Experience conditioning in modal choice modelling – Does it make a difference?
Access status:
Open Access
Type
Working PaperAbstract
It is often suggested that real experience counts when making choices and that choice responses are likely to be different after an individual has experienced specific alternatives. This has links with hypothetical bias associated with stated choice (SC) experiments as well as with ...
See moreIt is often suggested that real experience counts when making choices and that choice responses are likely to be different after an individual has experienced specific alternatives. This has links with hypothetical bias associated with stated choice (SC) experiments as well as with the concerns over the reliability of revealed preference (RP) data obtained on alternatives that individuals have had little if any exposure to through use or consumption. If we had a way of identifying experience with each alternative in a choice set, constructed from RP or SC data, might we obtain different estimates of key behavioural outputs such as willingness to pay and elasticity? What if we weighted each alternative associated with each observation by some measure of overt experience? A reason why this might be important is that it enables us to recognise that as more individuals use particular alternatives, especially new infrastructure such as toll roads and public transport, the forecasts of benefits are likely to converge closer to estimates of utility in which the parameter estimates (and hence implied willingness to pay and elasticities) are conditioned on experience, on the reasonable hypothesis that future predictions reflect growing levels of exposure and hence experience with specific alternatives. We use an RP data set collected in Sydney on commuter mode choice to investigate this phenomenon, comparing the value of travel time savings, and cost and travel time elasticities associated with car and public transport under a model in which we do not condition the time and cost parameters on experience, and a model in which we do, using frequency of use as a proxy indicator of overt experience. The differences are very marked, and suggest, albeit from a single study, that this topic appears to have great merit as a candidate source of potential errors in forecasts. We suggest a way forward in which this evidence on experience might be applied to obtain revised forecasts of modal shares.
See less
See moreIt is often suggested that real experience counts when making choices and that choice responses are likely to be different after an individual has experienced specific alternatives. This has links with hypothetical bias associated with stated choice (SC) experiments as well as with the concerns over the reliability of revealed preference (RP) data obtained on alternatives that individuals have had little if any exposure to through use or consumption. If we had a way of identifying experience with each alternative in a choice set, constructed from RP or SC data, might we obtain different estimates of key behavioural outputs such as willingness to pay and elasticity? What if we weighted each alternative associated with each observation by some measure of overt experience? A reason why this might be important is that it enables us to recognise that as more individuals use particular alternatives, especially new infrastructure such as toll roads and public transport, the forecasts of benefits are likely to converge closer to estimates of utility in which the parameter estimates (and hence implied willingness to pay and elasticities) are conditioned on experience, on the reasonable hypothesis that future predictions reflect growing levels of exposure and hence experience with specific alternatives. We use an RP data set collected in Sydney on commuter mode choice to investigate this phenomenon, comparing the value of travel time savings, and cost and travel time elasticities associated with car and public transport under a model in which we do not condition the time and cost parameters on experience, and a model in which we do, using frequency of use as a proxy indicator of overt experience. The differences are very marked, and suggest, albeit from a single study, that this topic appears to have great merit as a candidate source of potential errors in forecasts. We suggest a way forward in which this evidence on experience might be applied to obtain revised forecasts of modal shares.
See less
Date
2016-03-01Department, Discipline or Centre
ITLSShare