Studies of early-type galaxy evolution compare the properties of early-type galaxies at high redshift to those of nearby ones. The implicit assumption is that the high redshift galaxies are drawn from the same distribution as the low redshift galaxies, so that their properties can be compared directly.

However, the morphologies of galaxies may evolve, such that some present-day early-types were spiral galaxies or merger systems at high redshift. These young early-types are included in the low redshift sample, but drop out of the sample at high redshift. Therefore, the high redshift sample is a biased subset of the low redshift sample, containing only the oldest progenitors of today's early-types.

The effect of this "progenitor bias" is that observations underestimate the true evolution of early-type galaxies:

As we show in van Dokkum & Franx (2001) the observed redshift evolution of the mean luminosity and color of early-types, and the scatter in these properties, can mimic that of a stable population formed at very high redshift even if young early-types are continuously added to the sample. One of the manifestations of progenitor bias is that a sample of early-type galaxies at z=1 has a higher mean formation redshift than a sample of early-types at z=0.5, and that the scatter in age can decrease at higher redshift:

In van Dokkum & Franx (2001) we quantified the effects of progenitor bias with a model, aiming to minimize the number of free parameters. This calculator uses this model to create a table of mean luminosities and colors, and the scatter in these properties, as a function of redshift. It also provides the early-type galaxy fraction as a function of redshift, and the mean formation redshift of the early-types present at redshift z.