Crowding risk is one of the more elusive threats a long/short fund carries. A manager can size positions independently, hedge factor exposure carefully, and still find the book moving in lockstep with a dozen other funds nobody at the firm has ever spoken to. When one of those funds unwinds, the correlation that was invisible in normal markets becomes the dominant force in a drawdown. The problem has always been measurement: crowding tends to show up in P&L after the fact, rarely before.
13F filings offer one of the few systematic ways to see it coming. Every institutional manager with more than $100 million in qualifying US equity assets has to disclose long positions quarterly, within 45 days of quarter end. The data is backward-looking and incomplete, no shorts, no derivatives, a built-in lag, but it is also the only standardized, public record of what large pools of capital actually own. Aggregated across the right peer group, it becomes a usable proxy for consensus positioning.
We define Situational Awareness funds as strategies with AUM exceeding $20 billion that file 13Fs, a category that has grown large enough this year that it is hard for allocators and risk officers to ignore. These are funds big enough that their positioning decisions move markets on their own, and correlated enough with each other that overlap between any two of them is rarely coincidental. If a stock shows up across ten of these portfolios at similar weights, that is a data point about market structure, not just about the stock.
To make this usable, we built two baskets from the latest 13F holdings of Situational Awareness funds: .SITAWARE, weighted by the AUM of the filing fund, and .SITAWARE_EW, an equal-weighted version. The distinction matters because an AUM-weighted basket tends to be dominated by whichever two or three funds are largest, which tells you more about a handful of managers than about the category as a whole. The equal-weighted version treats each fund's positioning as one vote, surfacing names that are broadly held across the peer group even when no single mega-fund is particularly overweight. Toggling between the two gives a cleaner read on whether a crowded name is crowded because everyone owns it, or because one large fund owns a lot of it.
A simple overlap count, the number of names two portfolios share, is a blunt instrument. Two funds can hold the same twenty names and carry very different risk if the position sizes and betas differ. So we regress each client portfolio against .SITAWARE and .SITAWARE_EW and compute a few things at the position level: overlap weighted by portfolio exposure, the beta-adjusted contribution of shared names to portfolio volatility, and return correlation between the client's book and the basket over trailing windows. The output is not a single crowding score. It is a decomposition showing which specific positions are driving co-movement with the peer group, and how much of the portfolio's volatility that co-movement explains.
13F data itself is quarterly and lagged, which is precisely why treating the resulting basket as a static, quarterly reference undersells it. The basket composition gets rebuilt each quarter as new filings arrive, but the correlation and beta analysis against a client's current portfolio should run daily, because the client's book changes every day even when the peer basket does not. A position that looked lightly correlated with .SITAWARE in April can drift into high correlation by June simply because the client added to it, with no change at all in the underlying 13F data. Running the regression daily catches that drift as it happens, rather than at the next filing deadline.
We ran this analysis for a long/short client earlier this year and found that roughly 30% of the portfolio's realized volatility over the trailing quarter was explained by names also held across .SITAWARE_EW, concentrated in a handful of positions the manager had sized independently for idiosyncratic reasons. None of the individual positions looked crowded in isolation. The exposure only became visible once beta-adjusted overlap was aggregated across the basket. It did not change the manager's thesis on any single name, but it did change how the aggregate book was hedged heading into the next earnings cycle.
Crowding cannot be eliminated, and it should not be treated as inherently bad. Some of it is simply good ideas being good ideas, discovered independently by more than one manager. The value is in knowing how much of it a portfolio is carrying and where it sits, rather than finding out during the next unwind. If you would like to see how your portfolio measures up, get in touch.
