Does low trust in workers discourage firms from hiring? We conduct an experiment in Ghana with real entrepreneurs who have the option to hire anonymous workers for a trivial but tedious task. Shirking attracts no penalty and completion of the task is an indicator of trustworthiness. We elicit employers' expectations and study how they change with random signals of workers' previous behaviour. We find that employers underestimate workers' trustworthiness, which reduces hiring and profits. Negative signals lower employers' expectations, while positive signals do not affect them. This asymmetry can help to sustain an equilibrium with limited experimentation and biased beliefs.