Judge Jonathan Hager and Judge Michael A. Pappas each ruled on about a thousand claimed conditions in our dataset. Hager granted 55.4% of them. Pappas granted 14.8%. The two judges sit at the same Board, applying the same statutes.

I want to be careful about what I do and don't say with that statistic. I'm not going to call either judge a villain. I'm not going to call either one a hero. I'm going to walk through what the data shows about how much the disposition of a BVA appeal depends on which judge picks up the file, what the structural reasons for that variance are, and what you can and can't do with that knowledge as a veteran. The judges aren't applying different laws. They're making different interpretive choices, in different procedural styles, on caseloads that the data can't tell me are equivalent. The 40-point gap is real. The explanation is more interesting than the headline.

TL;DR

  • Across the 14 highest-volume BVA judges in Claim Raven's dataset (each with 740 or more condition records), grant rates range from 14.8% to 55.4%, a spread of 40.6 percentage points.
  • The two extremes. Jonathan Hager (55.4% grant, 10.2% remand) and Michael A. Pappas (14.8% grant, 51.4% remand), represent different procedural styles, not different statutes.
  • Pappas's remand rate of 51.4% is five times Hager's 10.2%. A remand isn't a win. It's a procedural reset that often adds years to the claim.
  • Denial rates vary more than I first reported. Across the top 14 they run from 30.4% to 52.8%. Pappas is the one judge whose low grant rate is explained by remands rather than denials.
  • The data cannot tell me cases are randomly assigned to judges. Docket type, hearing requests, and geography all influence which judge picks up which file.
  • You don't get to pick your judge. But knowing the spread exists changes how you should think about appeal route, evidence front-loading, and whether to request a hearing.
What evidence wins a VA claimFree

Strong medical opinions won 89.7% of the time in our analysis. Weak ones won just 3.7%.

  • Know what makes a medical opinion harder to dismiss.
  • Use your own statement where it can carry real weight.
  • Catch common C&P exam problems before they hurt your claim.
Get the free guide

The Hager-Pappas Gap

Two judges, similar volume, 40-point spread

I want to start with the two extremes, because the comparison is the cleanest in the data.

Our dataset holds 101,518 condition records drawn from 49,876 Board decisions, so one row is one claimed condition inside one decision, not one appeal. On that measure the two judges look almost identical. Hager has 1,037 condition records. Pappas has 1,039, a difference of two.

Count decisions instead and the resemblance disappears. Hager's 1,037 records come from 559 decisions, an average of 1.86 issues each. Pappas's 1,039 come from 411 decisions, an average of 2.53. Pappas hears appeals with about a third more issues in them, and he hears about a quarter fewer of them. I originally called these near-identical samples. They are not, and the difference matters for the rest of this piece.

Here's how each one disposed of those cases.

  • Jonathan Hager, 1,037 condition records. 55.4% grant. 34.4% denial. 10.2% remand.
  • Michael A. Pappas, 1,039 condition records. 14.8% grant. 33.8% denial. 51.4% remand.

Same court. Same statutes. Same regulations. A 40.6-point spread in grant rate.

The gap does not go away when you count decisions rather than issues. Of Hager's 559 decisions, 64.9% contain at least one grant and 14.0% contain at least one remand. Of Pappas's 411, 21.2% contain a grant and 52.6% contain a remand. Measured that way the spread is 43.7 points, slightly wider.

That number, sitting on its own, looks like an indictment of someone. I want to be explicit that I don't think it is. I'll explain why.

Notice the denial column. Both judges deny at about the same rate, 34.4% for Hager, 33.8% for Pappas. Within a percentage point of each other. The denial behavior is essentially identical. Where the two judges diverge is what they do with the issues they don't deny outright. Hager grants them. Pappas remands them.

That is true of this pair. It is not true of the Board generally, and I get to that below.

Pappas remands 51.4% of the issues in front of him. Hager remands 10.2%. That is five times the rate on a procedural disposition, and it is the story of this particular gap.

The Remand Mirror

A remand isn't a denial. It also isn't a grant. It's the Board sending the case back to the regional office or to a VA examiner with specific instructions about what additional development needs to happen before the Board can decide on the merits.

The veteran doesn't lose. The veteran doesn't win. The veteran waits.

How long? In my reading of remanded cases, the answer ranges from months to years. The case has to go back, the development has to happen, a new C&P exam, a private medical opinion, additional service treatment records, whatever the Board specified, and then the case has to come back to the Board for a new decision. The total elapsed time from remand to final BVA decision routinely runs 12-24 months, sometimes more, and during that period the veteran is in legal limbo.

So when Pappas remands 51.4% of his issues versus Hager's 10.2%, what that means in practical terms is that Pappas's veterans wait longer for finality. Pappas grants 15% of the issues he decides, denies 34%, and sends 51% back to the regional office for more work. Hager grants 55%, denies 34%, and sends 10% back.

A veteran in front of Pappas is about five times as likely to be told "your file isn't ready, we need more development" as a veteran in front of Hager. A veteran in front of Hager is about three and a half times as likely to get a grant in this round.

Both are legitimate ways to operate. The procedural rules give judges discretion on whether to remand for additional development. Some judges use that discretion aggressively. Others decide on the record in front of them. Both styles are within the bounds of the statute.

The downstream cost to veterans is different. That's what the data shows.

The Whole Top 14

I want to widen the lens. Here's the disposition data for the 14 highest-volume judges in our dataset, each with 740 or more condition records. The table sorts by volume, not by grant rate, so you can see how much variance exists across judges who all sit on the same Board.

  • Michael A. Pappas, 1,039 records over 411 decisions. 14.8% grant / 33.8% denial / 51.4% remand.
  • Jonathan Hager, 1,037 records over 559 decisions. 55.4% grant / 34.4% denial / 10.2% remand.
  • B. Mullins, 947 records over 466 decisions. 30.6% / 43.3% / 26.1%.
  • J. Parker, 885 records over 409 decisions. 39.9% / 52.2% / 7.9%.
  • Steven D. Reiss, 869 records over 472 decisions. 49.0% / 32.2% / 18.8%.
  • Eric S. Leboff, 835 records over 427 decisions. 33.9% / 45.1% / 21.0%.
  • Michael Martin, 825 records over 418 decisions. 24.8% / 52.8% / 22.3%.
  • Martin T. Mitchell, 817 records over 414 decisions. 37.3% / 44.3% / 18.4%.
  • L. Chu, 797 records over 421 decisions. 24.1% / 46.8% / 29.1%.
  • B. D. Watson, 787 records over 412 decisions. 28.2% / 49.2% / 22.6%.
  • Michael J. Skaltsounis, 784 records over 411 decisions. 46.4% / 30.4% / 23.2%.
  • David L. Wight, 775 records over 404 decisions. 28.1% / 44.5% / 27.4%.
  • Matthew W. Blackwelder, 742 records over 395 decisions. 26.4% / 52.3% / 21.3%.
  • Lesley A. Rein, 740 records over 409 decisions. 29.5% / 42.7% / 27.8%.

Each row reads: condition records, decisions those records came from, then grant rate / denial rate / remand rate. Rates are per condition record.

The cutoff at fourteen is close to arbitrary. The 15th and 16th judges hold 739 and 738 records, one and two records below the 14th. Read the list as "the highest-volume judges," not as a meaningful boundary.

Look at the spread.

Grant rates range from 14.8% (Pappas) to 55.4% (Hager). A 40.6-point span across 14 judges. The median grant rate sits at 30.1%. The Board-wide average is 30.6%, so the median high-volume judge grants at almost exactly the system rate.

Denial rates are tighter. 30.4% to 52.8%. A 22.4-point span, with a median of 44.4%. Compared to grant and remand rates, denials cluster more. They are not, however, flat, and that turns out to matter.

Remand rates are the wildest. 7.9% to 51.4%. A 43.5-point span on a single procedural disposition, with a median of 22.5%. One judge remands more than 30% of the issues in front of him (Pappas, at 51.4%, more than twice the median). Two remand less than 13% (J. Parker at 7.9%, Hager at 10.2%). The remand-heavy and remand-light judges represent different operating philosophies for how to handle an incomplete record.

I want to flag something about that table, and to correct what I said about it when I first published this piece.

Grant and remand do trade off against each other. Across the 14 judges the correlation between grant rate and remand rate is -0.74. The judges who grant a lot (Hager 55.4%, Steven D. Reiss 49.0%, Michael J. Skaltsounis 46.4%) tend to remand less than the judges who grant a little.

What I got wrong was the next step. I originally wrote that denial rates don't differ much and that the structural variance is in grant-versus-remand rather than grant-versus-denial. That claim does not survive the corrected table. Denial rates span 22.4 points, and the correlation between grant rate and denial rate is -0.46 across all 14 judges. Drop Pappas from the set and it strengthens to -0.77, stronger than the grant-remand relationship over the same 13 judges.

In other words, Pappas is the exception I mistook for the rule. He is the one judge in the top 14 whose low grant rate is absorbed almost entirely by remands. For the other thirteen, a low grant rate mostly shows up as a high denial rate, which is the ordinary pattern you would expect. J. Parker grants 39.9% and remands only 7.9%, but denies 52.2%. Michael Martin grants 24.8% and denies 52.8%. Matthew W. Blackwelder grants 26.4% and denies 52.3%.

So there are two distinct things in the table. Most of the spread is judges disagreeing about whether the evidence gets there, which shows up as grant versus denial. A smaller and stranger part of it is one high-volume judge who resolves that disagreement by sending the file back instead.

What Causes the Spread: Procedural Style, Not Bias

A working theory and what the data can't prove

I've spent enough time reading these decisions to have a working theory about why the grant-vs-remand split varies so much between judges. I want to lay it out and then immediately tell you what I can't prove.

The theory is about how a judge handles an incomplete record. Imagine a BVA appeal where the C&P exam from 2019 was thorough but missed a specific Correia-required range-of-motion measurement. The medical opinion in the file addresses nexus but doesn't address aggravation. The veteran's lay statement is detailed but the Board could read it as either competent lay evidence or inadmissible medical opinion. None of these issues, by themselves, is dispositive. Each of them is a judgment call.

A judge could look at that record and conclude: "On balance, with the benefit-of-the-doubt rule applied per Lynch v. McDonough and the lay evidence considered under Buchanan v. Nicholson, the evidence supports service connection. Grant." That's an exercise of judicial discretion in favor of deciding the case now, on the record present, applying the doubt rule when the evidence is in equipoise.

The same judge could look at the same record and conclude: "The 2019 exam was inadequate under Correia. The medical opinion failed to address aggravation under Allen v. Brown. Remand for a new exam and a supplementary opinion." That's an exercise of judicial discretion in favor of fuller development before deciding.

Both decisions are defensible. Both can be supported by case law. The first one is faster for the veteran in the short run. The second one is sometimes better in the long run, because a thinner record at the Board can produce a denial that's hard to appeal further. A remand can also surface evidence that strengthens the case.

The 40.6-point spread between Hager and Pappas, in my read, is the visible cost of two judges operating at opposite ends of that procedural-style continuum. Hager appears to decide on the record in front of him more often. Pappas appears to send cases back for more development more often. Both are working within their judicial discretion. Neither is breaking the rules.

What I cannot prove is that this is the entire story. Some of the spread might be driven by other things, assignment differences, docket composition, hearing-request rates, condition mix across the caseload. I'll talk about those next.

Cases Are Not Randomly Assigned

I want to be explicit about a structural caveat in this data that limits how strongly I can draw conclusions.

BVA cases are not randomly assigned to judges. Cases are routed based on a mix of factors that include docket type (legacy vs AMA), hearing request status (videoconference, in-person, virtual), geographic considerations for some hearing requests, and team assignments within the Board. The Board operates with judges organized into teams handling specific dockets, and some judges hear disproportionately more of one kind of case than another.

That matters because different docket types have different baseline grant rates. AMA Direct Review cases are decided without hearings on the existing record. AMA Evidence Submission cases give the veteran 90 days to submit new evidence before a decision. AMA Hearing cases include a live hearing with the judge. Legacy appeals come in with their own decades of accumulated procedural history.

If Judge A hears mostly AMA Hearing cases and Judge B hears mostly AMA Direct Review cases, their grant rates can differ for reasons that have nothing to do with how either judge approaches the evidence. Hearings let the veteran address weaknesses in the file directly. Direct Review doesn't.

I cannot fully control for this in the data. The case_analysis dataset doesn't always cleanly tag the docket type, and even when it does, the case-counts at the judge level are too small to do a robust within-docket comparison.

I originally wrote that I had looked for this effect and that it didn't appear large enough to explain the Hager-Pappas spread. I was too confident. There is a measurable composition difference between exactly these two judges, and it runs in the direction that would inflate the gap.

Pappas averages 2.53 claimed conditions per decision. Hager averages 1.86. That is a real difference in what lands on each docket, and it interacts with how remands work: one unresolved issue sends the whole decision back, and in a table where every issue is its own row, a single remanded multi-issue decision produces several remanded rows. A judge hearing denser appeals will register a higher remand rate for that reason alone.

I want to be careful not to overclaim in the other direction. Across all 14 judges the correlation between issues per decision and remand rate is +0.58, which looks like support. But remove Pappas and it flips to -0.41. The relationship is Pappas, not a Board-wide pattern, which means I cannot use it to explain him away either. What I can say is that his density and his remand rate are confounded in this data and I cannot separate them. The gap survives at the decision grain, 64.9% of Hager's decisions carrying a grant against 21.2% of Pappas's, so composition is not the whole story. How much of the remainder is procedural style, I cannot precisely say.

There's another assignment effect worth naming. Cases that are obviously going to grant, the strong-nexus, well-developed appeals that should never have been denied at the regional office, get appealed at lower rates because veterans accept the favorable RO outcome. Cases that get to the Board are by definition closer calls. But within that closer-call population, the cleanest cases may still cluster differently across judges based on which team they land on. The data can't fully separate that.

What Remand Actually Costs

I've been hinting at this. I want to land it.

A remand is procedurally neutral on the merits. The Board didn't deny the claim. The Board sent it back to fix something specific. In the medium term, many remands eventually result in grants once the development is complete.

But remands have time costs that the topline statistics don't capture.

The veteran has to wait for the regional office or the AMC (Appeals Management Center) to schedule the additional development the Board ordered. If a new C&P exam is required, contractor scheduling can take three to six months. If a supplemental medical opinion is required, that can take similar time. The actual production of new evidence can take three to twelve months from the remand order.

Once development is complete, the case goes back into the BVA queue for a new decision. Depending on the docket, that can be another six to eighteen months. The veteran's lawyer or VSO may need to file additional briefs. The veteran may need to participate in another hearing.

The total elapsed time from the original BVA remand to the final BVA decision routinely runs 12-24 months. I've seen cases run longer.

During that time, the veteran is not service-connected for the condition (unless they have a separate effective grant). They're not collecting compensation for it. They're not eligible for VA healthcare for it (unless they have other access). They're in legal limbo.

The 51.4% remand rate from one judge versus the 10.2% remand rate from another doesn't just describe a procedural difference. It describes a substantial difference in how long the average veteran waits for finality.

This isn't an argument that low-remand judges are better than high-remand judges. Sometimes the development the high-remand judge ordered is genuinely necessary. Sometimes a low-remand judge decides on a thin record that produces an unappealable denial. Both styles have costs and both have benefits. The cost of remand falls primarily on the veteran's clock, and that's worth naming when we look at the spread.

Hearings as a Lever

There's one variable veterans actually control in the BVA process, whether to request a hearing. Most veterans don't.

In our data, the hearing route appears to lift grant rates for certain condition types. In adjustment disorder cases, board_hearing produces a 53.2% grant rate versus 36.9% for board_direct review. That's a 16.3-point lift, on the largest single condition where I can run the comparison cleanly. Other conditions show similar but variable hearing-route lifts.

Why hearings help is, in my read, partly about evidence presentation and partly about which judge picks up the file.

On the evidence side, a hearing lets the veteran address weaknesses in the file directly. The veteran can describe symptoms, explain gaps in the medical record, respond to the C&P examiner's findings, and contextualize lay statements. The judge gets to ask follow-up questions. The record at decision time is more developed than it would have been on direct review.

On the judge-assignment side, hearings get scheduled with the judge who will decide the case. The veteran doesn't pick that judge, but the judge who agrees to hold a hearing is often one who's more willing to engage with the file than the judge who decides solely on the record.

There's a tactical implication for veterans facing a thin record or a contested nexus opinion. The hearing route can lift the grant rate. It also adds time, hearings have to be scheduled, which adds months to the appeal timeline. The trade-off is grant-rate-now versus speed-of-decision.

If you're being assigned a judge with a high remand rate, requesting a hearing may not change much, the judge may still remand for additional development after the hearing. If you're being assigned a judge who decides on the record, the hearing is your one opportunity to present evidence directly. The data can't tell you which judge will pick up your file, but the structural lever exists.

What I Cannot Tell You From This Data

I want to be explicit about the limits before I close, because the topline Hager-Pappas number is the kind of statistic that invites overreading.

I cannot tell you which judge will hear your appeal. The Board does not publish assignment in advance. You won't know until the decision arrives.

I cannot tell you that any individual judge is biased against veterans. The data shows variance in procedural style and disposition mix. It does not show evidence of any specific judge applying the wrong law or violating the benefit-of-the-doubt rule. Judges who remand a lot are not "anti-veteran." Judges who grant a lot are not "pro-veteran." They're working within the discretion the statute gives them, with different operating philosophies.

I cannot tell you that the grant rate for any individual judge is the grant rate you'll experience. The aggregate is across all conditions, all dockets, all hearing requests, all evidence profiles. Your specific case has its own particular facts that may push the outcome higher or lower than the judge's caseload average.

I cannot fully separate selection effects from interpretive-style effects. Some of the variance between judges may come from differences in which kinds of cases land on each judge's docket. I don't have clean enough data to fully control for this. The variance is real. The size of the variance attributable to procedural style specifically is harder to pin down.

I cannot tell you that the 40.6-point Hager-Pappas spread is the worst in the system. It's the widest spread I found in the top 14 by volume, where the sample sizes are large enough to be meaningful. Below that volume cutoff, judges with smaller caseloads can show larger apparent spreads that may not survive a larger sample.

What I can tell you is that judge variance at the BVA is substantial, structural, and partially driven by procedural style choices that fall within judicial discretion. The system has more slack than most veterans assume. Knowing that helps you think about appeal route and evidence preparation without making you a victim of a judge you can't choose.

How to Use Judge Knowledge Without Trying to Game It

There's a tactical question that comes out of all this. If judge assignment matters as much as the data suggests, what should you do with that knowledge?

A few patterns I see in cases that came out well despite judge variance.

Front-load the evidence. The thinner-record judge styles are most punishing when the record is genuinely thin. If you can get the C&P exam, the private medical opinion, the lay statements, and the relevant service treatment records into the file before the BVA decides, you give yourself the best chance regardless of which judge picks up the file. A complete record is harder to remand and easier to grant.

Anticipate the Correia and Jones issues at the C&P stage. Inadequate C&P exams drive a lot of remands. If you can get a private medical opinion that does what the VA exam should have done, range-of-motion measurements at the point of pain onset, an assessment of functioning without medication, a discussion of lay evidence, you reduce the surface area for remand.

Consider whether to request a hearing. The data shows hearings lift grant rates for some condition types. Hearings are a structural lever you control. They add time, but they also add evidence-presentation surface.

Think about appeal lane choice carefully. Direct Review is fastest but provides no chance to submit new evidence. Evidence Submission gives you 90 days. Hearing adds a live presentation. If your record is strong and you want a decision quickly, Direct Review may be right. If your record has gaps, Evidence Submission or Hearing gives you more chance to fill them.

Do not try to game judge selection. You can't. The Board doesn't publish assignment, and procedural maneuvers like withdrawing and refiling don't reset judge assignment in any reliable way. The energy spent trying to manipulate which judge hears your case is better spent making the file strong enough that the judge's style matters less.

The 40-point spread is real. It's a fact about the system. But the response to it isn't to find the magic judge. The response is to build a file that doesn't depend on a magic judge to win.

Bottom Line

Two BVA judges who each ruled on about a thousand claimed conditions can have grant rates 40.6 percentage points apart, and the gap holds when you count decisions instead of issues. Across the top 14 the spread is mostly judges disagreeing about whether the evidence gets there, which shows up as grant against denial. One judge, Pappas, resolves that disagreement by remanding instead, and his docket also carries about a third more issues per decision than Hager's, which I cannot fully separate from his remand rate. Cases are not randomly assigned, so some of the variance reflects docket composition and case selection. But the variance is too large to attribute entirely to assignment effects. Procedural style, how a judge handles an incomplete record, is doing real work in the data. You don't get to pick your judge. You do get to control the completeness of your record, the choice of appeal route, and whether to request a hearing. The structural slack in the system is a fact. The response to that fact is to make your file as strong as you can, so the judge's style matters less.

Methodology and Limitations

  • Data source: Claim Raven's analysis of 101,518 condition records drawn from 49,876 BVA decisions in the production case_analysis dataset. One row is one claimed condition inside one decision, so a decision claiming four conditions contributes four rows. Judge-level rates in this piece are per condition record unless the text says decisions.
  • Sample size: the 14 highest-volume judges, holding between 740 and 1,039 condition records each. The two judges in the direct comparison hold 1,037 and 1,039 records, drawn from 559 and 411 decisions respectively.
  • Grant rate definition: grant rate is favorable outcomes (granted, partial grant, mixed) divided by decided records (favorable plus denied plus remanded). Dismissed, withdrawn, and moot records are excluded from both halves because the Board never decided the issue. This is the same definition used everywhere else on the site. An earlier version of this piece used raw granted counts over all records, and quoted a Board-wide grant rate of 29.5%; the canonical figure is 30.6%. For the 14 judges in the table the two definitions give the same rates, because none of their records are dismissed.
  • Judge identity: judge names are grouped case-insensitively. The Board prints a judge's name in capitals in the signature block and in title case in the body, so the same judge otherwise appears under several spellings. An earlier version of this piece grouped them case-sensitively, which split 117 judges across multiple entries and put seven judges in the top 14 who did not belong there. Names are printed here as the decisions print them; where the Board itself is inconsistent (it writes both "Eric S. Leboff" and "Eric S. LeBoff") the more common spelling is shown.
  • Classification approach: Grant, denial, remand, and dismissal categories are taken from the BVA's own disposition field for each record. The judge assignment is captured from the decision's named author. Records whose author field holds a multi-judge panel (22 such values, 34 records between them) are left in the data under that combined label and therefore never enter any single judge's row.
  • Limitations:
  • BVA cases are not randomly assigned to judges. Docket type, hearing-request status, and team assignments influence which judge hears which file. The dataset does not include enough granular metadata to fully control for these assignment effects.
  • The judge-level caseloads in this dataset are a snapshot, not the complete career history of any judge. Judges with longer tenure may have caseloads in the dataset that reflect only part of their work.
  • Grant, denial, and remand rates in the table sum to 100% because the denominator is decided records only. None of the 14 judges has any dismissed, withdrawn, or moot records.
  • The Hager-Pappas direct comparison uses two judges whose caseload sizes happen to be statistically similar. It does not imply these two judges hear identical cases. Differences in docket composition could explain part of the spread.
  • I cannot prove from this data that procedural style accounts for the entirety of the observed variance. The estimate that procedural style is "doing real work" is qualitative, based on my reading of decisions across both judges' caseloads, not a quantitative decomposition.
  • Aggregate patterns are not predictions for any individual appeal. Your specific case has its own facts that may push the outcome above or below any judge's average.
  • This analysis does not characterize any judge as biased, prejudiced, or operating outside their judicial discretion. The variance described is structural, not pejorative.

Disclaimer

I'm not accredited by VA, not a lawyer, not a VSO. This is data analysis, not claim advice. These are patterns from cases that made it to the BVA, they don't predict individual outcomes. If you need help with your claim, work with an accredited representative.

Where to go next

-Landon Founder, Claim Raven | U.S. Army Veteran