I’m sorry but the research you cite is simply irrelevant. Putting C++ on the same graph as those other languages simply doesn’t make any sense. You simply don’t use that language for the same kind of tasks as the others! Generalizing this argument, I wouldn’t be surprised to find a correlation between the difficulty of the programming task and the likelihood of choosing a language with a stronger type system. Statistics is just using numbers to lie in an elaborate manner, unless you rigorously take hidden variables into account.
On an orthogonal note; people usually relate dynamic/static types as a trade-off between productivity/safety. So, then, it’s relevant to compare how writing a lot of tests to compensate for the lack of types impacts productivity. I also find it much more pleasing to read and write type signatures rather than tests.