Just another data point: I have written largely the same tetris program both in Haskell and C++.
At first, I was blown away how short the Haskell one was, but later someone wrote that actually, world-count-wise they were largely the same, and indeed he was right.
So in certain cases, ML languages are not terser, just more wide, rather than long.
Length of code is probably the wrong metric. What I mean is readability or intelligibility or code. Long code is rarely readable (too many files etc) but short code can be hard to read too (e.g. code golf). However, the most readable solution will tend to be short.
At first, I was blown away how short the Haskell one was, but later someone wrote that actually, world-count-wise they were largely the same, and indeed he was right.
So in certain cases, ML languages are not terser, just more wide, rather than long.