Existing models of regret aversion assume that individuals can make an ex-post comparison between their choice and a foregone alternative. Yet in many situations such a comparison can be made only if someone else chose the alternative option. We develop a model where regret-averse agents must decide between the status quo and a new risky option that outperforms the status quo in expectation, and learn the outcome of the risky option, if unchosen, with a probability that depends on the choices of others. This turns what was previously a series of single-person decision problems into a coordination game. Most notably, regret can facilitate coordination on the status quo - an action that would not be observed if the agents were acting in isolation or had standard preferences. We experimentally test the model and nd that regret-averse agents behave as predicted by our theory.