It is recognised that a better understanding of the job preferences of health workers is needed to inform policies intended to retain them in their roles. This has led to a growing interest in advancing and applying methods to study preference heterogeneity in order to acknowledge differences in health provider characteristics that are known to affect their labour market choices. Many countries dealing with human resource shortages depend on community-based workers to improve healthcare coverage. This thesis analysed the job preferences of community-based healthcare workers in Ethiopia and Ghana, to understand heterogeneity in their preferences and how it can be modelled using multiple methods. Primary data were collected to qualitatively explore the job preferences of community health workers in Ethiopia. Secondary datasets with unlabelled discrete choice experiments reproducing the roles of community health workers in Ethiopia and community volunteers in Ghana were analysed to elicit stated preferences for financial and non-financial job attributes. Sources of heterogeneity in preferences were explored by incorporating individual characteristics and psychological constructs in choice models. The difference between decision making heuristics and preference heterogeneity in the analysis of discrete choices was also examined. Community health workers in Ethiopia were found to strongly prefer non-financial incentives in their jobs, and motivation was found to be an important source of preference heterogeneity. The need for characterising well-defined and relevant attributes in a DCE was also highlighted, to ensure that heuristics in decision making do not get confused with preference heterogeneity. Similarly, in Ghana, non-financial incentives were also found to be very important to respondents. Three groups of health workers with heterogeneous job preferences for role incentives were identified in the dataset. This thesis contributes to the limited evidence on the job preferences of community-based healthcare workers, alongside sources and ways to model preference heterogeneity. This knowledge is important to inform policies in resource-constrained settings that rely heavily on lay workers for primary healthcare delivery.