why PFOF is problematic (in options)

Intro

The topic of payment for order flow (PFOF) has been discussed to death, so let’s do it again. PFOF entails a broker not sending an order to directly to an exchange, but to a market maker who pays them for the privilege. As long as the market maker executes this order within the NBBO, “internalizing” an order like this is fine and legal. Proponents of this practice (coincidentally often executives of trading firms who participate in this process) cite the better execution these retail orders get versus the NBBO as price improvements that they get for retail. They see the alternative as the retail trader sending a market order to an exchange that executes on NBBO.

This is a bad assumption for 2 reasons. Firstly, you don’t have to send a market order. As a very small trader you can put an order a tick above mid on exchange and there will be a good chance you will get filled. Excluding maker/taker fee structures, market makers prefer to take a small order like that versus having it hit a passive quote (as they could be getting picked off on their quote), which is why they will be “sniping” orders like this at tighter levels than their quotes. Secondly, trading firms engaging in PFOF do so because it’s profitable to trade against these small, mostly uncorrelated orders. As these profitable orders move away from exchanges, exchange spreads will widen as market makers on exchange will have to cope with a more toxic subset of orders compared to a world where all orders go to an exchange. The result is that NBBO in a PFOF world is artificially wide.

So at best the benefits of PFOF are overstated, at worst the existing structures make it impossible for anyone to determine whether retail is actually better off.

Equities versus options

Now the reality is, the above is an interesting philosophical argument, but from a practical standpoint, I don’t really care about it in equities. Spreads on stocks / ETF’s are incredibly tight nowadays, and it has never been cheaper for a random retail investor to transact in stocks. The cost/benefit analysis above is in the order of basis points or fractions of basis points. It will certainly not materially affect investing outcomes for someone whether he does his monthly VTI buy in a PFOF or no-PFOF world.

However, this is not the case for options, at all. In dollar terms, option bid-ask spreads are very wide (say ~10% for OTM single stock calls). This causes the arguments above to be much more practically significant than for equities. The way in which PFOF distorts option markets might actually have serious financial consequences beyond the intra-industry P&L distribution of trading firms. In 2021, total PFOF payments for options totaled $2.4 Billion versus just $1.3 Billion for equities¹, despite the latter obviously representing orders of magnitude more dollar volume.

Before anyone mentions this, I am aware that PFOF option orders are technically crossed on exchange, but the end result is still roughly the same. I’m not going to go through the nuances of that whole process here.

Market Maker perspective

Say I am a market maker on options on a random stock XYZ which is currently trading at $100. I have an accurate estimate for the current implied volatility (say 50%) and therefore know that the fair value of a 105 strike, 1-month call is $3.80. I am willing to sell above this number and buy below it, but by how much?

Market making options is a very tricky business. It involves quoting thousands of options on any single underlying and all of those are exposed to a variety of different risks all at the same time. The stock might move and you can get picked off on the delta component the option, the dividend might get cancelled and you can get stuffed on some bad forward trades, some other market maker might come in and lift you on the vega component because he is hedging a huge trade on a similar stock, thereby pushing up the implied volatility. I can go on and on about the plethora of risks an options market maker faces, but the end result is that passive quotes generally have to be pretty wide (on a dollar basis) to compensate for this. And I haven’t even tackled the fact that option pricing and volatility fitting is a relatively expensive operation (from a compute perspective) which causes your latency on all of these decisions to be much, much worse than for e.g. futures market making.

Let’s say our passive quote is 3.60 at 4.00 for the 105 call. Now what if the stock isn’t moving and a 3.90 bid for 1 lot pops up? I would be fairly comfortable aggressing on this bid and selling at 3.90. Firstly, because I know this isn’t someone trying to pick me off (if you want to pick someone off on e.g. delta you generally have to be aggressing as it’s a speed game). Secondly, I might reasonably assume that there is no larger order behind this that will move the implied volatility. If it was truly small retail, and there are more orders like this, uncorrelated in their direction, then I would be more than happy to collect the 10 cent edge this one offers. The buyer of the call cut his slippage compared to NBBO in half, but still paid 10 cents of spread, a sizable percentage.

There’s quite some assumptions in the paragraph above. My willingness to trade as a market maker is in large part dependent on the fact that this order is indeed a small retail order. As a trading firm engaging in PFOF you know that these incoming orders are small retail, which is why you can give them the “price improvement”. But the fact is, it is not even clear that this price improvement only happens because of PFOF. My example above is completely on exchange, I am not paying anyone to do this trade, I might even be able to provide a tighter effective spread because there is no PFOF cost to my firm. The pro-PFOF argument here is that via PFOF, you can directly discriminate retail orders and therefore give them a better price compared to large institutional orders that move the market against you (instead of building out logic determining which flow is which in your trading systems). This is directionally true but it is essentially impossible to estimate the magnitude.

2021 was an illustrative year in that regard, it’s not unrealistic to assume that in a lot of stocks, the majority of option trading was done by retail traders. If all the incoming orders are from retail, knowing it is retail by paying their broker for the order adds no value from a trading perspective and you are simply paying for it so that you can get a first look. This is great for trading firms engaging in PFOF because they don’t have to compete on exchange for these orders, where other firms are competing for the same trades, all with amazing technology to be able to do this as fast and efficient as possible. However, this is obviously bad for the retail investor. If there is no informational value in PFOF, then the $2.4 Billion dollars is simply the amount PFOF firms pay brokers to avoid trades going directly to exchanges, even though they would have gotten the same / a better fill on exchange. And these are serious amounts, trading is a large business, but not a huge one. $2.4 Billion dollars for options alone represents a significant part of the profit pool of the industry. Obviously this is an extreme example, but it might not be that far from the truth in 2021.

The above paragraphs are a long winded way of mostly explaining the first point in the intro. Even in a PFOF world, the effective level you can get filled at is much tighter than NBBO. Go ahead, put in a 1 lot AAPL call buy a couple cents above mid, I guarantee you will get filled almost immediately. Furthermore it is unclear whether the fact that PFOF enables better discrimination of these orders outweighs the extra cost, especially in inherently retail driven markets.

My second point is a bit more straightforward. Think about our example market of 3.60 at 4.00, with some small orders coming in here and there. What if every single harmless small order disappeared into the inventory of PFOF firms? The only orders left would be larger ones that are more likely to move the market against me. My 3.60 at 4.00 has some reasoning embedded that a harmless order might hit these quotes and I would simply earn the full 20 cent spread. If instead every single order that hits them is so large that it might move the fair value of the option against me, I need to widen out the bid-ask spread to make sure that I am compensated enough for this, to say 3.50 at 4.10. Now the perverse thing here is that this will increase the so called “price improvement” number that PFOF trading firms tout. They siphon off the harmless orders, which makes the average on exchange trade more toxic, which means the exchange spreads need to be wider. Then they turn around and compare against these widened exchange spreads to calculate “price improvement”.

concluding

All and all it’s quite clear that the benefits of PFOF are very overstated and the reality is we have no real way of knowing whether it is a net good or bad thing for retail traders as a whole. As a free market guy, if I can’t see clear benefits to a system where some large market players pay some other parties to specifically arrange trades away from exchanges, I’m not a fan of it. Furthermore, PFOF has allowed retail brokers to not charge commissions anymore. Making the commission implicit instead of explicit makes retail think they are trading for free, which incentivizes them to overtrade, something that we have obviously seen happen the last couple years. And I’m pretty sure the excessive options trading hasn’t exactly lead to the best financial and social outcomes for the average Robinhood user.