hopefully they look at the data and it substantiates the testimony and that ends the matter. But if it doesn't, then there is the opportunity to submit a memorandum to the ALJ. And I will also say specifically on the timing issue, nationally there are 164 hearing offices. The average amount of time to process a claim is 536 days. For the hearing offices within the Seventh Circuit, based on my calculation which I did last week, and I don't imagine the number has changed significantly, it's 529 days. So it's actually quicker in the Seventh Circuit. Now, these are obviously large numbers, but the basic point is this procedure which has existed since 2002 has not been shown to slow down the process. The other thing I'll say is that the government lawyer says, well, you don't know what particular questions are going to be asked at a hearing, so how do you know what sources to have? But if you don't know particular questions and you don't have the specific data-driven answers, you can't be answering the question in the first place. So by definition, if they give an answer that there are 3,000 nut-sorter jobs or 6,000 bench assembler jobs, there must be a data-driven basis for that, which is exactly what the expert again here identified as her source, not her experience, but a labor market survey that she then refused to permit. In sum, Your Honors, what Petitioner is asking for here is a reaffirmation of a basic rule of administrative procedure, which is that an agency cannot make a determination based upon testimony that is premised on secret data without ever giving that data to a requesting claimant. That's all the more so in a case like this, where this was the sole basis upon which my client was denied benefits for the relevant time period. This rule has worked without disruption in the Seventh Circuit since 2002, and it is entirely consistent with the very policy the Social Security Administration recognizes as good practice for vocational experts. Ginsburg-Ginzburg The government says you're asking for more than the Seventh Circuit position. Is that so? That is not, Justice Ginsburg, we are asking for exactly the same rule, which is, succinctly put, that when a vocational expert testifies and the expert identifies data sources that she has relied upon, if you request those data sources and they are not provided, the say-so of the vocational expert alone cannot constitute substantial evidence of the other work available to an applicant. For these reasons, Your Honors, we ask that the decision of the Sixth Circuit be reversed. Roberts-Garza Thank you, counsel. The case is submitted.