Cosmic Star-Formation History, as traced by Radio Source Evolution

I briefly review our current knowledge of the cosmological evolution of radio sources, and show that the redshift distributions of new complete samples of radio sources confirm the existence of the high-redshift decline in comoving number density (or `cutoff') beyond $z \simeq 2.5$ first deduced by Dunlop & Peacock (1990). Taken at face value these new data favour a luminosity dependent `cutoff', in which the decline is least drastic for the most luminous radio sources. I demonstrate, however, that regardless of the precise form of radio source evolution, the evolution of radio luminosity density is well determined, and appears uncannily similar to the evolution of ultra-violet luminosity density from star-forming galaxies in the universe (e.g. Madau 1997). I convert radio luminosity density into an estimate of black hole fueling rate per Mpc$^3$, and conclude that radio source evolution is a good tracer of the star-formation history of the Universe; at any epoch, for every 10$^7$ M$_{\odot}$ of material converted into stars, approximately 1 M$_{\odot}$ appears to be consumed by radio-loud active galactic nuclei. If this is indeed true at all epochs, then this would imply that star-formation activity in the Universe peaked at $z \simeq 2 - 2.5$, and that the values of star-formation rate density currently derived from Lyman-limit galaxies at $z \simeq 2.8$ and $z \simeq 4$ are under-estimates by a factor $\simeq 4$. Finally I briefly speculate as to why radio source evolution might trace global star-formation activity as manifested primarily in the disc/spiral population, whereas the hosts of the radio sources themselves generally appear to be well-evolved elliptical galaxies.