We study how women's choices over labor activities in village economies correlate with poverty and whether enabling the poorest women to take on the activities of their richer counterparts can set them on a sustainable trajectory out of poverty. To do this we conduct a large-scale randomized control trial, covering over 21,000 households in 1,309 villages surveyed four times over a seven year period, to evaluate a nationwide program in Bangladesh that transfers livestock assets and skills to the poorest women. At baseline, the poorest women mostly engage in low return and seasonal casual wage labor while wealthier women solely engage in livestock rearing. The program enables poor women to start engaging in livestock rearing, increasing their aggregate labor supply and earnings. This leads to asset accumulation (livestock, land and business assets) and poverty reduction, both sustained after four and seven years. These gains do not crowd out the livestock businesses of noneligible households while the wages these receive for casual jobs increase as the poor reduce their labor supply. Our results show that: (i) the poor are able to take on the work activities of the non-poor but face barriers to doing so, and, (ii) one-off interventions that remove these barriers lead to sustainable poverty reduction. JEL Classification: J22, O12. -5001). All errors remain our own.Understanding whether and how governments can take up these programs and whether they can be adapted to urban settings are all unknowns that will have a critical bearing on whether this idea spreads and scales. The juxtaposition of the goal of eliminating extreme poverty by 2030 and the promising set of initial results in this and related papers does, however, suggest that taking up these research challenges would be a worthwhile endeavor.Notes: All figures are derived using the baseline household survey and present statistics on the three main occupations: domestic maid (red), agricultural labor (blue), livestock rearing (green), and other (white). Panel A shows the share of hours devoted to the different occupations by BRAC branch, ordered by the share of hours devoted to casual labor in agriculture. Panel B shows the share of hours devoted to the different labor market activities by wealth class. Panel C shows the hourly returns to the different occupations by BRAC branch, ordered by returns to livestock rearing. For each activity, earnings per hour are calculated as total earnings from that activity divided by total hours worked in the activity, both defined over the year prior to the baseline survey for individuals who had positive hours and non-missing earnings in that activity. Panel D graphs local polynomial regressions of the hourly returns to activities by the value of livestock owned. The vertical lines correspond to the average value of livestock owned by the ultra-poor pre-and post-intervention. All monetary amounts are PPP-adjusted USD terms, set at 2007 prices and deflated using CPI published by Bangladesh Bank. In Notes: Quantile treatment...