I am not a cryptoanalyst but books on the subject typically assert that an N bit hash is superior to the concatenation of n-bit and m-bit hashes where n + m = N. I will try to look up more info later...

Update: Slightly misunderstood OP (but hey, length is just a weak hash digest anyway right!). There is a lovely example involving PostScript files which display distinct letters when viewed or printed but have same md5sum and same length.

More generally, here is a paper that proves the claim that concatenation of "different" hash digests will not increase security over a single hash digest of length equal to the concatenated length.

Jonathan J. Hoch and Adi Shamir, Breaking the ICE - Finding Multicollisions in Iterated Concatenated and Expanded (ICE) Hash Functions, proc. FSE 2006