We infer the redshift distribution of the faint blue galaxy excess (FEE) at mB = 23.5 by subtracting the predicted distribution of giant/normal galaxies from the observed N(z) distribution for all types. This is possible because of the recent deep Hubble Space Telescope (HST) WFPC2 morphological number counts that have convincingly demonstrated that little evolution of the giant population is seen to mB = 26.0. The mean redshift of the FBE at mB -23.5 is found to be 〈Z〉FBE= 0.40 ±0.07 with upper and lower quartiles defined by Z0.75 = 0.58 ±0.05 and z0.25 = 0.28 ±0.05, respectively. We compare this inferred FBE N(z) distribution to the predictions from three generic faint galaxy models: dwarf dominated (no-evolution), pure luminosity evolution, and evolving dwarfs. The inferred FEE N(z) distribution strongly supports a hybrid evolving dwarf-rich model wherein a large population of dwarfs present at z = 0.5 has subsequently faded to obscurity. The total integrated number density of dwarfs (down to MB = -11) is estimated to be a factor of 20 times greater than that of E-Sc galaxies and the estimated fading to be 1.0 < Δm < 1.4 mag. Thus, the dwarf population is estimated to be responsible for ∼30% of the luminosity density locally, rising to ∼57% at z = 0.5.