On average, how much slippage (measured in lost % return potential) is typical for an operating quant fund that trades in, say, major U.S. equities?
It depends almost entirely on the size of the positions the fund is trading. More specifically, it is a function of the liquidity of the stock.
ex-post liquidity can be observed from volume in aggregate and time & sales, in particular. ex-ante liquidity can be estimated (often reasonably well using a bucketed look-back approach)
It also depends on whether the fund is trading to provide liquidity(against momentum), or remove liquidity(momentum following). In the case of the former the slippage can often be negative. In the latter, slippage increases as trade size increases. Slippage can be lowered by using an algo like vwap, IS, twap, etc... But trades take longer to place under these block break-up approaches.
For large US equities, you can consistently trade in the low thousands of shares multiple times per day, with little to no impact on slippage.
This guy is generally considered the expert on this topic: http://www.courant.nyu.edu/~almgren/ his papers describe some novel approaches to minimizing slippage for large blocks.
The slippage is going to largely depend on the benchmark used and vary wildly depending on what kind of order working is being done (all day vwaps, pair orders working for a few minutes at a time)
Given that in my experience I've seen strategies that turn over infrequently with long running volume driven orders (hours, pov/vwap) using vwap as benchmark give up 20-30% in annual returns.
On the other hand I've seen strategies that turn over frequently and have short hold times have no slippage relative to the benchmark used (derived from tick/order book data).
But that is anecdotal, going to be hard to find good numbers since it requires disclosure of information that most want to keep private.
I can chime in with some figures and what it means for slippage to be significant.
First, some figures, as discussed by Zoltan Eisler form CFM. CFM refers to some of its strategies as "slow alpha". These models make directional (~0.1%) predictions over horizons from 1/2 to few hours.
According to these parameters, their US long-short equity fund has a pure annual PnL of 30%/year. On a daily turnover of 0.5% of the market's, their impact costs are close to 15%/year (as estimated by the Barro's square-root law). Clearly, these numbers live somewhere in between the realms of fiction and reality, but are nonetheless illustrative.
More details on this back-of-the-envelope calculation are here.
Second, slippage is also significant in the sense that it should be optimal. Indeed, some quant funds target an optimal slippage. Assume that the in-house best execution strategy is being used (exclude pure passive/market-making executions), so that there are no easy improvements there. Intuitively, if your slippage is too low, you should kick the market a bit more, and throw some more volume at it. If your slippage is too high, you should be a bit more gentle.