Self‐employment income is believed to be understated in economic statistics but there is debate about the extent of under‐reporting. This paper refines the widely used method of Pissarides and Weber (Journal of Public Economics, Vol. 39, No. 1 (1989), pp. 17–32) that relies on discrepancies between food shares and reported incomes. Our panel data approach disentangles under‐reporting from fluctuations in transitory income and gives a point estimate of the under‐reporting rate. Previous studies just give an interval estimate and also make the unlikely assumption that under‐reporting is independent of transitory income fluctuations. Panel data from Korea and Russia are used to illustrate the method, and suggest that in both countries almost one‐quarter of the income of self‐employed households is not reported.