Every letter on this site is a real comment submitted to the SEC docket. A docket this size also picks up duplicates: the same person submitting the same letter twice, sometimes by accident, sometimes on purpose. This page explains what counts as a duplicate here, what does not, and what happens to each.
Here are the patterns that show up in this docket:
The docket name is the name on file with the SEC for that submission — whatever the filer typed into the comment form's name field, or wrote on the envelope. It is what shows up in the docket listing and in this site's own name column. The signed name is whichever name the letter is actually signed with, inside its own body text — a signature block, a sign-off line, a closing "Sincerely,". Most of the time the two match, so this distinction rarely matters. Docket names alone never establish a cross-name resend — it takes the letter's own text: the signed name on one copy has to match the docket name on the other copy of the same letter. That match, found in the letters' own words rather than in how either was filed, is what links the two.
For the first three to count as a duplicate, the text has to be very similar between submissions, close to word for word — for a cross-name resend, a similar docket name proves nothing on its own; it's the shared text, confirmed by the letters themselves, that counts. If the wording differs meaningfully between two letters from the same person, it falls into the repeat-commenter category instead.
Where exactly that line sits. Two submissions from the same person are the same letter when the only differences are incidental: punctuation or capitalisation, a title or signature or e-mail footer added or dropped, a typo or a different form of the same word, an added interjection or courtesy phrase, or an artifact of how the mail was sent. They are two different letters as soon as the second one carries a real change of substance — a new argument raised, an existing one dropped, a different word chosen to make a point, or a clause rewritten. The test is whether the writer said something new, not how many characters moved.
One letter containing another is still one letter. If the second copy is everything the first one says, plus an added line that does not raise a new argument — a slogan, a sign-off, a restatement of the same point — the two are still the same letter, not two. The dividing line is containment: if one copy says X and the other says X plus Y, and Y is not a new point, the fuller one (X + Y) is the version that counts, the same as any other duplicate. Two letters are only separate when each raises something the other does not — one makes point X, the other makes point Y — even if the two overlap.
Identical text from two different people is not a duplicate. When the same wording turns up under two unrelated names, that is a template someone circulated, not one person submitting twice, and both letters count. Duplicates are always the same person twice.
In both duplicate cases, one copy counts toward the headline totals — the fuller version when the two differ. The other copy stays visible on the letters page, marked excluded duplicate, with a link to the copy that counts. Nothing is deleted, the letter is just excluded from the tally so the same person's comment is not counted twice.
Short excluded duplicates show their full text right on the page, the same as any other letter. When the letter is long, the text collapses behind a "show it" link instead, so a person with several near-identical resends does not repeat a long paragraph down the page five times. Here is an example.
A cross-name duplicate, in practice. This one would be invisible to a name match — the two copies carry different docket names, and the link only shows up in the text itself.
The SEC's own form letters are never merged into this. The SEC posts high-volume campaign letters separately, as a template text plus a count of how many people sent it. Those postings are the SEC's own record of the docket, so they are reported on their own terms, side by side with the individually classified letters and never added to them or folded into them. An individual letter that happens to match one of those templates stays an individual letter here; it is not reassigned to the campaign.
Repeat commenters are not held out. Michael Fox commented twice on this docket, in his own words each time, and both letters count toward the totals. Here's an example.
Does counting them every time inflate the totals? Not materially. Counting each repeat commenter once instead of once per letter would move the headline Oppose share by about two thousandths of a percentage point — largely because the repeats that would come out are, for the most part, duplicates that were never counted in the first place.
The full breakdown — how many letters sit in those groups, how many of the extras are already excluded, and what the remaining ones look like — is on the repeat commenters page. Since the number barely moves either way, I count every letter someone actually wrote and mark the pattern instead.
Repeat commenters are matched by name, and then the letters themselves have to back the match up. Two letters filed under the same name on different days are treated as one person when the text reads like one person writing. A bare first name is never enough on its own — "Barbara" appears in this docket dozens of times. Neither is a common full name: "David Smith" or "Susan Taylor" could easily belong to several people, and in this docket they do, which is why those letters stay separate and are tagged shared name instead:
Two things override the name. A pair of letters read in full and judged to be the same person is linked whichever way the names are spelled — that is how a letter filed under a handle gets tied to the same writer's letter filed under their full name. And a name that recurs many times is only treated as one prolific commenter when the letters themselves tie a good number of them together; otherwise it is left as what it is, several people who happen to share a name.