New Data on an Old Problem: Stanford’s Nora and David Freeman Engstrom Discuss their Study on Protective Orders and Court Secrecy

For decades, protective orders—orders that prevent litigants from disclosing information unearthed in discovery—have been the subject of fierce debate.  A parade of judges, rulemakers, and academics have argued, including before Congress, that protective orders are rare and also searchingly scrutinized by judges, who enter them only when they satisfy Federal Rule of Civil Procedure 26(c)’s exacting “good cause” standard.  Others disagree—or at least worry that the reality might be more complicated.

Until now, that debate has been grounded on guess work—not fact.  That’s about to change.  In “Secrecy by Stipulation,” available here, and soon to be published in the Duke Law Journal, Stanford’s Professor Nora Freeman Engstrom, Professor David Freeman Engstrom, Berkeley’s Professor Jonah B. Gelbach, SLS grad Austin Peters, and SLS 2024 grad Aaron Schaffer-Neitz offer the first rigorous accounting of stipulated protective orders in federal litigation, infusing a decades-old debate with much-needed empiricism.  Their findings are a jolt to conventional wisdom: Not only are stipulated protective orders far more common than previously thought, the authors find that the majority of federal judges over a ten-year span never rejected a single stipulated protective order request.

Nora and David Freeman Engstrom
Professors Nora and David Freeman Engstrom, co-directors of Stanford’s Rhode Center on the Legal Profession

Here, Nora and David explain their study’s methodology and key findings—and also discuss how their findings contribute to much broader debates.

Why did you choose to focus on protective orders? 

Nora: We have long heard stories of stipulated protective orders being used (often, by large corporate defendants) to conceal information from the public.  One example is the 2014 General Motors ignition switch debacle:  Hundreds of thousands of GM vehicles had faulty ignition switches that would cause vehicles to shut off while in motion and prevent the deployment of airbags.  Owing to this defect, nearly 400 people were injured or killed, and many lawsuits followed.  In an early case in Georgia involving the death of twenty-nine-year-old Brooke Melton, the plaintiffs, Ken and Beth Melton, learned damning information about the defect that caused their daughter’s death.  The Meltons wanted to disclose this crucial evidence. “We thought that people needed to know. There were still people out there driving those cars,” Brooke’s mother explained.  But, given a protective order, the Meltons were made to keep their mouths shut.  Chillingly, according to Kevin Vincent, previously Chief Counsel for NHTSA, the Agency was “stymied” by the protective order in the Melton litigation.

David: The Melton story shows that protective orders matter.  But, there’s more to it.  Partly driven by bracing accounts like the one involving GM’s defective ignition switches, reformers have for decades sought to beef up Rule 26(c), believing that some sort of fortification could better serve the public interest.  Those efforts have come in fits and starts for roughly forty years.  But those seeking to strengthen Rule 26(c) have always been stymied by a lack of evidence showing that reform is, in fact, needed.  Opponents of reform have long batted away reform efforts by arguing, first, that protective orders are rare and second that, in day-to-day practice, the entry of these orders is careful, not cavalier.  They insisted that judges already carefully consider the public interest before granting protective orders via Rule 26(c)—and so more consideration would be overkill.  As one federal judge testified before the House of Representatives in 2009: “what I want to hear is evidence of Federal courts . . . not doing what the rule says it should do, which is only grant protective orders for good cause shown.”  Reformers never could muster that evidence.  And in that empirical void, inertia won out.

That’s where your study comes in, although your study focuses on stipulated protective orders in particular.  What are those?

Nora:  Sometimes, a motion for a protective order is contested.  One side (frequently, the defendant), wants one, and the other side (frequently, the plaintiff) opposes the Rule 26(c) motion.  On many other occasions, though, motions for protective orders are not contested; they are, instead, “stipulated,” i.e., the product of party consent.  Given that it is generally defendants who are eager to keep information under wraps, it might, initially, seem strange that so many protective order motions are stipulated.  But, anecdotal evidence indicates that plaintiffs’ lawyers often recognize that, without a protective order, the defendant is apt to be less forthcoming—and, in the scheme of things, the plaintiff cares more about getting her hands on the documents and driving forward the litigation than she does about having the right to disclose.  Put another way, in an adversary system, there’s no constituency for transparency; there’s nobody to advance the right of the public to know.  Only the judge stands between the parties and the cloaking of information that might avert harm.

 Could you tell us a bit about your methodology? 

David: Our dataset contains docket-level information on virtually all civil actions filed in the federal district courts between January 1, 2005 and December 31, 2014—in all, 2.2 million cases and nearly 80 million docket entries.  To our knowledge, it is the most up-to-date and also the most comprehensive federal district court dataset available for research currently in existence.  In this piece, we zeroed in on court evaluation of “stipulated” protective orders.  In future work, we plan to analyze protective orders more broadly.

Nora: The sheer enormity of the dataset presented a tremendous opportunity, but it also presented some challenges.  Without going too deep into specifics, we trained supervised machine learning models to predict which of the 80 million docket entries in our dataset involved protective orders, stipulated or otherwise, and we also trained them to determine whether a judge granted or denied the parties’ protective order request.

David: That analysis allowed us to perform large-scale quantitative analysis.  But we also wanted to conduct a qualitative review of granted and denied stipulated protective orders.  To perform that granular analysis, we took a random sample of 300 granted stipulated protective orders and 100 denied stipulated protective orders.  Then, led by our terrific co-author, Aaron Schaffer-Neitz, a team of trained researchers manually reviewed and coded those orders to gather qualitative data.

Nora:  As should be clear, a study of this type is really resource intensive.  Virtually no one could do work like this solo.  This project relied on and greatly benefited from Jonah B. Gelbach’s utter brilliance as a lawyer and cutting-edged economist, Austin Peters’s skills as an SLS-trained lawyer and a top-notch political scientist with a machine learning toolkit (and further assisted by Stanford computer science students, Sreya Guha and Vrushank Gunjur), and Aaron’s doctrinal chops.  Other SLS students were also indispensable, including Ari Berman, Devin Flynn, and Jessica Seigel, all in the SLS Class of 2024.  This paper was a true team effort.

And what were your key findings?

David: Our findings are a lightning bolt in the decades-long debate over litigation transparency.  First, our study is the first to definitively determine the prevalence of stipulated protective orders in federal litigation.  Previous estimates suggested that stipulated protective orders were only entered in about 4,000 cases per year.  We found that they’re entered far more frequently—at more than double the rate previously assumed.  Critically, too, those numbers are increasing over time.  We saw a 50 percent increase in the number of stipulated protective orders from 2005 to 2012.

Nora: We also examined how judges assess Rule 26(c) motions, to gauge whether judges really are as careful as many insist.  We first calculated how often stipulated protective orders are granted.  Our results shocked us:  At least 95 percent of motions to enter stipulated protective orders are granted.  That’s an overwhelmingly high grant rate for a motion that, at least pursuant to the plain language of Rule 26(c), should be granted only when good cause is shown.

David: Next we narrowed our focus to a subset of 727 judges who had ruled upon at least twenty-five motions for stipulated protective orders from 2004-2012.  Zeroing in on these “experienced” judges, we found that over half had never once denied a stipulated protective order request.  From there, things got even more surprising.  Of the stipulated protective orders we reviewed, roughly two-thirds did not so much as mention Rule 26(c)’s good cause standard, while more than 80 percent contained no fact-based good cause analysis.  Put simply, our study shows that judges are not, by any measure, carefully analyzing motions to enter stipulated protective orders.  They grant them, no questions asked.

Given the previous lack of data on stipulated protective orders, it would seem that this study has the potential to shake up the conversation about their use.  Could you speak to the implications of your work?

Nora: Most obviously, our findings ought to invigorate the perennial debate surrounding Rule 26(c) reform—a debate that, to this point, has run aground on something like empirical exhaustion.  Of course, there are multiple perspectives and inputs that will also shape those discussions, and it may be that policymakers decide that no change is necessary.  But at a minimum, this data should discipline the debate.  It should sweep away reform opponents’ reassurances that protective orders are, in fact, rare and, when issued, carefully analyzed.  That’s a big deal.  Reform opponents have long agreed that there would be a problem if the entry of protective orders were “automatic or cavalier.”  Now, after examining 2.2 million cases, we show that that’s plainly the case.  Judges overwhelmingly rubber stamp stipulated protective orders without even paying lip service to the Rule 26(c) “good cause” standard.

David: The study also contains lessons regarding judicial discretion.  In an era when many worry that district court judges have ever greater discretion, our findings are provocative.  We assume judges’ discretion is cabined by procedural rules, but in this context at least, they act as free agents.  In fact, the regional U.S. Courts of Appeals differ, sometimes significantly, in their framing of the “good cause” standard, but this does not appear to translate into any noticeable difference in grant rates.

These insights have additional implications.  While one may assume that policymakers, including the Judicial Conference, are the ones to reform the rules through formal processes, our findings suggest that reform, at least sometimes, takes place sub rosa, through the daily exercise of judicial decisionmaking.  Perhaps, in a digital world characterized by the rapid and even viral dissemination of information, we should adjust the “good cause” standard to make it easier to shield information disclosed in the course of litigation.  But if such a step is to be taken, it should come transparently through the rulemaking process and only after candid discussion.  It shouldn’t happen secretly, through the steady drip of judicial discretion.

Do you have any further work concerning litigation secrecy in the words?

Nora:  We do!  This piece is the first of many papers Jonah, Austin, David, and I plan to write on litigation secrecy.  The litigation secrecy ecosystem involves not just protective orders but also orders to seal and secret settlements.  All these devices are deeply interconnected.  All contribute to the opacity of our courts.  And all, to this point, have been subject to surprisingly limited empirical scrutiny.  Stay tuned!

For more on the Rhode Center, which Nora and David Co-Direct, see here.  And, for a copy of “Secrecy of Stipulation,” see here.

Nora Freeman Engstrom is the Ernest W. McFarland Professor of Law at Stanford Law School. A nationally recognized expert in both tort law and legal ethics, she explores the day-to-day operation of the tort system, including the system’s interaction with alternative compensation mechanisms, such as no-fault automobile insurance and the Vaccine Injury Compensation Program.

David Freeman Engstrom is the LSVF Professor in Law at Stanford Law School. A far-ranging scholar of the design and implementation of litigation and regulatory regimes, Engstrom’s expertise runs to civil procedure, administrative law, constitutional law, law and technology, and empirical legal studies.