Laboratory
Fertility and Well-Being
At a Glance
Projects
Publications
Team
Project
Leveraging Micro-Level Computational Models to Advance Fertility Forecasting
Conducted by Daniel Ciganda; Julio César Cesano González; in Collaboration with Mikko Myrskylä, Julia Hellstrand (both: MPIDR)
Detailed Description
The capacity to accurately anticipate the evolution of fertility trends is one of the key instruments to avoid systemic crises and transform these institutions and services in ways that guarantee the improvement of living standards.
The problem we aim to solve is the overreliance of fertility forecasts on macro-level extrapolation methods. Currently, most national statistical offices and international organizations produce forecasts by extrapolating from aggregate-level indicators and in some cases by adjusting the extrapolated trend by taking the opinion of experts into account. This approach relies on the strong assumption that past fertility trends will extend into the future, regardless of how other relevant social processes are expected to evolve. Although extrapolation methods may be accurate in some contexts, they still give users no insights into the social, cultural, or technological processes that may be driving the forecasted trends.
In the past few years, we developed an approach that overcomes some of these limitations. This approach has been inspired by the search of what Keyfitz (1982) once referred to as a “truly behavioral way of estimating the future,” and by the idea that having a solid understanding of the microfoundations of demographic change is one of the keys to significantly improve our ability to forecast population dynamics.
The approach has only been made possible thanks to a series of fundamental advances in computing, statistics, and simulation approaches; these have converged to significantly expand the potential of individual-level computational models. Computing technology and infrastructure has drastically reduced the costs associated with the development and estimation of individual-level simulations. The popularization of agent-based modeling approaches has allowed us to incorporate preferences and decisions in demographic models, which are key to understanding contemporary fertility. Finally, the development of likelihood-free estimation methods has provided a framework that allows for inference and uncertainty estimation in large-scale computational models.
Our framework does not extrapolate past fertility but is driven by the evolution of its main determinants. Our model thus defines the mechanisms that relate individual characteristics (education and labor-force participation status) to fertility preferences and decisions and to other critical factors that affect the timing and final number of children. The information on these characteristics comes from empirical distributions, which means that the forecasts depend on assumptions about the future trajectory of these distributions.
As to the planning of public policies, this provides a valuable tool that can act as a laboratory to generate potential scenarios of the form: “If X trend moves in Z direction, we can expect Y fertility outcome.” This can significantly improve users' understanding of the social dynamics behind demographic forecasts, facilitating the communication of scientific insights to policy-makers and the general public and thus the implementation of proactive population policies.
Fertility Development, Projections and Forecasting
Publications
Ciganda, D.; Hellstrand, J.; Myrskylä, M.:
MPIDR Working Paper WP-2023-010. (2023)
Ciganda, D.; Todd, N.:
Population Studies 76:3, 495–513. (2022)
Ciganda, D.; Lorenti, A.; Dommermuth, L.:
MPIDR Working Paper WP-2021-016. (2021)
De-Armas, G.; Rodríguez-Collazo, S.; Álvarez-Vaz, R.; Carrasco, H.; Ciganda, D.:
IESTA Working Papers 2/20, Montevideo. (2020)