Tl,dr: "news" is sometimes a nebulous concept and pure rational expectations "has hands"

The paper claims that asymmetric information _alone_ cannot explain why monetary policy surprises are correlated with macro variables. Yet many papers build off the exact opposite premise. I thought that maybe this was a consequence of Karthik's model being static..but it's not

It was nice writing this blog post and not worrying about a character limit. I hope to do some more of these but it's tricky not to fall into the trap of tweaking everything and the balance of MB/MC of time spent being way off.

Original WP: papers.ssrn.com/sol3/papers....

My thoughts have been that the true "information effect" is not about better info but about vibes + institutional credibility. What I found supports this, or maybe it's just me forcing a confirmation bias. You can decide for yourself. I just hope the debate becomes more grounded

As far as what to infer from the paper, my initial impression is the evidence backing the overall claims is more suggestive than definitive.

Is it in bad taste to repost a critical thread word for word from twitter 🤔? Just in case I will link it here. x.com/paulbsqt/sta...

Another one to add -- IV estimation will be biased if our instrument is a function of multiple structural shocks. Evidence from Bauer and Swanson, Miranda-Agrippino, and the term structure paper from earlier in this thread suggest this is an issue with high-frequency MPS arxiv.org/abs/2208.11828

A final paper that captures the broad theme: there's nothing wrong with the premise "taking the literatures approach as given, here are some more implications". But there should be other work not taking the approach as given and "opening the black box" arxiv.org/abs/2505.12422

I have a related paper available on my website, any comments are very welcome and write about it down the road. These papers have really shaped virtually all my projects.

Many papers try to gauge if the effects of policy are non-linear by including various terms in a LP/VAR. This paper helps show when these approaches will be uncovering nonlinearities and what type (e.g., x^2 does not help you say anything about size)

x.com/pmbruera/sta...

This is a diagnostic we all should run. For example, take this gov spending series. Almost all of the weight is being put on positive shocks. So a LP/VAR with this series isn't really estimating the effects of general government spending, really it's spending _buildups_

Since the LATE etc. results came out important caveats has been provided: the weights may be negative and they need not be interpretable. This paper shows assumptions to ensure non-negativity and shows the weights can be easily estimated

Code: github.com/mikkelpm/non...

Kolesár and Plagborg-Møller extend RS with weaker assumptions and also touch on ID by heterosk. and non-Gaussianity. The most revelatory part of the paper to me is how they push work from the 90s on "regressions are a weighted average" (Angrist and others) arxiv.org/abs/2411.10415

Re: IRFs, true causal inference is more fragile than E[Y|x+\delta]-E[Y|x]. Rambachan and Shepard gave the first, broad overview of when various estimators and objects have a causal interpretation arxiv.org/abs/1903.01637