<p>Taxi is one of the most popular public transportation systems that is commonly available in many cities. It allows passengers to travel to their destinations conveniently. Nonetheless, traditional street-hail taxi systems are known to have low efficiency. Passengers often have to wait for a long time, while many vacant taxis spend a considerable amount of time searching for passengers. This inefficiency is a result of an imbalance between supply and demand. Understanding the statistical distributions of taxi supply and demand is vital to solving this imbalance problem. Moreover, to realistically simulate the number of vacant and busy taxis with a traffic simulator, it is necessary to know the types of distributions that can properly represent the empirical variation of taxi supply and demand. In our previous work, we have successfully characterized the temporal distribution of taxi demand. In this study, we investigate the supply side and characterize the temporal distribution of vacant taxis. Contrary to the conventional belief, we show that modeling the temporal distribution of taxi supply with a Poisson distribution is mostly invalid. Finally, we demonstrate that a geometric distribution is suitable for modeling the distribution of taxi supply in most scenarios.</p>