We present results of a search for bright Lyman break galaxies (LBGs) at 1.5 ≤ z ≤ 2.5 in the GOODS-S field using an NUV-dropout technique in combination with color selection. We derived a sample of 73 LBG candidates. We compare our selection efficiencies to BM/BX and BzK methods (techniques solely based on ground-based data sets), and find the NUV data to provide greater efficiency for selecting star-forming galaxies. We estimate LBG candidate ages, masses, star formation rates, and extinction from fitting PEGASE synthesis evolution models. We find that about 20% of our LBG candidates are comparable to infrared-luminous LBGs or submillimeter galaxies which are thought to be precursors of massive elliptical galaxies today. Overall, we can show that although BM/BX and BzK methods do identify star-forming galaxies at z ~ 2, the sample they provide biases against those star-forming galaxies which are more massive and contain sizeable red stellar populations. A true Lyman break criterion at z ~ 2 is therefore more directly comparable to the populations found at z ~ 3, which does contain a red fraction.