1The Economics of a Trading Firm

In 2020, a year of high volatility, a listed electronic market maker’s revenue rose from $1 517 million to $3 239 million, 2.13 times the year before. Its pay rose by 2.6% and its communication and data bill by 2.1%. The same cost base that had left it with a pre-tax loss of $116 million in 2019 left it with a pre-tax profit of $1 383 million in 2020. A listed quantitative asset manager, paid on the assets it manages and on the performance it delivers, lived a different cycle over the same years: its revenue rose when its trend-following funds did well and fell when they did not, whatever the volatility index did, and it paid a large share of each marginal dollar out in bonuses. This chapter reads both firms, and a bank’s markets division, through one standard set of lines, and asks of each the same two questions: which costs are fixed, and how far can revenue fall before profit is gone.

1.1 The income statement of a trading firm

Every trading firm, whatever it calls its lines, spends its revenue in the same four places: on costs that grow with the business it does (exchange and clearing fees, payments for order flow, the financing of the positions it holds), on pay, on the fixed machinery of the firm (technology, market data, premises, administration), and on what is left for its owners and the tax authority.

Net trading revenue is the market maker’s revenue capture (One Quant Book 11, chapter 1) summed over a year’s volume, and the right top line for comparing firms: two firms that earn the same after fees are the same business, whatever their gross revenue. A market maker that holds large inventories financed by securities lending reports both the interest it earns and the interest it pays; netting them is what makes its revenue comparable with an asset manager’s fees.

The mapping is a table from each filer’s reported lines to the standard ones (Listing 1.1). Three filers of three business models run through the chapter: an electronic market maker’s annual reports on Form 10-K for 2019–2025, a listed quantitative asset manager’s annual results for 2020–2025, and a bank’s markets segment for 2023–2025. Their figures, in $ millions as filed, are committed as small derived tables with the chapter’s code.

MAP_VIRTU = {
    "net_revenue": [("trading_income", 1), ("commissions_tech", 1), ("interest_div_income", 1),
                    ("brokerage_fees", -1), ("interest_div_expense", -1)],
    "volume_cost": [("brokerage_fees", 1), ("interest_div_expense", 1)],
    "comp": [("comp", 1)],
    "other_fixed": [("comm_data", 1), ("ops_admin", 1), ("dep_amort", 1)],
}
MAP_MAN = {
    "net_revenue": [("net_revenue", 1), ("asset_servicing", -1)],
    "volume_cost": [("asset_servicing", 1)],
    "comp": [("comp", 1)],
    "other_fixed": [("other_costs", 1), ("other_employment", 1)],
}
MAP_GS = {  # the segment does not split out volume-driven fees: all its revenue counts as net
    "net_revenue": [("net_revenues", 1), ("provision", -1)],
    "comp": [("comp", 1)],
    "other_fixed": [("other_opex", 1)],
}

The firm itself reports the same $2 145.3 million under another name, adjusted net trading income, and calls it a non-GAAP measure. Most listed trading firms publish such measures, because the accounting standards were not written for their businesses; the reader’s task is to find the reconciliation and check that the adjustments go both ways.

1.2 Costs that scale and costs that do not

The three firms spend their revenue in very different proportions (Figure 1.2). The market maker gives four dollars in ten to exchanges, clearing houses, brokers and lenders before it pays anyone; the asset manager’s volume-driven costs, its asset servicing, are five in a hundred, but it pays almost half its revenue to its staff; the bank segment reports no volume-driven line of its own, because its fees sit inside other operating expenses, and those are larger than its pay.

The split between costs that scale and costs that do not decides what a bad year does to a firm. Write net trading revenue as $\mathrm{Rev}$ (net of volume-driven costs), the fixed part of the costs, fixed pay included, as $C^{\mathrm{fix}}$, and let variable pay take a share $b$ of each dollar of net revenue. Operating profit is

Operating leverage is Book 11’s proposition for a market maker’s volume (chapter 1 there) written for the whole firm: there the fixed costs faced volume, here they face net revenue after every volume-driven cost. The one new ingredient is pay. A firm that pays a fixed share of every marginal dollar in bonuses shares its bad years with its staff; a firm whose pay is mostly salary does not.

1.3 Pay: the largest line, and how much of it is variable

A compensation ratio is not a cost structure. The asset manager’s ratio was 40% in 2021 and 2022 and 50% in 2023 (Man Group’s own figures, pay over core net revenue before asset-servicing costs); its stated policy is a ratio “between 40% and 50% of core net revenue, depending on the mix and level of revenue”. The ratio rose in 2023 because revenue fell by more than pay did: a ratio moves against revenue whenever part of pay is fixed. The slope of pay against revenue across years says how much of pay is in fact variable (Figure 1.3).

The market maker’s pay was between $376 million and $394 million in every year from 2019 to 2023, while its net trading revenue ranged from $975 million to $2 271 million; the slope is 0.048 with a standard error of 0.045, indistinguishable from zero. Its 2025 pay, $528 million, sits far above the line, the first sign that it had begun to change. The asset manager’s slope, 0.29 with a standard error of 0.094, says that about three dimes of each marginal dollar of its revenue went to pay. The bank segment’s three years give a slope of 0.20; three points make a line but not an estimate. The asset manager also reports the split directly for 2021: fixed compensation $208 million, variable $388 million.

1.4 Capital, financing and return on equity

The market maker earned net income, non-controlling interests included, of $912.3 million in 2025 on average total equity of $1 730 million: an ROE of 52.7%. In 2023, the quiet year, the same calculation gives 17.3%. The bank segment reports a return on average common equity of 16.4% for 2025, on attributed equity of $79.7 billion. The two numbers do not measure the same thing. A market maker’s equity is what its owners have left in the firm; it is bound by the capital rules for brokers and investment firms (chapter 2), and the balance sheet it finances with its brokers’ and lenders’ money is far larger than the equity. A bank’s segment equity is an allocation, set by whichever of the group’s capital constraints binds (chapter 5), and the group holds more of it than its markets business could ever lose in a year, by rule.

1.5 The cycle: volatility, volume and revenue

A market maker’s net revenue is volume times capture (One Quant Book 11), and both rise with volatility. Regressing the market maker’s net trading revenue on the year’s mean of the VIX index gives a slope of $52 million per index point (standard error 33) and a correlation of 0.58 over seven years. The asset manager’s net revenue has a correlation of $-0.27$ with the same index: its revenue follows its assets and its funds’ performance, not market volatility. Over the seven years the market maker’s pay and fixed costs moved within a narrow band while revenue swung by a factor of 2.3 (Figure 1.4).

Applying Method 1.7 at each firm’s latest year gives its break-even fall and operating leverage, with pay flexing at the estimated slope and with pay held at its latest level:

The market maker, the firm with the most rigid pay, is the most robust of the three at its 2025 revenue, because its margin is widest: a high margin is a large distance to break-even, however it was earned. The asset manager’s flexible pay is what keeps it alive in a bad year: held fixed, its pay would put break-even at a 32% fall instead of 46%. In 2019 the market maker’s margin was 21.8%, and its break-even fall at that year’s revenue, with the same pay slope, was 22.9%, against 59.0% in 2025. A firm’s robustness is a property of its cost base and of where it stands in the cycle.

1.6 Tutorial: three firms through one set of lines

What to change next. Drop one year at a time and watch each slope move (exercise 7); add a fourth filer from its annual report, writing only a new mapping table.

1.7 Build: the firm’s economics as code

1.8 Exercises

1.9 Problem: Three Firms, One Bad Year

1.10 Interview questions