Even though overall growth in India has recently accelerated, it has largely bypassed rural areas and agriculture; in fact it is agreed that the ratio of rural to urban poverty has increased. As a consequence, some of the marginalized groups in a society that is already characterized by a high level of inequality in opportunities and segregation along lines of, gender, caste, and social status, are widely reckoned to not have benefited from overall growth. To enable women and backwards castes to expand their livelihood opportunities, a vast range of government programs have been initiated and large amounts of resources are being channeled to poor areas. An increasingly popular approach to do so relies on the concept of Community Driven Development (CDD) whereby resources are made available to encourage formation of savings groups at the local level and, once they have attained a certain degree of maturity, the channeling to such groups of funds (either as a credit or a grant) which they can use for small projects aimed to improve their livelihood. This paper uses one of the earliest projects of this type, the Andhra Pradesh (AP) District Initiatives Project, a US $ 250 Mn intervention that was started in 2000, to provide an initial assessment of the strengths and weaknesses of the CDD approach. One of the interesting characteristics of the CDD approach is its desire to combine interventions to increase food security and deal with immediate needs with social mobilization to overcome long-standing patterns of discrimination and prejudices and attention to longer-term economic needs. Two interesting aspects of the AP program that greatly facilitate assessment of impact is the fact that only women's groups were eligible and that even in cases where it built on pre-existing group structures, the project aimed to establish second-tier organizations at the village and mandal (county) level which, by pooling the resources of individual groups aimed to significantly expand their ability to respond to shocks. It is therefore possible to compare between women and men in the project area, as well as between pre-existing groups in areas affected by the project and in areas that were not eligible, to assess project impacts. To do so, we rely on a comprehensive set of surveys that included 2,700 households (with separate questionnaires being administered to men and women in the household), 2,200 self help groups, and 200 village organizations in both project and control areas conducted in mid-2004. In addition to the standard issues, the questionnaire elicited retrospective information on most issues so as to be able to compare differences in differences between project and control areas. It also contained an elaborate section on social empowerment, participation in groups, and actual as well as hypothetical borrowing capacity. The issues to be addressed are, first, to what extent the project succeeded in improving the governance structure of the community organizations involved;, second to what extent women have been empowered to overcome social barriers that have traditionally stymied their economic advancement, and finally to what extent the project has led to an increase in levels of income and productivity. We use simple t-tests of differences in differences to assess whether areas or groups (women) targeted by the project have advanced significantly more than others. For such differences to translate into significant improvements for the poor, it is necessary that the program is able to target poorer segments in the overall population or those who are organized in self-help groups. Participation regressions indeed confirm that, by focusing on the poorest areas and by aiming to include poorer parts of the population, the project has been able to significantly expand the range of households organized in self-help groups and to target support for institutional strengthening towards the most needy. Comparison between pre-existing groups in intervention and control areas points towards significantly different rates of improvements in terms of group management and internal controls (e.g. members being fined for non-attendance, internal bookkeeping being of high quality). These, together with the federated structure, appear to have allowed groups in intervention areas to significantly expand the availability of credit to members and to access loans by other financial institutions. Second-tier institutions in self help group federations were able to take on completely new activities (e.g. taking over distribution of subsidized food grains from "fair price shops" which were had often remained out of the reach of the poor) that significantly improved participants' ability to smooth consumption. This is supported by evidence from the household level suggesting that, even though the incidence of shocks was higher for households in treatment areas than outside, households were able to deal with such shocks more easily than they had been able to in the past. The hypothesis that the improvement in access to credit can be attributed to the project is supported by the finding that the unambiguous and significant increase in the amount which women in areas eligible for the project as compared to those that were not, were able to borrow both from the formal and the informal sector is not matched by a commensurate increase in credit availability for men. To the contrary, credit access for men was higher in areas not eligible for the project (marginally significant) as compared to areas that were eligible. The logic of the project to use improved access to resources to empower women and overcome social barriers is corroborated by the fact that the change in the share of women who receive high respect in their family and who were not subject to domestic violence was indeed significantly lower in control than in intervention areas where women also have significantly higher participation in family matters relating to income generating activities, debt and savings, as well as family planning and the number of children. In fact, the improvements in women's participation seem to transcend the realm of the family and extent to the community level: the change in the share of women who always know of or participate in village assemblies, who are aware of other types of community institutions, and who are able to freely interact with government officials and villagers of other caste or religion is significantly higher in intervention than in control villages. While all of this suggests that the project has not only improved access to credit and risk diversification but also significantly increased women's empowerment, these significant effects were, however, not matched by increased beneficiary savings, possibly because the resources generated by productive activities initiated under the project are yet to mature. To explore this, a closer look at the extent to which the project has increased access of the poor to resources and/or enabled them to use these resources more effectively. Even though productivity of resource use is similar between areas eligible and non-eligible for the project, there is clear evidence that the project has helped to significantly expand the share of households who own productive assets. The key challenge to ensure sustainability and replicability of the intervention is thus to match the rather impressive performance in terms of social empowerment with an equally significant transfer of technical skills that would, by facilitating more productive use of such assets, put participants economic basis on a stronger footing. The ability to use the federated structure to provide effective technical assistance and to liaise with line ministries to ensure that services are available to the poor, the feasibility of which has been demonstrated in a number of individual cases, is likely to be a key issue in doing so.