Welcome to PAX

What PAX is for, and how a video becomes a measured result you can act on.

PAX is where political video gets made, tested and read. You write a script with me, record it, test it on a randomized panel, and see exactly how much it moved people, side by side with every other result on file. No guessing which ad works: you see it.

The loop

Everything here serves one loop. Each step has its own page.

1. Make a video. Write a script with me in Chat and record it on camera there, or record one in Clips.
2. Test it. A message test shows a randomized panel either one video or a placebo, then asks how they would vote. Staff set tests up in Tests, in Survey.
3. Field it. The Pollster in the Agentic Inbox mails each person in a contact group a link of their own, reminds whoever hasn't finished, and closes the test at its quota or its deadline.
4. Compute the lift. When a test closes, I weight the respondents to the population's targets and compute each video's lift against the placebo.
5. Read it. The result reads in Lift, on the video's panel in Clips, and in Reports.

A test and a survey

They sound alike, but they answer different questions.

• A test is a randomized experiment. Each respondent sees one arm, a video or the placebo. A video's lift is the difference between its viewers' answers and the placebo viewers' answers, in percentage points. Because the groups were assigned at random, that difference is the video's effect.
• A survey measures where people stand. Everyone answers the same questions, and the result is each option's share, with no video and no comparison group.

What a result reads against

America Thinks holds what America already thinks, in its own wording: the General Social Survey, ANES, CES, public polls, official results, the census, turnout and enrollment records. A new test's or survey's numbers read against that record, so you can tell a real shift from the noise.

Where to go next