Depending on the event, the answer is yes. Extreme events tend to be rare, but they occur on a regular basis. A case in point is the COVID-19 pandemic, which has been affecting our lives for the last year. For such events, we cannot predict exactly when they will occur, but they tend to exhibit a certain regularity, to such an extent that we can realistically determine things like how bad a 1-in-100-hundred-year event or a 1-in-1,000-year-event is likely to be. There is a well-developed branch of statistics called Extreme Value Theory that specializes in severe, high consequence events. A good application of this is building dikes or levees to prevent flooding. The Netherlands is a Western European country on the coast of the North Sea. About one third of the country is below sea level, so it depends on it system of dikes to prevent flooding. The Netherlands builds these tall enough to prevent a 1-in-1,000-year flood. They do not have thousands of years of data, but the size of floods has a regular enough pattern to make reasonable extrapolations given enough data. If the Netherlands were to build dykes to prevent average flooding events most of the country would be under water. See the graphic below for a comparison.
Some events have such wide variation in results that there is no typical member of the population. In such a case, sample means do not reflect the underlying population mean because there isn’t one! In such a case, extreme value statistics is the only meaningful comparison. This is the case with pandemics. The most meaningful representation of the data is as values that exceed specified thresholds. Such plots typically appear as lines when plotted on log-log graphs. See the graph below for the pandemic data in the tail. This is for 71 historical epidemics in which at least 1,000 people died. The data in the graph date back to 500 BC. The data have been adjusted to account for the increase in the world’s population over time. Note that the tail is so heavy that 40% of the historical pandemics appear in this tail.
Where does COVID-19 fit on this graph? It is in this tail. As of March 8, 2021, 2.6 million people around the world have died from COVID-19. Only one-third of historically recorded pandemics have resulted in higher fatality rates. Historically, in terms of global deaths, COVID-19 has killed more people than the Hong Kong flu of 1967 and 1968 that resulted in more than 2 million deaths (when adjusted for population growth), but not as many as the Asian flu of 1957 and 1958. So far, it has also resulted in far fewer deaths than the cumulative toll of the HIV/AIDS pandemic, with an estimated 30 million deaths (approximately 60 million when adjusted for population growth over time), or the Spanish flu, which is believed to have killed close to 200 million worldwide after taking population growth into account.
To predict what a 100-year event for pandemics looks like, we have to take into account that pandemics are becoming increasingly frequent. In the last thirteen years, there have been 11 pandemics that killed at least 1,000 people around the globe. This includes COVID-19 as well as two Ebola outbreaks in Africa, three episodes of cholera, and various flu and other infections disease outbreaks. The world is averaging approximately one pandemic each year. There are a variety of reasons attributed to the increasing frequency, with one of the primary ones being an increasing rate of global travel. In such a case, the 1-in-100-year-event is a pandemic that causes hundreds of millions of deaths, if not billions. Hopefully, this does not happen, but that is the kind of bad-day scenario that policies should guard against.