Please use this identifier to cite or link to this item: http://hdl.handle.net/10419/86760
Authors: 
van den Berg, Vincent
Kroes, Eric
Verhoef, Erik T.
Year of Publication: 
2010
Series/Report no.: 
Tinbergen Institute Discussion Paper 10-014/3
Abstract: 
It is a common finding in empirical discrete choice studies that the estimated mean relative values of the coefficients (i.e. WTP's) from multinomial logit (MNL) estimations differ from those calculated using mixed logit estimations, where the mixed logit has the better statistical fit. However, it is less clear under exactly what circumstances such differences arise, whether they are important, and if they can be seen as biases in the WTP estimates from MNL. We use datasets created by Monte Carlo simulation to test, in a controlled environment, the effects of the different possible sources of bias on the accuracy of WTP's estimated by MNL. Consistent with earlier research we find that random unobserved heterogeneity in the marginal utilities does not in itself biases the MNL estimates. Furthermore, whether or not the unobserved heterogeneity is symmetrically shaped also does not affect the accuracy of the WTP estimates of MNL. However, we find that if two heterogeneous marginal utilities are correlated then the WTP's from MNL may be biased. If the correlation between the marginal utilities is negative, then the bias in the MNL estimate is negative, whereas if the correlation is positive the bias is positive.
Subjects: 
Discrete Choice
Biases in WTP's
Multinomial Logit
Correlated Heterogeneous Marginal Utilities
JEL: 
C13
C25
Document Type: 
Working Paper

Files in This Item:
File
Size
440.92 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.