American courts, by tradition and for all the obvious reasons, reflexively tilt in a direction that favors public access to judicial records. As such, the law of sealing is "famously strict, at least on paper." In most Circuits, a trial court can grant a party’s motion to seal only if the party can demonstrate a compelling interest favoring secrecy. Moreover, a district court must articulate the basis for a decision to seal a document.
In [Sealed Document]: An Empirical Study of Sealing Orders in the Federal Courts, Nora Freeman Engstrom (Stanford) et al. bring data to how the law of sealing operates "on the ground." What they find, based on data from over 2 million federal civil cases, challenges the notion that sealing practices are operating as designed. Figure 3 (below) illustrates that among cases that reach summary judgment approximately 10% included at least one court-sealed document. To be sure, however, assessments of whether a 10% figure comports with a "famously strict" threshold are necessarily subjective. An excerpted abstract follows Figure 3.
Figure 3: Percentage of Civil Cases that Reached Summary Judgment with at Least One Sealed Document or Sealing Order by Filing Year of Case (2005-2011)
"... This Article offers overdue clarity—and our results are staggering. Using docket data from over 2 million federal civil cases, and combining state-of-the-art machine learning techniques with careful hand-coding, we find that at least 90% of motions to seal are granted. Further, although judges, when granting sealing motions, are duty-bound to supply specific on-the-record findings, most don’t. Most do not so much as cite the governing standard, and, when a standard is offered, it’s very frequently wrong. ..."