Prof. Burhan Can Karahasan (p), Assoc. Prof. Bilge Eris Dereli, Assoc. Prof. Burcu Düzgün Öncel, Assoc. Prof. Tolga Aksoy

Poverty has been extensively analyzed on monetary grounds. However, neglecting non-monetary dimensions of poverty can result in under-estimation of the actual poverty problem. Human capital, living standards, housing conditions and environmental concerns are the primary candidates to describe the non-monetary dimensions of poverty measurement. Meanwhile, there is a consensus that structure of production and industrialization have massive influence on income distribution and the poverty problem. In this research, we explore the 2014-2024 period and construct a unified multidimensional poverty index (MPI) for Turkish regions by using the Survey on Income and Living Conditions. Next, we analyze whether regional and temporal evolution of industrial development have influence on the distribution of multidimensional poverty. Our empirical design has two core steps: (i) We follow the AF methodology and construct the aggregate MPI and sub-MPI for different segments (based on gender, age, employment status, sector of employment and skill level) at the regional level (NUTS-2), (ii) Next, we explore the spatial distribution of MPI and its connection with regional industrialization patterns by using spatial panel models (SAR, SEM, SAC, SDM).
Our preliminary results suggest a declining trend for the MPI at the national level. However spatial differences follow the known west-east duality in Turkey. We also find out that different segments of the population suffer differently from the non-monetary poverty problem. Our results also show that regional MPI figures are subject to significant spatial autocorrelation, suggesting the existence of local externalities in terms of non-monetary deprivation. We also trace the local decomposition of spatial autocorrelation and detect limited mobility across different spatial clusters, suggesting strong spatial inertia in the MPI’s distribution. Our spatial modelling exercises show that industrial production has stronger poverty alleviation effects at the regional level. Moreover, we find out visible spatial networks for the MPI’s regional variation. Those regions which are clustered with high (low) MPI levels witness higher (lower) non-monetary deprivation. Finally, our sub-MPI analyses show stronger effects for industrial production as those regions with denser industrial production have lower propensity to suffer from the multidimensional poverty problem.
Our results show that, policies that focus on the structure of production can be more operative to combat with the poverty problem. Territorial disparities, deindustrialization trends and spatial externalities are the main pillars of our analyses carried out for Turkey. Therefore, findings from Turkey can be reference points for the construction of policies in developing countries with similar fundamental problems.

Prof. Marco Di Cataldo (p)

This paper develops a novel early-warning system to predict firm confiscation due to mafia infiltration in Italy. To do this, we construct a unique, dynamic corporate ownership network by tracing the first- and second-degree connections of all Italian firms confiscated over the last two decades. Using this dataset, we train a series of machine learning models that integrate firm-level financial data with a comprehensive set of network topology measures, successfully forecasting confiscation up to three years in advance. Our out-of-sample results demonstrate exceptionally high predictive power, with the integrated model significantly outperforming those relying on a single data source. We find that network features are crucial for comprehensive detection (high recall), while firm financials provide accuracy (high precision); our framework effectively balances this critical trade-off. A feature-contribution analysis using SHAP values reveals that firm size is the single most important predictor, showing that smaller and financially distressed firms face the highest confiscation risk. Network topology features collectively emerge as the second most important group of indicators. Taken together, our findings underscore the value of integrating ownership network intelligence with firm fundamentals to create a powerful and interpretable tool for risk-based supervision and enforcement against organized crime.