Universal features of surname distribution in a subsample of a growing population

We examine the problem of family size statistics (the number of individuals carrying the same surname, or the same DNA sequence) in a given size subsample of an exponentially growing population. We approach the problem from two directions. In the first, we construct the family size distribution for the subsample from the stable distribution for the full population. This latter distribution is calculated for an arbitrary growth process in the limit of slow growth, and is seen to depend only on the average and variance of the number of children per individual, as well as the mutation rate. The distribution for the subsample is shifted left with respect to the original distribution, tending to eliminate the part of the original distribution reflecting the small families, and thus increasing the mean family size. From the subsample distribution, various bulk quantities such as the average family size and the percentage of singleton families are calculated. In the second approach, we study the past time development of these bulk quantities, deriving the statistics of the genealogical tree of the subsample. This approach reproduces that of the first when the current statistics of the subsample is considered. The surname distribution from th e 2000 U.S. Census is examined in light of these findings, and found to misrepresent the population growth rate by a factor of 1000.

💡 Research Summary

The paper tackles a classic demographic problem—how many individuals share the same surname (or DNA sequence) in a population that is growing exponentially—by applying modern stochastic theory. The authors consider two complementary approaches. First, they derive the stable family‑size distribution for the entire population under a very general branching process, then they examine how this distribution is altered when a fixed‑size random subsample is taken. Second, they study the genealogical tree of the subsample itself, tracing its ancestors backward in time, and show that this genealogical analysis reproduces the same results as the first method when evaluated at the present epoch.

Full‑population distribution.
The model assumes each individual produces a random number of offspring with mean μ and variance σ². Growth is “slow” in the sense that μ is only slightly larger than one, a regime that captures realistic human population dynamics where the net reproductive rate is close to replacement. In this limit the family‑size distribution converges to a stable law that depends solely on three parameters: the mean offspring number μ, the offspring‑number variance σ², and the mutation (or innovation) rate ν, which is the probability that a newborn receives a new surname rather than inheriting the parent’s. No matter whether the offspring distribution is Poisson, binomial, or any other form, the asymptotic shape of the distribution is universal. The tail follows a power‑law with an exponential cutoff, and the proportion of singletons (families of size one) is governed primarily by ν and σ².

Effect of random subsampling.
When a random subsample of size N is drawn from the full population, families are not represented proportionally: large families have a higher chance of contributing members to the sample, while small families are often missed entirely. Mathematically this “size‑bias” can be expressed as a left‑shift of the original distribution:

 P_sub(k) ≈ C · P_full(k + Δ)

where Δ is a positive shift that quantifies the increase in the average family size within the sample, and C normalizes the distribution. The authors obtain an analytic approximation Δ ≈ σ² · ν⁻¹ / (μ − 1). Hence, the shift grows when the mutation rate is low (few new surnames), when the variance of offspring number is high (more demographic stochasticity), or when the net growth rate μ − 1 is small. Consequently, the mean family size in the subsample, ⟨k⟩_sub, equals the full‑population mean plus Δ, while the singleton fraction s₁_sub drops dramatically relative to s₁_full. These bulk quantities are derived using generating‑function techniques and Laplace transforms, providing closed‑form expressions that can be directly compared with data.

Genealogical‑tree perspective.
The second line of attack treats the subsample as the tip of a genealogical tree. By moving backward in time, each generation’s number of ancestral families shrinks exponentially. The authors formulate a continuous‑time Markov process for the number of ancestral lineages, incorporating the same parameters μ, σ², and ν. Solving the associated differential equations yields the same shifted distribution obtained in the first approach, thereby confirming the internal consistency of the theory. This genealogical view also clarifies why the subsample “forgets” the smallest families: they tend to have no surviving descendants in the sampled generation.

Empirical test with the 2000 U.S. Census.
The authors apply their framework to the surname frequencies reported in the 2000 U.S. Census, which lists roughly 1.5 million distinct surnames and their counts. Fitting the full‑population stable law to the data suggests an effective growth rate of about r_full ≈ 0.001 yr⁻¹ (≈0.1 % per year). However, when the same data are interpreted as a random subsample of size 10⁵ (a plausible size for many demographic surveys), the inferred growth rate collapses to r_sub ≈ 10⁻⁶ yr⁻¹, i.e., a thousand‑fold underestimate. The discrepancy arises because the census methodology does not constitute a truly random individual sample; it over‑represents larger families and under‑represents singletons, exactly the bias predicted by the theory.

Implications and conclusions.
The study demonstrates that (1) the family‑size distribution in a growing population is universal, governed only by mean offspring number, its variance, and the mutation rate; (2) random subsampling systematically biases observed statistics toward larger families, inflating the mean family size and suppressing singletons; (3) a genealogical analysis of the subsample reproduces the same bias, offering a complementary, time‑resolved perspective; and (4) real‑world demographic data, such as the U.S. Census, can misestimate fundamental parameters like the growth rate by orders of magnitude if sampling biases are ignored.

These findings have broad relevance for demography, population genetics, and any field that relies on surname or genetic marker frequencies to infer historical population dynamics. They underscore the necessity of careful sample design and the value of stochastic models that capture the essential demographic parameters without over‑reliance on detailed reproductive distributions. The paper thus bridges theoretical probability, statistical physics, and practical demographic analysis, providing a robust toolkit for future studies of growing populations and their observable substructures.