Using Selective Sampling for Binary Choice Models to Reduce Survey Costs
Marketing problems sometimes concern the analysis of dichotomous variables, like for example ``buy'' and ``not buy'' and ``respond'' and ``not respond''. It can happen that one outcome strongly outnumbers the other, for example when many households do not respond (to a direct mailing, for example). Standard econometric methods would imply the collection of many data to obtain precise estimates and this can be rather costly. To cut back costs, we propose to implement a non-random sampling scheme and to correct for the subsequent sample selection bias in the econometric model. In this paper we put forward the relevant method, which does not lead to a loss in precision. Our illustration suggests an opportunity to collect 60\\% less data points.
|Keywords||Outcome-dependent sampling, binary outcomes, logit model, sample size, survey design|
|JEL||Econometric and Statistical Methods: Other (jel C19), Statistical Decision Theory; Operations Research (jel C44), Business Administration and Business Economics; Marketing; Accounting (jel M), Marketing (jel M31)|
|Publisher||Erasmus Research Institute of Management|
|Series||ERIM Report Series Research in Management|
|Rights||Copyright 2001, B. Donkers, P.H. Franses, P. Verhoef, This report in the ERIM Report Series Research in Management is intended as a means to communicate the results of recent research to academic colleagues and other interested parties. All reports are considered as preliminary and subject to possibly major revisions. This applies equally to opinions expressed, theories developed, and data used. Therefore, comments and suggestions are welcome and should be directed to the authors.|
Donkers, A.C.D, Franses, Ph.H.B.F, & Verhoef, P.C. (2001). Using Selective Sampling for Binary Choice Models to Reduce Survey Costs (No. ERS-2001-67-MKT). ERIM Report Series Research in Management. Erasmus Research Institute of Management. Retrieved from http://hdl.handle.net/1765/131