Trial results for two COVID-19 vaccines suggest at least 90% efficacy against symptomatic disease (VE ). It remains unknown whether this efficacy is mediated by lowering SARS-CoV-2 infection susceptibility (VE ) or development of symptoms after infection (VE ). We aim to assess and compare the population impact of vaccines with different efficacy profiles (VE and VE ) satisfying licensure criteria. We developed a mathematical model of SARS-CoV-2 transmission, calibrated to data from King County, Washington. Rollout scenarios starting December 2020 were simulated with combinations of VE and VE resulting in up to 100% VE . We assumed no reduction of infectivity upon infection conditional on presence of symptoms. Proportions of cumulative infections, hospitalizations and deaths prevented over 1 year from vaccination start are reported. Rollouts of 1 M vaccinations (5000 daily) using vaccines with 50% VE are projected to prevent 23-46% of infections and 31-46% of deaths over 1 year. In comparison, vaccines with 90% VE are projected to prevent 37-64% of infections and 46-64% of deaths over 1 year. In both cases, there is a greater reduction if VE is mediated mostly by VE . The use of a "symptom reducing" vaccine will require twice as many people vaccinated than a "susceptibility reducing" vaccine with the same 90% VE to prevent 50% of the infections and death over 1 year. Delaying the start of the vaccination by 3 months decreases the expected population impact by more than 50%. Vaccines which prevent COVID-19 disease but not SARS-CoV-2 infection, and thereby shift symptomatic infections to asymptomatic infections, will prevent fewer infections and require larger and faster vaccination rollouts to have population impact, compared to vaccines that reduce susceptibility to infection. If uncontrolled transmission across the U.S. continues, then expected vaccination in Spring 2021 will provide only limited benefit.