Internal Research Seminars

Subjective poverty measures identify the income level below which individuals consider themselves unable to achieve an acceptable standard of living. Their estimation depends not only on observed income but on how subjective assessments are translated into a threshold. This raises the question of how subjective poverty lines (SPLs) can be identified transparently when the underlying information is ordinal.

 

This talk presents two complementary approaches, both using the EU-SILC "ability to make ends meet" (Ability) question as an alternative source for estimating SPLs. The first develops a probabilistic framework adapting the intersection approach traditionally used with the Minimum Income Question (MIQ). Rather than relying on a direct income vs minimum-income relationship, it models the probability of reporting financial difficulties across the income distribution and identifies the SPL where this probability matches the individual's cumulative position in that distribution. Parametric, semi-parametric and non-parametric strategies yield virtually identical SPLs. Ability-based lines are strongly associated with MIQ-based thresholds although the two can imply substantially different poverty rates.

 

In the second study, we develop a weighted Youden index framework in which the relative importance of correctly identifying the subjectively poor and non-poor is made explicit. Applying alternative weighting strategies and binary definitions to EU-SILC data for 32 European countries over 2010–2020 reveals substantial heterogeneity across countries and approaches: the global ROC-based approach produces sensitivity and specificity weights close to equal importance, whereas local approaches can favour identifying the subjectively poor. No single scheme dominates.

 

Together, the studies show that widely available ordinal measures of perceived financial well-being offer a viable basis for data-driven SPL estimation, while the resulting thresholds remain sensitive to methodological choices.