Evidence on the long-term impacts of microfinance credit is scarce. We use a unique four-round panel dataset on farm households in northern Ethiopia that had access to microfinance, observed on two key poverty indicators: household consumption and housing improvements. Fixed-effects and random trend models are used to reduce potential selection biases due to time-invariant unobserved heterogeneity and individual trends therein. Results show that borrowing indeed causally increased consumption and housing improvements. A flexible specification that takes into account repeated borrowings also suggests that borrowing has cumulative long-term effects on these outcomes, implying that short-term impact estimates may underestimate credit effects. Copyright 2010, Oxford University Press.