drew

As a *very general* rule of thumb, a MM should believe that price will trend down if lots of people cross the spread to sell, and that price will trend up if lots of people cross the spread to buy

In an effort to avoid this phenomenon (called adverse selection), MMs try to keep inventory as low as possible and stay hedged so as to avoid directional market risk. But it can’t be avoided completely

Ideally you want to have counterparties least likely to take advantage of stale quotes, i.e. the “least toxic” counterparties. This means that there is economic value inherent to non-toxic flow.

If you’re a retail trader, this means you! What makes retail nontoxic is that it tends to be substantially more *random* in the aggregate, which means it is less likely to attack a MMs stale quotes and lose them money.

So if MMs value nontoxic flow, payment for order flow represents *the monetization of that value*. What MMs are paying for is a counterparty that is statistically less likely to adversely select against them on a trade-by-trade basis.

In the highly competitive market of US equities MM, purchasing nontoxic retail flow is a loss reduction technique. So why is that good for a retail trader? The answer here is price improvement.

Essentially, part of the PFOF payment is set aside to give the retail trader a better price than the NBBO.

In the idealized form of this arrangement, the MM, broker, and retail investor all “win”. PFOF results in a tighter spread paid by retail traders for the service of instant liquidity, the brokers get a new revenue stream, and MMs get better, less toxic flow.

So why all the anguish over PFOF? Let's review arguments against such as they are.

the first argument against PFOF cuts directly against the argument for it: the argument is that price improvement figures are overestimated and that retail traders are not getting better prices.

The NBBO putatively represents the best bid and offer price for any given stock, but it turns to not necessarily be the case. Odd lot orders—orders for fewer than 100 shares— placed inside the NBBO are exempt from the NBBO calculation.

In addition, there is dark pool liquidity which is often priced at the NBBO midpoint and also not factored in.

The second argument against PFOF is that having retail on exchange would tighten spreads. One important thing to note about PI statistics is that they assume that the NBBO spread is static, but it isn’t. In fact it’s quite likely that the spread would change with retail flow.

Since lit market market makers have to interface with statistically more toxic flow, they have to quote wider spreads to compensate for the “cost” implicit in trading with that flow. So the argument is that making lit flow less toxic would cause MMs to quote higher spreads.

The third and final argument against PFOF is that there is a data/information advantage that is conferred to MMs because they get to see retail flow first.

For example, if a MM sees retail flow directionally tilted towards selling, they will quote their bid on-exchange lower than the NBB.

This has an effect on competition in the market making industry: other MMs without this data face disproportionately higher adverse selection when their orders execute at an NBBO price which is “crumbling” or otherwise quickly changing.

And it’s worth mentioning here that this factors into why there have been so few new retail equity MMs in the last ~20 years: MMs have to work to set up these relationships with retail brokerages and create commercial arrangements to buy flow.

The combination of the information advantage conferred by PFOF plus the duopolistic concentration in equities wholesaling means that the two largest firms in effect have strong knowledge of the direction of the NBBO, pricing others out of the business.

However, we should be clear that the information advantage does not equate to MMs “frontrunning” their clients, it merely means they are pricing differently than their competitors outside the NBBO and getting picked off at lower rates.