Hidden cost of free trading? US$34B a year, study says

The article is here .
The original paper is here .

As usual, the headline does not tell the full story. There seems to be large variations within paid brokers.

Can someone summarize the actual data? How does IBKR Pro compare to the others? How about Fidelity?

From the article:

“The six brokerage accounts have significantly different levels of price improvement. TD Ameritrade clearly has the most price improvement; 69% of our TD trades are executed at the midpoint price or better. Slightly below are the two curves for E*Trade and Fidelity, which are close to each other. Next comes Robinhood, with price improvement trailing those two brokerages. At the bottom are the two IBKR accounts. IBKR Pro has the lowest level of price improvement with only 16% of trades occurring at the midpoint price or better; IBKR Lite is slightly better, until PI=20%, at which point IBKR Pro has better performance.”

And this: " In contrast, IBKR Pro … translates into a cost of 62% of the spread. This is over 10 times more than TD"

Plus IBKR Pro has a commission.

Thank you Chaim.

I am assuming these are market orders?

Thank you! Looks like the option I use - IBKR pro really sucks. Wonder how the adaptive algos perform?

Hi RT,

I think that is correct:

“Our experiment generated approximately 85,000 trades over the December 21, 2021 to June 9, 2022 period by placing identical simultaneous market orders across various brokers.”



Whether algo’s are helpful probably depends on the algo and there are a good number of them. This paper mainly addresses the issue of “order price improvement” and I see no suggestion that any algos were ever looked at in this paper.

The only possible generalization that I could see is that some limit orders will be “marketable” limit orders in that they are between the BID/ASK spread and a broker can give “order price improvement” in some of those situations.

I guess I have no data on those limit orders but also no rational reason to think a broker that provides little price improvement for a market order goes out of its way to do so with a marketable limit order.

For a while IB and people at P123 were making claims that IB provided better " order price improvement" that other brokers. This should put that claim to rest at P123. IB may have already stopped making that claim. If not, the claim is less prominent on their web site. Any benefit for algos that may exist is a different topic that is not addressed in the paper. Other issues such as market maker/market taker, loaning shares, cost of commissions etc. also was not addressed, I think.


This is a great paper. And Jim’s parsing of it is very helpful indeed. I’ve never seen data like this, and it’s extremely eye-opening. It also confirms my anecdotal experience with Fidelity and IB. A big thank you to both Chaim and Jim.

Indeed, very interesting. I’ve used IB for more than a decade and in recent years I’ve felt that the executions just weren’t that great. Thought I was being paranoid but perhaps not. Also, IB’s high commissions and incessant requests for me to “verify my account information” compound the annoyance.

Regarding IB’s algos, I’ve never used the adaptive algo in particular, but I can generally say that I haven’t found them to be particularly useful. Some seemed no better than simple limit orders (accumulate/distribute), and others frankly seemed terrible (VWAP–the algo would sit idle when spreads were narrow, then trade aggressively once spreads widened). I trade mostly small, low-volume stocks so perhaps they work better on larger companies.

Looks like a change of broker might be in my future. Many thanks to Chaim and Jim for bringing this to my attention.

That’s interesting. Of the brokers I’ve used - I used to feel like IB had the best executions, but I hadn’t monitored it as closely in the past couple years.

Yuval, based on your comment I’m inferring that you’ve experienced degradation in IB execution performance relative to companies like Fidelity/Schwab? thanks,

For me, it was interesting to watch how people wanted to keep believing in Interactive Brokers and kept saying they were good at order price improvement despite a complete lack of evidence (other features like Algos, the ability to loan share etc aside).

For a while there has been absolutely no evidence that interactive brokers was better. for order price improvement and real data could be found on the subject. There have been posts on this topic before (with data). The data was not as definitive as this article but it is also true that there has been a complete lack of any data suggesting that IB was better for order price improvement.

Even IB has been unable to give any data to support their claim and this has been the case for a while. Here is their web site about “SmartRouting.”

They say their statistic are “impressive” in this screenshot. Possibly a term vague enough that it might not be technically false if shown to the SEC.

But then they do not seem to possess any statistics worth presenting on this page. None whatsoever (last line in the screen shot saying how good their statistics are). There are no statistics at all on this page. Link to full page where no actual statistics exist: [url=https://www.interactivebrokers.com/en/trading/smart_routing.php]https://www.interactivebrokers.com/en/trading/smart_routing.php[/url]

TD Ameritrade’s page on the subject for comparison. They have some statistics that they want to share.

Wait so people actually use market orders?

Some criticisms of the study in this thread on elitetrader:


And from “Fain”:[quote]
I’ve tested against 4 Brokerages in Canada and they came ahead each time.

These guys were too broke to do a proper study. Notice the trade sizes are all below a board lot. Could be internalizers get better odd lot fills but Increase the trade size and IB comes out ahead. Didn’t test options, small order sizes, only large cap companies, only US companies.

P.S. Reading that other forum reminded me once again how good the quality of the dialogue on this forum is! This forum is a treasure!

Yes, maybe the big brokers are better at getting fill improvements on large cap highly liquid US stocks. They might be worse at other things. IB has been used for a lot time in the professional day trading community despite being fairly non-intuitive and a clunky UI (imo), and I assume it’s for a reason. It probably comes down to the right tool for the right user on the specific job.

As mentioned, this study was for market orders. I wonder if limit orders would show different results.

Yes, my anecdotal experience is that I tend to get slightly better fills on Fidelity than on IB. But I may be wrong. I just executed VWAP sell orders on Fidelity and IB for the same stock at the same time and got almost exactly the same fill price. IB got a better fill by 2 basis points.

Matt Levine at Bloomberg has an interesting take on this paper:

It’s worth reading the piece in full if you can access it: Some Free Brokers Are Cheaper Than Others - Bloomberg

While I like the original paper’s data, its background regarding order flow and payment for order flow is lacking. Likewise, I think Bloomberg’s paper is overly simplistic.

Here is a page the describes order flow at Interactive Brokers for IBKR-Pro: IBKR-Pro Payment for order flow

You find one for any broker. There are a lot of “arrangements” in these legally required disclosures.

For example, Dark Pools which are mentioned in the link above (and used by other brokers). These can be an important source of order price improvement—for market orders, some limit orders and MOST VWAP orders. Some Dark Pools, for variousr reasons (including the level of participation in the Dark Pool), are better than the others.

IBKR may not have access to Fidelity’s Dark Pool (and vice versa). For sure they have no obligation to use each other’s Dark Pool.

But Dark Pools are just one thing that the quoted portion of the Bloomberg article (and the original article) did not mention as being possible causes for the difference.

You can sort through the disclosure yourself and make up your own mind on how some of the “arrangements” might affect the price you get. However it happens that you end up paying more (or less) with a broker, it is no accident.

And yes, with some brokers you can get really fast market orders if that is important to you. You are more likely to notice how large of an order you can make before the price you pay exceeds the NBBO. Some brokers will do better at that than others and last time I looked you can find some serious data on that too.

Yuval - is VWAP your preferred trading algo on IB? Thanks

It depends. For very large orders, yes. Otherwise I go with Relative NBBO Peg via P123.

Whichever broker actually does better, a VWAP order is a great opportunity for the broker and possibly for a P123 member.

Above, there is a discussion about the study and the smaller order size used in the study. Even a very large VWAP order will be broken up into smaller orders throughout the day. Often into 100 lot orders if I recall (I tend to use percent-of-volume orders now and I have not looked at the individual trade size). So it is not entirely clear that order size is a consideration for determining which broker has a better VWAP order. The study may be pertinent no matter how large your total VWAP order may be.

At IB, there is the potential for a rebate for being a market-maker on some of those trades, I believe. IB may share some of that rebate with you at times. You may find that IB is the better broker for you, in part because of that.

Presumably other brokers can get a rebate for being a market maker when executing market-maker trades through these exchanges. It is interesting that this does not count as payment-for-order flow it seems.

I have taken the time to look at some of the individual orders at Fidelity (a while ago) and there is order-price-improvement of some of the the individual VWAP orders going through at Fidelity. But it is difficult to quantitate how much total improvement one gets by the end of the day.

VWAP orders would present an opportunity for regular members who do not have computers that can execute tens of thousands of simultaneous orders with multiple brokers (as is done in the study above). The VWAP orders do that randomly (but not simultaneously) throughout the day for you. About 30 VWAP trades with 2 different brokers would be more trades than you would need to reach a conclusion.

So you can find your own definitive answer if you are investing enough for this to matter to you. There is no good reason for you to listen to any of my BS (or anyone else’s).

As far as real data that I have generated myself, I compared Folio Investing to Fidelity (with market orders) before I switched to Fidelity—alternating which broker I placed an order with first (at about the same time). Fidelity won by a huge margin. There was basically no order-price-improvement at Folio Investing for market orders when I looked a this.

My main takeaway from my little study: the difference can be huge which I think confirms what the paper said: there can be an order of magnitude (10X) difference according to the paper. Hmmmm……I think I am going to do a little more than say: “broker X has a lot of customers (and kind of a mystique), so I will use them.”

If anyone does the above study with VWAP trades and wants to share it, please include the commissions as well as any rebates.