In recent years, microfinance institutions are seen as beacons of hope to help eradicate poverty through, among others, providing credit to poor rural households. Availability of small but repeated loans is, in the long-term, expected to impact on poverty. However, decades after the introduction of microfinance institutions into many rural areas, there are still questions as to what extent such long-term benefits are realized. This is because evaluating the long-term impact of microfinance provision on household welfare is difficult due to difficulties in controlling for heterogeneities in the borrower pool and subsequent borrowing dynamics. We use four rounds of panel data on rural households in Ethiopia that had access to microfinance credit to assess the long-term impact of credit by explicitly accounting for such borrowing dynamics. In a panel data setting where only the outcome variable is time-varying, potential future paths of individuals in the control group are considered to account for counterfactuals. Matching is used to adjust for initial differences between participants and controls. These combined methodological innovations enable us to overcome biases due to selection as well as problems of accounting for dropouts and new participants in microfinance impact assessments. Results suggest that the timing of membership matters: the earlier the onset of membership the better the effect. Results are robust compared to standard matched pairwise effects. The overriding implication from this comparison suggests that not accounting for future counterfactuals, for the most part, can lead to biased estimates and overestimates of the impact of credit on household welfare.