We show how leakage differs, depending on the biofuel policy and market conditions. Carbon leakage is shown to have two components: a market leakage effect and an emissions savings effect. We also distinguish domestic and international leakage and show how omitting the former like the IPCC does can bias leakage estimates. International leakage is always positive, but domestic leakage can be negative. The magnitude of market leakage depends on the domestic and foreign gasoline supply and fuel demand elasticities, and on consumption and production shares of world oil markets for the country introducing the biofuel policy. Being a small country in world oil markets does not automatically imply that leakage is 100 percent or above that of a large country. We show leakage due to a tax credit is always greater than that of a mandate, while the combination of a mandate and subsidy generates greater leakage than a mandate alone. In general, one gallon of ethanol is found to replace only 0.35 gallons of gasoline – not one gallon as assumed by life-cycle accounting. For the United States, this translates into one (gasoline-equivalent) gallon of ethanol emitting 1.13 times more carbon than a gallon of gasoline if indirect land use change (iLUC) is not included in the estimated emissions savings effect and 1.43 times more when iLUC is included.