the way i phrase it is "static typing is a tradeoff, you get better refactoring tools in return for an upfront cost"
then i mumble "a cost of less expressive code" because god help me the people who harp on about static typing will get mad at me when i can point to row polymorphism, or threesomes with blame, or even just sage's hybrid type system to point out that your bog standard hindley milner is an appreciable subset of code amenable to deterministic static verification through abstract interpretation, rather than a "subset that just doesn't include type errors"
i don't say "it catches errors earlier" because there's type systems that handle dynamic typing (success typing) and also you catch errors in a dynamic language by just running the program, rather than abstractly evaluating the program
personally, the type errors that end up being caught late in a program's development are almost always implicit casts permitted by the type system
the irony is? i've hung around with haskell types like wadler and spj and mcbride and they don't seem to mind that i'm a dynamic type weenie
it's always the "i learned computer science and let me tell you how right i am about the tools i barely understand" types