Voluntary opt-in programs to reduce emissions in unregulated sectors or countries have spurred considerable discussion. Since any regulator will make errors in predicting baselines and participants will self-select into the program, adverse selection will reduce efficiency and possibly environmental integrity. In contrast, pure subsidies lead to full participation but require large financial transfers. We present a simple model to analyze this trade-off between adverse selection and infra-marginal transfers. We find that increasing the scale of voluntary programs both improves efficiency and reduces transfers. We show that discounting (paying less than full value for offsets) is inefficient and cannot be used to reduce the fraction of offsets that are spurious while setting stringent baselines generally can. Both approaches reduce the cost to the offsets buyer. The effects of two popular policy options are less favorable than many believe: Limiting the number of offsets that can be one-for-one exchanged with permits in a cap-and-trade system will lower the offset price but also quality. Trading ratios between offsets and allowances have ambiguous environmental effects if the cap is not properly adjusted. This paper frames the issues in terms of avoiding deforestation but the results are applicable to any voluntary offset program.