In household surveys, earnings data typically can be reported as point values, in brackets or as 'missing'. In this paper we consider South African household survey data that contain these three sets of responses. In particular, we examine whether there are systematic differences between the sample of the employed with earnings reported as point values and those with earnings responses in brackets; we compare five different methods of reconciling bracket and point responses so as to generate descriptive measures of earnings; and we investigate empirically how earnings measures differ by approach.