Income survey has to carry out the difficult task of estimating the income of families by measuring different parameters of life, as people are reluctant to reveal the actual income. The paper discusses a new approach towards the development of such an estimator. Let x1, x2, … , xn be a set of n variables which are linearly related to income (y). Therefore y = ao + a1 x1 + … + an xn. Collect the sample observations on these n + 1 variables from m families (m is larger compared to n). Here we use Goal Programming technique to estimate the values of ais. Each constraint equation should correspond to a family. ais would be the decision variables. Hence the information on family can be represented as a1 xi1 + a2 xi2 + … + an xin + di − d+i = yi. The Goal Programming model minimizes both the deviational variables (d−i, d+i) on each constraint according to a priority structure. It is best to set a priority structure according to the reliability of the information. In this model each family would correspond to a priority. All possible combinations of priorities are worked out. Select that combination (and the corresponding results) which gives the best correlation between the actual and the estimated income.
