A widening evidence base across low- and middle-income countries (LMICs) points towards mutually reinforcing linkages between poverty and mental health problems. The use of validated and culturally relevant measures of mental health outcomes is crucial to the expansion of evidence. At present, there is a paucity of measures that have been tested and validated in contexts of extreme poverty. Using data from adult women living in extreme poverty in rural Haiti this study assesses the cross-cultural validity of the widely used Rosenberg Self-Esteem Scale (RSES) and its applicability in assessing linkages between poverty and mental health outcomes. We find no evidence for a one-dimensional 10-factor structure of the RSES within our data and agree with other authors that the standard self-esteem model does not fit well in this cultural context. Comparisons with another widely used measure of mental health – the K6 measure – indicate that the RSES cannot be used as a proxy for mental health outcomes. We conclude that the use of the RSES in different cultural contexts and with samples with different socioeconomic characteristics should be undertaken with caution; and that greater consideration of the validity of psychosocial constructs and their measurement is vital for gaining robust and replicable insights into breaking the cycle between poverty and mental health problems.