Marking A Book: Market Versus Model

A liquid large-cap stock has a mark that nobody argues about: the last trade or the current mid, printed continuously all day by an exchange everyone can see. An illiquid convertible bond, a bespoke swap, or an option three strikes away from where anything actually traded today has no such mark — someone has to compute what it would be worth, using a model, and that computed number becomes the P&L that shows up on the desk report. The two situations look identical on the P&L screen. They are not remotely identical in how much you should trust the number.

Where model marking becomes unavoidable

Mark-to-market means using an actual observed, executable price — the last trade, the current bid/offer, or the mid between them, for an instrument trading actively enough that those prices are real and recent. Mark-to-model means using a pricing model — a discounted cash flow, an option pricing formula, an interpolated curve — because no recent, reliable trade exists to mark to directly. This is routine and necessary for large classes of instruments: most OTC derivatives, most corporate bonds outside the most liquid issues, private positions, and any exchange-listed instrument thinly enough traded that the last print is stale or the quoted spread is too wide to be a real price.

The risk is not that model marking is wrong to do — it is often the only option — but that the model has inputs the marker chooses, and choices that flatter the position are easy to make without anyone noticing for a long time, because there is no independent trade to check the mark against.

Worked example

A desk holds a corporate bond that last traded three weeks ago. Today's model mark depends on a credit spread assumption fed into the pricing model.

Credit spread assumption	Model price	Implied daily P&L versus yesterday's mark
180bp (independent pricing service, sector average)	96.20	-$14,000
165bp (trader's own estimate, unchanged from last week)	97.85	+$1,500

Both are defensible model outputs — nobody can prove either is "the" correct spread without an actual trade. But the trader marking their own book chose the input that produces a flat-to-positive day instead of the loss the independent service implies, and did so with no observed trade to be checked against. This is exactly why marking discipline separates who computes the mark from who holds the position: an independent pricing or valuation control function, not the trader, should own the model inputs for anything marked to model, precisely because the trader has an obvious incentive on the input that the market-marked instruments never present.

What good practice looks like

Independent price verification — a control function comparing the trader's marks against external sources (dealer quotes, independent pricing services, recent comparable trades) on a fixed schedule — exists specifically to catch the gap between the case above and reality before it compounds over months into a large, undiscovered mismarking. Any position marked to model should carry a visible flag distinguishing it from market-marked positions on the daily P&L report, so a reader of the report knows which numbers are observed and which are estimated.