There's been some work done on this in the US context (eg [0]) but not really enough. It seems like general large project issues (need to adapt to site specific conditions as they are discovered, difficulty ensuring work stays properly scheduled so that construction workers don't sit idle for large amounts of time, etc.) are a big part, and these are made worse because of the knock on impact when these interact with safety regulations/processes.

There's some hope that making designs smaller and more modular might enable less chance of big overspend at a slightly higher expected cost but that's speculation until we actually do it.

0. https://www.sciencedirect.com/science/article/pii/S254243512...