Is the radius valley a fossil of timing?

We queried the NASA Exoplanet Archive for planets with a measured radius, orbital period and stellar age, and kept 0.5–4 Earth radii and periods under 100 days. That gives 2,713 planets. We estimated the radius distribution in four age bins with a Gaussian kernel, located the minimum between the super-Earth and sub-Neptune peaks, and bootstrapped both the valley radius and its depth. The comparison of interest is stars younger than 1 Gyr against stars older than 5 Gyr.

• The valley moves the predicted way. Its radius falls from 2.24 R⊕ in the youngest bin to 1.93 R⊕ in the oldest. The difference is −0.24 R⊕ with a 95% bootstrap interval of −0.62 to +0.31 (p = 0.22).
• It deepens. Valley depth rises from 0.14 in the young bin to 0.37 in the old bin. The difference is +0.17, with an interval of −0.26 to +0.38 (p = 0.22).
• Neither shift is resolvable. Both point estimates are in the direction the fossil model predicts and the steady-state model does not. With 2,713 planets and literature ages that carry one-to-several-Gyr errors, the test does not separate the models.

This is the one result in the set that is neither a pass nor a failure. The distinctive prediction — age dependence — is visible in the point estimates, but the confidence intervals are wide enough to contain no effect. Calling it support would overstate the data; calling it a refutation would overstate the test. It is a question that a larger homogeneous age sample, or a cluster-based comparison, could decide.