Implicit Commissions

In the institutional trading world, folks generally think about trading costs in two terms: explicit costs and implicit costs. Explicit costs are the costs an investor directly pays for execution: i.e. the commission they pay to their broker. Implicit costs are all the other trading costs incurred, most notably the cost due to market impact. If I decide to buy a stock when the market is at $10, and I pay my broker $0.01/share in commissions, and my order moves the market a bit and my average price winds up being $10.02, then my total trading cost was $0.03/share or 30 bps: $0.01 explicit, $0.02 implicit. Explicit costs are easy to measure. Implicit costs are notoriously difficult: did my order actually push the stock or was that just market randomness? Was the broker sloppy in its execution leading to a worse price? Questions like these are extremely challenging to answer, although the hope is that the more data you have, the more the noise cancels itself out and the true implicit costs reveal themselves.

Explicit costs have shrunk dramatically over the past 25 years, although they have flattened out over the past 10, whereas implicit costs in US markets do not appear to be shrinking at all.

This decline in institutional commissions has not harmed brokers as much as one might think, as explicit commissions are not the only trading cost extracted by the broker. Brokers often make money on customer order flow in other ways, such as collecting rebates (or PFOF) from trading destinations. Some of these behind-the-scenes benefits to the broker are straightforward to measure, at least from the broker’s perspective, while other benefits may be intangible even to them. Tallying up rebates collected on a specific client’s order flow is easy. But if the broker is aggregating its clients’ data into some sort of alpha signal or data feed, and then re-distributing it to top-tier clients, it is far more difficult to quantify the value generated in good will. We think of all these indirect benefits collected by the broker as implicit commissions. Rarely are implicit commissions visible to the end investor, and this is no accident.

Ways brokers capture implicit commissions

Finance is an infuriatingly opaque industry. Even within a given firm, teams are siloed and largely kept in the dark about the business practices and economics of adjacent desks. As such, it is difficult to enumerate the various levers a broker has at its disposal to try to extract value from a client’s business.

Here is a list that comes to mind based on our experience in institutional US equity trading:

• Collecting rebates (or PFOF) from exchanges and SDPs.
• Capturing spread and/or post-trade reversion by trading as principal against client orders.
• Internalizing client order flow in the broker’s dark pool, which both generates additional commissions and avoids external trading costs. Also, liquidity begets liquidity.
• Cross selling other products (e.g. prime brokerage services like custody, margin, and stock loan; trading in other asset classes or regions, OMS/EMS, CSA management, etc.)
• Advertising client trading volumes to attract other client trading as well as investment banking business.
• Aggregating client data to provide alpha signals and/or historical position movements to other, (top-tier) clients.

It is hard to say how prevalent or material each of these practices is, and I am sure there are gaps in our knowledge, but you have got to start somewhere. Next, we’ll dive deeper into a couple of these examples to better illustrate the mechanisms of value extraction, as well as the cost to the client.

Example 1: Retail PFOF

The retail trading world provides a good illustration of implicit commissions as the dynamics are relatively straightforward and well publicized. Led by Robinhood, many retail brokers have stopped charging explicit commissions altogether. In this case, 100% of the retail broker’s profit comes from implicit commissions, most notably payment for order flow (PFOF). The retail broker sends the client’s order to a wholesaler, like Citadel or Virtu, who executes the trade and pays the broker for the opportunity.

Implicit commissions on a retail trade

The retail broker gets other benefits too, like net interest. Maybe it is not fair to consider net interest a “trading cost” per se, but it is at least an opportunity cost tied to being a Robinhood customer. A diligent retail trader might move their cash balance from Robinhood to an outside interest bearing account, but the average retail trader is paying this opportunity cost to Robinhood, another implicit commission.

Unlike an explicit commission paid to a broker, which is exactly equal to the explicit cost to the client, implicit broker commissions generally lead to outsized implicit costs. In other words, the impact to the client is often greater than the benefit to the broker.

Take the PFOF scenario: Robinhood is accepting a payment from Citadel, but Citadel is no charity; they must be making a profit from the trade as well, for example through spread capture. Citadel also has costs: any flow they are unable to internalize must be offloaded into the market, and they do incur exchange access, infrastructure, and data costs.

The actual implicit cost to the retail investor is Robinhood’s implicit commission + Citadel’s profit + Citadel’s trading costs. If Citadel is forced to offload its retail exhaust to another market maker; you could further break down Citadel’s implicit costs on that trade as the sum of the market maker’s profits and trading costs (for that trade), and you can continue down the chain (with the numbers likely getting smaller at each subsequent step).

Example 2: Central Risk Books

These days, most large institutional broker-dealers operate central risk books — effectively an internal proprietary trading team that interacts with client order flow. This poses an obvious conflict of interest, but most brokers actually try to position this arrangement as a value-add to the client. “Our central risk book allows us to connect client flows across different trading desks and asset classes, allowing us to internalize the risk and avoid paying trading costs on the street.” If you have one client buying a future, and another client selling the underlying, the central risk book can provide liquidity to both of them and just hold on to these non-fungible instruments with offsetting risk, allowing both clients to avoid trading costs in the open market. If the central risk book operates roughly at cost or at a loss, then this logic checks out — internalization without profit extraction seems like it would in fact be of value to the client. But if the central risk book is a profit center for the broker, then it is probably net-harmful or at best net-neutral to clients. Again, the conflict of interest is clear: when a broker provides liquidity to a client, the broker is the one who chooses the price and time to transact at. It stands to reason that the broker will choose the best possible price for themselves and the worst possible price for the client, within the confines of what the regulation allows.

One common practice for central risk books is to provide liquidity for the final portion of a client’s order. For example, see this FAQ recently published by Barclays. Barclays calls this practice “Parent CapComm” (capital commitment):

• “Parent CapComm — facilitates the remaining portion of an algorithm parent order, in a single fill, once the pre-determined facilitation threshold is crossed (typically 70% of the entire order in shares).”

The purported benefit to the client is to avoid additional impact and trading costs with the final piece of their order, but now the broker actually has an incentive to cause market impact with the first 70% of the client order. The greater the initial impact, the better the price will be for the broker on this final principal transaction, and also the greater the reversion will likely be afterwards. All of this spells greater profit for the broker’s central risk book.

Unless the client is closely monitoring and measuring what the broker is doing, one would not expect the broker to give them better prices out of the goodness of their heart. As such, many buyside firms do keep a close watch on any interaction with their brokers’ principal flow, while others opt-out of such interaction altogether.

In addition to potential trading PnL from direct interaction with clients, central risk books also collect data from their agency trading colleagues. The Barclays FAQ specifically describes what data their Systematic Principal Liquidity (SPL) group has access to:

• “On trade date the SPL group will not have any client identification or parent order information.
• On trade date the SPL group will only have visibility into the positions that were filled principally after the client’s order is complete.
• On T+1 the client’s identity and their parent order information will remain anonymous to the SPL group.
• On T+1 only the orders executed by the SPL group will be identifiable by the SPL group as executions that originated from an algorithm or SOR that was placed directly by a client through Barclays electronic trading desk.”

Supposedly, there is a wall between the SPL group and the agency desks that prevents free flow of information, but as long as Barclays sticks to the letter of the above disclosure, they are probably okay. This language is pretty clunky though, and it has some holes. For example, it appears there is some client information that the SPL group cannot see in real-time, but that does become available to them on T+1, at least for the client orders that the SPL group executes against. Perhaps the SPL group uses this T+1 data to do counter-party filtering to categorize or even blacklist certain “toxic” clients (e.g. clients who repeatedly reload and run over the principal desk following a Parent CapComm print). Such counter-party filtering would help ensure that they only interact with clients with whom it is profitable to do so. It stands to reason that central risk books are not only profitable overall; but they are likely profitable on a client-by-client basis.

Additionally, this FAQ language does nothing to preclude the SPL group from collecting real-time aggregated/anonymized parent order data from the agency trading desk. It seems possible, given this language, that the SPL group could consume a data feed of impending client order flow on an aggregate basis, which would allow them to preposition ahead of said flow (e.g. “in order to better provide liquidity,” of course). The SPL could potentially use this information to put on profitable trades, based on probabilistic models of impeding client impact, that wouldn’t necessarily even involve interacting with clients. I don’t know if any broker central risk books are actually doing this, but if so, it would allow these desks to capture greater profits but would clearly be harmful to the broker’s clients.

Reducing Trading Costs

While value in the stock market generally grows over the long-term, in the short-term it is a zero-sum game. In other words, investing has a positive expected return but trading is zero-sum. Any profit systematically collected from the trading process is a trading cost borne by another market participant. And I’m not just talking about value extracted by prop trading firms. Brokers, exchanges, data vendors, EMS/OMS providers, even the DTCC: any profits they earn linked to trading are a diffuse cost borne by other market participants.

Now all of these types of businesses, save for the DTCC, are highly competitive. In theory, natural competitive forces should squeeze these profits down to zero, which would in turn reduce trading costs dramatically as well. Unfortunately this trend does not appear to be happening in a meaningful way. That’s not to say trading costs and profits haven’t shifted in recent years. For example, exchanges have become incredibly proficient at capturing profits previously collected by HFT firms (via rent-seeking technology services).

There are a few major obstacles preventing trading costs from being squeezed down to zero: opacity, conflicts of interest, and effective monopolies with large moats. Another major problem in the institutional trading world is simply how difficult it is to measure execution performance. You’re trading 5% of a company’s float over the course of month? Good luck predicting your impact. Most funds compare their execution brokers’ performance, but as order sizes and time horizons increase, it is nearly impossible to separate meaning from noise.

In our conversations with sophisticated buyside firms, we have seen two approaches for keeping brokers honest and keeping trading costs in check:

1. For the lucky few asset managers who have enormous amounts of relatively standardized order flow, they can just run a randomized wheel and compare broker performance directly. If you have enough data in each bucket of comparable orders, simple slippage vs. arrival at the parent order level can be a clean, meaningful, stable metric over time. These firms have the luxury that they may not even need to dive into the specific practices that each broker employs; the data should speak for itself, and any harmful practices quickly become apparent and get a broker kicked out. Unfortunately though, as parent order sizes grow, and as the number of parent orders shrinks, the noise quickly overwhelms the meaning, even for large asset managers.
2. For everyone else, the common best practice seems to be a combination of careful measurement and deep inquiry. Parent order stats may be noisy, but they are a good sanity check; and mid-to-low level stats, while straightforward and robust, each only provide a window into a narrow aspect of the overall trading experience. Many firms supplement their measurement of execution performance with a deep dive into what exactly each of their brokers is doing behind the scenes. It requires a great deal of vigilance to ensure you’re asking the right questions and getting wholly truthful answers. Some asset managers employ blanket policies to disallow broker practices they perceive to be harmful or conflicted, for example opting out from interaction with principal desks or blacklisting venues they consider toxic. The main goal of such inquiries and policies is to prevent the broker from monetizing the client’s order flow in ways that increase the client’s implicit costs.

Explicit US institutional equities commissions are roughly $9B/year, although about half of this number is indirect payment for broker services like access to banking deals/IPOs, research, and corporate access. It would be very hard to estimate how much institutional brokers earn from implicit commissions — I would imagine most brokers couldn’t tell you even if they wanted to (and they certainly don’t want to). But it is safe to say that the buyside’s implicit costs are likely several times higher than their explicit costs. Is it possible to bring both of these costs down to zero? Once again, we believe trading is a zero sum, so reducing trading costs inherently means transferring money from all of the middle men across this industry back into the hands of end investors. Now, reducing trading costs to literally zero is probably impossible — and these middle men do provide at least some value to justify their profits — but what about cutting trading costs by say 80%? We believe this would be a wonderful shift, and we would love to play a part in making it happen.

What do we suggest?

Potential regulatory interventions

In general, we find regulatory interventions to be helpful but slow to come to market. Also, no matter how well rule language is drafted, banks will always find ways to get cute and obey the letter of the law while skirting its true intentions. That said, we are always a fan of increasing mandated transparency and eliminating or at least exposing conflicts of interest and monopolistic practices.

Increase transparency around:

• Broker routing practices
• Broker economics (e.g. disclose price discrimination across clients)
• Not-held 606 reports
• Broker use of client data

Examine conflicts of interest and monopolistic practices:

• Exchange oversight/control over public market data feeds
• Client flow interacting with principal desks
• Internal dark pool overweighting
• Venue rebates/payments that don’t flow back to the client
• Exchange price gouging around technology offerings such as proprietary market data, data center co-location, and cross-connects

It also seems like it would be a great disincentive to introduce individual executive liability for cases where a regulator finds a particularly egregious and nefarious rule violation.

Things we’re doing voluntarily at Proof

Given that a regulatory panacea is not imminent, we try to run our business as an example of how we think brokers should operate. Our number one goal is best possible execution — in other words, reducing overall trading costs to be as low as possible. We do charge a trading commission to our clients, which of course is a trading cost, but we try to keep our rate modest and fair (e.g. equal across clients). We intend to publish more detail about how we price our algos in the coming months, but the basic premise is that we target the same net-margin with all of our clients. That is, we don’t price discriminate, and we incorporate trade-ex costs (venue fees/rebates) into this calculation to align interests, even for non-cost plus clients.

We also operate with extreme transparency as a mechanism for clients to hold us fully accountable. We try to be open about everything including our research, technology, product design, and even financials.

We have no other business lines besides institutional US equity trading (i.e. nothing to cross-sell), and our only potential implicit commission, exchange rebates, is already factored into our clients’ rates. The only scenario where we trade as principal is when testing our algos, and we never trade against nor alongside our client order flow.

When it comes to client data, we have self-imposed a legally binding Trading Data Privacy Agreement, which grants legal control of trading data we generate to that respective client. Among other things, we are legally prohibited from ever selling client data.

Now in theory, maybe the best possible way to reduce client trading costs is to actually collect certain implicit commissions and then pass them back through to the client or charge lower commissions. This seems like just a convenient broker excuse to be shady, but if we did find convincing evidence that it was the case, I guess we would have to pursue this avenue. In that case, I think we would need to double down on our commitment to transparency; I for one would not trust any broker who made this argument who wasn’t 100% forthcoming about all of their economics.

Closing thoughts

We say it over and over again: we believe harmful practices flourish in the shadows, but sunlight is the best disinfectant. Many industries have been dramatically disrupted for the better by the transparency brought on by the internet age, but finance as an industry has lagged behind. The same embedded companies remain at the top. The specific mechanisms have evolved, but the inherent conflicts of interest persist.

In finance, most of the power is concentrated in a small number of people, almost all of whom benefit greatly from the status quo. And the harm from questionable practices is ultimately highly diffuse and mostly impacts folks outside the industry. This is a bad recipe for change. Collectively, we need to shine a brighter and brighter light on all the dark corners of this industry, dig deeper, ask questions, and better understand how all these interwoven pieces interact.