Suppose we then have attrition correlated with D. Lee Bounds logic suggests we should drop the individuals that pull up the ratios the most for one bound and individuals who pull down the ratios the most for the other bound.

Is that kosher? Is there a better approach?

Specifically, let D be binary treatment. We estimate preferences over binary options Y_1 and Y_2 with attributes (X_1,Z_1) and (X_2,Z_2) as follows:

P(Y_1>Y_2)=Beta_1*(X_1-X_2)+Beta_2*D*(X_1-X_2)+Beta_3*(Z_1-Z_2)

We are interested in Beta_1/Beta_3 and Beta_2/Beta_3

And there it is

They're now saying it's CNBC's fault

Weird preview. Should be

Reuters is repeating it, but doesn't say which interview

www.reuters.com/markets/us/f...

But maybe not? This whole thing is insane

bsky.app/profile/cono...

Ha! Thanks. Yea, I've been trying to mostly stay off of socials. We'll see if that holds long term or not.

I’d rather just go to sleep, but no way I’ll be able to