The article aims at isolating the effect of group participation on women empowerment using primary data on 1,500 individual women collected during 2007–2008 from two districts of West Bengal, India, namely, Hooghly and South 24‐Parganas. Since the impact evaluation exercise typically suffers from the problem of counterfactual, in the absence of biologically identical observations proxy has been constructed in terms of pairing statistically identical observations by applying matching techniques based on propensity scores. It is observed that mere inclusion in a self‐help group is not sufficient for any woman to enjoy the benefits of better connectivity with the social capital in terms of enhanced empowerment. Here both the duration of membership and the quality of participation matter. Moreover, in terms of a comparison of the probability of inclusion in the program with the actual proportion of inclusion for subjects with the same p‐scores, the extent of program mismatch has been assessed. This indicates a bias from the supply side where the more likely agents are being included in the absence of special effort to reach out for the relatively more vulnerable ones.