Abstract:
One of the strategies to improve governments’ allocation of resources, particularly when these are insufficient to meet all the public needs, is the use of targeting indices.
Generally, a targeting index is a linear combination of wellbeing indicators (for example years of education, dwelling conditions, and so forth) that orders the members of the population according to their living conditions to classify and identify the beneficiaries of social programmes.
The weighting method that we propose maximises a function that depends on the number of poor beneficiaries, by income and/or expenditure, given some normative
restrictions. Therefore, the indices designed with this method target poor people taking into consideration not only the income dimension, but also other dimensions, such as health, nutrition, and dwelling, although they are not correlated with income and/or correlated with one another. The method for taking these dimensions into account is to specify weighting restrictions by an indicator of wellbeing. In other words, we establish maximum and minimum values for the weights according to the number of indicators and the needs of social programmes. These values maintain the balance between income-based weights and normative weights. Because of these characteristics, we call our method Maximising Poor Beneficiaries with Normative Restrictions (MPBR). The algorithm behind MPBR can maximises or minimises any function. For instance, in the first exercise of this study, it exclusively counts the number of expenditure -poor beneficiaries, and in the second it contemplates both income-poor and expenditure-poor beneficiaries, given double weight to those that meet both conditions - thus the poorest of the poor households become the first beneficiaries of social programmes-.
We find that when the function to maximise counts exclusively the number of income-poor (or expenditure-poor) households before a cutoff, it is equivalent to minimising the IE (Inclusion error) and EE (Exclusion error). In fact, we theoretically and empirically demonstrate that the models that minimise the IE and EE
(considering income-poor or expenditure-poor households) are probabilistic models. Therefore, we conclude that probabilistic models are a useful tool to determine approximately the minimum IE and EE that can be obtained with a set of indicators. Applying MPBR, we find that when the increase in an indicator of wellbeing statistically increases the probability of not being expenditure-poor, the weight of the indicator converges to a specific positive value to maximise the number of expenditure-poor beneficiaries; in contrast, if the increase in an indicator statistically decreases this probability, the weight of the indicator converges to the minimum allowed weight. In the case in which the indicator is not statistically significant in the probabilistic model, its weight does not converge to a specific value. However, when the weighting restrictions become stronger (for example a greater minimum weight), most of the weights converge to a value. In a comparison of MPBR with other weighting methods (for example PCA and normative method), we conclude that the index estimated with MPBR has a smaller IE and EE and distributes the weights between indicators more equitably than the indices calculated with other methods.
Citación recomendada (normas APA)
Lina María Sánchez-Cespedes, "Minimising the Inclusion and Exclusion Errors to Design Targeting Indices: Between income-based weights and normative weights", Colombia:-, 2018. Consultado en línea en la Biblioteca Digital de Bogotá (https://www.bibliotecadigitaldebogota.gov.co/resources/3711621/), el día 2025-08-13.
¡Disfruta más de la BDB!
Explora contenidos digitales de forma gratuita, crea tus propias colecciones, colabora y comparte con otros.
