6.10.2 OTC Market Structure

Basic Prerequisites: 01 · From Zero.

---

1. Intuition & Practical Objective

Having established why OTC markets exist, this page maps how they are organised - the institutional anatomy a practitioner must know before modelling anything. The objective: understand the four layers of an OTC market and where each risk lives.

1. The customer layer. Asset managers, pension funds, insurers, corporates, and hedge funds trade with dealers. Quotes are requested (often by RFQ - request for quote, frequently to several dealers at once), sized, and negotiated. Trades are customized: a specific bond CUSIP, a specific swap tenor and notional schedule.
2. The dealer layer. Dealer banks (the “sell side”) stand ready to buy and sell, holding inventory on their balance sheet and earning the spread. A dealer’s price is firm for a size and for a moment - there is no continuous book.
3. The interdealer layer. Dealers do not hold positions forever: they offload in an interdealer market - historically voice brokers, now increasingly electronic (e.g. the swap-execution-facility regime for interest-rate swaps). This is the frictionless “outside option” $M$ in the model, and it is what lets a dealer quote without bearing unlimited risk.
4. The clearing layer. Since the post-2008 reforms, most standardized OTC derivatives are centrally cleared through a central counterparty (CCP), which sits between the two sides - but much of the cash bond market (e.g. corporate bonds) remains bilaterally cleared.

The one-sentence essence. “An OTC market is a dealer-intermediated search market with an interdealer wholesale layer: customers negotiate with dealers, dealers recycle risk among themselves, and (post-2008) a central counterparty stands in the middle of standardized derivatives - each layer existing to solve a different part of the search-and-risk problem.”

---

2. Mathematical Ground Truth & Derivations

The dealer’s two-sided quote and the interdealer price. Formally, a customer meeting a dealer bargains over a price with the dealer’s outside option being the interdealer price $M$ (the price at which the position can be recycled frictionlessly). With the customer’s gain $L$ from selling and $H$ from buying, and the dealer holding bargaining power $z$:

$$B=zL+(1-z)M,A=zH+(1-z)M.$$

The dealer’s per-trade profit is the difference between the price he charges the customer and the interdealer price he can recycle at:

$${\pi }_A=A-M=z(H-M),{\pi }_B=M-B=z(M-L).$$

In the benchmark equilibrium (Duffie–Gârleanu–Pedersen, Condition 1: the market is on the buyer’s side, ${\mu }_{lo}<{\mu }_{hn}$), the interdealer price equals the buyer’s reservation value, $M=A=H$, so the ask is exactly the buyer’s reservation and the spread collapses to

$$A-B=z(H-L).$$

Where does the “interdealer market” enter the price? Through the interdealer price $M$ and, in the general model, through the term $\rho (1-z)$ in the denominator $D$ - the accessibility of the dealer to the customer. This is the structural result: a customer’s price is better (spread tighter) when the dealer has more competition - either because the customer can find another dealer (higher $\rho$ with $z<1$) or because the dealer’s own recycling is cheap (higher $M$, i.e. a liquid interdealer market).

Trade reporting and transparency. A core structural variable is whether trades are reported publicly. Let ${\tau }_{\text{rep}}\in [0,1]$ measure the fraction of trades visible to all dealers. More transparency → each dealer can infer the market’s true state faster → lower informational rents → tighter spreads, but slower execution for large/illiquid trades (the price-dissemination trade-off). This is why the post-2008 regime (Dodd–Frank in the US, EMIR/MiFID II in the EU) mandates trade reporting for swaps while preserving anonymity of the counterparties.

Central clearing as a structural transformation. Bilateral exposures $\{W_{ij}\}$ are replaced by exposures to a CCP. Each dealer’s net position to the CCP is

$$n_i=\sum _j{W}_{ij}-\sum _j{W}_{ji},$$

and multilateral netting reduces total system exposure from the gross ${\sum }_{i,j}{W}_{ij}$ to the netted $\frac{1}{2}{\sum }_i|n_i|$. In a dense network the reduction is dramatic (see §3) - this is the quantitative case for central clearing, developed in 06 · Advanced Extensions.

---

3. Computational Implementation - the netting benefit of a CCP

We build a random bilateral OTC exposure network of $N=50$ dealers and compute the gross exposure versus the multilaterally netted exposure after a CCP steps in. Ran and verified.

The netting benefit grows with network density: in a dense interdealer market the CCP nets almost all offsetting positions away, cutting system exposure by $2.4\times$ to $15.5\times$. This is the structural reason clearing is attractive (and, symmetrically, why the CCP becomes a concentrated single point of failure - the risk-trade-off in 05 · Failure Modes).

---

4. Failure Modes & First-Principles Breakdowns

1. The “one price” illusion. OTC markets have no single price. Marking a portfolio at a “mid” that nobody traded at is a modelling convenience, not a market fact - and in stress it becomes a large valuation error (bridged to Counterparty Risk & XVA).
2. Interdealer liquidity is the hidden support. A dealer quotes a tight spread only because a liquid interdealer market lets it recycle risk. If that market freezes (2008), the dealer’s outside option $M$ deteriorates and customer spreads blow out - the propagation channel from a wholesale shock to retail prices.
3. RFQ gaming. When customers request quotes from several dealers at once, dealers learn that they are in competition and may shade quotes; the “search” is now an equilibrium design problem, not a pure friction.
4. Clearing concentrates risk. Netting reduces bilateral exposure but transfers it to a single CCP; a CCP failure is a systemic event precisely because it is the netting hub. Central clearing trades distributed contagion for concentrated exposure - not free.
5. Fragmentation. Multiple venues, RFQ protocols, and reporting regimes can fragment liquidity, so the same instrument trades at different prices on different systems - a structural source of price dispersion.

---

5. References

• Duffie, Gârleanu & Pedersen (2005), Over-the-counter markets, Econometrica 73(6)
• Duffie (2012), Dark Markets, Princeton UP
• Duffie (2010), Asset price dynamics with slow-moving capital, JF 65(4)
• Hasbrouck (2007), Empirical Market Microstructure
• Foucault, Pagano & Röell (2013), Market Liquidity, OUP

---

6. Connected Graph Bridges

• Back: 01 · From Zero
• Forward: 03 · The Search-and-Bargaining Model · Index Hub
• Sibling: Limit Order Book Mechanics (the exchange alternative) · Counterparty Risk & XVA