TLDR: A research paper details GEPOC Parameters Version 2.0, an open-source framework for analyzing and forecasting Austria’s population dynamics. It outlines comprehensive data processing methods, including advanced disaggregation algorithms, to harmonize diverse demographic data from Statistics Austria. The framework supports agent-based modeling of population, births, deaths, and internal/external migration. Validation against historical and forecast data shows the model’s high accuracy, particularly for the full regional internal migration model, demonstrating its utility for population-level research questions.
A new research paper introduces GEPOC Parameters Version 2.0, a comprehensive framework for understanding and forecasting population dynamics in Austria. GEPOC, which stands for Generic Population Concept, is a collection of models and methods designed to analyze population-level research questions. This latest version focuses on providing stable and reproducible data processes to generate valid and ready-to-use model parameters specifically for Austria, relying entirely on publicly available data.
The core of this work lies in its detailed description of data-processing methods. These methods are crucial for computing model parameters, especially for GEPOC ABM (Agent-Based Model), a continuous-time agent-based population model. The paper outlines all algorithms used for tasks like aggregation, disaggregation, data fusion, cleansing, and scaling of various data sources. It also provides a clear description of the resulting parameter files, ensuring transparency and reproducibility.
One of the key challenges in population modeling is dealing with data that comes in different resolutions—temporal, spatial, age, and sex. The GEPOC framework addresses this by harmonizing all data to the finest possible resolution. For instance, historical and forecast census information, which often have lower resolutions, are disaggregated to match the detailed current data. This process prevents information loss and enhances the model’s accuracy.
The paper details several disaggregation algorithms. For instance, ‘proportional disaggregation’ is used when the exact distribution needs to be maintained, while ‘integer-valued disaggregation’ (like the Huntington-Hill method) is applied for parameters that must be whole numbers, such as initial population counts or immigration figures. For more complex scenarios, such as estimating internal migration patterns where data is coarse in different dimensions, advanced techniques like ‘Iterative Proportional Fitting’ (IPF) are employed.
The primary data source for GEPOC’s parametrization is Statistics Austria, utilizing various open data platforms. This includes detailed records on population status, forecasts, births, deaths, and migration patterns. The paper meticulously lists each data source, its contents, resolution (time-frame, regional-level, sex, age), and licensing information, emphasizing the open-source nature of the parametrization.
The calculation of parameters for GEPOC ABM covers a wide range of demographic events. For population, data from different sources and time periods are merged and disaggregated to create a continuous time-series from 1962 to 2101, with fine-grained regional and age resolution. Births are calculated based on age-specific fertility rates, with a unique approach to forecast future birth rates using a Gaussian bell curve fit. The probability of a male child at birth is also determined from historical data.
Deaths are similarly processed, using a modern version of Farr’s Death Rate Formula to compute probabilities from census data. This involves carefully handling age-cohorts and ensuring compatibility with Statistics Austria’s mortality tables. Emigration and immigration figures are also harmonized, with immigration numbers being balanced against population changes, births, deaths, and emigrations to ensure consistency across the model.
Internal migration is a particularly complex aspect, requiring the integration of data that often lacks complete information (e.g., origin-destination data without age, or age-specific data without origin/destination). The paper describes how 2D and 3D Iterative Proportional Fitting algorithms are used to estimate these intricate flows, creating parameters for interregional, biregional, and fully regional internal migration models.
A rigorous quantitative validation study is presented, comparing the GEPOC ABM simulation results against the reference data from Statistics Austria. The model demonstrates high accuracy in depicting total population, sex distribution, and age-class populations over a 25-year period (2000-2025), with deviations generally within 0.15% for total population. For longer-term forecasts (2026-2050), the model maintains reasonable accuracy, though some underestimation trends are observed, particularly due to the lower resolution of forecast data.
The validation also scrutinizes births, deaths, and migration figures. Births and deaths generally remain within a ±5% range of the reference data, with some age-specific deviations noted for births due to forecasting assumptions. Emigration shows higher fluctuations, especially during periods like the COVID-19 pandemic, but the model captures the overall trends. Internal migration models are evaluated for their ability to accurately represent origin-destination flows and age distributions, highlighting the strengths of the full regional model over simpler approaches.
Also Read:
- A Dual-Agent Framework for Aligning LLMs with Human Travel Behavior
- Modeling Realistic Pedestrian-Driver Interactions with Human-Like Constraints
The research acknowledges funding from several agencies, including the Austrian Research Promotion Agency (FFG), Austrian Science Fund (FWF), Vienna Science and Technology Fund (WWTF), and the Society for Medical Decision Making (SMDM), which supported the development of these methods. For more in-depth technical details, readers can refer to the full research paper available here.
