I find this traditional treatment of referencing contractionary periods to be quite misleading. For instance, notice how the 2007-2009 Great Recession is shown to be as important as the 1981-1982 recession. But the 1981-1982 recession was only mild by comparison to the 2007-2009 recession, so it is a mistake to encode these two as equally important.
The problem with the traditional approach is that these areas encode the length of the recession and not their severity. As an alternative approach, please see my analysis at Graph Workflow on how one can encode both the length and the severity of a recession: https://graphworkflow.com/2019/11/04/contractions-recoveries/