The answer depends on the system. If your system is the He atom or the Li+ ion, then CCSD and CCSDT are exactly the same (they are both equivalent to FCI for 2 electrons). So the bigger basis set will provide you more accuracy than the higher-order coupled cluster method.
In general though, DZ basis sets are not considered very accurate on their own.
I don't know what your system is, but in my applications it would be more accurate to do CCSD/cc-pVDZ and CCSD/cc-pVTZ and then do a (DZ,TZ)->CBS extrapolation, than to do CCSDTQP/cc-pVDZ or even FCI/cc-pVDZ on its own.
However, since you say that TZ is beyond the effort that you can afford, everything that I said is moot. You can do CCSDT/cc-pVDZ and say that you didn't try CCSD/cc-pVTZ because it was too expensive.
Also, since this question isn't about MRCC, you might find more answers at MMSE:
mattermodeling.stackexchange.com/
with the "coupled cluster" tag. There's now more than 6100 users there!
