Congestion on urban freeways often creates a need to increase freeway capacity by adding an additional lane. Although adding a lane by widening the existing roadbed is often difficult and expensive, restriping the traveled way with narrower lanes, converting all or part of the shoulder to a travel lane, or a combination of both often is a practical solution. An observational before-and-after evaluation with the empirical Bayes method was done to examine the safety effects of projects involving narrower lanes or shoulder conversions on existing urban freeways in California with four or five lanes in one direction of travel. The evaluation found that projects converting four lanes to five lanes resulted in increases of 10% to 11% in accident frequency. Projects converting five lanes to si lanes resulted in smaller increases in accident frequency. These increases in accident frequency may be the result of accident migration caused by relocation of traffic operational bottlenecks.