Starting with 1.3, hash codes are cached. Given they they need to be calculted to put something into the Hashtable, there is no speed difference on retrieval of long vs. short keys. Prior to 1.3, they weren't cached and there would be a difference.

As another person mentioned, assuming you don't need the thready safety, HashMap is quicker as it is unsynchronized.

The speed of the Java hashing algorithm for strings should be roughly proportional to the string length, so shorter keys will buy you some speed. However, I would guess that a small fraction of the processing time is spent on the hashing algorithm in a typical case. If you need speed, you could probably gain more by switching from a synchronized Hashtable to an unsynchronized HashMap, given the relatively long time that the synchronization process takes.