Increasing traffic leads to increasing severity, spatial extension and duration of congestion. Congestion has two immediate consequences. One is that travel times increase on average. Another is that travel times become increasingly variable and unpredictable. When performing economic appraisal of transport policies it is important to account for both. There is a well-established practice of accounting for changes in average travel time. The concept is clear, average travel time is comparatively easy to measure and predict and the underlying economic principles are widely accepted. We are well able to account for the economic consequences of congestion as far as the effect on average travel is concerned. At present, there is no similarly well-established practice of accounting for changes in the variability of travel times. This is a major short-coming of current economic appraisal methodology, since the economic costs of variability are likely to be large. The objective of this study is to remedy this situation by doing three things. o Establish a definition of travel time variability and its value that is o theoretically sound o possible to estimate from individual preferences o applicable with existing or realistically foreseeable traffic models o Provide short term recommendation for including valuation of travel time variability in Danish practice for economic appraisal of transport projects o Outline a Danish study of the valuation of travel time variability The first task is motivated simply by the fact that so far a definition of travel time variability and its associated value has not existed that satisfies the above criteria. Various approaches have been proposed but all have serious short-comings relative to one of the criteria. This report proposes a theoretical economic model as the basis for defining and valuing travel time variability. The model says that the value of travel time variability, generally known as the value of reliability, can be defined in terms of scheduling preferences of individuals, the costs of being early or late and the value of time per se, and the travel time distribution summarised by its standard deviation.