Thread by @0xfdf

@__paleologo Okay, I read it. I'll summarize it and provide some commentary. In brief I think it's a useful and credible paper, with specific empirical results, and I'd explore it for further research. But it's not groundbreaking.

@__paleologo So they start with the standard expected price impact we all know and love. Linear in vol, polynomial in participation rate. Just stage setting.

@__paleologo They define position illiquidity using the expected impact function. Their first claim is that the position illiquidity can be decomposed into position concentration, a measure from Pastor 2020, and (a function of) fund size.

@__paleologo Concentration C is a quadratic function of the portfolio position weight normalized by the position "liquidity weight" in the portfolio universe. Liquidity weight is the position ADV x position vol, divided by the sum of all such ADV x vol in the position universe. okay, fair.

@__paleologo "Fund size" is AUM, but also divided by the sum of all positions' liquidity. They show how this position-level decomp will roll up to fund level. So we have a portfolio decomp into the Pastor concentration and size measures from liquidity, and that gives us portfolio illiquidity.

@__paleologo We briefly come up for air with salient examples: SPY is highly liquid, XNDA is not. XNDA is a tiny biotech ETF, and biotech is notoriously illiquid. So, duh :). A lot of formalism for not much yet but let's continue.

@__paleologo Their next major claim is the self-inflated ETF return from its own flow driven trading is estimable as a function of its fund-level illiquidity J multiplied by its price impact, theta and its relative flow f. Okay.

@__paleologo We arrive at the first bit useful for crowding: that the authors will estimate fund-level price impact caused by ETF flows. This is what carries us to the crowding-related pieces later on.

@__paleologo They proceed to a decomposition of ETF return into uninformative (flow-driven) and informative (fundamental) price movement, conceptually similar to decomposing price return systematic versus specific risk.

@__paleologo They estimate the theta term (impact) in the foregoing using a fixed-effects OLS pooled on (fund, stock, time). Parameter estimates follow. Itsy bitsy R squareds.

@__paleologo They show how to roll up the expected price impact to the fund level. tada, an ETF's flows drive up the prices of its constituents. more parameter estimates and R squareds. moving along.

@__paleologo We arrive at the brass tacks: reversal. that's what we care about right? are these dumb ETF flow-driven price movements going to revert? Will my poor portfolio be hit by the dreaded crowding degross cascade?

@__paleologo So, they estimate reversal, and they find that initial price impact reverts 5 - 10 days with a long-run impact of 0.4. I am skeptical, but this is a refreshingly specific claim in crowding literature, so kudos for that.

@__paleologo Here's that more closely.

@__paleologo Okay, another systematic vs specific style decomp of ETF returns, this time into self-inflated from price pressure and fundamental. So conceptually, how much is susceptible to a feedback loop?

@__paleologo This brings us to the provocatively named "Ponzi flow", which is a feedback loop of continuous flow-driven trading induced by the price pressure of the fund's previous flows. Any risk manager's next logical question, does this cause a bubble and crash in the fund?

@__paleologo Finally, they find a statistically significant effect. BUT conclude that timing it is very difficult. And so I throw my hands up in the hair and weep, because this is basically all crowding research.

@__paleologo Crowding is something I think about often. This paper is interesting to me because it's modelable on higher resolution (days, at least) than 13F. And there are some empirical results that are decent (though I'm skeptical of the power/significance).

@__paleologo However, when I think about crowding I care less about the "what" and more about the "when". I think it's nice that we can confirm the intuition that frothy flows precipitate drawdowns, but I can't use that information without much better timing specificity than is shown here.

@__paleologo In any case, hope that helps. Like I said - interesting research direction, not groundbreaking, but I think it's worth a read and pondering.

@__paleologo And @systematicls I think you'll like some of this too

@__paleologo @systematicls Also @choffstein this might be relevant to your interests (probably less on the crowding side, more the ETF scene).