Abstract
Let us say that a coincidence involving two events, where one seems to predict the other, happens a number of times. How many times until it can be considered not only a coincidence, but a statistically significant pattern? We propose a framework to answer this question. Using the framework, we find that the number of times required is 7. We illustrate the practical application of our framework in the context of a very important phenomenon: When the percentage difference between 10-year and 3-month U.S. Treasury yields falls below zero, a U.S. recession appears to occur within the next 18 months.