Bayesian inference of prehistoric population dynamics from multiple proxies: a case study from the North of the Swiss Alps

Martin Hinz, Joe Roe, Julian Laabs, Caroline Heitz, and Jan Kolář, 2022. Bayesian inference of prehistoric population dynamics from multiple proxies: a case study from the North of the Swiss Alps. Manuscript submitted for publication. doi:10.31235/osf.io/dbcag

Abstract

Robust estimates of population are essential to the study of human–environment relations and socio-ecological dynamics in the past. Population size and density can directly inform reconstructions of prehistoric group size, social organisation, economic constraints, exchange, and political and social institutions. In this pilot study, we present an approach that we believe can be usefully transferred to other regions, as well as refined and extended to greatly advance our understanding of prehistoric demography.

Here, we present a Bayesian hierarchical model that uses Poisson regression and state-space representation to produce absolute estimates of past population size and density. Using the area North of the main ridge of the Swiss Alps in prehistoric times (6000–1000 BCE) as a case study, we show that combining multiple proxies (site counts, radiocarbon dates, dendrochronological dates, and landscape openness) produces a more robust reconstruction of population dynamics than any single proxy alone. The model’s estimates of the credibility of its prediction, and the relative weight it affords to individual proxies through time, give further insights into the relative reliability of the evidence currently available for paleodemographic research. Our prediction of population development of the case study area accords well with the current understanding in the wider literature, but provides a more precise and higher-resolution estimate that is less sensitive to spurious fluctuations in the proxy data than existing approaches, especially the popular summed probability distribution of radiocarbon dates.

The archaeological record provides several potential proxies of human population dynamics, but individually they are inaccurate, biased, and sparse in their spatial and temporal coverage. Similarly, current methods for estimating past population dynamics are often simplistic: they work on limited spatial scales, tend to rely ona single proxy, and are rarely able to infer population size or density in absolute terms. In contemporary demography, it is becoming increasingly common to use Bayesian statistics to estimate population trends and project them into the future. The Bayesian approach is popular because offers the possibility of combining heterogenous data, and at the same time qualifying the uncertainty and credibility attached to forecasts. These same characteristics make it well-suited to applications to archaeological data in paleodemographic studies.

Metrics

Updated: