Caustic area biases and how to deal with them

Abstract

Quadruply imaged strongly lensed point sources (quads) are routinely used for measurements of the expansion rate of the Universe with time delays. It has recently been suggested that any quad lens is subject to a Malmquist-like bias that causes the inferred area enclosed within the tangential caustic to be systematically underestimated, and that such a bias might translate into a corresponding bias on the expansion parameter. In this work we extended that analysis by also considering the effect of Eddington bias. We find that the sign and amplitude of the combined bias depend on the functional form of the caustic area distribution of the lens population and on the noise associated with the caustic area estimation process. Based on simulations, we estimate that the corresponding impact on $H_0$ is of the order of a percent or smaller. If the likelihood of the lensing data is known, then the bias can be accounted for when modelling the lens population. However, ignoring the criteria used to select a quad might lead to a bias at the lens modelling stage that causes the inferred caustic area to be overestimated. Such a bias disappears for lens models that are well constrained by the data.