The cost-benefit analysis of marijuana legalization heavily relies on three factors that are not only unknown but for different reasons are also largely unknowable: how much legalization will increase marijuana use, how legalization will influence alcohol use, and the value of an hour of marijuana intoxication.  This presentation highlights the difficulties associated with generating credible ranges, let alone good point estimates.  We demonstrate that – a leaving aside the alcohol connection—benefit-cost analysis of marijuana legalization could boil down to a rather simple proposition: If you think marijuana intoxication is, on average, a good thing – counting both the happy controlled users and the unhappy dependent users – then a benefit-cost analysis done in a way that reflects your values will probably conclude that legalization improves social welfare.  Conversely, if you think marijuana intoxication is, on average, a bad thing, then an analysis that reflects your values will probably conclude legalization harms social welfare – because the dominant outcome of legalization will be more marijuana use.  Likewise, when considering indirect effects on alcohol and excluding user’s direct utility or disutility from consumption, if you think marijuana use substitutes for alcohol, then legalization will lead to a net reduction in social costs, whereas if you think they are complements, legalization will lead to a net increase in social costs.