Abstract
Sources of specification as well as measurement error in previous estimates of the value of time are examined, and a new set of estimates avoiding these errors is presented. The sources of specification error being analyzed are (1) failure to take account of drivers' uncertainty in making their choices between alternative routes of modes, which was found empirically to bias the value of time estimate 50% or more; (2) the differential in comfort and similar intangible characteristics between two modes; and (3) the inability to account for cross-time substitutions by travelers in peak-hour conditions.