How markets actually work

Market makers

M5 — The industry & the firms

M0.1 promised that when we reached Jane Street and Citadel Securities, their business would already be familiar. It is. A market maker is M1.7's equation, industrialised:

   MAKER'S EDGE PER UNIT TRADED

     +  spread captured          M0.4
     +  rebates                  M1.7
     −  adverse selection        M1.2
     −  inventory cost           M1.3
     −  fees & technology        M1.6
   ────────────────────────────────────────
     ×  an enormous number of units

There is nothing else to the business. The whole competitive game is making each term slightly better than the next firm, on billions of transactions.

The defining structural facts

They trade their own capital. No outside investors, no management fee, no redemptions. This is the single most consequential fact about how these firms behave — it means no client reporting, no marketing, no explaining a bad quarter to anyone, and therefore almost no public disclosure. It also means the partners keep the profits, which is why compensation at the top is extraordinary (M5.5).

They are flat overnight, mostly. Inventory is a cost, not a position (M1.3). The business is capturing spread, not having views.

They are technology firms. The edge is latency (M1.6), models, and infrastructure. Headcounts are small relative to profits, heavily engineering-weighted, and hiring is famously competitive with big tech.

They are long volatility. This is the one to internalise, and unlike everything else in this module it can be checked against audited accounts.

The one open window: Virtu

Almost every major market maker is private. Virtu Financial is publicly listed, so it files audited accounts — the only place you can see a market maker's P&L directly. From its SEC filings:

   year   revenue    consolidated    net      average
                     net income     margin      VIX
   ─────────────────────────────────────────────────────
   2015    $0.80bn      $0.20bn      24.8%      16.7
   2016    $0.70bn      $0.16bn      22.5%      15.8
   2017    $1.03bn      $0.02bn       1.8%      11.1
   2018    $1.88bn      $0.62bn      33.0%      16.6
   2019    $1.52bn     −$0.10bn      −6.8%      15.4
   2020    $3.24bn      $1.12bn      34.6%      29.3
   2021    $2.81bn      $0.83bn      29.4%      19.7
   2022    $2.36bn      $0.47bn      19.8%      25.6
   2023    $2.29bn      $0.26bn      11.5%      16.8
   2024    $2.88bn      $0.53bn      18.6%      15.6
   2025    $3.63bn      $0.91bn      25.1%      18.9

2019 → 2020 is the headline. Revenue more than doubled (+113%) as average VIX went from 15.4 to 29.3. In the year markets fell apart, the market maker had its best year to that point.

Across all thirteen available years, revenue correlates +0.63 with average annual VIX. Positive and strong, but not a straight line — and the reasons it isn't are instructive.

Why volatility pays a maker. Spreads widen when volatility rises (M0.4), and volumes rise too, so the maker earns more per trade and does more trades. The risk is higher, which is exactly what the wider spread compensates. This is why a market maker's revenue looks like a long option position — and why it is the natural counterparty to everyone in M4.6 who is short one.

Three honest caveats, because the table above is not as clean as it looks:

  • Acquisitions contaminate it. Virtu bought KCG in 2017 and ITG in 2019, so revenue growth is partly M&A, and the terrible 2017 and 2019 margins are substantially integration costs rather than trading performance. The 2017 row — lowest VIX in the sample and rising revenue — is the acquisition, not a counterexample to the volatility story.
  • Virtu is not a pure market maker. ITG brought a large agency-execution business, which earns commissions rather than spread and behaves differently.
  • It is the listed one for a reason. Firms go public when it suits their owners; Virtu's economics need not represent Jane Street's or Citadel Securities'.

Even with all that, the direction is unambiguous and it is the only audited evidence available. Market making is a long-volatility business.

The margins are also worth noting for what they deflate. Twenty to thirty-five percent net margin in good years is an excellent business — and a loss-making year in 2019. The popular image of infallible money-printing does not survive contact with the filings.

Who's who

Approximate and reported, not audited — read these as "known for", which is the roadmap's actual ask:

   firm                      known for
   ───────────────────────────────────────────────────────────────────
   Citadel Securities        the largest US retail wholesaler — the
                             counterparty on a very large share of US
                             retail orders (M0.3's PFOF). Separate
                             company from Citadel the hedge fund, same
                             founder.

   Jane Street               ETFs above all, plus an unusual public
                             identity — writes OCaml, publishes a tech
                             blog, and discloses financials only because
                             it issues bonds.

   Optiver / IMC /           options and ETF market making, Amsterdam
   Flow Traders              heritage. Optiver and IMC are major
                             exchange-floor descendants.

   Susquehanna (SIG)         options; famous for teaching new traders
                             poker as decision-making under uncertainty.

   Jump Trading /            latency-driven HFT (M1.6). Jump is
   Hudson River Trading /    notably secretive; HRT is more openly
   Tower Research / DRW      research-and-engineering flavoured.

   Virtu Financial           the public one. Read its 10-K.
   ───────────────────────────────────────────────────────────────────

The names shift; the business model does not.

Why the criticism lands where it does

Market makers are the most publicly criticised firms in the industry, and M1.6 and M1.7 already gave you the tools to sort the real arguments from the noise.

Weak criticisms: that they "front-run" retail (they mostly fill retail inside the quoted spread — M1.7's segmentation), or that HFT has widened spreads (spreads have collapsed since the 1990s).

Serious criticisms: that the resources burned on the latency arms race are socially wasted rents rather than competition on price (M1.6, Budish–Cramton–Shim); that PFOF and segmentation leave the lit book more toxic and therefore wider for everyone else (M1.7); and that the market's liquidity now depends on firms with no obligation to quote in a crisis — makers withdraw precisely when μ spikes (M1.2), which is when liquidity is most wanted.

Being able to make those distinctions is roughly what the whole course has been for.

Source: Virtu Financial’s 10-K filings, via SEC EDGAR — the XBRL company-concept API serves the numbers above as JSON, and reproducing them takes about ten lines. Jane Street’s tech blog is the roadmap’s pick for firm culture. For the criticisms, revisit M1.6 rather than the popular press.