Metal nanoparticles and diffractive nanostructures are widely studied for enhancing light trapping efficiency in thin-film solar cells. Both have achieved high performance enhancements, but there are very few direct comparisons between the two. Also, it is difficult to accurately determine the parasitic absorption of metal nanoparticles. Here, we assess the light trapping efficiencies of both approaches in an identical absorber configuration. We use a 240 nm thick amorphous silicon slab as the absorber layer and either a quasi-random supercell diffractive nanostructure or a layer of self-assembled metal nanoparticles for light trapping. Both the plasmonic and diffractive structures strongly enhance the absorption in the red/near-infrared regime. At longer wavelengths, however, parasitic absorption becomes evident in the metal nanoparticles, which reduces the overall performance of the plasmonic approach. We have formulated a simple analytical model to assess the parasitic absorption and effective reflectivity of a plasmonic reflector and to demonstrate good agreement with the experimental data.