Payment for Order Flow: A Prisoner’s Dilemma

Many thanks to everyone who has read our payment for order flow (PFOF) research release or our blog post last week. We’ve gotten so many comments and questions, I wanted to clarify one additional point. Some commenters pointed out that individual retail brokers have indeed chosen to send orders to exchanges in the past and did not experience the 25% price improvement we suggested retail orders would experience if all retail volume were traded on exchanges. (Some of these discussions were on twitter click here to follow me on twitter.)

Our paper does not suggest that retail orders will experience 25% price improvement on exchanges. The 25% number in our paper is not the "price improvement" but rather the reduction in the NBBO spread itself if all retail flow moved to exchanges due to the substantially lower toxicity associated with that flow. 

This confusion between relative costs and absolute costs is what makes the topic of wholesale market structure extremely confusing and controversial. When shopping for a car, would you rather pay $20,000 for it when everyone else pays $25,000? Or would you rather pay $18,000 for the car when everyone else also pays $18,000? In the first scenario there is "price improvement" of $5,000 because you pay $5,000 less than everyone else. But in the second scenario, while there does not appear to be any price improvement, the actual price for all those buying the car is far less than any party paid in the first scenario.

The first scenario is what we currently have. Retail gets price improvement over everyone else. The second scenario is what we are likely to have if all retail flow moved to a centralized venue where multiple market makers and institutional investors could compete for every individual market order sent by retail traders. Because retail flow is less toxic than the rest, more firms would be competing for that flow on an order-by-order basis lowering the spreads. In addition, the information asymmetry discussed at length in our paper between wholesalers and other market makers would be greatly reduced, increasing competition and lowering spreads even further.

However, it is NOT in the best interest of retail investors, if their broker alone moves their market (or marketable limit) order flow away from wholesalers and onto exchanges. The flow from one retail broker alone will not reduce the NBBO spread enough to compensate for the price improvement currently provided by wholesalers over the current NBBO spread. It is the shift of order flow from all retail brokers that would narrow the spread by (a lot) more than the price improvement they receive today.

In addition, it would be a comparative disadvantage for an individual retail broker who routes orders away from wholesalers to exchanges, because relative to other retail brokers, its customers would receive poorer pricing. This retail broker's order flow would decrease the NBBO spread only slightly on exchanges. And other retail brokers (sending order flow to wholesalers) would still receive better prices than the retail broker that shifted its flow to exchanges.

I believe this is exactly the reason that even retail brokers who do not accept PFOF (for example, Fidelity) still route order flow to wholesalers. They do not want to receive poorer pricing for their clients relative to other brokers. No retail broker can force others to send their flow to exchanges (which would be optimal for all retail investors), so they have no choice but to send their own order flow to wholesalers. A retail broker cannot be blamed for the fact that absolute spreads are worse than they could be. They are simply playing the game as they should, optimizing for just their customers rather than the entire market.

But this is indeed the Prisoner’s Dilemma, a frequently referenced situation in classical game theory. In the game, two prisoners are each given the option to remain silent or to betray one another by telling of their involvement in a crime. The rules of the game are known to both. If both remain silent, each gets only a 1-year sentence. If both betray each other they each receive a 2-year sentence. However, if one betrays the other while the other remains silent, the silent prisoner gets 3 years and the betrayer goes free. Clearly the optimal answer for the prisoners would be to remain silent and receive only a 1-year sentence, but they both end up betraying each other with the fear that the other one will betray, and the outcome would be far worse. 

The “game” (i.e., wholesale market structure) cannot be changed by any individual retail broker and they cannot be blamed for not putting themselves at a comparative disadvantage. Regulators control the rules of wholesale market structure, and if they want a better outcome for all market participants they should update regulations to incentivize/compel retail brokers to send order flow to a centralized venue.

I watched the testimony of new SEC Chairman Gary Gensler during the House Financial Committee's hearing on Thursday. Mr. Gensler, an economist, seemed to appreciate that the issue is not just in the trenches, but rather at the much broader market structure level. While he acknowledged the conflicts that PFOF represents, he stayed focused on "broad market structure review" as a priority. He spoke about concentration as one of the main issues in the wholesale market structure (which has little to do with PFOF itself) and said that "capital formation works better when there is intense competition in the middle of the market". I appreciated that his comments were heavily focused on lowering absolute costs rather than relative costs.