Do long duration gamma ray bursts follow star formation?

We compare the luminosity function and rate inferred from the BATSE (Burst and Transient Source Experiment) long bursts peak flux distribution with those inferred from the Swift peak flux distribution. We find that the BATSE and the Swift peak fluxes can be fitted by the same luminosity function and the two samples are compatible with a population that follows the star formation rate. The estimated local long GRB rate (without beaming corrections) varies by a factor of five from 0.05Gpc^-3yr^-1 for a rate function that has a large fraction of high redshift bursts to 0.27Gpc^-3yr ^-1 for a rate function that has many local ones. We then turn to comparing the BeppoSax/HETE2 and the Swift observed redshift distributions and compare them with the predictions of the luminosity function found. We find that the discrepancy between the BeppoSax/HETE2 and Swift observed redshift distributions is only partially explained by the different thresholds of the detectors and it may indicate strong selection effects. After trying different forms of the star formation rate (SFR) we find that the observed Swift redshift distribution, with more observed high redshift bursts than expected, is inconsistent with a GRB rate that simply follows current models for the SFR. We show that this can be explained by GRB evolution beyond the SFR (more high redshift bursts). Alternatively this can also arise if the luminosity function evolves and earlier bursts were more luminous or if strong selection effects affect the redshift determination.