Estimating Poverty Dynamics from Bracketed Income Data: A Comparison of Pseudo-Panels and Synthetic Panels Using Tuscany Cross-Section Data

Measuring poverty dynamics is essential to understand movements into and out of poverty and to design effective policy interventions. However, the limited availability of longitudinal data often constrains such analyses. To address this issue, the literature has developed methods to estimate poverty transitions from repeated cross-sectional surveys, including pseudo-panels and, more recently, synthetic panels, which overcome some limitations of pseudo-panels. Although synthetic panels are widely used with consumption data, their application to income, especially when collected through bracketed questions, remains limited. This paper applies the synthetic-panel approach to bracketed income and compares its accuracy with the pseudo-panel approach. Using three repeated cross-sectional surveys conducted in Tuscany between 2021 and 2024, we implement four scenarios combining the two methodological frameworks with two income specifications: bracketed and continuous income. Continuous income is derived from bracketed responses via Survey-to-Survey Imputation using Generalized Additive Models for Location, Scale and Shape (SSIT-GAMLSS) under a Generalized Beta of the Second Kind (GB2) distribution. Precision and reliability of poverty transition estimates are evaluated via bootstrap resampling. Results highlight substantial differences across methods and income specifications, identifying the most accurate framework and the population groups most exposed to poverty.