This study experiments with Instrumental Variables (IV) estimation of engel elasticities of items like clothing, from Indian National Sample Survey (NSSO) household budget data, in order to eliminate the effects of seasonal and other fluctuations affecting data collected for a short reference period. Such data generally relate to a moving reference period of '30 days' preceding the date of interview (also referred to as the last month reference period). The result being that seasonal and other short-run fluctuations thus get superimposed on the true variation across households exaggerating the inequality in the size distribution of population by per capita total consumer expenditure, the widely used measure of level of living as well as other poverty measures. Further the use of such a reference period introduces errors in observation in both the regressor and the regressand leading to biased estimates in the standard method of engel curve estimation. This problem has been largely ignored in the econometric literature. In this paper the problem is treated in an errors-in-variables (EIV) framework. However, the standard techniques of dealing with the EIV problem offer no easy solution in this case, especially because the errors in the regressor and the regressand are correlated. The possibilities of Instrumental Variable (IV) estimation have been tried out. Instruments were chosen from item groups whose consumption is relatively less affected by seasonal and short-run factors. Though the choice among the IV's is not very clear, different IV's give a range of estimates which may cover the true elasticity and which appear to be better than the conventional WLS estimates based on last month data.