Asking the annotator to select from a dropdown of 200 labels for every single annotation can easily slow down your process by a lot, and it'll make it much more difficult to get consistent data. It's very easy to forget a label, or spend a lot of time thinking about some fine-grained decision that doesn't even matter very much. The top-level categories are usually the ones that are most important – if you get those right, the fine-grained decisions will become much easier. If you get them wrong, the fine-grained decisions are often wasted.
I've explained a bit more about it in this thread: