Price impact in the opening/closing auction

[jufils]

I trade through IB and the majority of my positions are opened/closed same day using LOO/MOC orders. I typically trade in relatively liquid stocks, and my trades usually represent anywhere from 0.01 - 0.05% of the daily volume. I've been doing this for a while with moderate success but I can never quite pin down the proportion of my expected return that is being eroded through price impact. I estimate I'm losing around 0.1% of daily return on an average round-trip transaction, but I don't have much confidence in the accuracy of that figure.

IB allows you to observe some basic details of the opening auction for all US equities (e.g. reference price, volume) and for NYSE-listed specifically you can see the extent of the imbalance. Basic calculations using a) this observed imbalance volume and b) the departure from reference price suggest that the square root law holds approximately at auction for stocks with daily dollar volume of less than USD 15 million or so but above that it seems to depart quite rapidly from that estimate. In fact, the impact seems to be much less than what the square root law would dictate for stocks > $15m daily dollar volume.

A few questions:

Does anyone have a more reliable estimate of impact at auction, either theoretically or through experience?
Does anyone know why impact would depart significantly from the square root law from stocks with > $15m daily dollar volume?
Would the fact that both my LOO and MOC orders are usually placed the night before have any significant (ideally, mitigating) effect on the price impact my trades incur?

 

[Robert Morse]

WHy do you think you are changing the auction price? WIth your volume, I would say it is not material and you have no other option if that is when you want to enter/exit.

 

[d08]

A more relevant number is the percentage of the auction, not the daily volume. Sometimes the ratio of daily volume/auction volume fluctuates wildly.

 

[sle]

Does anyone know why impact would depart significantly from the square root law from stocks with > $15m daily dollar volume?

Participation in the auction depends on many things. Sometimes you have stocks that have well over a billion in ADV and an opening auction with an average participation of under million.

Would the fact that both my LOO and MOC orders are usually placed the night before have any significant (ideally, mitigating) effect on the price impact my trades incur?

No real difference, as long as you participate in the opening cross it will do whatever it does.

Does anyone have a more reliable estimate of impact at auction, either theoretically or through experience?

IIRC, there are a few public models around. E.g. there is an old paper by Madhavan with an auction model that you can (most certainly) improve on. It's a very tricky thing to model properly especially now when there is concurrent live trading.

 

[jufils]

WHy do you think you are changing the auction price? WIth your volume, I would say it is not material and you have no other option if that is when you want to enter/exit.

Well there's strictly no such thing as zero price impact if we are talking about expected values, is there? And even 0.1% is a big deal if you are entering and exiting positions on a daily basis.

Participation in the auction depends on many things. Sometimes you have stocks that have well over a billion in ADV and an opening auction with an average participation of under million.

A more relevant number is the percentage of the auction, not the daily volume. Sometimes the ratio of daily volume/auction volume fluctuates wildly.

I understand that the proportion of daily volume that is traded at auction is very stock-dependent, but I was under the impression that price impact models generally require you to input the total volume for the period you are interested in, i.e. if I am interested in price impact per day, I should input the daily volume, even if I expect my order to be filled entirely instantaneously. If that differs when considering auctions then that's useful to know.

IIRC, there are a few public models around. E.g. there is an old paper by Madhavan with an auction model that you can (most certainly) improve on. It's a very tricky thing to model properly especially now when there is concurrent live trading.

Thanks, I'll check that out. I've read a couple on the subject but there seems to be very little consensus.

Right now I use the square root law as a guide. I guess I just want to be sure that there is a conservative measure rather than a reckless one...

 

[jtrader33]

Well there's strictly no such thing as zero price impact if we are talking about expected values, is there? And even 0.1% is a big deal if you are entering and exiting positions on a daily basis.




I understand that the proportion of daily volume that is traded at auction is very stock-dependent, but I was under the impression that price impact models generally require you to input the total volume for the period you are interested in, i.e. if I am interested in price impact per day, I should input the daily volume, even if I expect my order to be filled entirely instantaneously. If that differs when considering auctions then that's useful to know.


Thanks, I'll check that out. I've read a couple on the subject but there seems to be very little consensus.

Right now I use the square root law as a guide. I guess I just want to be sure that there is a conservative measure rather than a reckless one...

https://www.itg.com/assets/ITG-Trading-Around-The-Close.pdf

It’s been years since I’ve read it but from memory I believe the conclusion was that you’ll slip only the bid/ask spread regardless of the size of the imbalance. So for your model, the expectation depends on the percentage of time you’re trading with/against the imbalance and how often your size flips the imbalance. Please feel free to update here if I’m misremembering.