Estimates of Semiparametric Equivalence Scales Thanasis Stengos, Yiguo Sun and Dianqin Wang ∗ Department of Economics, University of Guelph February 11, 2003 Abstract Within the semiparametric framework introduced by Pendakur (1999) we introduce a new loss function to estimate equivalence scales. This loss function uses all available information from the total ex- penditures of both the reference and nonreference households and as such it produces more reliable estimates. We also use the correlation coefficient to directly test the hypothesis of shape invariance of the nonparametric Engel curves, which is a necessary condition for base independence. Using Canadian family expenditure data for 1996 we find shape invariance for the food share equation but not for fuel and clothing. ∗ The first author would like to acknowledge SSHRC of Canada for financial support. We would like to thank P. Lyssiotou, K. Pendakur, T. Crossley and participants of the Workshop on Consumption Behaviour and Welfare Measurement at the Individual and Household Level, March 15-16, 2002, Verona Italy.