Do you ever find yourself sitting through yet another big-studio blockbuster, wondering when it’s going to end? I have been feeling this feeling a lot lately. Is the 2+ hour running time a standard since movies immemorial, or have they been slowly getting longer? Naturally, I had to do some data analysis to find out.
Using publicly-available data from Wikipedia, I scraped and compiled all top box office movies released in the United States and charted them.
This scatter plot shows clearly that most movies cluster within the 80-120 minute range, and that there is a stronger “floor” to running time than a “ceiling,” which may be due to the fact that in order to be called a “feature film” a movie must have a minimum running time (but no maximum).
However, there may be some hidden dimensions to these markers. Namely, how many movies are clustered within any given area? To cut through the noise, I need to plot the average running time.
It’s fairly clear that running time has mostly stayed consistent over the decades. The month-to-month oscillations in time may be due to just one or two movies with outlier values that drag the average up or down. That difference is mostly erased across years.
Only in the last several years has the annual average begun to change noticably. So, when I first began noticing this around 2015, average movie time was indeed climbing!
Before I could be fully satisfied, I had to check one more dimension of the data: the monthly and yearly median.
Notice how the average running time is consistently longer than the median running time, by several minutes? When average and median are different, it generally means that there are data points that “drag” the average value away from the “true middle” of the median. Looking at this, my interpretation is that even though most movies fall in the 1-2 hour range, there are always a few super long ones that “drag” the running time closer to 2 hours. Looking at the dots, it appears to scan: the area below the median line is denser with dots than the area above.