Conference Agenda

We extend the classical Lee-Carter model by adding spatial components to analyze subnational variation in age-specific mortality in Italy over time. A Bayesian framework is used, incorporating smoothing priors to borrow strength across regions and time periods, improving model accuracy. The model is estimated with the inlabru package, which extends the INLA methodology for non-linear models. Mortality data for males and females from 107 Italian provinces between 2002 and 2022 are analyzed, revealing significant geographical and age-specific mortality disparities. For example, the well-known North-South socio-economic divide is reflected in higher mortality rates among children aged 0-10, while elevated mortality is observed among working-age men in industrialized regions in the North-West and certain Southern provinces. The findings demonstrate how Bayesian extensions of the Lee-Carter framework can enhance understanding of variations in mortality, by explicitly incorporating a geographical component which can account for different underlying socio-economic or environmental risks. Future work may involve incorporating cause-specific mortality or including socioeconomic covariates, allowing for more targeted public health measures to address mortality disparities across Italy’s provinces.

Smart surveys combine data provided by the respondent in a traditional way via a web questionnaire or app with data collected by sensors embedded in smart devices. This reduces the burden on respondents and improves the quality and timeliness of the results, especially in complex surveys.

The aim of the study is to give an overview of the main findings from an experimental survey conducted in Italy in order to assess the respondents’ attitudes towards the new ways of collecting data in smart surveys.

Logistic regression models are used to identify the factors that most influence the participation in the survey and the performance of smart tasks. The profiling of the individuals that emerges from the application of the models proves useful in defining appropriate strategies and solutions to increase both the response rate of smart surveys and the quality of the collected data.

The prevailing view in segregation scholarship is that White/Black segregation in the United States has modestly declined over recent decades, while White/Hispanic and White/Asian segregation have remained stable. This consensus is based on the assumption that measures of unevenness, such as the ubiquitous dissimilarity index, are independent of underlying racial compositions, a property known as compositional invariance. This property is critical to segregation analysis as population compositions differ substantially across locations, and have undergone rapid racial change over the last several decades. In this work, we demonstrate that this common interpretation of dissimilarity's compositional invariance is fundamentally flawed and propose an easily implementable adjustment factor to facilitate meaningful segregation comparisons over space and time. The implications of this adjustment are profound: we find that White/Hispanic and White/Asian segregation have substantially decreased in the last several decades, with Hispanic segregation declining more rapidly than Black segregation.

The Eco_Pop_ER Database, a core component of the MEMOREC project, is an open-access resource designed to analyze the complex interplay between demographic dynamics and climate change in Emilia-Romagna. Central to the project’s philosophy is adherence to FAIR principles (Findable, Accessible, Interoperable, Reusable), promoting open science and data activism. By integrating over 5.5 million observations and 516 indicators spanning demographic, climatic, and territorial variables from 2000 to 2024, the database offers opportunities for analyses of climate impact on population trends.

Eco_Pop_ER engages diverse audiences, from researchers to the public, through its citizen science approach. Initiatives like the "Data4Resilience" challenge encourage University of Bologna students to explore the database, fostering data literacy and providing peer-reviewed feedback to enhance metadata and inclusivity.

Applications of the database include clustering analysis, principal component analysis, spatiotemporal modeling, and exposure analysis, addressing topics like climate vulnerability and hydrogeological risks. The database encompasses a wide range of indicators, such as population characteristics, climate data, environmental risks, and socioeconomic variables.

Eco_Pop_ER sets a benchmark for open-access data platforms, bridging academic research with societal impact. Future expansions will enhance data continuity and foster collaborations with local offices, ensuring the database's for climate resilience and demographic studies.

We examine the predictive power of fertility intentions using Random Forest, a robust machine learning (ML) technique. Utilizing data from the Generations and Gender Surveys, which allow us to link the intention to have a child within the next three years to actual births during that period, we estimate the predictive accuracy of short-term birth intentions compared to other determinants of fertility. We adopt a cross-country approach across nine low-fertility, high-income countries (Austria, Bulgaria, France, Germany, Georgia, Hungary, Lithuania, Netherlands, Russia). Our findings suggest that while the predictive power of fertility intentions is moderate to high, other factors such as age, household income, and gender attitudes hold greater predictive significance. In addition, fertility intentions do not consistently rank among the top predictors of fertility outcomes across countries. This empirical evidence has important implications for the utility of survey data in forecasting short-term fertility trends.