Vaccination is a critical tool, along with suppression and treatment, for controlling epidemicssuch as SARS-CoV-2. To maximize the impact of vaccination, doses should be allocated to thehighest value targets, accounting for health and potential economic benefits. We examine whatallocation strategy is optimal and how to translate that strategy into actionable procurementdecisions in the context of India. We compare 3 different allocation strategies (oldest first,highest contact rate first, random order) across 4 outcomes (lives saved, life-years saved, value ofstatistical lives saved, value of statistical life-years saved). We make 3 methodologicalcontributions. First, we estimate the incremental health benefit of vaccination using novel, localseroprevalence data from India. Second, we estimate the value of statistical life-years usingdisaggregated, monthly data on consumption during the pandemic. Third, and most importantly,we estimate social demand curves for vaccines that can practically guide governmentprocurement decisions. Our analysis yields 4 novel findings. First, the need to speed-upvaccination does not justify deviation from elderly-first prioritization. Second, much of the valueof vaccination comes from improvements in consumption rather than longevity. Moreover,vaccination increases the value of a life year because it increases consumption. Third, socialdemand for vaccination falls over time as natural immunity from infections increases. Therefore,the slower a country vaccinates its population, the fewer doses it should procure. Fourth, there isenough variation in consumption and infection risk that it makes sense to vaccinate some areasbefore others. Our approach of connecting epidemiological models and data on health andconsumption to economic valuation methods generalizes to other infection control strategies,such as suppression, and public health crises, such as influenza and HIV.