Matthew Miller's reaction to the new Washington Post-ABC News poll at the weekend caught my eye. The poll showed just 18 per cent of Americans believed US President Donald Trump should pardon Paul Manafort.
"This is the 'shoot someone on 5th Avenue' caucus, and it's much lower than Trump would have you believe," the former Obama-era Justice Department official tweeted.
That reference, of course, is to Trump's famous 2016 campaign claim that his supporters would stick by him even if he pulled out a gun and shot someone on Fifth Avenue in Manhattan.
The implication was that Trump could do pretty much whatever he wants and his devotees wouldn't blink.
It doubles as an ego trip for Trump and a handy way for journalists and pundits to describe how Trump skates past controversies, no matter how jaw-dropping they are, because his base shrugs.
Which got me thinking: How big is the "shoot someone on Fifth Avenue" caucus in America?
The answer I arrived at as anywhere from one in eight to one in five Americans - between 12 and 20 per cent, around where Miller pegged it.
These are the people who seem prepared to justify and/or forgive pretty much anything Trump has done or even has threatened to do. It represents half or less of Trump's overall supporters.
Here's how I arrived at that number.
The Post-ABC poll shows 53 per cent of American adults strongly disapprove of Trump, while 24 per cent strongly approve. Those numbers are slightly worse for Trump than other recent polls, but it's usually somewhere around 2-to-1 strong-disapprove to strong-approve. Generally, about a quarter of Americans strongly approve of Trump.
It's notable that the opposition to Trump is more likely to feel that way "strongly" than his support.
That suggests that, despite Trump's approval hanging tough at around 40 per cent for his entire presidency, a significant portion of that isn't completely thrilled with him.
Those folks could feasibly at some point be convinced he's gone too far - whether by shooting someone or for some other reason.
But that 24 per cent also probably oversells the true Trump-or-die supporters. So I sought out other poll findings to see what portion of that 24 per cent might be immovable (or something approximating it).
Here's a summary:
12 per cent say it's "acceptable for a presidential campaign to obtain information on a political opponent from a hostile foreign power" ( Quinnipiac University )
This probably undersells the Trump-or-die caucus, given Donald Trump jnr maintains he didn't actually obtain useful information at the Trump Tower meeting with a Kremlin-tied lawyer. Republicans may also be okay with the meeting given many of them don't view Russia as "hostile."
26 per cent say Trump should be able to shut down news outlets for "bad behaviour" ( Reuters )
This oversells the Trump-or-die caucus, given 20 per cent of Democrats (!) back a president's ability to do this, and that proportion of Democrats most certainly doesn't back Trump. If you isolate just GOP-leaning voters who favour this approach, it's 16 per cent of the total population. Of course, they might just be anti-First Amendment more than pro-Trump. (But still ...)
12 per cent say it's not a "big deal" if Russia interferes to help Republicans ( Yahoo-SurveyMonkey )
This recent poll asked whether it was acceptable for Russia to try to help the GOP in elections, as the intelligence community says Russia did in 2016.
Fully 40 per cent of Republicans said it was either "appropriate" or "not appropriate, but wouldn't be a big deal" if Russia did that - despite foreign involvement in US elections being clearly illegal.
Given Trump has said accepting info from Russians would simply be par-for-course in Americans politics, this seems like a telling number as far as how many Americans accept that defence.
22 per cent say if Trump shot someone on Fifth Avenue, they would approve of his job performance ( Democratic pollster Public Policy Polling )
Points for asking the question directly, I guess. But it's also a pretty ridiculous hypothetical that respondents probably recognise as such.
There have to be some Trump supporters who just answer "yes" out of devotion to Trump, and given what he's said on the topic. Also, "approve" isn't quite the same as saying you would vote for a murderer/gunman.
15 per cent say there is almost nothing Trump could do to lose their support ( Public Religion Research Institute )
Credit to PRRI for the best way of asking this question.
What Trump is really saying is that he could do pretty much anything and not lose his supporters, and this poll suggests 15 per cent say that exemplifies them.
Whether they'd actually follow through if Trump killed a guy? That's less certain, for the reasons described in the previous item.