Is the data type `numeric(p,s)` accurate?

I know that numeric is an "exact" data type, representing numbers as accurately as possible so that there are no floating-point errors such with float or real data types. But what about numeric(p,s)? It must approximate a number because of the fixed scale, but does it have floating-point inaccuracies or is it “safe”? I could not find any information about it. Maybe it is obvious, but I would like to understand this aspect better.