The Market Maker's Capital Problem

Primary dealers held about 24 billion dollars of net corporate securities in the weeks before 4 March 2020. By 11 March the figure was 15.6 billion, a contraction of 35 per cent into the worst credit-market stress in a decade. Six weeks later, after the Federal Reserve announced its corporate credit facilities, it averaged 33.7 billion. Inventory moved against customer need on the way in and recovered only once the terms of holding it had changed.

Figure 1 tracks that series from 2013 and then zooms into the first half of 2020. The chronology is suggestive and it is not an identified causal effect, which Section 03 takes seriously rather than in passing.

Section 01Who provides liquidity now

Non-bank electronic firms supply a material share of liquidity in equities, futures, options, foreign exchange and parts of the Treasury market. Their model differs from traditional relationship dealing: inventory is turned over rapidly, risk is managed by speed and hedging rather than by warehousing, and capital is deployed against a book measured in minutes rather than weeks.

That difference matters for what capital is for. A firm that holds inventory overnight needs capital to absorb the loss while it does. A firm that flattens by the close needs capital to survive the interval between taking a position and hedging it. Both are real; they are not the same requirement, and a rule calibrated for one will misprice the other.

When volatility and adverse selection rise, any rational liquidity provider may widen, reduce size, hedge harder, or step away. Capital affects how much loss and inventory a firm can absorb before it does. It does not determine that the firm will choose to.

Section 02The capital argument

That distinction became an explicit UK policy question in December 2025, when the Financial Conduct Authority opened an engagement process on market-risk capital for investment firms, noting that rules inherited from a banking framework may not fit firms whose risk profile is a rapidly turned trading book rather than a loan portfolio.

The prudential counterargument does not ask for every market maker to be capitalised like a bank. It observes that lower private failure costs do not imply zero market-function costs. If capital requirements fall, competition may pass the saving into tighter spreads in normal conditions while leaving less loss-absorbing capacity in the conditions that matter. Those two effects show up in different states of the world, and only one of them is visible in a normal-times execution-quality statistic.

Public data cannot resolve this. A firm's regulatory capital position, its internal risk limits, and the relationship between the two are not observable from outside. What is observable is the quantity those constraints operate on, which is why Figure 1 measures inventory rather than capital.

measured: primary dealer net outright positions in corporate securities
          weekly, as-of Wednesday, net of dealer short positions
          series PDPOSCS-TOT, Federal Reserve Bank of New York

    2019 mean                       18.7 bn
    mean, 1 Feb to 4 Mar 2020       24.1 bn
    trough, 11 Mar 2020             15.6 bn      -35% into the stress
    mean, 15 Apr to 30 Jun 2020     33.7 bn     +116% from the trough
    2026 mean                        8.9 bn      48% of the 2019 mean

NOT measured anywhere in this article:
    regulatory capital, internal risk limits, credit exposure taken
    through derivatives, positions held elsewhere in the same group

Section 03What capital buys in a crisis

Figure 2 separates normal-time quote quality from stress behaviour because the two need not share the same ordering across firms. A provider can quote tightly because of superior technology, internal flow, or hedging access rather than because of a large capital base, and the same firm can withdraw quickly when its hedges become expensive.

The 2020 chronology is the closest thing to a natural experiment the public record offers, and it is not close enough. Inventory fell as spreads widened and rose after the facilities were announced, which is consistent with the argument that intermediaries reduce principal risk exactly when customers most want to transfer it. It is also consistent with dealers responding to price, to hedging cost, or to customer flow, and no counterfactual exists. The most that can be said is that the sequence is what the argument predicts, which is weaker than the argument being demonstrated.

The longer series carries a fact that is not about any crisis. Dealer net corporate inventory averaged 18.7 billion dollars in 2019 and 8.9 billion in 2026, roughly half. Whether that reflects capital rules, balance-sheet cost, a shift of risk to other intermediaries, or a change in how corporate credit trades, this series cannot say. It does say that the buffer available for warehousing credit risk is smaller than it was, in a market that has grown.

A trading firm can be systemically important without being a bank.

Volume I closes on the distinction between private buffers and public liquidity. Capital regulation is designed around the resilience and orderly failure of firms. Market liquidity is an emergent outcome of many firms choosing, in the same conditions, whether to keep quoting. A rule that makes each firm safer while making all of them more likely to withdraw at once would satisfy the first objective and fail the second, and nothing in the public record would show the trade-off until it was tested.

This research is analysis and commentary for general information. It is not investment advice, an offer, or a solicitation, and it contains no price forecasts. Quantities in the text are the author's own calculations from the official source cited; regulatory positions summarise the publications named.