| Abstract: | As a basis for judging how public policy affects the poor, this article explores how "poor" families may be defined and how well such families can be distinguished from other families in the less developed countries. This is done by seeking proxies for poverty which are relatively easy to measure, accurate in discriminating between the poor and the non-poor, and relevant to public policy. To this end, a highly parsimonious model is developed, based on truncation and regression procedures, using only family size and number of wage earners in addition to either income or an education-age combination. Application of this model to data from household surveys in three major cities of Latin America shows that the model is highly effective in pinpointing poverty households, although the pattern of errors is not random, the most frequent type of error being to classify poverty households as non-poor.
Especially significant is that the model is nearly as effective for discriminating poverty households from others when financial variables are excluded as when they are included. This would suggest that a good deal of flexibility exists in deciding what variables to include in future studies of this type. The results also suggest that even better results should be possible if more complete information is obtained on the employment status of the different members of the household and on the contribution of each to household income. Ideally, the data collection and model development should proceed in an iterative manner since there are numerous possible variables as well as alternative model formulations. |
