Estimating commuters' "value of time" with noisy data: a multiple imputation approach

We estimate how motorists value their time savings and characterize the degree of heterogeneity in these values by observable traits. We obtain these estimates by analyzing the choices that commuters make in a real market situation, where they are offered a free-flow alternative to congested travel. We do so, however, in an empirical setting where several key observations are missing. To overcome this, we apply Rubin's multiple imputation method to generate consistent estimates and valid statistical inferences. We also compare these estimates to those produced in a "single imputation" scenario to illustrate the potential hazards of single imputation methods when multiple imputation methods are warranted. Our results show the importance of properly accounting for errors in the imputation process, and they also show that value of time savings varies greatly according to motorist characteristics.