In a previous paper we showed that the radio sources selected by combining large areas radio and optical surveys , present a strong deficit of radio emission with respect to 3CR radio-galaxies matched in line emission luminosity .
We argued that the prevalence of sources with luminous extended radio structures in high flux limited samples is due to a selection bias .
Sources with low radio power form the bulk of the radio-loud AGN population but are still virtually unexplored .

We here analyze their photometric and spectroscopic properties .
From the point of view of their emission lines , the majority of the sample are Low Excitation Galaxies ( LEG ) , similar to the 3CR objects at the same level of line luminosity .
The hosts of LEG are red , massive ( 10.5 \lesssim { log } M _ { * } / M _ { \odot } \lesssim 12 ) Early-Type Galaxies ( ETG ) with large black holes masses ( 7.7 \lesssim { log } M _ { BH } / M _ { \odot } \lesssim 9 ) , statistically indistinguishable from the hosts of low redshift 3CR/LEG sources .
No genuine radio-loud LEG could be found associated with black holes with a mass substantially lower than 10 ^ { 8 } M _ { \odot } or with a late type host .
The fraction of galaxies with signs of star formation ( \sim 5 \% ) is similar to what is found in both the quiescent ETG and 3CR/LEG hosts .
We conclude that the deficit in radio emission can not be ascribed to differences in the properties of their hosts .
We argue that instead this could be due to a temporal evolution of the radio luminosity .

A minority ( \sim 10 \% ) of the sample show rather different properties , being associated with low black hole masses , with spiral galaxies , or showing a high excitation spectrum .
In general these outliers are the result of the contamination from Seyfert and from galaxies where the radio emission is powered by star formation .
For the objects with high excitation spectra there is no a clear discontinuity in either the host or nuclear properties as they span from radio-quiet and radio-loud AGN .