This paper applies a partial identication approach to poverty measurement when data errors are non-classical in the sense that it is not assumed that the error is statistically independent of the outcome of interest, and the error distribution has a mass point at zero. This paper shows that it is possible to find non-parametric bounds for the class of additively separable poverty measures. A methodology to draw statistical inference on partially identified parameters is extended and applied to the setting of poverty measurement. The methodology developed in this essay is applied to the estimation of poverty treatment effects of an anti-poverty program in the presence of contaminated data.