Timetable risk, priced: modelling approval calendars

A pharmaceutical or veterinary position typically depends on a decision by an authority at a date nobody controls. Common practice is to assume the date in the plan of the sponsor, apply a probability of success and discount the result. That treatment gets the average roughly right and the shape entirely wrong, because the cost of delay is not linear. Cash burn continues, competitor filings advance, and exclusivity periods run from a fixed start regardless of when approval arrives.

The approach here is to model the date as a distribution built from observed review durations. Regulators in the jurisdictions we track publish target timetables and, in most cases, statistics on how often those targets are met. Reviews cluster: a straightforward filing in an established class behaves very differently from a first-in-class application or one carrying a manufacturing change. We fit separate distributions by class and by jurisdiction and hold them to a documented origin.

The second term is the information request. Most delays are not refusals. They are questions, and a question restarts a clock. Our model treats a request as an event with its own probability and its own duration, and it allows for two of them. That single change moves the tail of the distribution far more than any adjustment to the probability of eventual approval, and the tail is where a financing structure either holds or fails.

Correlation deserves more attention than it usually receives. An institution holding four development-stage exposures is not holding four independent bets if three of them share a manufacturing site, a class of product or a single jurisdiction. Site-level and class-level correlation is the reason a portfolio of individually modest positions can produce an outsized loss in one quarter. We aggregate exposure by site and by review pathway as well as by counterparty, and limits are set on the aggregate.

Model output does not replace the investment judgement, and the division does not present it as such. The output is a range, a set of dates at which the position needs funding, and a statement of what the position is worth if the slowest quartile of outcomes occurs. The Investment Committee decides. The position of the firm on model-assisted work is settled: a named person answers for every decision a model informs, and no client decision is taken automatically.

The reason this work sits with the Quantitative division rather than with a sector desk is independence. A team that has originated a transaction should not also set the distribution that values it. The separation is a control and not a comment on anyone's judgement. It carries a second benefit: because the distributions are maintained centrally, every pharmaceutical position in the firm is valued on the same assumptions, and a change in review times applies everywhere on the day it is observed.

Published 2022-06-22 by the Quantitative Strategies division. Research is prepared for eligible counterparties and does not constitute advice.