Market making and bond trading: spread, inventory, axes, RFQ, cross trades and matrix pricing

Advanced11 min read · 24 sources

Market making means standing ready to buy and sell at any time and earning the bid-ask spread for doing so. In the process, the market maker risks being stuck with inventory or selling to better-informed traders. In bonds, which trade almost exclusively over the counter, this shapes the whole market: dealers show axes, investors request quotes via RFQ, funds trade with each other via cross trades and value rarely traded bonds using matrix pricing. Here you will find the mechanics, the key rules and two worked examples.

How market makers earn money: spread, inventory and adverse selection

A market maker quotes bid and ask prices and earns the spread as long as purchases and sales balance out. Research breaks the spread down into three parts. First, processing costs. Second, inventory risk: Ho and Stoll (1981) show how a risk-averse dealer who does not know when orders will arrive or how their inventory will develop sets prices around their own indifference price. Third, adverse selection: according to Glosten and Milgrom (1985), the market maker loses to better-informed traders and recoups this through the spread charged to uninformed ones, even without costs or risk aversion.

An example: if the value is equally likely to be 99 or 101 and 20% of orders come from informed traders, the market maker sets the ask at 100.20 and the bid at 99.80. The spread of 0.40 follows from information alone; the more informed traders there are, the wider it becomes.

Inventory and optimal quotes according to Avellaneda and Stoikov

Avellaneda and Stoikov (2008) build inventory risk into a tractable model. A market maker with inventory q shifts their indifference price against their position: if you are long, you want to sell and lower both quotes. The spread around this price depends on risk and on how quickly orders become rarer the further the quotes are from the mid.

Dealer or broker: risk trade, agency and central risk book

As a dealer (principal), a bank buys or sells for its own account and takes on the risk; the client gets a firm price, a risk trade. As a broker (agency), it executes the order on the client's behalf, the market risk stays with the client, and a commission is paid (Harris 2002). Some firms pool the risks from client business across different desks in a central risk book: offsetting flows are netted internally, and only the remainder is hedged in the market. This can bring tighter quotes, but requires clear rules on how information from client orders is handled.

Bonds over the counter: RFQ, all-to-all and portfolio trading

Bonds are mostly traded over the counter, with search costs, dealer networks and limited transparency (Bessembinder et al. 2020). Traditionally, the investor calls dealers. Electronically, they send a request for quote to several dealers at once and take the best price; on all-to-all platforms, other investors can respond too. O'Hara and Zhou (2021) find for US corporate bonds that electronic RFQ trading is still small and fragmented, but changes costs and execution quality in voice and interdealer trading as well.

In portfolio trading, an investor trades an entire basket of bonds as a single risk unit with a market maker. Meli and Todorova (2026) show that this is closely linked to bond ETFs: ETFs help the market maker to price, hedge and unwind the risk of the basket, much as in an ETF's primary market (see order types and execution).

Bid and ask axes: what dealers show

An axe is a specific interest in buying or selling a security (ICMA and FIX Trading Community 2025). For dealers, there is usually a position behind it that they want to reduce or cover, or a client order they are working. A dealer with too many bonds in inventory is axed to sell and shows an ask axe (offer axe); one who is short or has a buy order is axed to buy and shows a bid axe. Axes are valuable for investors because they allow trading inside the bid-ask spread and often in larger size than usual.

• Axe under the ICMA standard: one side (bid or ask), size as a mandatory field, price optional, firm interest until withdrawn.
• Run or price list: usually two-sided prices and sizes, indicative or tradable depending on the dealer.
• Tradable price: binding for size and time, for example as a response to an RFQ.

Quality is a perennial issue: buyers complain about unreliable axes, some dealers show axes on both sides, and because selection algorithms reward flagging, axes are now barely distinguishable from ordinary price lists. The most common reason for not distributing axes electronically is concern about information leakage (ICMA and FIX Trading Community 2025). Hendershott et al. (2026) show that non-binding quotes have an effect too: indicative dealer quotes lower trading costs in corporate bonds, dealers use them to manage their inventory, and clients use other dealers' quotes to negotiate better prices.

Liquidity and transparency: TRACE and MiFIR

How expensive bonds are to trade depends heavily on size and transparency. For US corporate bonds from 2003 to 2005, Edwards et al. (2007) find that costs fall significantly with order size, are lower for highly rated, recently issued and soon-to-mature bonds, and decline once the TRACE reporting system publishes prices. Bao et al. (2011) show that the illiquidity of corporate bonds is greater than bid-ask spreads can explain, and that it explains a large part of the variation in the yield spreads of highly rated bonds; according to Dick-Nielsen et al. (2012), its share of the spread rose sharply with the subprime crisis.

In the US, dealers report corporate bond trades to TRACE within 15 minutes at most. A reduction to one minute for electronic trades, approved in 2024, has not been implemented by the self-regulatory organization FINRA, which proposed in 2025 to stay at 15 minutes (SEC, TRACE 2025). In the EU, the MiFIR regulation also requires post-trade transparency for bonds, with possible deferrals of publication; a new deferral regime has applied since March 2, 2026 (ESMA, bond transparency 2026), and in July 2025 ESMA selected the first consolidated tape provider for bonds (ESMA, Consolidated Tape 2025).

Cross trades: transactions between funds of the same manager

In a cross trade, one fund sells a security directly to another fund or mandate of the same manager. Both save the spread and market impact, especially for rarely traded bonds. The catch is the conflict of interest: the manager is on both sides, and any price favors one of the funds. The rules therefore require a fair price that can be independently verified.

• USA: Rule 17a-7 allows cross trades by registered funds only for cash, at the independent current market price (for securities not listed on an exchange or Nasdaq, the average of the highest independent bid and the lowest independent offer), without commission, in line with investment policy, under procedures adopted by the board with quarterly review and with six years of record retention (SEC Rule 17a-7 2005). Because the rule requires readily available market quotations and valuation Rule 2a-5 understands these to mean only quoted prices in active markets, most bonds have effectively been excluded since September 8, 2022. A proposal by the SEC dated October 9, 2026 would permit securities valued using observable inputs (Level 2), but not Level 3; the price would be the next NAV or a price from independent pricing sources such as pricing services (SEC, Cross Trading 2026).

Research shows that caution is justified. Eisele et al. (2020) analyze four million equity trades: fund companies set the prices of internal cross trades strategically to shift returns between sister funds, and many cross trades are backdated. Gaspar et al. (2006) had already found that fund families favor funds with high fees or good past performance, among other things through opposing trades between the funds.

Matrix pricing and evaluated pricing

Many bonds do not trade at all on a given day. Funds then value them using matrix pricing, i.e. through their relationship to quoted benchmark bonds rather than through prices of their own (IFRS 13 2011). A typical approach is a matrix of rating, sector and maturity: you read off the credit spread of similar bonds, interpolate across maturity, add the yield of the government bond with the same maturity and discount the payments (see bond mathematics).

In practice, pricing services supply such evaluated prices. The SEC requires US funds to oversee pricing services that provide evaluated prices, matrix prices and similar estimates, and cites as a conflict that a service might value more generously in order to retain clients (SEC, Rule 2a-5 2021). An evaluated price is not a price at which you can trade: it says roughly what the bond should be worth, not what a dealer will pay today. Cici et al. (2011) show that different funds value the same corporate bond differently at month-end, all the more so the more illiquid the bond and the more turbulent the market, with patterns consistent with smoothed returns. Under stress, the market price of a bond ETF can therefore be well below its NAV, as in March 2020 (Aramonte and Avalos 2020).

Advanced: what links the models

• The spread as a sum: processing costs, inventory risk (grows with γ σ²) and adverse selection (grows with the share of informed traders). The parts can only be separated empirically with additional assumptions.
• Information via size or direction: in Kyle (1985), information acts through the size of the order flow (lambda), in Glosten and Milgrom (1985) through the direction of individual orders; in both models, transaction prices become the best estimates of value.
• Inventory and axes: a dealer who reduces positions slowly and aligns quotes and axes with their inventory is following the logic of Avellaneda and Stoikov (2008).
• Matrix pricing as curve estimation: instead of two points, entire spread curves can be estimated for each rating and sector, for example with the Nelson-Siegel model. This reduces noise, but also smooths out genuine differences between individual issuers.
• Smoothing in metrics: smoothed valuations lower measured volatility and increase autocorrelation, as with private credit.

What research shows

The theory of the spread rests on inventory risk (Ho and Stoll 1981; Avellaneda and Stoikov 2008) and information (Glosten and Milgrom 1985; Kyle 1985). For bonds, Bessembinder et al. (2020) summarize that search costs, dealer networks and low transparency shape costs, and that electronic trading is advancing only slowly. Greater transparency lowered costs after the introduction of TRACE (Edwards et al. 2007); illiquidity explains a large part of the variation in yield spreads (Bao et al. 2011) and grows in crises (Dick-Nielsen et al. 2012). More recent work shows the effect of indicative quotes (Hendershott et al. 2026) and the role of ETFs in portfolio trading (Meli and Todorova 2026). For cross trades and valuations, the evidence is a warning: strategic prices (Eisele et al. 2020) and inconsistent valuations of the same bond (Cici et al. 2011).

Limitations and common mistakes

• Confusing an axe with an offer: an axe shows interest, not a guaranteed price; according to complaints from the buy side, some axes are not genuine.
• Taking an indication as a price: prices on screens and in runs are often non-binding for bonds; only the response to a request is tradable.
• Treating an evaluated price as liquidity: a smooth NAV does not tell you at what price bonds can be sold under stress.
• Assuming a cross trade is free: without an independent price, it can favor one fund at the expense of the other.
• Using the matrix unchecked: rating and maturity do not capture everything. Subordination, call rights, covenants and issuer news shift the fair spread.

How Portfolio Lab calculates

Portfolio Lab does not trade bonds over the counter and does not value individual bonds using a matrix. Bonds appear via funds and ETFs; the Backtest uses their prices, which for ETFs are exchange prices, not NAVs. You enter the bid-ask spread as a percentage of the price: purchases are made half a spread above the closing price, sales half a spread below it. For bond ETFs with thin markets or for periods of stress, a higher spread is worthwhile. Price, yield, duration and spreads of individual bonds are worked through in bond mathematics.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.