In this paper, we improve upon the Pissarides-Weber (PW) method for estimating tax evasion among the self-employed by utilizing unique register-based consumption measures from the Swedish and Finnish mandatory registers for pleasure boats. Our main contributions are twofold: i) The PW method crucially relies on the assumption that the intrinsic consumption preferences are the same for self-employed and wage earner households. Thanks to our rich data, we are able to test this assumption directly by contrasting the Engel curves between wage earner households and households that will become self-employed in the future. We find that the group that will become self-employed is only marginally more prone to own a boat before they actually become self-employed. This lends support to the assumption of similar intrinsic preferences between the two groups. ii) The large sample size that comes with register-based consumption measures allows us to challenge the standard functional form assumptions in previous PW applications. The estimated sizes of income underreporting in absolute monetary amounts are almost constant over reported income levels, whereas previous studies have assumed that the underreporting is proportional to income.