A paper in the Journal of Public Administration Research and Theory by Morgeson, Van Amburg & Mithas (which can incidentally be found here) infers that “although e-government may help improve citizens’ confidence in the future performance of the agency experienced, it does not yet lead to greater satisfaction with an agency interaction nor does it correlate with greater generalized trust in the federal government overall”. The research was employing data from the American Customer Satisfaction Index study.
The authors argue that improving the customer experience now will improve longer term usage and that the way to achieve this is by measuring website satisfaction and using the information to drive improvement. Personally, I believe this is a necessity for all channels, if there is to be channel shift, since not all services can or will be delivered successfully by self-service.