Our concern with this paper is to develop a deeper understanding of what we can and cannot say about intra-household allocation based on the household survey data used in Browning and Lechene (2001) [henceforth: BL]. Our humble contribution can be summarised thus: i) Accepting the BL framework, we show that their conclusions on the presence of caring can be strengthened by using a better GMM estimator ii) where BL cite difficulties in coping with endogeneity in a semiparametric framework, we venture to integrate GMM technique into a semiparametric framework. iii) Using bootstrapped confidence intervals of the semiparametric fit to measure spline fits, we show that the data imposes severe limits on how many splines should be allowed in the unrestricted parametric model. iv) Combining the insight from parametric and semiparametric analysis, we find evidence that the way BL divide their data sample in splines favours their conclusion that agents exhibit caring v) Using our semiparametric fit as an unrestricted model against which to test linear restrictions, we show that all versions of the collective model fit well. vi) In the semiparametric framework with optimal and adaptive bandwidth we show, that which collective model fits the data best depends on the choice of adaptor. Our conclusion is, that data is not good enough to discriminate between caring and non-caring. Till a better dataset is found, we are not convinced that anyone cares - especially not men.