This piece from Parapundit discusses an empirical study by Peter Swan and Michael Belzer which suggests that privatising roads can lead to traffic diversions which worsen external costs.
The general argument is one I have commented on before. Pricing on some roads - for example on major highways alone - can lead to traffic diversion onto smaller roads where congestion costs and quality of life costs impact on local residents and where road maintenance costs are greater. Traffic accident costs are also greater.
The diversions itself are socially costly also because drivers are not taking the route that would be efficient.
The authors demonstrate these conclusions for a privatised road which sets monopoly prices. The result is thus to be expected. These roads should be priced at something less than the direct social marginal cost of congestion were there no substitution possibilities. This will almost always be much less than the monopoly price.