It is known that the polymerization reaction or crosslinking process is faster with increasing concentration of initiator. Faster reaction can cause defects and local shrinkage during crosslinking. These defects can lead to roughening of the surface which in turn leads to lower gloss.

Thank you, Subramaniam, for your answer.

The previous answer on roughness is correct.  To expand a bit, gloss depends on basically two properties:  refractive index (RI) of the polymeric medium, and degree of planarity of the surface on a micro scale. 

The refractive index issue is not simple, because mixtures can be employed, and because the size of the secondary component can have a significant effect on the perceived overall refractive index.  So one needs to know the refractive index of the homopolymers of the various components, including cross-linkers.  if any of them are significantly different, that will degrade the attainable gloss.  Note that it doesn't matter whether the difference is positive or negative -- only the magnitude of the difference matters.  One can estimate the effect of monomer mixes by a weight-average of the components' respective RI.  As long as the domain sizes remain below a few hundred nano-meters, then domain probably doesn't matter much.  Bottom line of all this:  think of a compositional-roughness factor, which is separate from a physical-roughness factor. 

The physical roughness factor is separately important, as already mentioned. 

It would be useful to record gloss at various incident angles:  20 deg, 60 deg, 85 deg, all measured from the vertical.  The lower the angle, the greater will be the effect of the two "roughness" factors described above and mentioned earlier.  Commonly, a given industry will focus on one of these measurements -- for example, automotive focuses on the 20 deg reading plus an additional distinctness-of-image measurement that is only relevant to very high gloss coatings. 

I would advise that you measure the compositional-heterogeneity separately from the physical roughness.  The latter can be easily measured optically by using interference measurements of reflected images from the surface.