"oversampling" is a good thing

Trump approval in recent polls: UMass Amherst/YG: 33% (-5 from Jul '25) Fox: 41% (-2 from early March) Qpac: 38% (+1 from early March) AP-NORC: 38% (+2 from Feb) Reuters/Ipsos 36% (-4 from mid-March) CBS/YG: 40% (unchanged from Feb)

[edmond dantes] but you make one pan fried, egg dipped ham and cheese sandwich, what do they call you?

[tess] but you meet one d'urberville and all of a sudden, what do they call you

New Likert scale @aedwardslevy.bsky.social

astronomical winter, meteorological winter, punxsutawney phil, whatever; my only two arbiters of when springtime starts are the Capital Weather Gang and @darthbluesky.bsky.social

1. See alarming chart of federal workers' confidence falling between 2024 and 2025 2. See that survey went from officially-administered to convenience sample, and survey website warns "Results presented here should not be directly compared with the results of previous federal employee surveys."

have I mentioned recently how earnestly and unironically I love survey research

followed quickly by "are you sure you don't want to link the pdf of the full survey"

if you write a headline that says "More X say that Y," clippy's nerd cousin should pop up and start screaming until you clarify whether you mean "more X say Y than previously," "more X say Y than say Z" or "X are more likely than some other group to say Y"

I feel increasingly keenly that the thanks given are inadequate in light of my attention to the matter.

how has nobody done a nonresponse or panel attrition paper titled "I would very much like to be excluded from this sample"

nobody talks about how fun variable naming is

no matter which of the film endings you get, it's still a fun, lighthearted classic vote clue, no matter who(dunnit)

I am predisposed to be skeptical of "pollz: r they broken??" pieces (regardless of conclusion), but this is a really thoughtful essay about some of the current trends we're seeing

to make your homemade vanilla extract, start by simply unzipping the vanilla folder

Four new polls, one consistent finding: Most Americans oppose the United States’ strikes in Iran.

This is a really interesting read (tangentially, it's still kind of wild to me that we now have ~20-year longitudinal trends that are entirely online-mode; relevant here in that how hard you push indies to lean can matter, so it's nice to have apples-to-apples data)

As the United States turns 250, the country’s population is more racially and ethnically diverse than in previous decades. Three-quarters of U.S. adults see racial and ethnic diversity as a good thing for the country. Read more about Americans’ views on diversity here: www.pewresearch.org/...

if you're not inherently excited for this event, please note that I recently subjected @baseballot.bsky.social and @pbump.com to an impromptu round of "name that pollster based on the crosstab formatting" and they got them all right

i'm not answering these questions without a lawyer

repost = endorsement approval != vote intention 2024 GE electorate != a sample of all adults or all registered voters

The "being struck by lightning" item here is a fun little way to benchmark whether respondents are exaggerating perceived risk on other items. democracy.psu.edu/poll-report-...

do the kids these days even know about ^_^

also: so much data so little timeseries

if you're pulling randomly from a dataset of every State of the Union speech, is that an address-based sample

Setting the field of polling back by a millennia because when my buddy proposed "Prompt Rotation" I said "lol this guy thinks the questions have to take turns."

little-known piece of trivia*: the byrds' "turn! turn! turn!" is actually about the then-novel practice of rotating survey response options to prevent primacy effects *bluesky disclaimer: the reason this is little-known is because I just made it up

This distinction is much less discussed than the basic prob/non-prob distinction. But it's going to be the key one going forward: Did the entity claiming to be a respondent approach the survey firm (responding to ads or on a website) or did the survey firm find the respondent and initiate contact?

i want to share one of the coolest data collection methods i employed for a story, from over a decade ago. it was used for this story in what i still think might be the best series i have ever worked on:

A) if you want a rough guess about what people will think of a new question based on demos/past polls, you can literally just ask anyone who spends time with this sort of data B) there's a reason new polls are conducted anyway C) survey research is fundamentally a way for people to talk to people

In light of this, allow me to promote my own research showing that federal judges are really bad at statistics. onlinelibrary.wiley.com/doi/full/10....

maybe the potato is wearing a jacket because the potato is cold

never has there been an ad less successfully targeted to me

Sharing this for anyone with reading to recommend! (FWIW, my feelings on this are generally built around taking indifference/uncertainty seriously as an attitude in public opinion, but I think both DK/neutrals and forced choice give useful info and have a place.)

generally a fan of neutral options but I Do Not like how they are shoehorned into this scale

i'm sure this is my fault but I don't remember how

a-side: the birds are back in town b-side: girls just wanna have direct sun

lingthusiasm is always fun, but this episode is *extremely* fun

writing a black mirror where it's just like our world but closing metro doors make the slack message sound and vice versa

just flagging this latest press release from the world vexillology council

judging by the cumulative number of "for your awareness" emails I have personally received over my lifetime, I believe that I may now be aware of all things

I have an election law joke but I can't recount it.

select all that apply

(all the composers look glum except Samuel Barber)

there should be a show called the bach-elor(ette) where a bunch of aspiring classic composers have to, etc.

they hate to hear my "As Per My Previous Email"

the problem for Siena and High Point is that polling schools have to win by a margin of at least 3 points or it's technically too close to call

i'm going to be a very specialized ghost who only haunts people using "Americans" and "voters" interchangeably in discussing polling data

Neat new project from Pew, ⚡#poweredbyIPUMS Time Use 👀