It’s not great. There are ~2 broad categories of language based on their use of affixes: analytic (which doesn’t make much use of affixes, instead relying on word order), and synthetic (which does use affixes). Synthetic has multiple sub-categories depending on how the affixes are used.

This approach completely ignores analytic languages which put their complexity in word order instead of affixes, and it ignores the distinctions within synthetic languages such as agglutinating, which stacks up affixes vs fusional, which combines multiple affixes into a single new affix.

It also misinterprets what people mean when they say languages which collide “reduce complexity” which really means they tend to become more analytic. Nobody expected the Scandinavian languages to have this characteristic because they’ve been in the process of diverging from a common language.

The process they’re referencing applies to languages which come from the collision of distinct languages, such as pidgins and creoles. The hypothesized tendency is for languages which have had consistent boundaries and populations for a long period of time to be more agglutinating then eventually fusional, and eventually polysynthetic like some Native American languages, where a single word can be translated to a complete sentence in English for instance

Edit: corrected for distinction between fusional and synthetic