Why are there rebate tiers? And other bedtime stories …

The first stories we read to children have pretty simple plots. There are good characters and bad characters, a struggle between them, and ultimately the good characters prevail. Good and bad are extremal points on opposite sides of a single axis. It is understandable, then, that we human beings often view real life conflicts through this lens: expecting that there is a clear and unassailable notion of what is “good” with an equally clear opposite of “bad.”

What this typical story structure fails to capture is situations where “good” is not an extremal point, but rather a delicate balance of unavoidable tensions. Markets are like this. There are reasons why we want markets to be kind of centralized: would-be buyers and sellers will find each other more easily when they spend more time in the same places. There are reasons why we want markets to be kind of de-centralized: if there are multiple markets with multiple ways of catering to buying and sellers, then we will get innovation in market design and competition between markets that should help control costs. What this means is that neither extremal point is ideal. A single marketplace may be good for “liquidity” (ease of completing transactions), but is bad for innovation. An extremely fragmented marketplace may be good for innovation, but is bad for liquidity. This is a Goldilocks kind of situation — which is perhaps not too surprising (it would be hard to come up with a character name in nursery rhymes that is more finance-y than “Goldilocks.”)

And so we cannot evaluate the “goodness” or “badness” of market design changes in isolation. We cannot meaningfully say things like: “this would be good for liquidity, so it is good for markets.” We have to evaluate proposed changes in context: what is the current state of fragmentation and competition between marketplaces now, and would the proposed change move us closer to a better balance of competing goals, or a worse one?

Some of the people who have the very boring, adult job of answering these kind of questions work for regulators like the SEC, as an important role of regulation in the US equities market is to define the tools that venues like stock exchanges are allowed to use in order to compete with one other. We should want these tools to be effective enough to enable healthy competition, but not so powerful that they blast apart the marketplace into a gory, overly fragmented mess. We must also be cognizant of the degree to which these tools are more or less effective in the hands of well-established vs. up-and-coming entrants. Essentially, regulators should be like fairy godmothers. They should give you what you need as a new competitor to make it in the door for the main event, but the rest is on you.

Into this necessary contextual landscape, let us introduce a more controversial character: the rebate. In the US equities market, rebates are a tool that exchanges use to compete with each other. If you ask market participants whether rebates are “good” or “bad,” you are likely to get a wide variety of answers. For example, you can read IEX’s opinion here, NYSE’s opinion here, an academic commentary here, some recent press coverage here and here, and older press coverage here. If you are really into this stuff (no judgement!), you can even read all of the comment letters to the SEC concerning the transaction fee pilot here. On one end of the spectrum, you will hear excoriations of rebates as a bastion of corruption. On the other end, you will hear impassioned defenses of rebates for being as American as apple pie.

To unpack what rebates are, what effects they have, and where these contrasting views come from, we must consider the market structure version of the cold start problem: starting a new exchange and gaining traction is hard. If you can’t attract a lot of traders to your exchange from day one, the small number of traders you do attract will not find willing counterparties, and hence will have a bad experience and go away. To attract business on day one and then to grow/maintain it going forward, exchanges compete on their methods of matching buyers and sellers (their rules, their technology, etc.) as well as on the prices they charge for executing trades and other services, like providing market data feeds. The term “rebate” refers to the practice of charging a negative commission on trades to certain traders in certain circumstances. In these cases, an exchange pays a trader for the privilege of executing the trade, rather than the trader paying the exchange. This can serve as a powerful tool for attracting traders to a new venue.

There is a reason this feels unsettling to some people. We know well to be suspicious of things given away for free (poison apples, cursed spinning wheels, usb sticks at cybersecurity conferences, etc.). “What’s in it for the exchange?” we wisely ask. If exchanges are paying traders to trade on their venues, rather than the other way around, how do exchanges make money? Part of the answer is that the rebate given to one trader is counter-balanced by a fee charged to another trader. Every trade involves two parties, so there are various criteria an exchange can use to charge a fee to one of the parties and give a rebate to the other. The fee is typically larger than the rebate, allowing the exchange to still collect a net fee on each matched trade. Another part of the answer is that exchanges can subsidize their trade matching services with revenue they collect from other services, such as their market data products. This kind of subsidization makes it difficult to see from the outside what sorts of costs and margins exchanges have for the various services they provide.

Subsidization is one aspect of what feels sketchy about rebates, but it isn’t really fair to blame rebates for this phenomenon. Exchanges could subsidize their core trade matching service with their other services just as easily without the device of a rebate. In the extreme, they could charge nothing for matching trades and only charge for ancillary services. But in this case, what’s true in fairy tales is also true in trading: no matter where it temporarily hides, the total bill does always come due. Either when the clock strikes midnight, the three wishes are exhausted, or the monthly billing cycle resets.

Another thing that makes rebates feel sketchy is the conflict of interest that agency brokers face in routing trades to venues on behalf of clients. When a client asks a broker to buy stock, for example, the client wants to broker to find the lowest possible price, but they also want the broker to do so quickly. Balancing these competing goals already means it is not obvious how the broker should behave in each circumstance. A broker may be tempted to trade on a exchange that will pay the broker a rebate, rather than on the exchange that is likely to achieve the lowest price for the customer. This is a classic example of what economists call an “agency problem.” Human and corporate entities tasked with acting “in the best interest” of someone else face a deep and natural temptation to shortchange their clients when their personal interests diverge from their duty. This is why the most selfless fantasy sidekicks are always non-human: even children might be credulous that Tinker Bell would drink the milk to save Peter Pan if she was a human.

This conflict of interest is a real problem, but once again it isn’t really fair to entirely blame rebates. This conflict arises to some extent if exchanges have different pricing in any ways at all. If one exchange charged the broker 15 mils per share to trade, and another charged 20 mils per share, the broker could be tempted to trade at the cheaper venue at the expense of the client. Naturally, to the extent that rebates exacerbate the differences between the highest and lowest costs a broker may incur, they exacerbate the conflict of interest. But if you wanted to get rid of this conflict entirely, removing rebates would be insufficient. You would need to instead mandate that all exchanges have identical pricing in all situations, or you would need to mandate that all transaction costs are passed through from brokers to clients.

But of course, the phrases “pricing” and “transaction costs” here are still hiding some problematic dimensions of complexity due to subsidization. Should the fees that exchanges charge for data services and related products also be forced to be identical or passed on? If not, is the problem really solved?

It’s not hard to imagine a fable beginning with this premise: “…and so the SEC decreed throughout the land that all exchanges would have to charge identical prices for all services.” And in the picture book version there would be illustrations of a sad gaggle of traders, all dressed in identical gray suits with identical gray ties, stifled in an atmosphere devoid of meaningful choice and innovation. And out of this grim setting there would emerge a plucky hero, who would cleverly skirt the rules somehow and demonstrate the joys of greater freedom. And the SEC would have a change of heart, lift the decree of identical pricing, and a grateful population would revel in the fruits of healthy competition once again.

But of course, this story has a sequel. At some point, when the dust has settled, and things have been peaceful and boring again for too long, a shrewd observer at one of the exchanges realizes: “since prices can vary across exchanges, who’s to say that prices cannot vary across customers? It would be even easier to compete with other exchanges if we start setting different prices for different customers.”

It might start innocently enough, with prices that vary based on a customer’s role in the trade at the moment. The most common example of this is “maker-taker” pricing, where the trader who acted first (essentially proposing the trade) is called the “maker” and given a rebate, while the trader who acted second (essentially agreeing to the proposed trade) is called the “taker” and charged a fee. A natural variant of this is “taker-maker” pricing, which gives a rebate to the “taker” and charges a fee to the “maker.” What’s important here is not that one is a rebate and one is a fee — in a qualitative sense, it would be analogous to charge both a fee, but make the fee for one party less than the fee for the other. What’s important is that the pricing for a trade is a function of the trader’s behavior in that particular trading interaction.

A standard rationale put forward for maker-taker pricing is that it compensates the party to the trade who was willing to put themselves out there first, thereby incentivizing published prices. A standard rationale put forward for taker-maker pricing is that a “maker” can increase their probability of trading by posting in a venue that sweetens the deal for a “taker.” Both rationales fit nicely with common capitalist storylines: bravery as a behavior to be rewarded, paying to cut the line as an inalienable right.

Both rationales are also a little weak. (Aside: shouldn’t belief that the trade price is profitable be motivating both the maker and the taker, rather than these lower-order incentives?) What’s more compelling is what these different pricing structures can do as tools of competition between exchanges: they can segment the potential customer pool. Though it is by no means a perfectly clean division, there are some traders that more commonly act as a makers, and some traders that more commonly act as a takers. By varying their pricing models on the maker-taker axis, exchanges can compete for more targeted segments of the trading flow. After all, pricing structures are decided by exchanges, not by traders. So it’s not too surprising that rationales for pricing structures that are formulated in terms of what effects they have on traders are missing a crucial driver in pricing design: the effects that pricing structures have on competition between exchanges.

But why stop at segmenting traders only in this limited way? Exchanges have lots of information about customer behavior beyond the confines of a single interaction. What if the rebate given to a customer (or the fee charged) on a particular trade could be a function not just of the customer’s role in that particular trade, but rather a function of the customer’s behavior pattern over time? Just as we see these days in the evolution of digital advertising, the ability to target and customize pricing based on patterns of customer actions over time is a powerful tool for influencing behavior. And it is a tool that is most powerful in the hands of dominant incumbents, who have the richest data and the most leverage.

But one thing stands in the way of incumbent exchanges behaving like jealous kings who lavishly reward their loyal subjects and devise punishments for dissidents: the fair access rule. Unlike ATS’s whose market share is below a a certain threshold, exchanges are legally required to provide “fair access” to their markets, meaning that they cannot say things like “You can’t trade here because I don’t like you.” More formally, the Securities Exchange Act of 1934 specifies that exchanges cannot capriciously deny membership to brokers that want to trade, and must have rules that “are not designed to permit unfair discrimination between customers, issuers, brokers, or dealers…” Bummer. Charging highly individualized prices sure sounds a lot like “unfair discrimination.” Perhaps all the venues who have have been tempted to go down this path have stopped at this and said “aw shucks, what can you do,” and moved on.

But we all know that’s not how these stories go. Instead, we have the introduction of the rebate pricing tier. The main difference between pricing tiers and simple maker-taker style pricing policies are that tiers look at trading behavior over an extended period of time (typically one month) to determine a price, rather than looking only at the behavior inherent in a single trade. At first perhaps, this feels like a small shift in the air. A few traders begin walking around with a new spring in their step, buoyed by the knowledge they are getting larger rebates than others. The tiers at an individual exchange begin to divide traders into finer and finer groups, based on how and how much they trade at the exchange on a monthly basis. Traders are rewarded for loyalty: trading a higher percentage of their volume on a particular exchange allows them to climb the tiers and obtain a higher rebate (or a lower fee). Exchanges try to capture customers by offering them increasing rewards to send more and more order flow to the exchange. Implicitly, this means exchanges are also punishing customers who partially divert their flow elsewhere with worse pricing for the remaining flow.

Over time, the definitions of the tiers grow increasingly complex. The variables of behavior that they refer to expand in number and deepen in specificity. Traders are lined up and measured against yard sticks like “Consolidated Average Daily Volume,” “Average Daily Added Volume,” “NBBO time,” “NBBO size time,” and interrogated about their personal histories. How many displayed orders did you execute in tape A stocks in the month of April, 2015? How much liquidity did you take, relative to the total consolidated volume and relative to your average daily volume, in the month of September, 2019? How many orders did you submit in the opening auctions of tape C stocks while reading Reddit while pretending to read the Wall Street Journal? (That last one is made up, but the real tier definitions are not much simpler. For example, see here, here, and here.)

Some traders begin to feel suspicious of the invasive definitions that increasingly divide them. Some question the need for 3,762 different variables in the determination of pricing, and 1,023 different possible paths for price calculations. Some begin to suspect that exchanges are employing a classic colonial strategy: if you want to control a large and unruly population that you cannot subdue by force, simply anoint some of them as “superior” to the others, and the anointed ones will join with you in suppressing the others and crush all rebellions, as they want to preserve their own coveted status.

Pricing tiers at the current level of complexity basically allow a exchange to present a different rebate or price (or at least a different pricing rationale) to each trading firm. At this point, we might as well let each exchange keep a spreadsheet that has a row for each trading firm and columns for the rebates/fees that will be assessed in each trading scenario. The exchange could fill out this spreadsheet however they wanted to — there would be no constraints preventing them from being arbitrarily and capriciously inconsistent, other than the overall cap that each individual fee must be no greater than 30 mils.

This effect is a rather unsurprising consequence of allowing exchanges a rich language to describe rebates and fees. There ultimately are not that many brokers (at least, not that many who maintain a significant market share), and so a language that contains many differentiating variables can easily be warped into a reasonable-sounding sentence that effectively isolates a single broker. Allowing exchanges to combine an arbitrary number of such sentences into a pricing “schedule” completes the effect of potentially isolating every broker.

Exchanges often argue that their pricing tiers are non-discriminatory and do not violate fair access because theoretically anyone could qualify for a particular pricing tier. But this is like arguing that the “princess and the pea” test isn’t intended to discriminate against commoners — after all, couldn’t anyone somehow magically develop the sensitivity to detect the presence of a pea? Obviously a test that is chosen from a rich, though abstractly formulated set of possibilities is not a perfect defense against unintended qualifiers, but it is pretty clear that the game is tilted, and by a bit more than the effect of a pea. (Sorry, I couldn’t resist.)

The extent to which all of this disturbs the natural behavior of brokers is also a function of the maximum difference in the rebates/fees that are possible. If the maximum per-trade fee charged to any broker by any exchange under any circumstances is 30 mils (which it is by regulation), and the maximum rebate is 45 mils (which is not currently capped by regulation, but this rebate at NYSE American is the highest I happened to find in a cursory glance at the pricing documents), then the range of -45 to +30 mils (a total range of 75 mils) controls how significant rebates/fees can be to the overall economics of a broker-dealer. Comparing this range to the average spreads in various categories of stocks, for instance, can give us a sense of how significant rebates/fees should be for profitability for market makers. It is notable in particular that the range of 75 mils is less than 1 cent, which is the typical minimum spread for equities. Hence claims that regulation banning rebates will result in dramatically wider average spreads are a bit silly: at worse, an average spread of X cents could be expected to grow to an average spread of about X + 0.75 cents (and likely the real growth would be considerably less, as we are considering the extremal endpoints of the rebate/fee ranges, not the average cases). Not to mention that exchanges could simply lower the market data fees and other sources of revenue they collect from market participants in proportion to the rebates they no longer pay out, and this would enable market makers to recoup some of that lost revenue without further widening spreads. To be fair, this wouldn’t be a perfect balance: the lower market data fees would be less targeted to market makers than the rebates were. But nonetheless, there are many reasons to suspect that the ultimate impact on spreads would be relatively mild. Chicken Little’s predictions that “liquidity is falling” and “the spreads will grow to consume us all!” are unlikely to come to pass.

With all of this in mind, we can see why capping or banning rebates, or reducing the maximal allowed fee would result in a reduction of the influence that pricing may exert over broker behavior. If a regulatory reform results in a compressed range of variance between extremal rebates and fees, than the economic relevance of this pricing relative to other variables, such as spreads, is reduced, and hence conflicts of interest, etc. are better controlled.

But this sort of reform does not address the ability of exchanges to design bespoke pricing per customer. Regardless of whether this was originally the intent of the structures that have been put in place to modify fee and rebate pricing over time, it is clearly the reality of where we are now. If we wanted reform to limit exchange ability to tailor pricing down to the granularity of individual trading firms, we would need to think about how to limit the language of variables that exchanges can use to define pricing, as well as the number and type of functions of those variables that can be combined into the full pricing policy. We need a regulatory regime that effectively considers this holistically, rather than an evaluation process that looks at a single incremental proposed changed and attempts to gauge its reasonableness in isolation.

We could have an open debate about whether bespoke pricing capability is healthy as a tool for competition between exchanges in our current market state. Proof certainly has a stake in such an argument, as small brokers are likely subject to the worst pricing since they don’t have enough flow to be juicy target customers for exchanges. But what I find most frustrating about much of the current debate over rebates and pricing issues is not the conclusions that are being reached, but that we are not even having the debate on the right terms. I support the proposed transaction fee pilot because I think it would reduce the extent of the conflict of interests and help get the markets closer to a healthy state of competition between exchanges. But as discussed above, this is because it would reduce the total range of pricing difference, not because rebates are inherently bad. The histrionics on both sides of the current debate are unjustified: zero is not some magic point on the pricing line that divides good from evil. Nor is tinkering with a few small variables to dynamically rebalance the mechanisms of competition between exchanges likely to lead to any drastic consequences. But what I’m really hoping you’ll take away from this tale is not a particular policy position, but rather this:

Sadly the real world is much more boring, and ultimately more exhausting, than the world of bedtime stories. Maintaining healthy markets is less like fighting dragons and more like the painstaking task of keeping the arrow of a compass pointed toward true north. We must give ourselves the freedom to constantly correct ourselves in small steps towards a better balance. And we must do this by being honest and clear-sighted about where we are now, and where we want to be. We must let go of the absolutist terms of childhood, and find a way instead to agree upon more nuanced terms of debate.

Ugh. Now that I write that it sounds so tedious. And probably I’ve successfully put you to sleep.