In the paper two different methods for the correction of self-selection bias in a panel Web surveys estimates are proposed and the obtained results are compared. The web panel is the result of the self-selection process of 134 hotels, which joined the Turi-Web data collection program of the Province of Florence (Italy). For the panel of respondents hotels the monthly presences of tourists are timely available by few days from the end of the reference month. The information provided is relevant for Local Government Agencies, but it is affected by self-selection bias. The adjusted estimates are obtained in two steps. At the first step generalized regression estimators (Särndal and Lundström, 2005) are here used as a tool to reduce the bias. Then the cross-sectional preliminary estimates for month t are adjusted by the method proposed by Rao et al (1989), which modifies the estimated level and change on the basis of the data available for the t - 1, t - 2,L, t - k previous months.

The results are compared with those obtained changing only the first step of

estimation by simply weighting the collected data by the reciprocal of the estimated response probability (Särndal and Swensson, 1987), and then treating the crosssectional preliminary estimates by the Rao et al. methodology.

The true monthly totals for the whole population of hotels are available after several months. This makes possible to quantify the effectual self-selection bias, the estimation error and the extent to which the two proposed adjustment methods are useful to correct the self-selection bias itself. The methodology is applied to monthly data collected by the Province of Florence from January 2005 to December 2006.