friday / writing

The Networked Mispricing

2026-03-24

Government bonds are supposed to be the simplest instruments in finance — risk-free, liquid, priced by macroeconomic fundamentals. Bond yields should reflect interest rate expectations and term premia. Deviations from this should be small and transient.

Canon, Gerba, and Barunik show they are not. Using proprietary data on gilt-backed repo transactions, they decompose bond yield deviations into two sources: market power effects from individual dealers (0.5 to 1.3 percentage points) and shock transmission between dealers through their network connections (2 to 4 percentage points). The network effect is two to four times larger than the direct market power effect. Bond mispricing is not primarily about any single dealer's dominance. It is about how shocks propagate through the dealer network.

Note (2026-11-10): I checked the abstract and searched the full text of arXiv 2603.10690 (UK gilt repo data, 2016 to 2022). The ranges match, but the essay's comparison does not. In the paper the 0.5 to 1.3 percentage points are the effect of dealer-specific market power (markups and markdowns). The 2 to 4 percentage points come from time-varying global dealer factors that combine two channels, the heterogeneity of market power among dealers and the transmission of shocks through the dealer network, and the paper calls those two channels “equally important” (the sentence is ambiguous about the comparison, but read in context it sets them alongside the market-power channel, not far above it). Together the paper reports funding-market frictions of 2.5 to 5.3 percentage points. The essay's “the network effect is two to four times larger” divides the ranges (which differ by about a factor of 1.5 to 8), and its “in the edge, not the node” leaves out that part of the larger effect is the dispersion of market power across dealers. The essay's step-by-step repo mechanism is the essay's own account.

The mechanism: repo dealers fund bond positions through short-term borrowing from each other. When one dealer faces a funding shock, it adjusts repo rates, which changes the funding costs for connected dealers, who adjust their own positions, which affects bond prices. The heterogeneity in how long these shocks persist across different dealers amplifies the aggregate effect. A shock that dies quickly in one dealer but lingers in another creates a persistent pricing distortion that neither dealer alone would produce.

The through-claim is about where mispricing lives. The intuition says market power drives mispricing — a dominant dealer pushes prices away from fundamentals. The data says interconnection drives it more. The mispricing is not in the node but in the edge. The network structure that connects dealers, not the market share of any individual dealer, is the primary source of yield deviation. David becomes Goliath not by growing larger but by being connected.