In relational contracting the threat of punishment in future periods provides an incentive to not to cheat. However, to what extent do people actually carry out this punishment? We compare relational contracting patterns in Ghana and the United Kingdom by conducting a repeated principal agent lab experiment, framed in a labour market setting. Each period, employers make offers to workers, who can choose to accept or reject this offer and, after accepting and being paid, what effort to exert. The employers and workers interact repeatedly over several periods. While in the UK, in line with theoretical predictions and previous experiments (e.g. Brown et al., 2004, 2012), high effort is rewarded and low effort punished, we do not find evidence for the use of such incentives in Ghana. As a result, employers fail to discipline a subgroup of “selfish” workers, resulting in a low average effort and low and often negative employers’ earnings. Set identification of Fehr-Schmidt preferences of the Ghanaian and British workers shows that the share of “selfish” workers in our experiment in Ghana is not substantially different from the UK. Introducing competition for workers or a reputation mechanism does not significantly improve workers’ effort.