How Payment for Order Flow Can Change Spreads for Different Traders
Summary
The document explains a market making mechanism through a distinction between informed traders, whose orders may predict price moves, and noise traders, whose orders are treated as less informed. Without payment for order flow (PFOF), exchange market makers face both types and set spreads based on the combined adverse selection risk. If noise trader orders are routed directly to firms paying for that flow, exchange market makers face a more informed mix and may widen their quotes.
The firms receiving the routed orders may offer price improvement because they see less informed flow. Whether noise traders benefit depends on whether that improvement exceeds the spread widening on the exchange. The explanation frames this as an empirical comparison, suggesting a randomized trial or a natural experiment across markets as ways to assess the effect. Its informed-versus-noise distinction is simplified, and the document supplies no market data or measured outcome, so it does not establish that PFOF improves or worsens execution in practice.
Key ideas
- Order flow with greater adverse selection risk can lead market makers to quote wider spreads.
- Routing less informed orders away from an exchange may change the mix of flow remaining there.
- Firms receiving routed orders can use price improvement to attract those orders.
- Noise traders benefit only if price improvement exceeds the spread widening caused by the changed flow mix.
- The effect must be measured empirically, and the informed-versus-noise classification is an approximation.
Tags
Full text
# PFOF how it works # PFOF how it works I am wondering how PFOF (Payment For Order Flow) works and why people say that hedge funds make wider spread using PFOF. If a client sends an order A then the broker is going to redirect that order to a hedge fund B. Then B is not allowed to offer a worse price than what's available in the market. Hence if on the exchanges the tighter spread is $s$ then B can makes at most $s$, but B needs to pay a percentage of $s$ to the broker. Hence I don't see how PFOF widen the spreads, and what are typical algos employed by hedge funds to execute these orders. ## Answer by Chris Taylor (score 3, accepted) https://quant.stackexchange.com/a/82026 Consider a scenario with two kinds of trader: - Informed traders, who are price sensitive and trade when they expect the price to move in their favour - Noise traders, who do not have any edge and trade randomly. You can map these onto "institutions" and "retail" for the purposes of this answer, which isn't perfect (some institutions have no edge and some retail traders do have an edge) but is directionally accurate. What would happen if there was no PFOF and all orders were routed to the exchange? Market makers would face a mix of informed flow and noisy flow, and their spreads would reflect the mix of flow that they face. More informed flow would lead them to quote wider spreads, and more noisy flow would lead them to quote narrower spreads. Say that the spread on the exchange in this scenario is $S$, and all traders (both informed traders and noise traders) pay the same spread. Now introduce PFOF, which means that noise trader orders can be routed to market making firms directly, and informed orders still go to the exchange. Then a market maker on the exchange faces a greater fraction of informed flow, so they will need to quote a wider spread, $S'>S$. The market makers who are paying for order flow only see noisy orders, so they will be able to offer price improvement, and quote a spread $S'' = S' - I$ where $I$ is the amount of price improvement. Under which scenario do noise traders pay less? In the first scenario they pay $S$, and in the second scenario they pay $S'-I$, so they do better in the second scenario if $$ S' - I < S $$ which can be written as $$ I > S' - S $$ In order for noise traders to get tighter spreads under PFOF, the amount of price improvement needs to be greater than the amount by which spreads needed to widen as a consequence of routing a greater fraction of informed flow to the exchange. Whether this is true or not is an empirical question that can only be answered with data (either a randomised trial, which is unlikely to ever happen, or more likely a natural experiment where e.g. you compare spreads in a market which has PFOF to one which does not have PFOF).
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.