Run a supercritical branching process for T generations. Sample k individuals from generation T. Trace their genealogy backward. The most recent common ancestor lived at some generation t — how is t distributed?
Reader note (2026-11-17): Checked against the full text of arXiv:2603.11990 (Krasnowska, Jenkins and Johansen, University of Warwick; the version served, v1 of 12 March 2026). The essay says the multi-type coalescence time has been analyzed only in the critical case. The paper's abstract says critical multi-type processes were studied more recently, and its introduction cites an earlier result (its reference 21) giving the distribution of the generation of the most recent common ancestor in a supercritical multi-type model with finitely many types that cannot die out; the paper states that its formula generalizes that result to a process that can become extinct and can have countably infinitely many types. So the supercritical multi-type case was not open before this paper; the extension to extinction and countably many types is what the paper adds. The paper's exact formula is a statement in the limit as the number of generations T tends to infinity, with bounds in terms of harmonic moments; for finite populations the paper reports numerical results showing the distribution can be approximated effectively in practical settings, so the essay's closing line that it is computable for finite populations goes beyond the paper's wording.
For single-type processes, this is well understood. For multi-type processes — where individuals have types that affect their offspring distributions — the coalescence time distribution has been analyzed only in the critical case. The supercritical multi-type case yields a formula expressing the distribution function of the MRCA generation in terms of the limiting distribution of the normalized population size (arXiv:2603.11990).
The decay of this distribution — how quickly the probability that the MRCA is at generation t approaches certainty as t grows — is bounded using harmonic moments of the population size. Harmonic moments (expectations of negative powers of population size) are harder to control than ordinary moments because they are sensitive to the probability of the population being small. In a supercritical process, the population grows exponentially on average, but individual realizations can be small by chance, and these rare events dominate the harmonic moments.
The Harris-Sevastyanov transformation converts the multi-type harmonic moment problem into a moment problem for a transformed process at the first generation. The transformation changes the measure so that what was a harmonic moment becomes an ordinary moment — tractable by standard tools.
The practical message: coalescence time in structured populations is computable, not just asymptotically but for finite populations with multiple types. The genealogy of a sample from a complex population is no more mysterious than the branching process that generated it — once the right transformation is applied.