Multidimensional poverty with missing attributes

•We minimize the probability of misclassifying individuals as being poor or not poor.•The method works when data on some attributes are missing.•The method requires an estimate of the population distribution of attributes.•We take the method to data from India.

We show how to minimize the probability of misclassifying individuals as being poor or not poor when data on some of their relevant attributes are missing, but an estimate of the population distribution of attributes is available.