dierre: We did a lot of that kind of stuff at my previous job, but eventually found out that all scientific methods would essentially rise the cost of such migrations so much that it is actually cheaper to just do it. Simple, not completely scientific method which more or less worked for us was this: Take a feature, give it to several developers to develop independently from each other, half of those developers work using "old" technology, the other half uses the new one, measure whatever is your KPI, average within same technology. But key here is, the new technology group has already to know the technology for the results to be meaningful.
If you want to also measure ease of learning the new technologies, I'm not too optimistic, everyone learns with different speed, it is heavily influenced by past experience and attitude towards the new technology itself. And other fun quirks 🙂