Protected areas and Community Forest Management (CFM) are among the most widespread interventions to conserve forests in tropical countries. In addition to their impacts on forests and the biodiversity they contain, these interventions also affect human well-being, particularly that of the local communities who are often poor and politically marginalized and whose livelihoods depend directly on the forest resources being conserved. To develop effective interventions, practitioners need to have credible, strong and scientifically rigorous evidence on their impacts on forests (or the biodiversity they contain) and human well-being. However, while cientifically rigorous impact evaluation of programs is well advanced in fields such as development, health and education, it is rare in nature conservation. The rare existing studies focus mostly on protected areas and other interventions, such as CFM, are relatively untouched by scientifically rigorous impact evaluation. Different challenges account for the limited adoption of rigorous impact evaluation in nature conservation. Among these are the identification and elimination of rival explanations: factors other than the intervention that can explain the observed relationship between the intervention and the outcome. Potential rival explanations are factors that can confound impact estimates by affecting both assignment of units to intervention and the outcome. Another potential rival explanation is baseline outcome data that should have been collected before the intervention was implemented. Baseline data are often missing in conservation studies. Another challenge is the heterogeneity of management practices within and units exposed to the same intervention. A challenge pertaining particularly to studies on human well-being impacts is the multi-dimensional nature of human well-being. In this thesis, I aim to investigate the impacts of different conservation interventions on environmental and human well-being outcomes while addressing the challenges to conservation impact evaluation discussed above. My case studies are CFM and strict protection in Madagascar; one of the world’s hottest biodiversity hotspots. I have three specific objectives which are addressed in three manuscripts with the following titles: i) Effectiveness of CFM at reducing deforestation across Madagascar (manuscript 1): With colleagues, I investigated the impacts of CFM on deforestation at the national scale between 2000 and 2010 using matching to control for factors confounding impact estimates. We did not detect an impact of CFM, on average, when CFM areas were compared to non-CFM areas, even when the sample was restricted to only where information suggests effective CFM implementation on the ground. However, impacts were heterogeneous conditional on whether CFM permits commercial use of forest resources. No CFM impact was detected where commercial use of natural resources is allowed. However, we did detect some reduction of deforestation in areas managed under CFM that does not permit commercial use, when compared to non-CFM or CFM permitting commercial use. Our findings suggest differentiating among types of CFM is important for estimating the impacts of this conservation approach. ii) Impacts of CFM on human economic well-being across Madagascar (manuscript 2): In this manuscript, we investigated impacts on household living standards across Madagascar as measured by per capita consumption expenditure. We used matching to control for confounding factors and addressed the issue of missing baseline values of household consumption expenditures using an approach known as the placebo test. We cannot statistically reject the hypothesis of zero impact, but we can credibly reject the hypothesis that CFM has had substantial negative impacts on economic well-being across CFM communities in Madagascar. There were heterogeneous impacts, with a mixture of positive and negative impacts, conditional on household proximity to forest and education level. In conclusion, the impacts of CFM vary with household characteristics: some may lose while others may gain. iii) The potential of the Global Person Generated Index (GPGI) for evaluating the perceived impact of conservation interventions on subjective well-being (manuscript 3): In this study, we used the GPGI, a subjective and multidimensional well-being instrument, to investigate the relative impacts of strict protection and CFM on human well-being in sites in eastern Madagascar. We used a participatory approach to establish the cause-effect relationship between the interventions and the outcomes (i.e., to eliminate rival explanations). We did not detect statistically significant difference, on average, between the two approaches in three measures we used to examine the magnitude of their relative impacts on subjective well-being. However, we found some differences in the characteristics of subjective well-being component domains impacted by the strict protection and CFM and in the priority domains that could be targeted by increased resource allocation to improve well-being in locally meaningful ways. Combined with the participatory approach to establish cause-effect relationship, we suggest GPGI provides highly relevant insight that can be used to design policy seeking to increase local participation and develop more positive local attitudes towards conservation. The first two manuscripts (1 and 2) involve analyses at the national scale, objective indicators (deforestation and consumption expenditure) and rigorous quantitative causal inference designs making them of value to external stakeholders, such as government agencies and donors, seeking to know the magnitude of impacts to inform large scale conservation policy. However, these large scale studies may be of limited use for project managers who want to build locally legitimate interventions or those who want a deeper understanding of how conservation interventions affect local people. In the third manuscript, we used a subjective measure of well-being (the GPGI) in combination with participatory approach to establish cause-effect relationship between interventions and locally meaningful outcomes. This has limited value for quantitatively measuring the magnitude of impacts; but holds some promises for project managers who seek local participation and social sustainability. Conservation has long suffered from poor quality evaluation of its impacts. This thesis shows that methods for impact evaluation are available, but the appropriate method that should be applied depends, among other things, on the purpose of the evaluation.