This is some dumb academic question, right? Coursework?

You can't just decide that bitmap isbetter without knowing how the 100 new rows a month are inserted, nor how frequently the values change. Knowing that each credit rating changes "infrequently" tells you nothing, when there's a million rows in the table.

In fact, an index might not even be beneficial in this case. What % of the rows are "BAD" or "F", and what % of blocks do they represent.?

I hate these stupid questions. Anyway, I'd guess that a function-based index might also be best. Or a b-tree.