Tuesday, July 21, 2026
This post discusses the [U.S.] Supreme Court Database, which was used to help build this hub. It is simultaneously both a invaluable resource and an occasional source of frustration, as discussed below.
Note that as of this post, this site is using version “2025 Release 01” of the Modern Database and version “SCDB Legacy 07” of the Legacy Database. Per the database’s citation guidelines:
Harold J. Spaeth, Lee Epstein, Michael J. Nelson, Andrew D. Martin, Jeffrey A. Segal, Theodore J. Ruger, and Sara C. Benesh. 2024. Supreme Court Database, Versions 2025 Release 01 and Legacy 07. https://scdb.psu.edu
It’s also worth noting that, in 2024, the database’s home moved to Penn State College of Liberal Arts. While the new site looks great, it does lack some of the features of the previous Washington University Law site, like the Analysis page, and in the two years since the move, nothing much has changed – except of course the 2025 database release. It’s also strange that the previous site continues to update its releases to match those at Penn State, while making no mention of the database’s move, nor any indication of what might happen to the supremecourtdatabase.org domain.
On this site, under Collections, you’ll find a group named Outstanding Issues, and in there, you’ll find a number of other interesting groups, but the one we’re going to focus on first is named Cases with Incorrect SCDB Dates.
This is an automatically-generated list of all cases where we have flagged an issue with one more of the dates that SCDB has recorded for those cases; specifically, their dateArgument, dateRearg, and dateDecision fields.
Since we rely on the U.S. Supreme Court’s own meticulous research, “DATES OF SUPREME COURT DECISIONS AND ARGUMENTS,” any deviations from that research are suspect. And there are numerous SCDB deviations.
For example, Miller v. Kerr (1821) is documented by the Court as being argued on March 13, 1821 and decided two days later on March 15, 1821.
SCDB claims that it, too, is relying on that data; regarding dateDecision, it says:
This variable contains the year, month, and day that the Court announced its decision in the case. For volumes 2-107 of the U.S. Reports (1791-1882), we relied on Dates of Supreme Court Decisions and Arguments, prepared by Anne Ashmore of the Library of the Supreme Court, because many early reporters do not list the date of decision.
And yet, inexplicably, SCDB set dateDecision to March 15, 1822 instead of 1821. So either this was a typo or someone concluded they had found a better source, but nowhere is there any notation, either in the case record or anywhere else on the website, of what this better source might have been.
If you examine Miller v. Kerr (1821) on our website, you’ll see that the U.S. Reports heading is “FEBRUARY TERM, 1822,” which might lead someone to believe this case must have been decided in 1822… until you read a bit further, where it says:
This cause was argued and determined at the last term, but omitted to be reported.
And so it goes.
Originally posted on February 18, 2019 on lonedissent.org and reproduced below with some minor changes and link updates
The Supreme Court Database, aka SCDB*, is a enormously valuable resource. Other sites have even referred to it as “the gold standard for high-quality legal information.” It owes much of its reputation to Harold Spaeth, a political science professor who created “The Original U.S. Supreme Court Judicial Database (nickname: ALLCOURT)” decades ago, and worked with the SCDB folks to help produce the modern version. Sadly, Harold passed away in 2017.
So, regarding the SCDB: does it really meet the gold standard? I would certainly concede bronze or silver status perhaps, but not gold. And this is not because I think there’s another, better, or more deserving database out there. I simply feel that the Gold Standard should be reserved for a database that is not only without peer, but is also well-designed, well-maintained, rigorously audited, and open-source, embracing transparency and community participation, and continuing to evolve to solve old problems and new challenges. At the moment, SCDB has lots of room for improvement.
Here are some examples of SCDB docket numbers for Original Jurisdiction cases:
"5, Orig."
"126, ORIG."
"10 Original"
"8 (Original)"
"15 orig."
"6 ORIG"
"8 original"
"ORIG" and " ORIG"
"15 ORIG ORIG" (just to be sure?)
"No. 12, Original"
"No. 137, Orig."
"22O142"
Yes, the human eye can easily discern that these are all original jurisdiction docket numbers, but databases are designed to be consumed by computers, not humans, and one of the presumptions for any database is well-defined and strictly-adhered-to data formats. Even after consulting SCDB’s online codebook regarding the docket field, this is all we’re told about such docket numbers:
Cases invoking the Court’s original jurisdiction have a number followed by the abbreviation, “Orig.”
Well, apparently, there’s more to it than that.
There’s also a small problem with “Miscellaneous” cases; SCDB usually appends a single letter (“M”) to the docket number, but sometimes there’s a space (eg, “61 M”) and sometimes not (eg, “133M”).
The U.S. Supreme Court often “consolidates” multiple cases from lower courts into a single case. For example, the docket number of the “lead” case in McConnell v. Federal Election Commission (540 U.S. 93) is 02-1674. However, the complete list of consolidated cases, by docket number, looks like this:
02-1674,02-1675,02-1676,02-1702,02-1727,02-1733,02-1734,02-1740,02-1747,02-1753,02-1755,02-1756
and if you download SCDB’s “Cases Organized by Docket” and search for 540 U.S. 93, you will indeed see all 12 cases listed.
So what’s the problem? Consolidated cases are not consistently included.
For example, look at East Texas Motor Freight System, Inc. v. Rodriguez (431 U.S. 395). Three cases were consolidated:
75-718,75-651,75-715
but even when using SCDB’s “Cases Organized by Docket” files, all you’ll find is 75-718.
There may be some rationale at work here. For example, it’s possible that the disposition of the “non-lead” cases did not differ in any material way from the “lead” case, so the other cases were deemed superfluous. But there are numerous examples where the exact opposite is true (i.e., all consolidated cases recorded even when they all had the same disposition), so that would be a rationale of convenience rather than of principle.
The SCDB website simply says:
Multiple docket numbers under a single case citation almost always contain the same issue as the lead case and differ only in the parties to the case and its origin and source.
And this isn’t a trivial problem. When you look for the transcript in the East Texas Motor Freight System on the Supreme Court’s website, it’s only listed as Teamsters v. Rodriguez, No. 75-651. Not as 75-715 or 75-718, but as 75-651 – a docket number which you will not find in the SCDB.
For a case’s Date of Decision, the SCDB online codebook says:
This variable contains the year, month, and day that the Court announced its decision in the case. For volumes 2-107 of the U.S. Reports (1791-1882), we relied on Dates of Supreme Court Decisions and Arguments, prepared by Anne Ashmore of the Library of the Supreme Court, because many early reporters do not list the date of decision.
Importing dates from a Supreme Court document should have been an error-free process, yet it wasn’t. Take the case of United States v. McDowell (8 U.S. 316). SCDB claims it was decided on March 7, 1807, but the Supreme Court’s “Dates of Supreme Court Decisions and Arguments” document – which SCDB says it relied upon – indicates March 7, 1808. I have found dozens of similar mistakes.
And these kinds of mistakes aren’t just limited to those older cases. Look at Perry v. Leeke (488 U.S. 272). It was decided January 10, 1989, but SCDB lists the decision date as “1/1/1989”. Even worse, SCDB lists the argument date for Brown v. Board of Education (347 U.S. 483) as “12/8/1952”, but in fact, arguments began on December 9, 1952 and lasted three days.
There is also another, subtler problem with cases listed in the “Dates of Supreme Court Decisions and Arguments” document: the decision date of a number of cases could not be precisely identified, even by the Supreme Court’s librarian, so only the date of the term was listed. This occurred, for example, in Welsh v. Mandeville (9 U.S. 321), where the decision date was listed as “Feb. term 1809”.
Unfortunately, SCDB appears to have morphed such dates into the first day of the first month of the term, resulting in a date (e.g., February 1, 1809) that appears to be precise but is almost certainly incorrect.
NOTE: As a public service, I have extracted all the decision dates and argument dates from the Supreme Court’s Dates of Supreme Court Decisions and Arguments and produced an easy-to-use spreadsheet. I recommend using this file instead of the one on the Free Law website, because the last time I checked, the dates in their file were badly scrambled, and it didn’t include any argument dates. The dates on the first few lines of their file:
2 U.S. 401|West v. Barnes|2|401|1791-08-17
2 U.S. 401|Vanstophorst v. Maryland|2|401|1791-08-17
2 U.S. 401|Oswald v. New York|2|401|1792-02-14
...
clearly do not match those provided in the Court’s PDF.
Here’s a list of all the corrections we’ve made to dateDecision in SCDB thus far, with links to the source material used, so that they can all be verified. This is a degree of transparency that you will not find on the SCDB website.
Some cases are argued over a period of multiple days (and not necessarily consecutive days). An early example of this is Talbot v. Janson (3 U.S. 133), which was argued over the course of ten days:
Thursday, August 6, 1795
Friday, August 7, 1795
Saturday, August 8, 1795
Monday, August 10, 1795
Tuesday, August 11, 1795
Wednesday, August 12, 1795
Thursday, August 13, 1795
Friday, August 14, 1795
Tuesday, August 18, 1795
Wednesday, August 19, 1795
This also occurs with some regularity in the “modern” era. See American Trucking Assns., Inc. v. Atchison, T. & S. F. R. Co. (387 U.S. 397), which was argued:
Thursday, April 13, 1967
Monday, April 17, 1967
However, recording all the dates of an oral argument (or even just the number of argument days) didn’t seem to interest Harold Spaeth much, because his “ALLCOURT” database (SCDB’s predecessor) provided only an ORAL field for the first date of argument.
Despite my best efforts ten years ago to convince SCDB to consider broader research interests and to at least enable the coding of all argument dates for a case, all they did was rename Spaeth’s variable to dateArgument and continue the old practice, without justification:
On some occasions, oral argument extended over more than a single day. In such cases, only the first date is specified.
NOTE: For the record, SCDB incorrectly reports that 387 U.S. 397 was argued on March 13, 1967, so we have more than a completeness problem – we have the usual accuracy problems as well.
Then there’s the problem of multiple rearguments. Once again, the Spaeth “ALLCOURT” database dealt with this, but in the same limited fashion, by providing a single REORAL field, and SCDB followed suit with its dateRearg variable:
On those infrequent occasions when the Court orders that a case be reargued, this variable specifies the date of such argument following the same day, month, and year sequence used in the preceding variable (dateArgue [sic]).
The limitation here is even worse than before, because not only can a reargument span multiple days, but there can also be multiple rearguments. Take a look at Boyle v. Landry (401 U.S. 77). The second reargument on November 16, 1970 is nowhere to be found in SCDB.
SCDB also doesn’t comprehensively list cases that were granted, argued, and then dismissed without an opinion. This can happen when the Court “DIGs” (dismisses as improvidently granted) a case, or when it dismisses a case that has later become moot. To be clear, I’m referring to cases that were fully briefed and argued and then dismissed, which makes them significantly different from the many petitions that are routinely denied, as well as the occasional petition that is granted and then dismissed before argument.
This is not to say that SCDB doesn’t track any DIG’ed cases, but merely that its recording of them is haphazard. For example, Stiles v. United States (393 U.S. 219), argued November 20, 1968, is not listed in SCDB, while Ford Motor Co. v. McCauley (537 U.S. 1), argued October 7, 2000, is listed. The failure to record all such cases frustrates a variety of research, such as the accurate tracking of oral argument activity, the frequency of DIGs, etc.
As an aside, it’s also not a simple matter to identify just DIG’ed cases. SCDB has a caseDisposition variable that is generally set to 9 (“petition denied or appeal dismissed”) in such cases, but that value is also used in other cases, such as Schwarz v. National Security Agency (526 U.S. 122), where the case was granted and a per curiam opinion was issued denying petitioner’s motion.
A Natural Court, as the SCDB online codebook explains, is:
[A] period during which no personnel change occurs. Scholars have subdivided them into “strong” and “weak” natural courts, but no convention exists as to the dates on which they begin and end. Options include 1) date of confirmation, 2) date of seating, 3) cases decided after seating, and 4) cases argued and decided after seating. A strong natural court is delineated by the addition of a new justice or the departure of an incumbent. A weak natural court, by comparison, is any group of sitting justices even if lengthy vacancies occurred.
Although one could quibble with the SCDB’s natural court definitions (which I’m sometimes tempted to do), the larger problem is the accuracy of the dates for the courts that SCDB has chosen.
For example, it lists the transition between the Warren and Burger courts like so:
1411 Warren 11 May 14, 1969 - June 22, 1969
1501 Burger 1 June 23, 1969 - June 08, 1970
However, SCDB also lists a series of decisions handed down on June 23, 1969:
1969-06-23: North Carolina v. Pearce [413,418] (395 U.S. 711)
1969-06-23: Chimel v. California [770] (395 U.S. 752)
1969-06-23: Benton v. Maryland [201] (395 U.S. 784)
1969-06-23: Von Cleef v. New Jersey [837] (395 U.S. 814)
1969-06-23: Shipley v. California [540 Misc.] (395 U.S. 818)
1969-06-23: Moya v. DeBaca [996 Misc.] (395 U.S. 825)
And while it would be very impressive for the Court to hand down six decisions on the first day under a new Chief Justice, the reality is that June 23, not June 22, was Chief Justice Earl Warren’s last day.
And this mistake with the “Warren 11” court isn’t an isolated “one-off”. There are similar problems with the “Warren 4”, “Warren 5”, “Warren 6”, and “Warren 7” courts, not to mention “Stone 2” or “Rehnquist 1”, among others.
And this isn’t merely a problem with the natural court dates. Numerous cases are filed under one natural court even though they were decided under another.
Look at Braverman v. United States (317 U.S. 49). It was argued on October 21, 1942 and decided on November 9, 1942, which would put it squarely in SCDB’s “Stone 1” natural court. But it’s coded in SCDB as being in the “Stone 2” (1202) natural court.
How SCDB defines the Term in which a case was decided is problematic: it uses a simple number (a year), which is insufficient to properly identify the actual term in which a case was decided.
Specifically, until 1802, there were two terms per year. This is why my project has adopted a string format for Supreme Court terms (“YYYY-MM”) rather than an ambiguous numeric format (YYYY).
The ambiguity didn’t stop in 1802, either. In 1844, there were two terms as well, because after the normal January 1844 term began, the Act of June 1844 changed the start of subsequent terms to December. Apparently out of habit, U.S. Reports still called these terms “January Terms”, but that didn’t change the fact that, beginning in December 1844, the Court started churning out new opinions.
SCDB, on the other hand, ignores the actual dates that the Court operated, and instead pretends that the Court’s work started every January – up until 1850, when U.S. Reports finally changed its “term-inology”. As a result, SCDB implies there two terms in 1850, when in fact, there were not.
SCDB apologists could argue that, as long as the ambiguity of the Term variable is properly documented, researchers can work around its limitations by also examining the dateDecision variable and checking for all the above conditions. Of course, the logical extension of that argument would be to eliminate the Term variable entirely, because obviously the precise term of any case can be determined by applying a complicated set of rules to dateDecision.
Harold Spaeth’s TERM variable didn’t suffer from this ambiguity, because his “ALLCOURT” database didn’t deal with cases before the Warren Court; however, there have also been a number of Special Terms, both before and after the Warren Court, which even the “ALLCOURT” database failed to properly deal with.
As I’ve previously documented, there are some variables, such as caseOrigin containing undocumented values (e.g., 157, 158, 161, etc).
And then there’s lawMinor, a free-form string that has become very problematic. Here’s a small subset of the values, to give you a sense of the problems:
"unidentifed act of congress",
"unidentifed act of congress, 1828",
"unidentifiable",
"unidentified",
"unidentified 1807 act of congress",
"unidentified act of congeress",
"unidentified act of congress",
"unidentified act of Congress of 1824",
"unidentified act of congress, 1824, sec. 32",
"unidentified acts of congress",
"unidentified federal statute",
"unidentified law",
"unidentified patent law, sec. 15",
"unidentified RS",
"unidentified sdtatute",
"unidentified statute",
"unidentified US laws",
"unidentified US statute",
"Uniformed Services Former Spouses' Protection Act 10 U.S.C. 1408",
"unknown",
"unpaid opium tax",
"unrestricted sale of allotments",
"unspecified",
"UNSPECIFIED",
"unspecified act of congress",
There are lots of duplicate values, varying only in form, not in substance, as well as lots of typos.
When cross-referencing the cases in SCDB with other reputable sources (eg, data extracted from the Supreme Court’s Case Citation Finder), I’ve also come across a number of cases which, even though they were considered “cite-worthy”, do not appear in SCDB.
I’ve logged some of those instances on my website (e.g., missing cases and unknown citations). I realize there are many “back of the book” cases that don’t merit attention (e.g. denials of cert), but that’s not true in all such cases, so perhaps SCDB should consider creating a second much simpler table of cases that cites all the cases it has deliberately omitted.
This is a broad category, encompassing every field of every record, and it’s best illustrated with a simple example.
The case Ableman v. Booth (59 U.S. 479) is recorded in SCDB with an argument date of “1856-01-04”. This is at odds with the argument data reported in the Supreme Court’s “Dates of Supreme Court Decisions and Arguments” document, which reports no argument date for that case. Remember, that’s the document that SCDB explicitly says it relies on for dates in early cases such as this.
One interpretation is that this is simply an SCDB error, in which case the argument date should be deleted from the next release. However, typos typically manifest themselves as a mistake in one or two digits, not as an entirely new value appearing out of nowhere.
Another interpretation is that SCDB, relying on some other (unspecified) primary source, discovered that the case had indeed been argued on January 4, 1856.
And we have no idea which is the correct answer. It’s also quite likely that, at this moment, no one working on SCDB knows the correct answer, either.
Some years ago, one of the SCDB principals, Andrew Martin, told me that detailed differentials, change logs, etc, don’t matter, because end-users “can perform the differentials just as well as we can.” That is absolutely true, and also absolutely beside the point, because it isn’t just the “differentials” we care about, but also where the new data came from. Without that, we run the risk of future researchers continually tripping over previously discredited bits of data (e.g., dates in U.S. Reports that turned out to be typos).
Without any explanation to accompany the growing number of corrections that are being made over time, the database becomes increasingly impossible to validate, because the sources of the underlying data, as well as any changes made to that data and the reasons for those changes, are kept private – if they are kept at all.
I long ago advocated for greater transparency in what SCDB chooses to add or correct in its database, including change logs with every release. These days, an even better step forward for SCDB would be to do what I’ve done here, which is to create an open-source repository containing copies of all the data sources being used, along with the scripts used to process them.
Issues like those with Terms arose simply because SCDB didn’t fully consider the impact of older cases on a design that it inherited from Harold Spaeth’s “ALLCOURT” database. Other issues, like those with Argument and Reargument Dates, could be considered a failure of imagination – except that we discussed dating issues with SCDB principals (Andrew Martin and Troy DeArmitt) at least ten years ago, and the only headway we were able to make was a vague commitment to consider “database extensions” that would allow groups like Oyez to add more comprehensive oral argument information (e.g., dates, names of advocates, etc). As far as I can tell, that never happened.
In any event, it’s never too late to fix problems. Instead of making excuses, justifications, or brushing off good suggestions as “too much work”, SCDB should start acknowledging problems and create a roadmap for improving and evolving the database, defining new variables to address old issues and new features, deprecating problematic variables, and above all, adding rigorous data validation rules and cross-checks to eliminate mistakes and prevent future errors.
I think if SCDB really wants to be a Gold Standard, it also needs to stop being a siloed operation, performing all its updates behind a veil of secrecy, and making the world wait with bated breath for each new release. As an academic endeavour, more knowledge – not less – should be one of the goals, as well as encouraging cooperation and participation among all interested parties.
Or, SCDB can quietly extract whatever it wants here, without telling me or anyone else, and then roll out its next release with the usual (and completely useless) notation: minor corrections. And neither I nor anyone else will have any incentive to help again.
Since we use the SCDB, we shall cite it. In fact, we shall go one step better, and recite their instructions on how one should cite it:
To cite to the Supreme Court Database, please employ either of the following:
Harold J. Spaeth, Lee Epstein, Andrew D. Martin, Jeffrey A. Segal,
Theodore J. Ruger, and Sara C. Benesh. 2018 Supreme Court Database,
Version 2018 Release 02. URL: http://Supremecourtdatabase.org
Harold J. Spaeth, Lee Epstein, et al. 2018 Supreme Court Database,
Version 2018 Release 2. URL: http://Supremecourtdatabase.org
Please be sure to include the specific Version Number; e.g., ‘Version 2018 Release 02’ in your citation as this will indicate the particular version of the database being employed at the time of your reference. This matter is of great importance as the database will be updated with newly announced decisions, corrections, and the addition of new data for existing cases.
Note that indicating which release you are using is a matter “of great importance”.
Which is puzzling, since SCDB consistently refuses to describe, list, or otherwise explain exactly how any release differs from any other release. The differences are invariably described as nothing more than “minor corrections” – which can’t be right if they are also “of great importance”.