Propuesta de índices multidimensionales para la evaluación del programa Ser Pilo Paga

Abstract

In different studies statistical units are described in two types of variables: quantitative and qualitative. A common example where these types of variables are observed are the information collection instruments known as surveys, in general, in these instruments there are closed questions that are usually gathered together in sets that may be of a quantitative or qualitative nature. A multivariate technique that analyzes this type of variables is the factorial analysis for mixed data (FAMD), this methodology consist in transform the quantitative variables into qualitative variables, breaking down the range of variation in the classes producing a homogeneous table that allows you to implement an analysis Multiple Correspondence (MCA), however, the notion of grouping variables is not an FAMD property. If the instrument complies with this characteristic (the grouping of variables in the individuals of the analysis) it is considered necessary to use multiple factorial analysis (MFA). Because this kind of analysis it allows to analyze quantitative and qualitative variables being broadly useful when the number of variables from each group is very different.

Extending the MFA to groups of qualitative or mixed variables (of both natures) is known as a multiple factorial analysis for mixed data (MFAMD), which combines both the MCA, the FAMD and the MFA, this technique highlights the main variability factors of the individuals, described these, in a balanced way by the groups of variables. Applying the MFAMD to the instrument of the baseline of the program Ser Pilo Paga, four multidimensional indices were created that summarize 9 thematic groups composed of 111 variables of quantitative and qualitative type, characterizing a total of 1,487 young people, which 682 are eligible to be part of the program and 805 are not. An estimator of difference in difference applied to the four multidimensional indices is proposed for the impact evaluation.