More than two hundred classical cepheids were measured by the Hipparcos astrometric satellite , making possible a geometrical calibration of the cepheid distance scale .
However , the large average distance of even the nearest cepheids measured by Hipparcos implies trigonometric parallaxes of at most a few mas .
Determining unbiased distances and absolute magnitudes from such high relative error parallax data is not a trivial problem .

In 1997 , Feast & Catchpole announced that Hipparcos cepheid parallaxes indicated a Period-Luminosity scale 0.2 mag brighter than previous calibrations , with important consequences on the whole cosmic distance scale .
In the wake of this initial study , several authors have reconsidered the question , and favour fainter calibrations of cepheid luminositites , compatible with pre-Hipparcos values .

All authors used equivalent data sets , and the bulk of the difference in the results arises from the statistical treatment of the parallax data .
We have attempted to repeat the analyses of all these studies and test them with Monte Carlo simulations and synthetic samples .
We conclude that the initial Feast & Catchpole study is sound , and that the subsequent studies are subjected in several different ways to biases involved in the treatment of high relative error parallax data .
We consider the source of these biases in some detail .
We also propose a reappraisal of the error budget in the final Hipparcos cepheid result , leading to a PL relation – adapted from Feast & Catchpole – of

M _ { V } = -2.81 ( { assumed } ) \log P - 1.43 \pm 0.16 ( { stat } ) ^ { +0 } _ { -0.03 } ( { % syst } ) We compare this calibration to recent values from cluster cepheids or the surface brightness method , and find that the overall agreement is good within the uncertainties .

We conclude by commenting on the mismatch between the cepheid parallax distance scale and kinematical determinations , for cepheids as well as RR Lyrae .