Most jurisdictions in North America have some version of graduated driver licensing (GDL). A sound body of evidence documenting the effectiveness of GDL programs in reducing collisions, fatalities and injuries among novice drivers is available. However, information about the relative importance of individual components of GDL is lacking. The objectives of this study are to calculate a summary statistic of GDL effectiveness and to identify the most effective components of GDL programs using a meta-analytic approach. Data from 46 American States, the District of Columbia and 11 Canadian jurisdictions are used and were obtained from the Fatality Analysis Reporting System (FARS) for the U.S. and from Transport Canada's Traffic Accident Information Database (TRAID) for Canada. The timeframe of this evaluation is 1992 through 2006, inclusive. Relative fatality risks and their 95% confidence intervals were calculated using fatality counts and population data for target and comparison groups, both in a pre-implementation and post-implementation period in each jurisdiction. The target groups were 16-, 17-, 18- and 19-year-old drivers. The comparison group was 25-54-year-old drivers. The relative fatality risks of all jurisdictions were summarized using the random effects DerSimonian and Laird model. Meta-regression using Restricted Maximum Likelihood (REML) and Markov Chain Monte Carlo (MCMC) Gibbs sampling was also conducted. Strong evidence in support of GDL was found. GDL had a positive and significant impact on the relative fatality risk of 16-year-old drivers (reduction of 19.1%). Significant effects were found for meta-regression models with 16-, 18- and 19-year-old drivers. These effects include length of night restriction in the learner stage, country, driver education in the learner stage and in the intermediate stage, whether night restrictions are lifted in the intermediate stage for work purposes, passenger restriction in the intermediate stage, whether passenger restrictions in the intermediate stage are lifted if passengers are family members, and whether there is an exit test in the intermediate stage. In conclusion, several GDL program components have an important effect on the relative fatality risk of novice drivers. These results help understand how such effects are achieved.