Evaluations on list undercoverage bias and possible solutions: the case of ISTAT CATI survey "Trips, holidays and daily life"

The list of fixed line telephone numbers, used as sampling frame for CATI surveys on households, suffers of an increasing undercoverage rate. Some methodological studies conducted on ISTAT CATI household surveys have pointed out the risk of biased estimates for particular phenomena of interest due to this type of frame imperfections. A research group in Istat has been set up and has highlighted a risk of estimation bias due to the undercoverage effect for the Trips, holidays and daily life survey, for which data are collected by means of CATI technique on a sample selected from the phone directories. In the paper two possible approaches to pursue unbiased estimates are described. The first one is based on the enhancement of the estimator aiming at taking into account the frame imperfections. An experiment on real survey data shows a good improvement on the undercoverage bias of the estimates choosing more properly the calibration variables. The second approach defines a different survey strategy, based on a different frame not suffering from undercoverage (or most likely less affected by undercoverage), the set of municipal population registers, together with a new sampling design and a mixed mode data collection.