Titolo della tesi: Essays on ex-ante analysis of risk and welfare
In an era marked by unprecedented global challenges, characterized by pandemics, conflicts and political and economic instability, there is a growing significance placed on empirically assessing the impact of exposure to risks on multifaceted dimensions of welfare. Moreover, the imperative to identify and focus on individuals and groups susceptible to future impoverishment has never been more critical for policymakers.
This thesis represents a nuanced and comprehensive exploration of vulnerability and economic fluctuations, focusing keenly on the integral role of risk at both macroeconomic and microeconomic levels. The research undertakes a deep dive into the intricate dynamics of risks associated with consumption patterns, food insecurity, and poverty.
This work builds upon the comprehensive theoretical literature on poverty, vulnerability, consumption, and risk, and the recent methodological one on machine learning algorithms. On the theoretical front, the study delves into the existing discourse on poverty, vulnerability, and consumption, unravelling intricate relationships and latent dynamics. Simultaneously, methodological advancements are introduced by incorporating risk factors and integrating theoretical models with state-of-the-art machine learning techniques. Specifically, this work makes three key contributions. First, it quantitatively assesses the impact of fluctuations on consumption; second, it tests the accuracy of existing models of vulnerability and critically scrutinises their limitations; and third, it proposes a novel hybrid measure of vulnerability, merging theory-based and data-driven models holding the potential for more effective policymaking and targeted interventions. The ultimate aspiration is to contribute substantive insights that can inform the development of robust risk management strategies, fostering resilience in the face of a myriad of challenges.
This dissertation encompasses three essays, which, although fully self-contained, are closely interconnected.
Essay 1 investigates the topic by encompassing macroeconomic considerations. It delves into the enduring debate surrounding the influence of macroeconomic fluctuations on long-term consumption prospects. Using Penn World Table data, the study establishes a significant association between ex-ante permanent risk and consumption behaviour and scrutinises the Precautionary Savings channel to unravel additional layers of complexity in the relationship between risk and welfare across 183 countries from 1970 to 2019. The identified "extra savings" resulting from this association have permanent negative effects on welfare, particularly in the context of the poorest countries.
Moving further, Essay 2 takes a microeconomic lens using the exogenous shock of the COVID-19 pandemic to retrospectively evaluate the validity of theory-based methods for targeting household vulnerability to food insecurity. The empirical applications extend beyond theoretical frameworks, incorporating out-of-sample assessments to enhance the reliability and applicability of the findings. Employing the World Bank's multi-topic surveys for Ethiopia and Nigeria, the study reveals that existing vulnerability measures exhibit poor out-of-sample performance compared to a data-driven routine, emphasizing the need for methodological improvements. The findings stress the importance of enhancing targeting approaches for identifying vulnerability hotspots and underscore the role of data availability in constraining predictive models in data-scarce environments.
Finally, Essay 3, builds upon and extends the research trajectory of Essay 2. It proposes a hybrid methodological approach that integrates machine learning into the theory-based vulnerability-to-poverty framework. By bridging the gap between theoretical literature and data-driven methodologies, the study tests the out-of-sample forecasting ability of the indicator using household data from Nigeria (LSMS-ISA). The results demonstrate the indicator's effectiveness in anticipating households experiencing poverty. Empirical applications draw on publicly available data, providing a practical and accessible avenue for policymakers and researchers to leverage in their pursuit of more effective risk management strategies.
In conclusion, despite acknowledging methodological caveats in each paper, the three essays offer a comprehensive understanding of the phenomenon from different perspectives contributing to the formulation of a methodological proposal capable of modernising standard vulnerability analysis and targeting methodologies with innovative approaches. The findings contribute not only to academic discussion but also carry practical implications for policymakers, emphasising the importance of empirical validation in predictive modelling and advocating for innovative approaches to enhance the resilience of vulnerable communities in an ever-changing global landscape.