Anatomy of a High Frequency Market Maker: Menkveld Opens the Books

For years, the debate about high frequency trading was conducted almost entirely in the dark. HFT firms are private, secretive, and disclose nothing. Everyone had opinions about what they do. Almost nobody had data.

Menkveld got the data. Not aggregate statistics, but the actual trading records of one large high frequency trader, identified in the data of a European exchange, trade by trade, day by day. And then he did the most valuable thing an empiricist can do with such a thing: he simply described, carefully and honestly, what it did and how it made its money.

The picture that emerged is not the villain of the newspapers, and it is not the hero of the industry's press releases. It is something more interesting: a market maker, doing exactly what market making theory says a market maker does, at a thousand times the speed.

The problem: what does an HFT actually do all day?

The public arguments were mostly about motive. Are they front-running? Are they manipulating? Are they providing genuine liquidity or "phantom liquidity" that vanishes on contact?

These are answerable questions if you have the trading record. You can see whether the firm posts passive quotes or aggressively takes liquidity. You can see whether it ends the day flat. You can compute where its profit actually comes from. Menkveld could do all of this.

The key idea via analogy: the market stall that never closes and never carries stock

Here is what the HFT in Menkveld's data looked like.

It traded constantly, on the order of well over a thousand times per stock per day. It was overwhelmingly passive: it posted quotes on both sides and waited to be hit, rather than crossing the spread. It kept its inventory tiny and mean-reverting, ending the day close to flat, and rarely holding a meaningful position for long.

Think of a market stall that buys and sells the same item hundreds of times an hour, always trying to hold nothing at the end of each minute, earning a sliver on each round trip.

Now, the profit decomposition, which is the heart of the paper and the part every aspiring market maker should stare at.

The HFT's profit split into two components that pull in opposite directions:

It made money on the spread. Every time it bought at its bid and sold at its ask, it earned the difference. This is the core market making revenue, and it was substantially positive.

It lost money on its positions. Whenever it was holding inventory, the price tended to move against it. Menkveld calls this the positioning component, and it was consistently negative.

That second finding is the profound one, and it is exactly what the theory predicted. Why does the price move against you when you are holding inventory? Because of how you got the inventory. You are long because people were selling to you. And people were selling to you, in part, because the price was about to fall. That is adverse selection, live and quantified, in the books of a real firm. The market maker cannot escape it. It can only make sure the spread it earns exceeds the losses it eats.

So the business model, in one sentence: earn a small, reliable margin on the spread, bleed a smaller amount to adverse selection, keep the difference, and do it enough times that the difference adds up. That is Glosten and Milgrom in an operating company.

There is a second finding, and it is arguably the more consequential one for market structure. Menkveld showed that the success of a new entrant exchange, Chi-X, critically depended on this HFT. The new venue needed liquidity to attract traders, but traders would not come without liquidity, the classic chicken-and-egg problem that kills most new exchanges. The HFT solved it by quoting aggressively on the new venue and simultaneously hedging on the incumbent, effectively knitting the two markets together. The HFT was the bridge that made fragmentation viable.

Why it mattered

• It replaced speculation with evidence. This is one of a small number of papers where someone actually looked inside an HFT's book. The finding that the firm was a passive, inventory-averse, spread-earning market maker, not an aggressive predator, cooled a great deal of overheated rhetoric.
• It validated classical microstructure in the machine age. The profit decomposition is a textbook confirmation that the old models were right about the fundamental economics. Spread revenue positive, adverse selection cost negative, inventory managed toward flat. Ho and Stoll would have recognized this firm instantly. Speed changed the scale and the tempo, not the logic.
• It explained how fragmentation actually happens. The finding that a new exchange needed an HFT to bootstrap its liquidity is a genuinely important piece of market structure economics. It explains why exchanges court HFTs with rebates and colocation, and it complicates the simple story that HFTs are pure extractors: in this episode, the HFT was the reason a competitor to the incumbent exchange could exist at all, which is the sort of thing that lowers costs for everyone.
• It reframed the liquidity question. If the HFT's positioning P&L is negative, then the HFT is, on net, absorbing adverse selection risk from the market. That is a service. Whether it is fairly priced is a separate question, but it is not nothing.

The honest limitations

• It is one firm, one market, one period. A single HFT on a European exchange over a limited window. Other HFTs run entirely different strategies: some are aggressive latency arbitrageurs, some are statistical arbitrageurs, some are momentum-igniting. Menkveld's firm was a market maker. You cannot generalize from this to "HFTs are market makers." The paper describes one animal in a diverse zoo, and it is careful to say so.
• Identification of the firm relies on data quirks. The ability to isolate one trader's activity depended on features of the specific dataset, and there is always some uncertainty in attributing trades to a single account.
• It says nothing about stress. The paper describes ordinary trading days. The great fear about HFT market makers is not what they do when things are calm, it is that they withdraw instantly when things are not. This dataset cannot address that.
• The welfare question is left open. The HFT earned a profit. Somebody paid it. Whether that payment was a fair price for a valuable service, or a toll extracted through a speed advantage that ordinary investors ultimately fund, is not settled by showing that the firm behaves like a market maker. Market makers can be both useful and overcompensated.
• The volumes and profits are specific to that era's fee structures. Rebates, tick sizes and fragmentation have all changed, and the profitability of pure passive market making has been competed down substantially since.

The one-line takeaway

Menkveld looked inside a real high frequency trading firm and found an old-fashioned market maker running at machine speed: it earned a steady margin on the bid-ask spread, consistently lost money on the positions it was forced to hold, kept the difference, and in the process became the liquidity bridge that let a new exchange survive against the incumbent.