coffee, these problems are a little harder than what you have in standard statistics. You may think about a problem: given 10 random events, I get 10 same outcomes, this is not likely BUT in this case, the criterion for selecting 10 random events is usually constructed post factum. That is, there are 100 ways to define events you plan to look at: those that were within 1%, or those in the South, or whatever. Then, the chance of having these 10 events having same outcome is subject to selecting these 10 according to some rule.
the way to test this is by dividing your data in (at least) two parts: use 1990s data to analyze and define how you select events, and then check ONCE on the data from 2000s.