Police corruption, especially in the form of bribery, is a severe social problem in many societies. However, neither the extent nor the factors contributing to police bribery are well understood because of data limitation issues. Understandably, it is incredibly challenging to observe and quantify such bribery, as it is usually considered illegal and/or unethical for police to accept and/or ask for bribes. Agent-based modelling can solve such data limitation issues because it allows for the realistic modelling of hidden behaviours. This study uses an agent-based modelling technique to investigate a threshold model of police corruption, more specifically, bribery. The authors assume that agents have a threshold regarding bribery, which may be conceptualised as either an honesty threshold or a risk threshold. The threshold value is a dynamic variable randomly assigned to each agent, and each interaction between citizens and officers possesses the potential to change the threshold of each agent.